Systems and methods for molecular quantum simulation using NMR hardware

A hybrid quantum-classical framework using NMR hardware and classical computers addresses the challenges of modeling chemically complex systems by optimizing quantum states across geometrical configurations, providing accurate chemical property predictions through reduced quantum-derived data and machine-learning calibration.

WO2026234575A2PCT designated stage Publication Date: 2026-11-12KIMCHY YOAV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/IB2026/054458
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-08-24
Filing Date
2026-05-05
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Existing computational methods struggle to accurately model chemically complex systems, such as those involving transition metals and correlated electrons, due to high computational cost and the inability to capture quantum mechanical effects like electron correlation and intermolecular interactions, leading to impractical and inaccurate predictions in fields like drug discovery and materials science.

Method used

A hybrid quantum-classical computational framework using NMR quantum hardware and classical computers to perform molecular simulations, where parameterized quantum states are optimized and measured across multiple geometrical configurations, generating reduced quantum-derived data transformed into orbital-basis data to predict chemical properties, incorporating machine-learning calibration for enhanced accuracy.

Benefits of technology

This approach enables efficient and accurate prediction of chemical properties by capturing electronic structure variations and intermolecular interactions, overcoming scalability and accuracy limitations of existing methods, particularly in drug discovery and materials science applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure IB2026054458_12112026_PF_FP_ABST
    Figure IB2026054458_12112026_PF_FP_ABST
Patent Text Reader

Abstract

A method and system for predicting a chemical property of a molecular or materials system using hybrid quantum-classical computation are disclosed. Electronic structure calculations are performed across a plurality of geometrical configurations to generate corresponding representations, including reduced active-space Hamiltonians. A parameterized quantum state is optimized using quantum hardware and evaluated across the plurality of configurations to obtain measurement data. Reduced quantum-derived data, including reduced density matrices or quantum observables, are generated and transformed into an orbital basis using electronic structure information. Features capturing both values and variation of quantum observables across configurations are extracted and used to predict the chemical property, optionally via a trained predictive model. In certain embodiments, fixed optimized parameters are reused across configurations to extract configuration- dependent quantum responses. The approach enables incorporation of correlation effects beyond reduced active spaces and supports scalable implementation using distributed quantum processors, including Nuclear Magnetic Resonance (NMR) systems.
Need to check novelty before this filing date? Find Prior Art

Description

ISUS Ref: 53524-006PCTSYSTEMS AND METHODS FOR MOLECULAR QUANTUM SIMULATION USING NMR HARDWARECROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Application No. 63 / 800,556, titled “A Hybrid Classic Quantum Computer,” filed May 6, 2025; U.S. Provisional Application No. 63 / 807,675, titled “An Enhanced Hybrid Classic Quantum Computer,” filed May 18, 2025; U.S. Provisional Application No. 63 / 812,051, titled “A Classically Coordinated Hybrid Quantum Computer,” filed May 26, 2025; U.S. Provisional Application No. 63 / 848,343, titled “Nuclear Magnetic Resonance Quantum Computing System for Accelerated Deuterated Drug Development and Optimization,” filed July 22, 2025; and U.S. Provisional Application No. 63 / 869,345, titled “Solid-State NMR Quantum Computing System for Materials Science Applications,” filed August 24, 2025, each of which is hereby incorporated by reference in its entirety.TECHNICAL FIELD

[0002] The present disclosure relates to quantum computing, molecular simulation, and computational chemistry in general, and to systems and methods for predicting molecular properties using Nuclear Magnetic Resonance (NMR) quantum hardware in combination with classically generated molecular Hamiltonians, in particular.BACKGROUND

[0003] Accurate prediction of molecular properties remains an important challenge in pharmaceutical research, chemical catalysis, and molecular design. In many practical settings, computational workflows are used to evaluate molecular structure, intermolecular interactions, and electronic behavior in order to support compound screening, optimization, and selection. Such workflows may be applied, for example, to drug discovery programs, catalyst development, and related chemistry-driven applications.

[0004] Existing computational approaches may include empirical force-field methods, density functional theory (DFT), hybrid quantum-mechanical / classical methods, and other electronic-structure techniques. These approaches can provide usefulISUS Ref: 53524-006PCTapproximations in many contexts, but the computational cost and modeling complexity may increase substantially for chemically complex systems, including systems involving transition metals, correlated electrons, bond-formation or bond-breaking coordinates, or other interaction motifs of interest in medicinal and catalytic chemistry.

[0005] Separately, quantum computing technologies have been investigated for use in molecular simulation and related computational tasks. Certain quantum processing platforms enable manipulation and measurement of quantum states under practical operating conditions. Such platforms may be used in combination with classical computing resources in hybrid computational workflows, in which at least part of a problem is represented in a form suitable for execution on quantum hardware.

[0006] There remains a need for improved computational frameworks that can combine classical molecular modeling with quantum-hardware-based measurements in a practical and scalable manner for molecular property prediction. There is also a need for such frameworks to be adaptable to different chemical classes and application domains, including pharmaceutical compounds, catalytic systems, and other molecular systems of interest.ISUS Ref: 53524-006PCTBRIEF SUMMARY

[0008] One exemplary embodiment of the disclosed subject matter is a method for predicting a chemical property of a molecular system. The method comprises performing, using a classical computer, an electronic structure calculation for the molecular system to obtain molecular orbital information corresponding to a plurality of geometrical configurations defined along at least one interatomic coordinate in the molecular system. The method further comprises generating a plurality of representations of the molecular system corresponding to the plurality of geometrical configurations. The method comprises optimizing, using a quantum hardware, a parameterized quantum state with respect to each of the plurality of representations, thereby obtaining a set of optimized variational parameters. The method comprises performing, using the quantum hardware, quantum measurements with respect to the plurality of representations based on the set of optimized variational parameters, thereby obtaining measurement data associated with the plurality of geometrical configurations. The method comprises generating, based on the measurement data, a set of reduced quantum -derived data associated with the plurality of geometrical configurations. The method comprises transforming, using the classical computer, the reduced quantum-derived data into orbital-basis data, based on the molecular structure information corresponding to the plurality of geometrical configurations of the molecular system. The method comprises extracting features based on the orbital-basis data, including variation of the features across the plurality of geometrical configurations. The method comprises predicting the chemical property based on the extracted features.

[0009] Optionally, the plurality of representations of the molecular system comprises a plurality of Hamiltonians. Each Hamiltonian represents an electronic structure of the molecular system at a respective geometrical configuration. Optionally, the plurality of Hamiltonians corresponds to a plurality of interatomic distances along a bond formation or dissociation coordinate. Optionally, each Hamiltonian encodes interactions derived from a full electronic structure of the molecular system and projected onto a reduced representation. Optionally, each Hamiltonian is represented as a linear combination of Pauli operators suitable for execution on quantum hardware.

[0010] Optionally, the Hamiltonians encode contributions from electronic degrees of freedom outside a reduced active space, including contributions associated with virtualISUS Ref: 53524-006PCTorbitals that are not explicitly included in the reduced active space but are incorporated into the Hamiltonians via an effective representation. Optionally, each Hamiltonian is constructed based on a reduced active-space representation of the molecular system, the reduced active-space representation being derived from a full electronic structure of the molecular system. Optionally, the reduced active-space representation comprises a subset of molecular orbitals including at least one occupied orbital and at least one unoccupied orbital. Optionally, the reduced active-space representation comprises at least one of a two-electron, two-orbital active space. Optionally, each Hamiltonian of the plurality of Hamiltonian has eigenvalues that correspond to energies of the molecular system at a corresponding geometrical configuration.

[0011] Optionally, the plurality of geometrical configurations comprises a plurality of interatomic distances along a bond formation or dissociation coordinate. Optionally, generating the plurality of Hamiltonians comprises performing electronic structure calculations for each geometrical configuration; and constructing a reduced active-space Hamiltonian. Optionally, the electronic structure calculations comprise Density Functional Theory (DFT) calculations configured to determine molecular orbital coefficients and electronic integrals for each geometrical configuration.

[0012] Optionally, the reduced active-space Hamiltonian comprises a CASCI(2e,2o) representation. Optionally, the Hamiltonians are mapped to qubit representations comprising Pauli operator coefficients. Optionally, the Hamiltonians encode contributions from electronic degrees of freedom outside the reduced active space. Optionally, the contributions from electronic degrees of freedom outside the reduced active space comprise virtual orbital contributions associated with unoccupied molecular orbitals that are not explicitly included in the reduced active space. Optionally, the virtual orbital contributions represent electron correlation effects arising from interactions between occupied orbitals in the reduced active space and unoccupied molecular orbitals outside the reduced active space.

[0013] Optionally, optimizing the parameterized quantum state comprises selecting a representation of the molecular system (e.g., a Hamiltonian) corresponding to an equilibrium configuration of the molecular system, wherein the equilibrium configuration corresponds to a configuration associated with a minimum energy of the molecular system along at least one interatomic coordinate; and optimizing, using the quantumISUS Ref: 53524-006PCThardware, a parameterized quantum state with respect to the selected Hamiltonian to obtain optimized parameters.

[0014] Optionally, using the parameterized quantum state to perform quantum measurements comprises maintaining the optimized parameters fixed; and for each of the plurality of representations of the molecular system (e.g., Hamiltonians) preparing a quantum state using the optimized parameters; and performing quantum measurements to obtain measurement data. Optionally, the optimized parameters approximate a ground state of the molecular system at the equilibrium configuration. Optionally, maintaining the optimized parameters fixed across the plurality of Hamiltonians enables extraction of configuration-dependent quantum responses. Optionally, optimizing the parameterized quantum state comprises iteratively updating parameters using a classical optimizer until convergence or until a maximum iteration threshold is reached.

[0015] Optionally, optimizing the parameterized quantum state the quantum state comprises performing a Variational Quantum Eigensolver (VQE). Optionally, optimizing the parameterized quantum state minimizes an expected energy of the form: (y(0)|H(R_eq)|\| / (0)). Optionally, the quantum hardware comprises an NMR quantum processor, wherein VQE is executed on the NMR quantum processor. Optionally, the NMR quantum processor performs analog quantum evolution based on nuclear spin interactions, thereby capturing contributions from electronic degrees of freedom beyond a programmed active space.

[0016] Optionally, optimizing the parameterized quantum state comprises generating the variational parameters via a classical Tensor-Train (TT) hypernetwork, the TT hypemetwork comprising a sequence of low-rank tensor cores {G_k}_{k=l}A{K} that map a classical input to the rotation angles defining the parameterized quantum circuit. Optionally, the trainable parameters reside entirely in the classical tensor-train cores, and the NMR quantum processor operates solely as a fixed forward-pass evaluator that provides mixed-state observables for evaluating the expected energy (y(0)|H(R_eq)|\| / (0)). Optionally, gradients with respect to the tensor-train cores are computed classically by backpropagation, while the analog NMR hardware functioning as a molecular quantum simulator supplies open-quantum-system measurements used to compute the loss, thereby decoupling optimization from the quantum hardware. Optionally, the low-rank tensor-train parameterization distributes gradient sensitivityISUS Ref: 53524-006PCTacross the tensor cores such that measurement-noise variance in the gradient estimator is attenuated by a factor that decreases with the product of qubit count and circuit depth, mitigating gradient vanishing and improving noise resilience as the simulated system size increases.

[0017] Optionally, performing quantum measurements across the plurality of geometrical configurations enables amplification of correlation information associated with electronic degrees of freedom not explicitly represented in a reduced active space. Optionally, the method comprises reconstructing, from the measurement data, reduced density matrices for each of the plurality of geometrical configurations.

[0018] Optionally, generating the set of quantum-derived data comprises obtaining a calibration set of two or more candidate systems sharing a chemically conserved activespace center with molecular system; for each system in the calibration set, and generating a plurality of representations corresponding to a plurality of geometrical configurations of the system defined along the at least one interatomic coordinate in the system. Optionally, predicting the chemical property comprises calibrating a regression model mapping the feature vectors of the calibration-set systems to corresponding experimentally measured values of chemical properties; and predicting the chemical property of the molecular system by applying the calibrated regression model to the feature vector of the molecular system.

[0019] Optionally, generating the set of quantum -derived data comprises extracting, from the measurement data, expectation values of one or more quantum observables for each of the plurality of geometrical configurations. Optionally, the one or more quantum observables comprises Pauli operators acting on qubits encoding the molecular system. Optionally, the Pauli operators comprise at least one of: single-qubit Pauli-Z operators, two-qubit Pauli-ZZ operators, and two-qubit Pauli-XX and Pauli-YY operators. Optionally, the expectation values are sufficient to reconstruct at least one of: a one-electron reduced density matrix (1-RDM); and a two-electron reduced density matrix (2-RDM) for each of the plurality of geometrical configurations. Optionally, the expectation values across the plurality of geometrical configurations form a spatial correlation profile along the at least one interatomic coordinate.

[0020] Optionally, the quantum measurements comprise quantum state tomography. Optionally, the measurement data comprises expectation values of Pauli operators.ISUS Ref: 53524-006PCT

[0021] Optionally, transforming the quantum-derived data comprises transforming the one or more quantum observables (e.g., reduced density matrices) into a molecular orbital basis defined by eigenvectors of the corresponding representations of the molecular system (e.g., Hamiltonians).

[0022] Optionally, transforming the quantum -derived data comprises applying a basis rotation defined by eigenvectors of the plurality of representations of the molecular system (e.g. Hamiltonians). Optionally, the eigenvectors are derived from electronic structure calculations. Optionally, transforming the quantum-derived data integrates classical electronic structure information and quantum measurement data. Optionally, transforming the quantum-derived data is performed separately for each geometrical configuration.

[0023] Optionally, extracting the features is performed based on at least one of: quantum observables derived from the one or more quantum observables (e.g. reduced density matrices), and variation of the quantum observables across the plurality of geometrical configurations. Optionally, the quantum observables comprise at least one of: orbital occupation, coherence, entropy, purity, cumulant, mutual information, diagonal dominance, entanglement, concurrence, negativity, spin correlation, natural orbital occupation numbers, gradient of a quantum observable across the plurality of geometrical configurations, integral of a quantum observable across the plurality of geometrical configurations, extremal values of a quantum observable across the plurality of geometrical configurations, coefficient of variation of a quantum observable across the plurality of geometrical configurations, and center of gravity of a quantum observable along the at least one interatomic coordinate. Optionally, the features further comprise statistical descriptors across the plurality of geometrical configurations. The statistical descriptors comprise at least one of: mean, variance, slope, range, or curvature.

[0024] Optionally, the extracted features are predictive in a molecular orbital basis but not in a computational qubit basis. Optionally, the features comprise a virtual -orbital trace deficit. Optionally, the features are extracted from an orbital-basis-transformed spatial correlation profile across the plurality of geometrical configurations. Optionally, transformation may be performed into alternative physically meaningful bases, including localized orbital bases, natural orbital bases, or symmetry-adapted bases. Optionally, theISUS Ref: 53524-006PCTfeatures comprise a combination of values at an equilibrium configuration, and variation across the at least one interatomic coordinate in the molecular system.

[0025] Optionally, the features comprise a subset of selected features having a cardinality below a predefined threshold k, wherein Aus a small integer, preferably k<3. Limiting the number of features in this manner constitutes a deliberate model design choice that reduces overfitting, improves generalization, and enhances interpretability. In particular, rather than utilizing the full feature set (e.g., approximately 242 features), the model is constrained to operate on a minimal subset identified as most informative based on empirical evaluation. Optionally, the selected features are determined based on a statistical relevance metric.

[0026] Optionally, generating the plurality of representations comprises generating representations for a plurality of spatially distinct active spaces within the molecular system. Optionally, features derived from the plurality of spatially distinct active spaces are combined. Optionally, predicting comprises applying a supervised machine-learning model. Optionally, the model is trained using compounds sharing a common bonding mechanism, such that the training data is restricted to a consistent chemical bond type. This mechanism-specific calibration improves predictive accuracy by ensuring that the learned relationships are not confounded by differing bonding behaviors across heterogeneous compound classes.

[0027] Optionally, training the model comprises leave-one-out cross-validation. Optionally, the model is trained using chemistry-specific calibration data and validated using permutation testing that re-executes feature selection. Optionally, the feature selection comprises evaluating combinations of a limited number of features. Optionally, the feature selection comprises evaluating combinations of k features, where k is limited to a predefined small number. Optionally, the model comprises ridge regression. Optionally, the validation comprises permutation testing, wherein permutation testing comprises repeating feature selection for each permutation to correct for multiple-testing bias.

[0028] Optionally, the chemical property comprises binding affinity, enantioselectivity, solid-state material impurity, battery electrolyte material, or battery electrode interaction. Optionally, the molecular system comprises a drug-target system, or a catalytic system. Optionally, predicting the chemical property comprises predicting a confidence metricISUS Ref: 53524-006PCTassociated with the chemical property. Optionally, predicting the chemical property comprises further comprises computing at least one of an uncertainty estimate, contributing features, measurement fidelity, and a confidence indication associated with an applicability domain.

[0029] Another exemplary embodiment of the disclosed subject matter is a system for predicting a chemical property of a molecular system. The system comprising a classical computer configured to perform an electronic structure calculation for the molecular system to obtain molecular orbital information corresponding to a plurality of geometrical configurations defined along at least one interatomic coordinate, and to generate a plurality of representations of the molecular system corresponding to the plurality of geometrical configurations. The system further comprises quantum hardware configured to optimize a parameterized quantum state with respect to each of the plurality of representations to obtain a set of optimized variational parameters, and to perform quantum measurements with respect to the plurality of representations based on the set of optimized variational parameters to obtain measurement data associated with the plurality of geometrical configurations. The classical computer further configured to generate, based on the measurement data, a set of reduced quantum-derived data associated with the plurality of geometrical configurations, to transform the reduced quantum-derived data into orbital-basis data based on molecular structure information corresponding to the plurality of geometrical configurations, to extract features based on the orbital-basis data including variation of the features across the plurality of geometrical configurations, and to predict the chemical property based on the extracted features.

[0030] Yet another exemplary embodiment of the disclosed subject matter is a method for predicting a chemical property of a molecular or materials system. The method comprises providing a calibration set of two or more candidate systems sharing a chemically conserved active-space center, and a target system to be predicted. For each system in the calibration set and for the target system, the method comprises generating a plurality of representations corresponding to a plurality of geometrical configurations of the system defined along at least one interatomic coordinate, and constructing, from each said representation, an active-space second-quantized Hamiltonian. The method further comprises optimizing, on an Analog Nuclear Magnetic Resonance (NMR) hardware functioning as a molecular quantum simulator and operating as an openISUS Ref: 53524-006PCTquantum system, a parameterized quantum state with respect to at least one of said representations of each said system, thereby obtaining a set of optimized variational parameters per system. The method further comprises preparing, on said analog NMR hardware, the parameterized quantum state under said optimized variational parameters with respect to each of said plurality of representations of each said system, and performing reduced-density-matrix measurements on said analog NMR hardware, thereby obtaining mixed-state measurement data comprising one- and two-particle reduced density matrices associated with the plurality of geometrical configurations of each said system. The method further comprises transforming said reduced density matrices into a canonical orbital basis derived from electronic-structure information associated with each said system, said transforming comprising rotation by eigenvectors of the active-space Hamiltonian. The method further comprises extracting, from the transformed reduced density matrices of each said system, a feature vector comprising statistics of said transformed reduced density matrices computed across the plurality of geometrical configurations of said system, said statistics including, in particular embodiments, integrals, ranges, coefficients of variation, diagonal -dominance measures, and one- or two-particle cumulant components. The method further comprises calibrating a regression model mapping the feature vectors of the calibration-set systems to corresponding experimentally measured values of the chemical property. The method further comprises predicting the chemical property of the target system by application of said calibrated regression model to the feature vector of the target system.

[0031] Yet another exemplary embodiment of the disclosed subject matter is a hybrid quantum-classical computing system comprising a plurality of quantum processors optionally comprising Nuclear Magnetic Resonance (NMR) quantum processors. Optionally, the quantum processors comprise at least one of: NMR quantum processors, superconducting qubit processors, trapped-ion processors, or photonic quantum processors. Each quantum processor comprising a set of nuclear spins configured as qubits and operable to perform quantum operations. The hybrid quantum-classical computing system further comprises a classical control system communicatively coupled to the plurality of NMR quantum processors. The plurality of NMR quantum processors is configured to operate independently and in parallel. The classical control system is configured to decompose a computational task into a plurality of sub-tasks; assign the sub-tasks to respective NMR quantum processors, coordinate execution ofISUS Ref: 53524-006PCTquantum operations across the plurality of NMR quantum processors, and aggregate measurement results obtained from the plurality of NMR quantum processors to determine a solution to the computational task. Optionally, the classical control system is configured to decompose the computational task into a plurality of sub-tasks based on at least one of Hamiltonian partitioning, spatial decomposition, observable grouping, circuit depth constraints, or hardware availability.

[0032] Optionally, the classical control system is further configured to obtain a plurality of representations of a molecular system corresponding to a plurality of geometrical configurations defined along at least one interatomic coordinate in the molecular system. The classical control system is further configured to assign to each NMR quantum processor a representation of the molecular system corresponding to a distinct geometrical configuration along the at least one interatomic coordinate. The classical control system is further configured to coordinate independent execution of parameterized quantum circuits on the respective NMR quantum processors. The classical control system is further configured to aggregate measurement results from the plurality of NMR quantum processors to construct a spatial correlation profile across at least one interatomic coordinate and apply cross-processor fidelity calibration to normalize measurement data obtained from different NMR quantum processors. Optionally, the classical control system is configured to aggregate the measurement results using at least one of: weighted averaging based on fidelity metrics, majority voting, Bayesian inference, or error-mitigated reconstruction.

[0033] Optionally, each NMR quantum processor comprises between 2 and 15 qubits. Optionally, the plurality of NMR quantum processors operates at room temperature. Optionally, the plurality of NMR quantum processors operates without cryogenic cooling. Optionally, quantum operations are implemented using radiofrequency pulse sequences.

[0034] Optionally, the plurality of sub-tasks comprises fragments of a larger computational problem, wherein the fragments are processed concurrently across the plurality of NMR quantum processors. Optionally, fragments correspond to subsets of variables or Hamiltonian terms. Optionally, the fragments overlap to enforce solution consistency. Optionally, the classical control system is further configured to assign a plurality of NMR quantum processors to perform redundant computation of the sameISUS Ref: 53524-006PCTsub-task, compare results, and determine a validated result based on statistical agreement. Optionally, at least three processors perform the same sub-task. Optionally, a result is validated based on majority voting.

[0035] Optionally, the classical control system is further configured to assign a plurality of NMR quantum processors to perform redundant computation of a same sub-task; determine a fidelity metric for each NMR quantum processor based on physicality constraints of the measurement data; and determine a validated result by weighting results from the plurality of NMR quantum processors according to their respective fidelity metrics. Optionally, statistical aggregation comprises at least one of outlier rejection, averaging, or Bayesian inference.

[0036] Optionally, the system further comprises a pool of pre-initialized quantum processors maintained in a known reference state. The quantum processors are preprepared and held in a ready state for rapid deployment. The system further comprises a classical control system configured to monitor operational fidelity of a currently active quantum processor and to detect an error condition. The error condition may comprise, for example, a deviation of measured observables from expected values, a drop in readout fidelity, loss of coherence beyond a predefined threshold, or failure to successfully complete a quantum circuit. In response to detecting the error condition, the classical control system is configured to replace the active quantum processor with a standby quantum processor selected from the pool of pre-initialized quantum processors.

[0037] Optionally, the system further comprises a pool of pre-initialized NMR quantum processors maintained in a known reference state, the processors being pre-prepared and held in a ready state. The system further comprises a classical control system configured to monitor performance of an active NMR quantum processor based on one or more fidelity metrics. An error condition is defined as a degradation in performance exceeding a predefined threshold, the degradation being indicative of, for example, RF coil drift, temperature fluctuation, or magnetic field inhomogeneity. In response to detecting the error condition, the classical control system is configured to replace the active processor with a standby processor selected from the pool of pre-initialized NMR quantum processors.

[0038] Optionally, the system further comprises a pool of standby NMR quantum processors, each maintained in a calibrated thermal equilibrium state. The classicalISUS Ref: 53524-006PCTcontrol system is configured to: continuously monitor a fidelity metric of each active NMR quantum processor during execution; detect degradation of an active NMR quantum processor when its fidelity metric falls below a predetermined threshold; reassign the sub-task of the degraded processor to a standby processor from the pool; and transition the degraded processor to a recalibration state.

[0039] Optionally, each NMR quantum processor is assigned a state from the group consisting of active, standby, and recalibrating; and wherein the classical control system dynamically transitions processors between states based on their respective fidelity metrics. Optionally, a processor in the recalibrating state undergoes a re-initialization sequence comprising thermal re-equilibration and reference pulse calibration, and upon successful recalibration is transitioned to standby state.

[0040] Optionally, the classical control system is configured to obtain partial measurement data from a subset of the plurality of NMR quantum processors, the partial measurement data comprising measurements of a reduced set of observables selected to provide an estimate of a target quantity with limited measurement overhead. The classical control system is further configured to determine a confidence level based on statistical agreement between results obtained from the subset of processors and / or based on variance relative to an expected distribution. In response to the confidence level exceeding a predefined threshold, the classical control system accepts the partial measurement data. In response to the confidence level falling below the predefined threshold, the classical control system selectively triggers additional measurements from one or more processors, including expanded measurement sets or full state tomography, to increase estimation accuracy and reduce uncertainty.

[0041] Optionally, the classical control system is configured to obtain, from each of a subset of the plurality of NMR quantum processors, a first set of measurement data comprising expectation values of diagonal quantum observables; determine, based on the first set of measurement data, whether off-diagonal quantum observables are expected to carry significant information; and selectively trigger, on processors where off-diagonal information is expected to be significant, a second set of measurements comprising off-diagonal quantum observables to obtain a complete density matrix reconstruction.

[0042] Optionally, the determination of whether off-diagonal observables carry significant information is based on a degeneracy measure of orbital occupations obtainedISUS Ref: 53524-006PCTfrom the diagonal quantum observables, such that when orbital occupations are neardegenerate, the second set of measurements is triggered.

[0043] Optionally, the selective triggering is performed independently for each geometrical configuration of the plurality of geometrical configurations, such that full density matrix reconstruction is performed only at geometrical configurations where diagonal measurements indicate near-degenerate orbital occupations. Optionally, at geometrical configurations where the diagonal measurements indicate non-degenerate orbital occupations, the quantum-derived data is generated from the first set of measurement data without performing the second set of measurements, thereby reducing total measurement time

[0044] Optionally, the classical control system is configured to iteratively update subtask assignments to the plurality of NMR quantum processors based on intermediate results. Optionally, correlations between sub-tasks are identified and incorporated into subsequent quantum operations. Optionally, the plurality of NMR quantum processors comprises liquid-state NMR systems. Optionally, the liquid- state NMR systems comprise ensembles of molecules in solution providing ensemble-average measurement signals.

[0045] Optionally, the liquid-state NMR systems comprise a solution containing a macroscopic ensemble of molecules, each molecule providing a corresponding quantum processor, such that quantum operations are performed simultaneously across the ensemble and measurement signals are obtained as ensemble-averaged observables representative of the collective quantum state. Optionally, quantum measurements are obtained without projective collapse using ensemble averaging.

[0046] Optionally, the liquid-state NMR systems operate at room temperature without cryogenic cooling. Optionally, qubits are implemented using nuclear spins in heteronuclear molecular systems. Optionally, quantum gate operations are implemented using radiofrequency pulse sequences. Optionally, the system further comprises hyperpolarization. Optionally, the processors operate at physiological temperature.

[0047] Optionally, the plurality of NMR quantum processors comprises solid-state NMR systems. Optionally, the solid-state NMR systems comprise nuclear spins in representative material samples. Optionally, the solid-state NMR systems are configured to apply a magic angle spinning (MAS) mechanism. Optionally, the MAS mechanism isISUS Ref: 53524-006PCTconfigured to rotate a sample containing the nuclear spin qubits at predetermined “Magic Angle” of 54.74 Degrees relative to an applied magnetic field, thereby reducing anisotropic spin interactions and enhancing coherence and spectral resolution for quantum operations. Optionally, the magic angle spinning is performed at a frequency in the range of 10-70 kHz. Optionally, the magic angle spinning increases coherence time of nuclear spin qubits.

[0048] Optionally, the solid-state NMR systems operate under controlled temperature and pressure conditions corresponding to target material environments. Optionally, the solid-state NMR systems comprise representative materials selected based on similarity to a target system being modeled. Optionally, the plurality of NMR quantum processors comprises both liquid-state and solid-state NMR systems. Optionally, liquid-state processors perform measurement-intensive tasks and solid-state processors perform environment-specific simulations. Optionally, results from liquid-state and solid-state processors are combined to determine the solution.

[0049] Yet another exemplary embodiment of the disclosed subject matter is a method for performing quantum-assisted computation using a distributed NMR quantum computing system. The method comprises: decomposing, by a classical control system, a computational task into a plurality of sub-tasks; assigning the sub-tasks to respective NMR quantum processors from a plurality of NMR quantum processors, each comprising a set of nuclear spins configured as qubits; executing, in parallel and independently, quantum operations at the respective NMR quantum processors in accordance with the assigned sub-tasks; coordinating, by the classical control system, execution of the quantum operations across the plurality of NMR quantum processors; obtaining measurement results from the plurality of NMR quantum processors; and aggregating, by the classical control system, the measurement results to determine a solution to the computational task.

[0050] Optionally, the method / system further comprises performing cross-processor calibration to normalize measurement data obtained from different quantum processors. Optionally, assigning the sub-tasks comprises selecting processors based on at least one of: coherence time, current fidelity, qubit count, or availability. Optionally, the aggregated measurement results are further processed to extract features for use in a machine-learning model configured to predict a chemical property.ISUS Ref 53524-006PCT

[0051] Optionally, the system is configured to tolerate continuous degradation of processor fidelity and to maintain computation accuracy via redundancy and dynamic reassignment. Optionally, the plurality of NMR quantum processors comprises between 2 and 100 processors. Optionally, at least a subset of the processors performs redundant execution of identical sub-tasks to enable statistical validation.

[0052] Although described in connection with NMR quantum processors, the disclosed methods may be implemented using other quantum computing platforms, including gatebased, analog, or hybrid quantum systems.

[0053] The disclosed method may be implemented as a data processing pipeline in which quantum-derived data is progressively transformed, aggregated, and mapped to predictive outputs.ISUS Ref: 53524-006PCTTHE BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0054] The present disclosed subject matter will be understood and appreciated more fully from the following detailed description taken in conjunction with the drawings in which corresponding or like numerals or characters indicate corresponding or like components. Unless indicated otherwise, the drawings provide exemplary embodiments or aspects of the disclosure and do not limit the scope of the disclosure. In the drawings:

[0055] Figure 1 shows a hybrid quantum -cl as si cal computing system, in accordance with some exemplary embodiments of the subject matter;

[0056] Figure 2 shows a block diagram of a molecular quantum simulation platform, in accordance with some exemplary embodiments of the disclosed subject matter; and

[0057] Figures 3A-3G show flowchart diagrams of a method, in accordance with some exemplary embodiments of the disclosed subject matter; and

[0058] Figure 4 shows a hybrid quantum -cl as si cal computing system in accordance with some exemplary embodiments of the disclosed subject matter.ISUS Ref: 53524-006PCTDETAILED DESCRIPTION

[0059] One technical problem dealt with by the disclosed subject matter is the inability of existing computational methods to accurately model chemical phenomena. Such modeling is essential for applications including drug discovery, prediction of binding affinity, identification of chemically relevant interactions, and evaluation of catalytic and material properties. Existing classical approaches, including force-field methods and density functional theory, rely on approximations such as fixed point-charge representations and single-determinant formulations, and therefore fail to capture quantum mechanical effects including strong electron correlation, quantum tunneling, spin-state multiplicity, and transition-state behavior. These limitations are particularly pronounced in chemically complex systems such as metalloproteins, covalent drug targets, asymmetric catalysts, and deuterated compounds, and further result in an inability to distinguish between chemically similar compounds or to capture compound-specific electronic structure variations relevant to experimentally observed properties.

[0060] Another technical problem dealt with by the disclosed subject matter is the loss of critical electronic structure information in reduced computational models. In practice, reduced representations, including active-space models, are employed in order to make quantum mechanical simulations tractable for realistic molecular systems. However, such representations inherently omit contributions from virtual orbitals and from the broader electronic structure outside the selected active space. As a result, important physical effects, including charge transfer, 7t-back bonding, and dynamic correlation, are not captured. Furthermore, existing approaches typically rely on Hamiltonian eigenvalues or energy-based outputs, which do not encode sufficient information to predict chemical properties such as binding affinity or material behavior. Consequently, reduced models often fail to produce compound-specific or predictive outputs, even when based on first-principles calculations.

[0061] Yet another technical problem dealt with by the disclosed subject matter is the prohibitive computational cost of accurate quantum chemistry methods. While high-accuracy methods, including multi -reference and wavefunction-based approaches, are capable of capturing the relevant electronic structure effects, such methods scale unfavorably with system size and require extremely large computational resources. For realistic systems, including drug candidates, catalytic systems, and materials, suchISUS Ref: 53524-006PCTmethods may require thousands to tens of thousands of computational hours per compound, rendering them impractical for large-scale applications such as virtual screening, drug development pipelines, or materials discovery workflows. This creates a fundamental trade-off between accuracy and scalability that is not resolved by existing approaches.

[0062] Yet another technical problem dealt with by the disclosed subject matter is the inability of existing approaches to accurately and efficiently model real-world chemical and material systems under realistic conditions. Accurate prediction of chemical and material properties requires simultaneous consideration of intramolecular electronic structure and intermol ecul ar interactions, including dispersion, electrostatics, polarization, and environmental effects such as temperature, solvent, or crystalline structure. Existing methods are unable to capture these effects simultaneously with sufficient accuracy and computational efficiency. In addition, many approaches attempt to model entire molecular or material systems, leading to inefficient use of computational resources, despite the fact that relevant quantum effects are often localized to specific active sites, bonds, or interaction regions. As a result, existing techniques fail to provide accurate and practical predictions for applications such as polymorph stability, crystal packing, adsorption, and biochemical interactions.

[0063] Yet another technical problem dealt with by the disclosed subject matter is the limitations and impracticality of existing quantum computing platforms for molecular simulation. Gate-based quantum computing systems, including superconducting and trapped-ion architectures, require cryogenic operation, exhibit limited coherence times, and do not currently provide sufficient qubit counts or error rates to simulate complex molecular systems. At the same time, conventional Nuclear Magnetic Resonance (NMR) quantum computing systems suffer from intrinsic limitations, including the use of pseudo-pure states that result in exponential degradation of signal -to-noise ratio with increasing qubit number, thereby limiting scalability. Additionally, NMR systems rely on low-energy radio-frequency transitions, which impose signal detection constraints and require large ensembles, limiting the ability to resolve individual quantum states. Collectively, these limitations prevent existing quantum computing platforms from being effectively applied to large-scale, chemically relevant molecular simulations.ISUS Ref: 53524-006PCT

[0064] One technical solution provided by the disclosed subject matter is predicting properties of molecular systems using a hybrid quantum-classical computation. In some embodiment, a plurality of molecular Hamiltonians corresponding to different geometrical configurations of a molecular system are generated using classical computation. Quantum computations are then performed based on said Hamiltonians to obtain measurement data. The measurement data is used to reconstruct reduced density matrices, from which quantum observables are extracted and used to predict properties of the molecular system.

[0065] In some exemplary embodiments, the plurality of Hamiltonians corresponds to different interatomic distances or geometrical configurations along a reaction coordinate or potential energy surface of the molecular system, thereby generating a configurationdependent representation of the molecular system.

[0066] Additionally, or alternatively, the reduced density matrices comprise one-electron and / or two-electron reduced density matrices, and the extracted quantum observables comprise at least one of orbital occupation values, coherence metrics, or correlation-related quantities including cumulant values derived from the reduced density matrices.

[0067] Another technical solution provided by the disclosed subj ect matter is performing a variational quantum optimization on physical quantum hardware and reusing the optimized quantum state across multiple Hamiltonians to generate a spatially resolved quantum response profile. In some exemplary embodiments, a Variational Quantum Eigensolver (VQE) is executed on a Nuclear Magnetic Resonance (NMR) quantum processor at an equilibrium configuration of the molecular system to obtain optimized circuit parameters defining a quantum state. The equilibrium configuration refers to a configuration corresponding to a local or global minimum in the potential energy surface. The optimized parameters are then maintained fixed, and the corresponding quantum state is evaluated against a plurality of Hamiltonians corresponding to different geometrical configurations, thereby enabling measurement of the system response to Hamiltonian perturbations without repeated optimization. This approach generates a multi-configuration dataset of reduced density matrices that capture geometry-dependent electronic structure variations that are not accessible from a single configuration.ISUS Ref: 53524-006PCT

[0068] In some embodiments, optimization of the parameterized quantum state is performed independently for each representation corresponding to each geometrical configuration. In other embodiments, optimization is performed with respect to a selected representation, such as an equilibrium configuration, and the resulting optimized parameters are reused across the plurality of representations. In further embodiments, a hybrid approach may be employed in which optimization is performed for a subset of configurations and interpolated or transferred across remaining configurations.

[0069] Yet another technical solution provided by the disclosed subject matter is performing multi-distance Hamiltonian scanning combined with orbital-basis transformation to extract chemically meaningful quantum features. In some exemplary embodiments, the measurement data obtained from quantum hardware is transformed from a computational qubit basis into a molecular orbital basis defined by eigenvectors of the corresponding Hamiltonians, such that the Hamiltonian provides a directional framework for interpreting the quantum measurements. The transformed reduced density matrices are used to compute quantum observables including virtual-orbital trace deficit, cumulant-based correlation measures, and coherence metrics, as well as spatial statistics of said observables across the plurality of geometrical configurations. This combination of Hamiltonian-derived orbital structure and hardware-derived quantum signal produces feature representations that encode electronic correlation effects beyond the reduced active space and enables accurate prediction of molecular properties through subsequent calibration.

[0070] Another technical solution provided by the disclosed subject matter is extracting quantum observables from reduced density matrices and computing spatial statistics across multiple geometrical configurations to generate a high-dimensional feature representation. In some exemplary embodiments, features such as orbital occupation values, coherence, entropy, cumulant tensors, and virtual-orbital trace deficit are computed at each configuration and aggregated using statistical measures including mean, range, slope, and curvature. These features capture both instantaneous and geometry-dependent electronic structure properties and provide a robust basis for predicting chemical properties.

[0071] Another technical solution provided by the disclosed subject matter is applying machine-learning calibration to map quantum-derived features to experimentallyISUS Ref: 53524-006PCTmeasurable properties. In some exemplary embodiments, the extracted features are used as input to a regression model trained on experimentally measured data for compounds of a given chemical class, thereby enabling prediction of properties such as binding affinity, reactivity, or enantioselectivity with accuracy exceeding that of classical simulation methods.

[0072] In some embodiments, prediction of the chemical property comprises applying a trained machine-learning model. In other embodiments, prediction may comprise applying a deterministic mapping, regression function, physics-based model, or statistical inference method that maps the extracted features to the chemical property.

[0073] In some embodiments, the extracted features are not derivable from classical diagonalization of the Hamiltonian alone and arise from interaction between the quantum state and the physical quantum hardware.

[0074] In some embodiments, the extracted features are robust to measurement noise due to aggregation across multiple geometrical configurations and / or statistical processing.

[0075] Another technical solution provided by the disclosed subj ect matter is performing multi-site active space profiling, wherein multiple independent Hamiltonians corresponding to different spatial regions of a molecular system are processed through the quantum-classical pipeline and their resulting feature vectors are combined. This enables capturing long-range and multi-site interactions that are not observable within a single reduced Hamiltonian representation.

[0076] Another technical solution provided by the disclosed subject matter is a hybrid quantum-classical computing system comprising a plurality of small-scale NMR quantum processors configured to collectively perform molecular simulation tasks. In some exemplary embodiments, each processor operates on a reduced active-space Hamiltonian and performs quantum measurements to obtain partial results, which are aggregated by a classical control system to generate a complete representation of the molecular system.

[0077] Another technical solution provided by the disclosed subject matter is distributing quantum computations across multiple NMR processors operating in parallel, thereby enabling scalable simulation without requiring a large monolithic quantumISUS Ref: 53524-006PCTcomputer. In some exemplary embodiments, the processors are coordinated by a classical controller that assigns computational tasks, collects measurement data, and combines results into a unified dataset for feature extraction and prediction.

[0078] Another technical solution provided by the disclosed subject matter is providing fault tolerance and reliability through redundancy and dynamic processor management. In some exemplary embodiments, a pool of pre-initialized NMR processors is maintained, and processors are dynamically assigned operational states including active, standby, and reset. Faulty processors may be detected and replaced with standby processors, and measurement results from multiple processors may be combined using voting or aggregation mechanisms to improve accuracy.

[0079] Another technical solution provided by the disclosed subject matter is exploiting the analog physical properties of NMR hardware as part of the computational process. In some exemplary embodiments, the quantum processor operates under a combined Hamiltonian including both programmed interactions and native hardware interactions such that the measured quantum states include contributions from physical electron-mediated effects. This feature is referred to in the literature as an open quantum system, or quantum reservoir. This feature enables extraction of quantum observables, such as nonzero virtual-orbital trace deficit, that reflect coupling to degrees of freedom outside the reduced active space and provide additional predictive information beyond classical computation.

[0080] Another technical solution provided by the disclosed subject matter is providing alternative hardware implementations, including liquid-state and solid-state NMR systems. In some exemplary embodiments, solid-state NMR techniques such as magic angle spinning are used to capture additional interactions relevant to materials and solid-state molecular systems, thereby extending the applicability of the system to prediction of properties of crystalline materials and extended molecular structures.

[0081] Referring now to Figure 1, schematically illustrating illustrates a hybrid quantum-classical computing system, in accordance with some exemplary embodiments of the subject matter.

[0082] In some exemplary embodiments, System 100 may be utilized for predicting a chemical property of a molecular or materials system. System 100 comprises an NMRISUS Ref: 53524-006PCTquantum hardware 110. NMR quantum hardware 110 may be an analog NMR quantum hardware or processor functioning as a molecular quantum simulator. In the illustrated embodiment, a single NMR quantum hardware 110 is depicted for clarity; however, in other embodiments, System 100 may comprise a plurality of such hardware units operating in parallel or distributed configurations. NMR quantum hardware 110 may be configured to implement quantum operations using nuclear spin qubits. NMR quantum hardware 110 may operate as an open quantum system in which nuclear spins of a physical molecular or material sample evolve under a physically realized Hamiltonian and are controlled and measured to obtain quantum-derived data. An "open quantum system" refers to a system that interacts with external degrees of freedom, resulting in non-unitary evolution and mixed-state measurements. In some embodiments, the quantum-derived data is generated using a subset of observables selected to approximate reduced density matrices without full tomography. Because the analog NMR hardware functions as an open quantum system, the Hilbert space accessible to the physical sample is strictly larger than the Hilbert space addressable by any closed unitary evolution that conserves particle number. For a representative active space comprising two electrons distributed among two spatial orbitals (i.e., a CASCI(2e,2o) projection mapped onto two qubits under Jordan-Wigner encoding), the closed-system Hilbert space respecting particle-number conservation has dimension C(4,2) = 6, whereas the full Fock space on the same four spin-orbitals has dimension 24= 16. The additional ten dimensions correspond to sectors of differing particle number (N = 0, 1, 3, 4), including particleaddition, particle-removal, and pair-process configurations, which are categorically inaccessible to any closed unitary evolution that conserves particle number — including classical configuration-interaction diagonalization and noiseless digital simulation of variational quantum eigensolvers — regardless of ansatz depth or classical compute budget. The analog NMR hardware populates these sectors because the physical Hamiltonian of the sample contains inter-sector couplings — including without limitation dipolar couplings to non-encoding nuclei, hyperfine contact terms, chemical-shift-tensor cross-terms, and radio-frequency control imperfections — that do not commute with the particle-number operator of the programmed active-space Hamiltonian. In some embodiments, the mixed-state observables measured at the conclusion of the variational protocol therefore carry a structured, molecule-specific imprint of bath physics that lies formally outside the closed-system active space, and this imprint provides the signal uponISUS Ref: 53524-006PCTwhich downstream orbital -basis rotation and statistical calibration operate. System 100 further comprises a Classical Control System 130 configured to coordinate operation of NMR quantum hardware 110. Classical Control System 130 may comprise one or more classical computing devices, such as a general -purpose processor, workstation, server, or distributed computing system, configured to perform control logic, data processing, and optimization routines associated with quantum -cl as si cal hybrid computation. Classical Control System 130 may be configured to generate representations of molecular or materials systems corresponding to a plurality of geometrical configurations. Additionally, or alternatively, Classical Control System 130 may be configured to construct active-space second-quantized Hamiltonians associated with said representations. Additionally, or alternatively, Classical Control System 130 may be configured to control preparation of parameterized quantum states on NMR quantum hardware 110. Additionally, or alternatively, Classical Control System 130 may be configured to process measurement data obtained from NMR quantum hardware 110 and perform data transformation, feature extraction, model calibration, and prediction of a chemical property.

[0083] In some exemplary embodiments, NMR quantum hardware 110 may comprise a magnet assembly including two powerful electromagnet, permanent magnets or superconducting Magnets 112 and 114. A Sample Tube 116 may be disposed within the magnetic field. Sample Tube 116 may comprise a molecular or material system whose nuclear spins define the quantum degrees of freedom (and utilized as qubits of the NMR quantum hardware 110). Magnets 112 and 114 may be configured to generate a static magnetic field across a sample region within Sample Tube 116. The magnetic field establishes the physical conditions under which the sample evolves as an analog quantum system, thereby realizing a Hamiltonian corresponding to the system under study. Additionally, or alternatively, the magnet assembly may define a quantization axis for nuclear spin qubits within Sample Tube 116.

[0084] Sample 116 comprises a physical instance of a molecular or material system. In some embodiments, Sample Tube 116 may comprise a liquid-state ensemble of molecules, enabling ensemble-averaged quantum measurements. For example, the sample may comprise a liquid solution containing molecules with spin- 1 / 2 nuclei such as 'H,13C,15N,19F, or31P, including heteronuclear systems (e.g., 'H-3IP pairs) exhibiting J-ISUS Ref: 53524-006PCTcoupling interactions suitable for implementing multi-qubit quantum gates. Example compounds include dimethyl phosphite or diethyl phosphonate, which provide well-defined two-qubit systems. In other embodiments, the sample may comprise a solid-state material. The sample constitutes the physical open quantum system on which the parameterized quantum state is prepared and measured. For example, solid-state NMR implementations may utilize representative material samples or chemically active sites within a material, optionally under Magic Angle Spinning (MAS) conditions, to enable quantum computation of material or catalytic properties.

[0085] An RF Coil 118 may be disposed in proximity to Sample Tube 116, such as surrounding Sample Tube 116. RF Coil 118 may be configured to apply radiofrequency (RF) signals to control evolution of nuclear spins in Sample 116 and detect response NMR signals emitted by the sample. RF Coil 118 thereby enables both preparation of parameterized quantum states and acquisition of measurement signals corresponding to quantum observables. In some embodiments, RF Coil 118 implements quantum gate operations via controlled RF pulse sequences.

[0086] RF Control Signal Generator 120 is configured to generate RF control signals, which are transmitted to RF Coil 118 to manipulate the nuclear spin qubits. RF Control Signal Generator 120 is controlled by Classical Control System 130 to encode a parameterized quantum state through a sequence of control signals corresponding to variational parameters. In additional or alternative embodiments, the RF control signals correspond to parameterized quantum circuits executed on NMR quantum hardware 110. Optionally, RF Control Signal Generator 120 is controlled by Classical Control System 130 to generate pulse sequences corresponding to variational quantum circuits, including sequences implementing single-qubit rotations and two-qubit entangling operations based on J-coupling evolution.

[0087] A radio frequency Output Receiver 140 is configured to receive NMR response signals from RF Coil 118. The received signals correspond to measured physical responses of Sample 116, from which quantum observables are derived. These signals constitute measurement data obtained from NMR quantum hardware 110, including expectation values of quantum observables.

[0088] The received signals are transmitted to Classical Control System 130, which is configured to process the measurement data and generate an NMR spectrum 135.ISUS Ref: 53524-006PCTClassical Control System 130 may be configured to process the measurement data to obtain measured one-particle and two-particle reduced density matrices associated with a plurality of geometrical configurations. The reduced density matrices represent mixed-state quantum data obtained through measurement of the open quantum system. In some embodiments, the plurality of geometrical configurations comprises a sparse sampling of configurations, and intermediate values are inferred using interpolation or regression techniques applied to the quantum -derived data.

[0089] In some exemplary embodiments, Classical Control System 130 may be configured to assign to NMR quantum hardware 110 a plurality of representations of each system corresponding to different geometrical configurations along at least one interatomic coordinate. Classical Control System 130 may further be configured to control optimization of a parameterized quantum state with respect to at least one representation based on measurement data and subsequently control preparation and measurement of the parameterized quantum state under optimized parameters across the plurality of representations.

[0090] Classical Control System 130 may further be configured to aggregate measurement results obtained from NMR quantum hardware 110. Such aggregation may include combining measurement outcomes, such as measured reduced density matrices across the plurality of geometrical configurations.

[0091] Classical Control System 130 may further be configured to transform the measured reduced density matrices into a canonical orbital basis derived from electronic-structure information associated with each system, including rotation by eigenvectors of the active-space Hamiltonian. Classical Control System 130 may further extract a feature vector comprising statistics computed across the plurality of geometrical configurations, including one or more of integrals, ranges, coefficients of variation, diagonal -dominance measures, and cumulant components.

[0092] Additionally, or alternatively, Classical Control System 130 may be configured to calibrate a regression model using feature vectors associated with a calibration set of systems and corresponding experimentally measured chemical properties. Classical Control System 130 may then predict a chemical property of a target system by applying the calibrated regression model to a feature vector of the target system.ISUS Ref: 53524-006PCT

[0093] In some embodiments, NMR quantum hardware 110 operates using ensemblebased measurement of a plurality of identical quantum systems. Measurement corresponds to ensemble-averaged observables obtained from the physical system rather than projective single-shot measurements.

[0094] In some embodiments, System 100 may operate at room temperature or physiological temperature.

[0095] Referring now to Figure 2, showing a Molecular Quantum Simulation Platform 200 for predicting a chemical property of a molecular system, in accordance with some exemplary embodiments of the disclosed subject matter.

[0096] Molecular Quantum Simulation Platform 200 may be configured to receive a molecular representation or a molecular structure of the molecular system, and generate a prediction of a chemical property of the molecular system based on coordinated quantum and classical processing.

[0097] In some exemplary embodiments, Molecular Quantum Simulation Platform 200 is configured to process the molecular system through a hybrid quantum-classical pipeline that captures electronic structure information across a plurality of geometrical configurations and derives predictive information therefrom. The platform may be configured to generate a representation of the molecular system that reflects variation along at least one interatomic coordinate, to obtain quantum-derived information associated with the molecular system using quantum hardware, and to integrate such quantum-derived information with classical electronic structure information.

[0098] In some exemplary embodiments, Molecular Quantum Simulation Platform 200 is further configured to derive a configuration-dependent quantum response of the molecular system, and to generate a representation of electronic correlation behavior across the plurality of geometrical configurations. The platform may further process the resulting data to obtain features that are indicative of chemical behavior of the molecular system.

[0099] In some exemplary embodiments, Molecular Quantum Simulation Platform 200 is configured to utilize the extracted information to predict a chemical property of the molecular system, including, for example, binding affinity, enantioselectivity, or otherISUS Ref: 53524-006PCTelectronic-structure-dependent properties, optionally using a trained machine-learning model calibrated on reference systems.

[0100] Accordingly, Molecular Quantum Simulation Platform 200 provides a unified system for predicting chemical properties of molecular systems by leveraging quantum measurements in combination with classical computation to capture information that is not accessible through classical methods alone.

[0101] Molecular Quantum Simulation Platform 200 may comprise an Input / Output (I / O) Module 205, a Representation Generator 230, a Quantum Hardware 210, an Optimization and Measurement Module 240, a Reduced Data Generation Module 250, a Transformation Module 260, a Feature Extraction Module 280, a Prediction Module 290, and a Classical Processing Unit 220. The components may be operatively coupled as illustrated, and may exchange data and control signals via Classical Processing Unit 220.

[0102] In some exemplary embodiments, Molecular Quantum Simulation Platform 200 may comprise a hybrid computational architecture including both classical computing resources (e.g., Classical Processing Unit 220) and quantum computing resources (e.g., Quantum Hardware 210). Classical Processing Unit 220 may comprise one or more processors configured to perform classical computations, and in some embodiments may include processors capable of executing quantum instructions or interfacing with quantum hardware. Quantum Hardware 210 may be configured to perform quantum operations, including operations associated with Optimization and Measurement Module 240, Reduced Data Generation Module 250, and Transformation Module 260, such as quantum state preparation, quantum evolution, and quantum measurements.

[0103] Classical Processing Unit 220 may be configured to orchestrate operations of Molecular Quantum Simulation Platform 200, including performing classical electronic structure calculations, executing optimization loops, processing measurement data, and coordinating data flow between modules. In some exemplary embodiments, Classical Processing Unit 220 may further be configured to perform classical data processing tasks, such as feature extraction and prediction operations associated with Feature Extraction Module 280 and Prediction Module 290. Additionally or alternatively, at least a portion of such operations may be performed using quantum hardware.ISUS Ref: 53524-006PCT

[0104] Quantum Hardware 210 may comprise one or more analog Nuclear Magnetic Resonance (NMR) quantum processors configured to execute quantum computations based on the generated Hamiltonians. In some exemplary embodiments, the NMR quantum processor may operate as an open quantum system and perform analog quantum evolution based on nuclear spin interactions. Additionally, or alternatively, Quantum Hardware 210 comprises a plurality (e.g., more than one) of analog NMR quantum processors. Each NMR quantum processor may comprise between 2 and 15 qubits, operate at room temperature without cryogenic cooling, and implement quantum operations using radiofrequency pulse sequences. Each NMR processor may comprise a set of nuclear spins configured as qubits, wherein the plurality of NMR quantum processors are configured to operate independently and in parallel.

[0105] In some embodiments, the plurality of NMR quantum processors may comprise liquid-state NMR systems, solid-state NMR systems, or a combination thereof. In some embodiments, Classical Processing Unit 220 may be configured to decompose a computational task into a plurality of sub-tasks, assign the sub-tasks to respective NMR quantum processors, coordinate execution of quantum operations across the plurality of NMR quantum processors, and aggregate measurement results obtained from the plurality of NMR quantum processors. The Classical Processing Unit 220 may further be configured to assign to each NMR quantum processor a representation of the molecular system corresponding to a distinct geometrical configuration along the at least one interatomic coordinate, thereby enabling parallel processing of different configurations.

[0106] In some exemplary embodiments, VO Module 205 may be configured to receive input data associated with a molecular system. The input data may comprise a representation of a target molecular system to be predicted. Additionally or alternatively, VO Module 205 may be configured to receive a representation of a calibration set of candidate systems sharing a chemically conserved active-space center with the molecular system. The calibration set being usable for training or calibrating a predictive model utilized by Prediction Module 290 for predicting the chemical property of the molecular system. The predictive model may be pre-trained, trained using a calibration dataset, or incrementally updated during operation.

[0107] In some exemplary embodiments, VO Module 205 may be configured to receive structural information of the molecular system, such as a molecular structure or otherISUS Ref: 53524-006PCTstructural descriptors, optionally including identification of a target interatomic interaction relevant to a chemical property. The molecular structure may be represented, for example, as a molecular string representing the structure of the molecular system. The molecular string may comprise a text-based representation of the structure of the molecular system being analyzed. The molecular string is configured to encode at least one of atomic composition, atomic connectivity, molecular geometry, or electronic structure of the molecular system. As a non-limiting example, the molecular string may be provided in a Simplified Molecular Input Line Entry System (SMILES) format. Additionally or alternatively, any other machine-readable molecular representation may be utilized, including, but not limited to, a graph-based representation, a coordinate-based representation, or an electronic representation of the molecular system.

[0108] In some exemplary embodiments, EO Module 205 may be configured to receive data defining at least one interatomic coordinate of the molecular system. Additionally, or alternatively, EO Module 205 may be configured to receive a plurality of geometrical configurations of the molecular system along said one interatomic coordinate, such as configurations corresponding to a bond formation or dissociation process. The plurality of geometrical configurations may comprise an interatomic distance scan spanning a range of interatomic distances between atoms participating in the target interaction.

[0109] Additionally, or alternatively, EO Module 205 may be configured to receive experimental property data associated with the calibration set, the experimental property data being usable by Prediction Module 290 for calibration, training, or validation of a predictive model.

[0110] Representation Generator 230 may be configured to generate a plurality of representations of the molecular system corresponding to a plurality of geometrical configurations defined along at least one interatomic coordinate, such as interatomic distances along a bond formation or dissociation coordinate.

[0111] In some exemplary embodiments, the plurality of representations may comprise a plurality of Hamiltonians, each representing an electronic structure of the molecular system at a respective geometrical configuration. The plurality of Hamiltonians may correspond to a plurality of interatomic distances and may be constructed based on electronic structure calculations, such as Density Functional Theory (DFT), configured to determine molecular orbital coefficients and electronic integrals for each geometricalISUS Ref: 53524-006PCTconfiguration. In some exemplary embodiments, the molecular representation of the system, such as a molecular string, may be processed to generate a three-dimensional molecular geometry, for example using cheminformatics tools, prior to performing the electronic structure calculations.

[0112] In some exemplary embodiments, each Hamiltonian may be constructed based on a reduced active-space representation derived from a full electronic structure of the molecular system, such as a CASCI(2e,2o) representation, comprising a subset of molecular orbitals including at least one occupied orbital and at least one unoccupied orbital. The reduced active-space representation may correspond to orbitals associated with a target interatomic interaction, and may reduce the full molecular Hamiltonian to a low-dimensional representation, such as a matrix corresponding to a two-qubit Hilbert space.

[0113] In some exemplary embodiments, the Hamiltonians may encode interactions derived from the full electronic structure and projected onto the reduced representation, including contributions from electronic degrees of freedom outside the reduced active space, such as virtual orbital contributions representing electron correlation effects. In some exemplary embodiments, each Hamiltonian may be represented as a linear combination of Pauli operators suitable for execution on quantum hardware, thereby enabling mapping to qubit representations.

[0114] In some exemplary embodiments, the Hamiltonians may be mapped to qubit representations, for example via a fermion-to-qubit transformation such as a Jordan-Wigner transformation, to obtain a set of Pauli operator coefficients suitable for execution on Quantum Hardware 210. The resulting plurality of Hamiltonians may comprise a set of distance-indexed Hamiltonians H(Ri), H(R2), ..., H(R_N), each expressed in terms of Pauli operators for quantum processing.

[0115] Optimization and Measurement Module 240 may be configured to optimize, using Quantum Hardware 210, a parameterized quantum state and to utilize the parameterized quantum state to perform quantum measurements. In some exemplary embodiments, Optimization and Measurement Module 240 may be configured, such as via Variational Quantum Eigensolver (VQE) Optimization Module 242, to optimize a parameterized quantum state with respect to each of the plurality of representations of the molecular system, thereby obtaining a set of optimized variational parameters.ISUS Ref: 53524-006PCT

[0116] Additionally, or alternatively, Optimization and Measurement Module 240 may be configured, such as via Measurement Module 244, to perform quantum measurements with respect to the plurality of representations based on the set of optimized variational parameters, thereby obtaining measurement data associated with the plurality of geometrical configurations. In some exemplary embodiments, the quantum measurements comprise quantum state tomography, expectation values of Pauli operators, or the like. It may be noted that the measurements are quantum measurements performed on or using Quantum Hardware 210. In some embodiments, the optimization and measurement operations may be executed independently and in parallel across the plurality of NMR quantum processors, each processing a respective sub-task or geometrical configuration.

[0117] In some embodiments, measurement acquisition may be selectively controlled such that a first set of measurement data comprising expectation values of diagonal quantum observables is obtained, and based thereon, a determination is made whether off-diagonal observables are expected to carry significant information. In response to said determination, a second set of measurements comprising off-diagonal observables may be selectively triggered, optionally only for geometrical configurations exhibiting neardegenerate orbital occupations, thereby reducing total measurement time.

[0118] In some exemplary embodiments, VQE Optimization Module 242 may be configured to select a representation of the molecular system corresponding to an equilibrium configuration, wherein the equilibrium configuration corresponds to a configuration associated with a minimum energy along at least one interatomic coordinate. The selected representation may be encoded for execution on Quantum Hardware 210 (NMR quantum processor, and a parameterized quantum state may be optimized with respect thereto by minimizing an expected energy of the form (y(9)|H(R_eq)|\| / (9)). The optimization may be performed iteratively using a classical optimizer until convergence or until a maximum iteration threshold is reached, thereby obtaining optimized variational parameters.

[0119] In some exemplary embodiments, the optimized variational parameters may approximate a ground state of the molecular system at the equilibrium configuration. The optimized variational parameters may be maintained fixed across the plurality of representations, and the parameterized quantum state may be prepared on QuantumISUS Ref: 53524-006PCTHardware 210 using said optimized parameters for each of the plurality of representations.

[0120] Quantum measurements may then be performed for each representation to obtain measurement data associated with the plurality of geometrical configurations, wherein maintaining the optimized parameters fixed enables extraction of configurationdependent quantum responses and amplification of correlation information associated with electronic degrees of freedom not explicitly represented in a reduced active space.

[0121] Reduced Data Generation Module 250 may be configured to generate, using Quantum Hardware 210 and based on the measurement data obtained as described with respect to Optimization and Measurement Module 240, a set of reduced quantum-derived data associated with the plurality of geometrical configurations. In some exemplary embodiments, the reduced quantum -derived data may be generated for the target molecular system and, additionally or alternatively, for each system in a calibration set of candidate systems.

[0122] The term “reduced quantum -derived data,” as used herein, refers to a representation of quantum measurement data in a compressed or physically meaningful form derived from raw measurement outcomes, including but not limited to reduced density matrices, expectation values of selected observables, cumulant representations, or other lower-dimensional representations capturing relevant quantum correlations. In some embodiments, the quantum measurement data encodes correlation effects arising from interactions with degrees of freedom not explicitly represented in the reduced active space, such that the resulting features contain information that is not recoverable from classical electronic structure calculations alone.

[0123] In some exemplary embodiments, for each representation corresponding to a geometrical configuration, the parameterized quantum state prepared using the optimized variational parameters may be utilized to obtain measurement data. The measurement data may comprise expectation values of one or more quantum observables, including Pauli operators acting on qubits encoding the molecular system.

[0124] In some exemplary embodiments, the Pauli operators may comprise single-qubit and two-qubit operators, including at least one of Pauli-Z, Pauli-ZZ, Pauli-XX, and Pauli-YY operators. The expectation values of the quantum observables may be sufficient toISUS Ref: 53524-006PCTreconstruct reduced quantum -derived data comprising one-particle and two-particle reduced density matrices (1-RDM and 2-RDM) for each of the plurality of geometrical configurations, thereby obtaining mixed-state measurement data.

[0125] In some exemplary embodiments, the reduced density matrices across the plurality of geometrical configurations may define a spatial correlation profile along the at least one interatomic coordinate. Said spatial correlation profile may capture configuration-dependent quantum responses and enable amplification of correlation information associated with electronic degrees of freedom not explicitly represented in a reduced active space.

[0126] In some exemplary embodiments, the generation of the reduced quantum -derived data may be performed for a plurality of representations corresponding to spatially distinct active spaces within the molecular system, thereby enabling downstream combination of information derived from multiple active-space regions.

[0127] Transformation Module 260 may be configured to transform, using Quantum Hardware 210 and / or Classical Processing Unit 220, the reduced quantum -derived data into canonical orbital-basis data based on electronic structure information corresponding to the plurality of geometrical configurations of the molecular system. In some exemplary embodiments, the transformation may integrate classical electronic structure information with quantum measurement data, thereby converting data obtained in a computational qubit basis into a representation that reflects molecular orbital characteristics of the molecular system.

[0128] In some exemplary embodiments, Transformation Module 260 may be configured to perform said transformation separately for each geometrical configuration of the plurality of geometrical configurations, thereby preserving configurationdependent information associated with variation along at least one interatomic coordinate.

[0129] In some exemplary embodiments, Transformation Module 260 may comprise an Orbital Basis Rotation Module 262 configured to perform a basis rotation of the reduced quantum-derived data. The transformation may comprise rotating one or more quantum observables, including reduced density matrices, into a molecular orbital basis defined by eigenvectors of the corresponding representations of the molecular system, such asISUS Ref: 53524-006PCTeigenvectors of Hamiltonians associated with the respective geometrical configurations. The eigenvectors may be derived from electronic structure calculations and may encode electronic structure information associated with the molecular system.

[0130] In some exemplary embodiments, the reduced quantum -derived data may initially be obtained in a computational qubit basis, and Orbital Basis Rotation Module 262 may be configured to transform said data into a canonical orbital basis defined by the eigenvectors of the Hamiltonians corresponding to each geometrical configuration. Said transformation may comprise applying a basis rotation defined by the eigenvectors of the Hamiltonians, thereby generating orbital-basis-transformed reduced density matrices for each geometrical configuration.

[0131] In some exemplary embodiments, the transformation performed by Orbital Basis Rotation Module 262 may enable alignment between quantum measurement data and electronic structure information, such that the resulting transformed data reflects both the Hamiltonian-defined electronic environment and the quantum-derived response of the system. The transformed data may thereby provide a basis in which extracted features are predictive of a chemical property, whereas corresponding data in the computational qubit basis may not be predictive.

[0132] Feature Extraction Module 280 may be configured to extract features from the transformed quantum-derived data. In some exemplary embodiments, the features may be based on quantum observables derived from the reduced density matrices and variation of such observables across the plurality of geometrical configurations. The quantum observables may comprise, for example, orbital occupation, coherence, entropy, purity, cumulant values, mutual information, and additional correlation-related quantities.

[0133] Additionally, the features may comprise statistical descriptors computed across the plurality of geometrical configurations, such as mean, variance, slope, range, integral, and curvature, thereby forming a multi-dimensional feature vector representing a spatial correlation profile.

[0134] In some exemplary embodiments, the features may further comprise a combination of values at an equilibrium configuration and variation across the interatomic coordinate, and may be filtered based on statistical relevance.ISUS Ref: 53524-006PCT

[0135] Prediction Module 290 may be configured to predict the chemical property based on the extracted features. In some exemplary embodiments, Prediction Module 290 may apply a supervised machine-learning model, such as ridge regression, to map the feature vectors to corresponding chemical properties. The model may be calibrated using a calibration set of candidate systems, wherein feature vectors of the calibration-set systems are mapped to experimentally measured values of the chemical property. The calibrated model may then be applied to the feature vector of the target molecular system to predict the chemical property. In some embodiments, the prediction may be based on aggregated results obtained from multiple NMR quantum processors and may incorporate processor-specific fidelity metrics and confidence measures derived from cross-processor agreement.

[0136] In some exemplary embodiments, the prediction may further comprise computing a confidence metric, an uncertainty estimate, contributing features, measurement fidelity, and an applicability-domain indication associated with the prediction.

[0137] It may be noted that in some embodiments, a plurality of NMR quantum processors (of the one or more Quantum Hardware 210) may be assigned to perform redundant computation of a same sub-task, and the Classical Processing Unit 220 may be configured to compare results and determine a validated result based on statistical agreement, optionally using majority voting, weighted aggregation based on fidelity metrics, outlier rejection, averaging, or Bayesian inference. The Classical Processing Unit 220 may be further configured to aggregate measurement results from the plurality of NMR quantum processors to construct a spatial correlation profile across the at least one interatomic coordinate and to apply cross-processor fidelity calibration to normalize measurement data obtained from different NMR quantum processors.

[0138] Molecular Quantum Simulation Platform 200 may thereby implement a hybrid quantum-classical pipeline in which a plurality of Hamiltonians corresponding to different geometrical configurations are generated, a parameterized quantum state is optimized and reused across configurations, quantum measurements are performed to obtain reduced density matrices, the data is transformed into an orbital basis, features capturing spatial variation are extracted, and a machine-learning model is applied to predict a chemical property of the molecular system.ISUS Ref: 53524-006PCT

[0139] The pipeline may further be executed in a distributed manner across a plurality of independently operating NMR quantum processors, with dynamic orchestration by the Classical Processing Unit 220, including reassignment of sub-tasks, selective triggering of additional measurements based on confidence levels, and replacement of degraded processors with standby processors from a pool of pre-initialized NMR quantum processors maintained in a calibrated reference state.

[0140] Referring now to Figure 3A showing a flowchart diagram of a method for predicting a chemical property of a molecular system using a hybrid classical -quantum pipeline, in accordance with some exemplary embodiments of the disclosed subject matter.

[0141] In some exemplary embodiments, Method 300 may be applied to different domains, including drug discovery (e.g., by predicting binding affinity prediction), catalysis (e.g., enantioselectivity prediction), the effects of impurities in solid state material, single molecule and bulk magnet design, battery electrodes and electrolyte optimization, catalyst effect predictions such as FeMo Cofactor, or the like, thereby demonstrating general applicability across chemical classes. Additionally, or alternatively, Method 300 may be applied to experimentally validated molecular systems. As one example, Method 300 may be performed to predict binding affinity between a drug molecule and a biological target, such as CYP51 azole inhibitor compounds interacting with a CYP51 enzyme, in which binding affinity values (e.g., Kd or pKo) are predicted based on quantum -derived features. As another example, Method 300 may be performed for quantum prediction of enantioselectivity (ee%), such as in Rh-catalysed asymmetric hydrogenation of methyl 2-acetamidoacrylate using a ligand library, in which enantiomeric excess is predicted based on quantum observables derived from reduced density matrices.

[0142] In Step 310, a plurality of Hamiltonians associated with a molecular system being tested may be generated. In some exemplary embodiments, the plurality of Hamiltonians may be derived from a representation of the molecular system and may correspond to a plurality of geometrical configurations of the molecular system defined along at least one interatomic coordinate, such as a target bond within the molecular system being tested. The plurality of geometrical configurations may be associated with a plurality of interatomic distances along a target bond coordinate or target interaction coordinate. TheISUS Ref: 53524-006PCTplurality of interatomic distances may correspond to a bond formation or bond dissociation coordinate within the molecular system, such as a metal-ligand interaction, and may span a range from compressed geometries to dissociated geometries, for example on the order of angstrom-scale variations. As an example, in the binding affinity prediction, the plurality of interatomic distances may correspond to Fe-N coordination distances spanning a bond-formation coordinate for azole inhibitor binding. As another example, in enantioselectivity prediction, the plurality of interatomic distances may correspond to Rh-P bond distances associated with catalytic ligand coordination.

[0143] In some exemplary embodiments, Step 310 may comprise generating, using one or more classical computational processes, a set of reduced representations of the molecular system that capture electronic structure information across the plurality of geometrical configurations. The generated Hamiltonians may encode interactions within the molecular system while reflecting contributions from a larger electronic environment and may be represented in a form suitable for subsequent processing by quantum hardware.

[0144] In some exemplary embodiments, the plurality of Hamiltonians may form a distance-indexed set H(R,), H(R2), ..., H(RN), thereby providing a configurationdependent representation of the molecular system along the interatomic coordinate. The plurality of Hamiltonians may collectively define a representation of a potential energy surface or interaction profile of the molecular system, enabling subsequent analysis of geometry-dependent electronic behavior.

[0145] In some exemplary embodiments, Step 310 may be performed as shown in Figure 3B.

[0146] In Step 311, a molecular string representing the structure of the molecular system may be obtained. The molecular string representation may be configured to encode at least one of the atomic composition, connectivity, geometry, or electronic structure of the molecular system. In some exemplary embodiments, the molecular string may be a textbased representation of the structure of the molecular system being tested. As an example, the molecular string may be provided in Simplified Molecular Input Line Entry System (SMILES) format, InChi (TUPAC International Chemical Identifier) representation, CML (Chemical Markup Language), PDB (Protein Data Bank), XYZ A simple, plaintext format listing atom types and their Cartesian coordinates, or the like. Additionally,ISUS Ref: 53524-006PCTor alternatively, any other machine-readable molecular representation may be utilized, such as but not limited to a graph-based representation, a coordinate-based representation, an electronic representation, or the like.

[0147] In Step 312, a target interatomic interaction may be identified, and a plurality of interatomic distances may be defined. In some exemplary embodiments, the target interatomic interaction may correspond to a chemically relevant interaction associated with a property of the molecular system to be predicted, such as for example, a specific bond in the interaction, such as Fe-N, C-S, Rh-C, or another target bond or interaction coordinate in the molecular system.

[0148] In some embodiments, the identification of the target interatomic interaction is defined based on an input representation of the molecular system and further guided by electronic structure analysis, including frontier molecular orbital (FMO) theory. According to FMO theory, reactivity is governed primarily by the Highest Occupied Molecular Orbital (HOMO) and the Lowest Unoccupied Molecular Orbital (LUMO), wherein the HOMO acts as an electron donor and the LUMO acts as an electron acceptor. Spatial distributions of the HOMO and LUMO may be analyzed to identify reactive sites and corresponding interatomic interactions, while a H0M0-LUM0 energy gap may indicate kinetic stability. Accordingly, the target interaction may correspond to a bond associated with dominant H0M0 / LUM0 contributions, such as a metal-ligand coordination bond (e.g., Fe-N in CYP51), thereby enabling selection of an active space that captures binding-relevant electronic behavior, including a reduced active space representation comprising frontier orbitals (e.g., CASCI(2e,2o)).

[0149] The plurality of interatomic distances may define a scan along the corresponding coordinate from a compressed geometry to a dissociated geometry. The defined distances may be selected to capture variations in electronic structure along a reaction coordinate, a binding coordinate, or a potential energy surface.

[0150] In Step 313, a plurality of three-dimensional molecular geometries of the molecular system may be generated. In some exemplary embodiments, each molecular geometry may correspond to a respective interatomic distance defined in Step 312, such that the plurality of geometries collectively defines a spatial scan of the molecular system. In some exemplary embodiments, a classical molecular modeling tool may be used toISUS Ref: 53524-006PCTgenerate three-dimensional geometry from the molecular representation. As an example, an RDKit or another cheminformatics library may be utilized.

[0151] In Step 314, a Density Functional Theory (DFT) calculation may be performed on the molecular system. In some exemplary embodiments, the DFT calculation may be performed for each of the plurality of molecular geometries thereby recomputing an electronic structure of the molecular system at each interatomic distance. In some exemplary embodiments, the DFT calculation may comprise determining molecular orbital coefficients, electron density distributions, and / or one-electron and two-electron integrals associated with the molecular system, which may subsequently be used to construct one or more Hamiltonian representation of the molecular system. The DFT calculation may employ a selected exchange-correlation functional and basis set, such as B3LYP / 6-31G*, although other functionals and basis sets may be used. The DFT calculation may be performed using a quantum chemistry software package, such as PySCF, Gaussian, ORCA, or any other suitable computational chemistry tool. In some exemplary embodiments, the DFT calculation may capture long-range electrostatic interactions and environment-dependent electronic effects of the molecular system, such that the resulting electronic structure information reflects contributions from atoms and orbitals beyond a subsequently defined reduced active space. Referring to the binding affinity prediction example, the DFT calculation may be performed for each of a plurality of Fe-N distances (e.g., approximately 5 distances spanning about 1.95 A to 3.00 A), thereby recomputing electronic structure along a binding coordinate. Referring to the enantioselectivity prediction example, the DFT calculation may be performed across a plurality of Rh-P distances (e.g., 5-7 distances spanning approximately 2.00 A to 4.50 A).

[0152] In Step 315, molecular orbital coefficients may be obtained based on the DFT calculation. In some exemplary embodiments, the molecular orbital coefficients may provide a basis for constructing a reduced Hamiltonian associated with the molecular system, and for defining a basis for subsequent transformations.

[0153] In Step 316, the molecular system may be projected onto a reduced active-space representation. In some exemplary embodiments, the active space may capture a subset of orbitals (e.g., 2 orbitals) selected to represent the primary bonding interaction, while the remaining orbitals, including virtual orbitals (which may number in the hundreds),ISUS Ref: 53524-006PCTare implicitly encoded through the DFT-derived Hamiltonian structure. The reduced active-space representation may comprise a CASCI(2e,2o) active space corresponding to orbitals most relevant to the target chemical interaction. Referring to the binding affinity prediction example, the selected active space may correspond to a metal-ligand bonding interaction (e.g., Fe dz2orbital and ligand lone-pair orbital), while hundreds of virtual orbitals remain outside the active space.

[0154] In some exemplary embodiments, classical diagonalization of the reduced Hamiltonian may fail to capture contributions from virtual orbitals, resulting in reduced predictive power for certain properties. However, the reduced Hamiltonian may still serve as an effective interface between classical and quantum processing stages. In such embodiments, classical diagonalization of the reduced Hamiltonian may produce trivial density matrices (e.g., Tr(y1)=2), thereby failing to capture virtual orbital contributions relevant to the predicted property.

[0155] In Step 317, for each of the plurality of interatomic distances, a corresponding qubit Hamiltonian may be generated based on the reduced active space. In some exemplary embodiments, the corresponding qubit Hamiltonians may be output as distance-indexed Pauli operator coefficients for execution on quantum hardware, or any other linear combination of Pauli operators. In some exemplary embodiments, the reduced Hamiltonian may be represented as a 4^4 matrix and mapped to a two-qubit Hilbert space.

[0156] In Step 320, a quantum state associated with the molecular system may be optimized. In some exemplary embodiments, an optimized quantum state corresponding to at least one of the plurality of Hamiltonians generated in Step 310 may be determined using quantum hardware.

[0157] In some exemplary embodiments, the optimization process may involve a hybrid classical-quantum procedure, such as in a Variational Quantum Eigensolver (VQE) framework, in which quantum measurements are used to evaluate a cost function and classical processing is used to update parameters of the quantum state. The optimization may be performed using a variational quantum algorithm (via VQE), with respect to a selected Hamiltonian corresponding to an equilibrium configuration of the molecular system, denoted H(Req). The equilibrium configuration may correspond to an interatomic distance Reqat or near an energy minimum of the molecular system along the selectedISUS Ref: 53524-996PCTinteratomic coordinate, such as a bond length associated with a stable or reference geometry. Referring to the binding affinity prediction example, the VQE optimization may be performed at an equilibrium Fe-N distance, such that optimized parameters 9* are subsequently reused across all other distances in the scan.

[0158] In some exemplary embodiments, the optimization may be performed with respect to the selected equilibrium Hamiltonian H(Req), such that the resulting quantum state approximates a low-energy state of the molecular system under the equilibrium configuration. The selected equilibrium Hamiltonian may be encoded as a linear combination of Pauli strings and transmitted to a quantum hardware, such as an NMR processor. The variational quantum algorithm (e.g., VQE) optimizes a parameterized quantum circuit to minimize a cost function, such as the expected energy (v(9)|H(Req)|v(9)). A classical optimizer may iteratively update circuit parameters until convergence or until a stopping criterion is reached. The resulting quantum state may be represented by optimized circuit parameters 9* that approximate a low-energy state associated with the selected Hamiltonian.

[9159] In some exemplary embodiments, the quantum hardware may comprise an analog or digital quantum processor, such as an NMR-based quantum system, and the resulting optimized quantum state may reflect interactions between the encoded Hamiltonian and physical degrees of freedom of the quantum system. The optimized quantum state may serve as a probe for subsequent measurements across the plurality of Hamiltonians, thereby enabling extraction of geometry-dependent quantum observables associated with the molecular system. In some exemplary embodiments, the quantum hardware may comprise an NMR-based system in which nuclear spin dynamics are coupled to electronic degrees of freedom, such that the optimized quantum state reflects interactions beyond a programmed Hilbert space.

[0169] In some exemplary embodiments, maintaining the optimized quantum state fixed across multiple Hamiltonians corresponding to different geometrical configurations may enable extraction of a configuration-dependent quantum response that is not obtainable from independent optimization at each configuration.

[9161] In some exemplary embodiments, Step 329 may be performed as shown in Figure 3C.ISUS Ref: 53524-006PCT

[0162] In Step 322, a selected Hamiltonian, such as an equilibrium Hamiltonian H(Req) may be encoded as a linear combination of Pauli strings. The linear combination of Pauli strings may be, for example of the form: H = Si Ci Pi. Each Pi is a tensor product of singlequbit Pauli operators selected from {I, X, Y, Z}, such that Pi E {I, X, Y, Z}®n, and Ci are coefficients derived from electronic structure calculations. In some exemplary embodiments, the Pauli strings comprise at least one of: single-qubit terms (e.g., Zo), two-qubit interaction terms (e.g., ZoZi, XoXi, YoYi), and higher-order multi-qubit terms. In some embodiments, the Pauli strings are obtained via a fermion-to-qubit mapping including at least one of: Jordan-Wigner, Bravyi-Kitaev, or parity mapping.

[0163] In other embodiments, electronic structure calculations may comprise Hartree-Fock, post-Hartree-Fock methods, semi-empirical methods, quantum Monte Carlo, or machine-learned electronic structure models.

[0164] In Step 324, a parameterized quantum circuit may be executed on quantum hardware, such as NMR hardware. In some exemplary embodiments, the quantum hardware may operate at room temperature.

[0165] In Step 326, circuit parameters may be iteratively optimized via a classical feedback loop to minimize a cost function, such as an expected energy associated with the selected Hamiltonian. In some exemplary embodiments, the cost function may be evaluated based on measurement outcomes obtained from the quantum hardware. In some exemplary embodiments, a classical optimizer may update circuit parameters based on measured results until convergence or until a stopping criterion is reached.

[0166] In Step 328, optimized parameters may be provided. In some exemplary embodiments, the optimized parameters may comprise a set of parameters 0* that approximate a low-energy state associated with the selected Hamiltonian. The optimized parameters may be maintained fixed and reused in subsequent measurements across the plurality of Hamiltonians corresponding to different geometrical configurations, such as across the plurality of interatomic distances.

[0167] In Step 330, the molecular system may be evaluated across a plurality of geometrical configurations by performing quantum measurements to reconstruct reduced density matrices (RDMs). For each Hamiltonian H(Rk) corresponding to a respective interatomic distance, the Hamiltonian may be applied to quantum hardware using theISUS Ref: 53524-006PCTequilibrium quantum state |y(9*)) and measured. Referring to the binding affinity prediction example, measurements may be performed across multiple distances (e.g., 5 distances per compound), using the same quantum state |y(9*)) measured at equilibrium, thereby generating a spatial correlation profile associated with the molecular interaction.

[0168] In some exemplary embodiments, the quantum measurements may comprise performing quantum state tomography to obtain expectation values of a plurality of observables, such as Pauli operators, from which one-electron and two-electron reduced density matrices (1-RDM and 2-RDM) may be reconstructed for each geometrical configuration. The resulting distance-indexed set of reduced density matrices may define a spatial correlation profile of the molecular system across the plurality of interatomic distances.

[0169] It may be noted that maintaining the optimized parameters 9* (derived at equilibrium distance) fixed while varying the Hamiltonian across geometrical configurations enables the quantum state to act as a probe of the molecular system under different interaction conditions. Because each Hamiltonian H(Rk) encodes electronic structure information derived from full-molecule calculations, the measurements obtained across the plurality of configurations may capture geometry-dependent quantum responses, including contributions associated with virtual orbitals that are not explicitly represented in the reduced active space. In some exemplary embodiments, measured reduced density matrices may exhibit properties indicative of coupling to degrees of freedom beyond the reduced active space, including reduced purity (Tr(p2)<l), reduced trace of the one-electron density matrix, and non-zero cumulant contributions.

[0170] In some exemplary embodiments, this multi -di stance measurement process may amplify latent correlation information that is not accessible through classical diagonalization of the corresponding Hamiltonians, thereby enabling generation of compound-specific signatures of the molecular system. Referring to the binding affinity prediction example, the resulting spatial correlation profile may encode compoundspecific signatures associated with binding affinity, which are not present in classical Hamiltonian eigenvalues.

[0171] In some embodiment of this invention, multiple VQE optimizations at multiple distances may be used together with their corresponding 1 RDM and 2 RDMISUS Ref: 53524-996PCTmeasurements to capture multiple interactions at multiple distances, and use these as the basis for the correlation training.

[0172] In some exemplary embodiments, Step 330 may be performed as shown in Figure 3D.

[0173] In Step 332, for each Hamiltonian H(R_k), a fixed quantum state may be prepared using 9* and quantum measurements may be performed. In some exemplary embodiments, the fixed parameters obtained during VQE optimization (the optimized parameters 9*) may be held constant while the Hamiltonian changes across the plurality of interatomic distances.

[0174] In Step 334, reduced density matrices may be reconstructed based on measurement outcomes obtained from the quantum hardware. In some exemplary embodiments, the measurement outcomes may comprise expectation values of Pauli operators obtained via quantum state tomography. In some exemplary embodiments, the reconstructed reduced density matrices may comprise a one-electron reduced density matrix (1-RDM) and a two-electron reduced density matrix (2 -RDM) for each geometrical configuration (e.g., interatomic distance).

[9175] In Step 336, a spatial correlation profile may be generated. In some exemplary embodiments, the spatial correlation profile may comprise a distance-indexed collection of reconstructed reduced density matrices or observables derived therefrom, representing variation of quantum properties of the molecular system across the plurality of geometrical configurations.

[9176] In Step 349, the reconstructed data may be transformed into an orbital basis. In some exemplary embodiments, density matrices data obtained from quantum hardware in a computational basis may be transformed into a molecular orbital basis defined by eigenvectors of a Hamiltonian H(Rk) corresponding to each geometrical configuration (distance). The eigenvectors of the Hamiltonian may be computed based on electronic structure information derived from classical calculations, such as Density Functional Theory (DFT), and may encode electrostatic and environmental effects of the molecular system beyond a reduced active space. The eigenvectors encode the full molecular electrostatic environment through the DFT calculation. Referring to the binding affinity prediction example, the Hamiltonian eigenvectors define an orbital basis in whichISUS Ref: 53524-006PCTbinding-relevant information becomes observable, whereas features in the computational basis may lack predictive significance.

[0177] In some exemplary embodiments, the transformation may comprise applying a basis rotation to reconstructed reduced density matrices, such that the transformed density matrices are expressed in a basis aligned with molecular orbitals associated with the molecular system. The transformation into the molecular orbital basis enables integration of (i) electronic structure information encoded in the Hamiltonian eigenvectors and (ii) quantum measurement data obtained from quantum hardware, thereby generating a representation that reflects both classical and quantum contributions.

[0178] In some exemplary embodiments, the Hamiltonian eigenvectors may define directions in an orbital space corresponding to chemically relevant interactions, while quantum measurements capture responses of a quantum state under the Hamiltonian. The transformation aligns the measured data with the orbital space defined by the eigenvectors, enabling extraction of geometry-dependent quantum features.

[0179] It may be noted that neither the Hamiltonian representation alone nor measurement data in the computational basis may be sufficient to generate predictive features, and the transformation into the molecular orbital basis enables extraction of information not accessible from either independently. Referring to the binding affinity prediction example, classical diagonalization provides directional information without signal, while raw hardware measurements provide signal without direction, and the transformation into the orbital basis combines both to produce predictive features.

[0180] In some exemplary embodiments, the transformed density matrices may reflect both classical electronic-structure information and hardware-based quantum measurements, and may be provided as orbital-basis density matrices corresponding to the plurality of geometrical configurations.

[0181] In some exemplary embodiments, Step 340 may be performed as shown in Figure 3E.

[0182] In Step 342, eigenvectors of the Hamiltonian H(Rk) corresponding to each geometrical configuration may be computed.ISUS Ref: 53524-006PCT

[0183] In Step 344, a molecular orbital basis may be defined based on the computed eigenvectors, such that the basis corresponds to orbitals associated with the molecular system at the respective geometrical configuration.

[0184] In Step 346, reconstructed density matrices obtained from quantum measurements may be transformed from the computational basis into the molecular orbital basis. The transformation may comprise applying a unitary rotation defined by the eigenvectors to the reconstructed reduced density matrices. Orbital-basis-rotated density matrices at each distance point may be generated.

[0185] In Step 350, features may be extracted from the orbital -basis-transformed density matrices by computing quantum observables at each geometrical configuration and deriving statistical descriptors across the plurality of geometrical configurations.

[0186] In some exemplary embodiments, a multi-dimensional feature vector may be generated based on quantum measurement data represented in an orbital-basis spatial correlation profile. Quantum observables may be computed from orbital-basis-rotated density matrices at each geometrical configuration, and spatial statistics may be derived across the plurality of geometrical configurations.

[0187] In some exemplary embodiments, the extracted features may comprise two categories of observables: (i) primary quantum observables computed at each geometrical configuration, and (ii) statistical descriptors characterizing variation of the observables across the spatial correlation profile. The primary quantum observables may include one or more of: orbital occupation numbers (e.g., nt, n2), quantum coherence, von Neumann entropy, purity, virtual -orbital trace deficit defined as n virt = N electrons - Tr(p’), Wick cumulant tensor (A2), concurrence, mutual information, correlation strength, natural polarization, pair occupation probabilities, measurement fidelity, or the like. The spatial statistics may include one or more of: mean, standard deviation, slope, minimum, maximum, range, equilibrium value, peak distance, integral, curvature, skewness of respective observables across the spatial correlation profile, or the like. The spatial statistics may be computed for each of the primary quantum observables across the plurality of geometrical configurations, thereby capturing geometry -dep endent variation of quantum properties of the molecular system. Referring to the binding affinity prediction example, extracted features may include cumulant-based correlation measures,ISUS Ref: 53524-006PCTequilibrium coherence values, and distance-dependent orbital occupation ranges, which correspond to physical quantities associated with electron correlation and orbital mixing.

[0188] In some exemplary embodiments, the features may be aggregated into a feature vector. The feature vector may comprise a plurality of features derived from combinations of the primary quantum observables and the spatial statistics, including equilibrium values associated with one or more observables. In some exemplary embodiments, the feature vector may comprise on the order of hundreds of features per molecular system, In some exemplary embodiments, the feature vector may comprise on the order of tens to hundreds of features per molecular system (e.g., approximately 1 GO-242 features depending on filtering and aggregation procedures, derived from combinations of observables and statistical descriptors, (about 22 primary observables x 10 spatial statistics + 22 equilibrium values), pre-filtered to the top 100 by Pearson correlation with training binding affinities). In some embodiments, Pearson correlation will not be used since it may discard nonlinear features.

[0189] In some exemplary embodiments, the feature vector may be pre-filtered to a subset of selected features, such as top-ranked features according to a statistical relevance measure, such as Pearson correlation or nonlinear selectors with training binding affinities. Referring to the enantioselectivity prediction example, feature extraction may additionally include face-dependent features (e.g., re-face and si-face observables) and differences therebetween, enabling prediction of asymmetric outcomes.

[0190] In some embodiments, the number of selected features is limited to a small subset, such as fewer than 10 features, fewer than 5 features, or fewer than 3 features, depending on dataset size and application.

[0191] In some exemplary embodiments, Step 350 may be performed as shown in Figure 3F.

[0192] In Step 352, quantum observables may be computed from density matrices. The quantum observables may comprise one or more of orbital occupations, coherence, entropy, purity, virtual-orbital trace deficit, Wick cumulant tensor, concurrence, mutual information, correlation strength, natural polarization, pair occupation probabilities, and / or measurement fidelity.ISUS Ref: 53524-006PCT

[0193] In Step 354, statistical descriptors across distances may be derived. The statistical descriptors may comprise one or more of mean, standard deviation, slope, minimum, maximum, range, equilibrium value, peak distance, integral, curvature, and / or skewness of respective observables across the spatial correlation profile.

[0194] In Step 356, the extracted data may be aggregated into a feature vector and feature selection may be performed. In some exemplary embodiments, the feature vector may comprise approximately 242 total features per molecular system and may be prefiltered to a subset of selected features, such as top-ranked features according to Pearson correlation with training binding affinities.

[0195] In Step 360, a property of the molecular system may be predicted using a machine-learning model based on the extracted feature vector. In some exemplary embodiments, a supervised machine-learning model may map the feature vector to a predicted chemical property of the molecular system, such as a binding affinity. In some exemplary embodiments, the machine-learning model may be trained according to a calibration pipeline comprising one or more of the following design principles: (1) Chemistry-Specific Calibration; (2) Exhaustive Combinatorial Feature Selection; (3) Regularized Regression; and (4) Cross-Validation and Statistical Validation. Referring to the enantioselectivity prediction example, the model may be trained on a ligand library (e.g., supraphos ligands) and validated on both samemechanism and different-mechanism compounds to assess generalization and mechanism specificity.

[0196] In some exemplary embodiments, the model may be trained using calibration data comprising compounds sharing a common bonding mechanism with the prediction target, such that calibration is specific to a given chemical interaction type. Models trained on one bonding mechanism may not be transferable to different bonding mechanisms. Referring to the enantioselectivity prediction example, predictive accuracy may be significantly higher for compounds sharing a common catalytic mechanism, and significantly reduced for compounds with different binding mechanisms.

[0197] In some exemplary embodiments, feature selection may comprise evaluating combinations of a limited number of features, such as combinations of k = 1, 2, or 3 features, to identify predictive feature subsets. Limiting the number of selected features may reduce overfitting in datasets having a limited number of samples.ISUS Ref: 53524-006PCT

[0198] In some exemplary embodiments, the machine-learning model may comprise a regularized regression model, such as ridge regression, with a regularization parameter selected based on cross-validation.

[0199] In some exemplary embodiments, model evaluation may comprise cross-validation, such as leave-one-out cross-validation (LOO-CV), to estimate predictive performance. Statistical validation may comprise permutation testing, in which the prediction pipeline is repeated on shuffled labels to generate a null distribution and estimate statistical significance.

[0200] In some exemplary embodiments, the permutation testing may comprise multiple iterations, such as on the order of hundreds or thousands of permutations and may include repeating feature selection to control for multiple-testing bias.

[0201] In some exemplary embodiments, the output may include a predicted chemical property, such as predicted pIC50 or pKo, together with associated confidence metrics. In some exemplary embodiments, the associated confidence metrics may comprise one or more of: an uncertainty estimate, contributing features, measurement fidelity information, and / or a confidence indication associated with an applicability domain of the model.

[0202] In some exemplary embodiments, Step 360 may be performed as shown in Figure 3G.

[0203] In Step 362, a supervised regression model may be trained and applied using selected features. In some exemplary embodiments, training may include cross-validation, such as leave-one-out cross-validation, feature selection procedures, or the like.

[0204] In some exemplary embodiments, the supervised model may be trained using calibration data comprising compounds sharing a common bonding mechanism with the prediction target, such that the model captures chemistry-specific structure-property relationships. Additionally, or alternatively, models trained on one bonding mechanism may exhibit reduced predictive performance when applied to compounds associated with different bonding mechanisms, thereby indicating that the learned feature representation encodes mechanism-specific information.ISUS Ref: 53524-006PCT

[0205] In some exemplary embodiments, feature combinations may be evaluated under cross-validation to identify predictive subsets of features, and regularization parameters may be selected during training. In some exemplary embodiments, combinations of a limited number of features, such as k < 3 features, may be evaluated to reduce overfitting in datasets comprising a limited number of compounds. The regression model may comprise ridge regression. In some exemplary embodiments, the regression model may comprise a regularized regression model, such as ridge regression with a regularization parameter selected based on cross-validation.

[0206] In some exemplary embodiments, statistical validation may comprise permutation testing in which the pipeline, including feature selection, is repeated using shuffled labels to estimate statistical significance. In some exemplary embodiments, permutation testing may comprise a plurality of iterations, such as on the order of hundreds or thousands of permutations, to construct a null distribution and control for multiple-testing bias. In some exemplary embodiments, predictive performance may depend on features derived from quantum hardware measurements in combination with Hamiltonian-derived representations, such that neither classical Hamiltonian analysis nor raw hardware measurements alone provide sufficient predictive accuracy. Referring to the binding affinity prediction example, predictive performance may be substantially improved when using quantum hardware-derived features (e.g., Q2~0.95) compared to classical Hamiltonian-based features alone (e.g., Q2~0 or negative), thereby demonstrating that the predictive signal arises from quantum measurement data.

[0207] In Step 364, a predicted chemical property may be generated with associated confidence metrics. In some exemplary embodiments, the predicted chemical property may comprise predicted pIC50, pKo, enantio selectivity (ee%), Gibbs free energy binding, or another chemistry-related property. The associated confidence metrics may comprise an uncertainty estimate, feature contribution information, measurement fidelity information, and / or a confidence flag indicating whether the prediction falls within a chemical applicability domain.

[0208] In some exemplary embodiments, the selected features contributing to the prediction may be associated with physically interpretable quantities, such as quantum coherence, entropy, or cumulant-based measures of electron correlation, thereby enabling interpretation of the predicted property in terms of underlying molecular interactions.ISUS Ref: 53524-006PCTPredictive accuracy may be higher for compounds within a defined chemical applicability domain and may degrade for compounds outside the domain, thereby providing an indication of model validity and scope.

[0209] Referring now to Figure 4, showing a hybrid quantum-classical computing system in accordance with some exemplary embodiments of the disclosed subject matter.

[0210] In some exemplary embodiments, hybrid quantum-classical computing System 400 may be configured to perform distributed quantum computation using a plurality of NMR quantum processors coordinated by a classical control layer.

[0211] System 400 comprises a Classical Control System 410 communicatively coupled to a plurality of NMR quantum processors 430A-430N. Classical Control System 410 is configured to receive a Computational Task / Problem Input 402 and to generate a Solution / Output 404 based on aggregated quantum computation results.

[0212] Classical Control System 410 includes Task Decomposition and Scheduling 412, which is configured to decompose the computational task into a plurality of sub-tasks. The plurality of sub-tasks may comprise fragments of a larger computational problem, including subsets of variables, Hamiltonian terms, or representations of a molecular system corresponding to different geometrical configurations along at least one interatomic coordinate.

[0213] Assignment of Sub-Tasks 414 is configured to assign respective sub-tasks to corresponding NMR quantum processors 430A-430N. Sub-task information is transmitted as Sub-Task Assignment / Hamiltonian / Parameters 422A-422N, which may include parameterized quantum circuits, Hamiltonian fragments, or optimization problem representations.

[0214] Orchestration and Synchronization 416 is configured to coordinate execution of quantum operations across the plurality of NMR quantum processors 430A-430N, enabling independent and parallel operation of the processors. In some embodiments, sub-tasks may be distributed such that fragments overlap to enforce solution consistency and enable correlation between sub-tasks.

[0215] Each NMR quantum processor 430A-430N comprises an NMR Cell 431A-43 IN, Nuclear Spins (Qubits) 432A-432N, and an RF Coil & Probe 434A-434N. The NMR Cell contains an ensemble of molecules in which nuclear spins serve as qubits. TheISUS Ref: 53524-006PCTNuclear Spins 432A-432N are configured to represent quantum states and may comprise spin-1 / 2 nuclei. In some embodiments, each processor comprises between 2 and 15 qubits.

[0216] RF Coil & Probe 434A-434N is configured to implement quantum operations using radiofrequency pulse sequences applied to Nuclear Spins 432A-432N. These quantum operations may include single-qubit rotations and multi-qubit operations enabled by spin-spin coupling interactions. The RF Coil & Probe 434A-434N is further configured to measure quantum states and generate measurement signals.

[0217] Due to the ensemble nature of NMR systems, each NMR Cell 431A-431N contains a macroscopic number of molecules performing identical quantum operations in parallel, resulting in ensemble-averaged measurement signals. These measurements provide expectation values of quantum observables without requiring projective collapse.

[0218] Each NMR quantum processor 430A-430N generates Measurement Data / Fidelity Metrics 424A-424N, which are transmitted to Classical Control System 410. Measurement Aggregation and Correlation 418 is configured to aggregate measurement data across the plurality of processors to determine a global solution. Aggregation may include combining expectation values, constructing correlation profiles, or integrating results from overlapping fragments.

[0219] Fidelity Calibration and Validation 420 is configured to evaluate the reliability of measurement data and normalize results obtained from different processors. In some embodiments, Classical Control System 410 assigns multiple NMR quantum processors to perform redundant computation of a same sub-task. Measurement data obtained from redundant processors is compared, and a validated result is determined based on statistical agreement, including majority voting, weighted averaging, or outlier rejection.

[0220] In certain embodiments, at least three NMR quantum processors perform identical sub-tasks, and validated results are determined using majority voting. In other embodiments, validated results are determined by weighting measurement data according to fidelity metrics associated with each processor.

[0221] Classical Control System 410 may further be configured to obtain partial measurement data from a subset of NMR quantum processors 430A-430N and determine a confidence level based on statistical agreement. When the confidence level falls belowISUS Ref: 53524-006PCTa threshold, additional measurements may be selectively triggered. In some embodiments, diagonal quantum observables are measured initially, and off-diagonal observables are measured only when required for complete state reconstruction.

[0222] In some embodiments, system 400 further comprises a pool of standby NMR quantum processors (not shown). Classical Control System 410 is configured to monitor fidelity metrics of active processors and to replace a degraded processor with a standby processor when the fidelity metric falls below a predetermined threshold. Processors may be assigned states including active, standby, and recalibrating, and may transition between states based on performance metrics.

[0223] In some embodiments, recalibration includes thermal re-equilibration and RF pulse calibration. Sub-tasks assigned to degraded processors may be reassigned to standby processors without interrupting overall computation.

[0224] The plurality of NMR quantum processors 430A-430N may comprise liquidstate NMR systems, solid-state NMR systems, or a combination thereof. Liquid-state NMR systems may provide ensemble-averaged measurements in solution, while solid-state NMR systems may operate under controlled conditions and may employ Magic Angle Spinning to enhance coherence and spectral resolution. In some embodiments, hyperpolarization techniques may be applied to enhance signal strength.

[0225] The hybrid architecture illustrated in Figure 4 enables scalable quantum computation by distributing computational tasks across a plurality of independently operating NMR quantum processors coordinated by Classical Control System 410. This approach leverages parallel execution, ensemble averaging, and classical coordination to overcome scalability limitations of conventional NMR quantum computing systems while maintaining high fidelity and operational robustness.

[0226] The embodiments described with reference to Figure 4 are illustrative and nonlimiting. Various modifications and alternative implementations may be employed without departing from the scope of the invention.

[0227] In some embodiments, the plurality of NMR quantum processors 430A-430N comprises liquid-state NMR systems, wherein each NMR Cell 431A-431N contains a solution comprising a macroscopic ensemble of molecules. In such embodiments, Nuclear Spins 432A-432N are implemented using nuclear spins within moleculesISUS Ref: 53524-006PCTdissolved in the solution, and quantum operations are applied uniformly across the ensemble using RF Coil & Probe 434A-434N.

[0228] In these embodiments, measurement signals obtained from each NMR quantum processor correspond to ensemble-averaged observables, thereby enabling quantum measurement without projective state collapse. The ensemble nature allows simultaneous execution of quantum operations across a large number of molecular instances, providing inherent signal amplification and statistical averaging. In some implementations, the solution may comprise heteronuclear molecular systems enabling selective addressing of qubits via distinct resonance frequencies. In other implementations, homonuclear systems may be employed with selectively shaped pulse sequences.

[0229] In alternative embodiments, the plurality of NMR quantum processors 430A-430N comprises solid-state NMR systems, wherein each NMR Cell 431 A-43 IN contains a representative material sample. In such embodiments, Nuclear Spins 432A-432N are implemented using nuclear spins within the solid-state material, and RF Coil & Probe 434A-434N is configured to apply quantum operations to the material sample.

[0230] In some embodiments, each NMR quantum processor 430A-430N further comprises a Magic Angle Spinning mechanism configured to rotate the material sample at approximately 54.74 degrees relative to an applied magnetic field. The spinning may be performed at frequencies in the range of 10-70 kHz to reduce anisotropic interactions and enhance coherence time of Nuclear Spins 432A-432N. In certain implementations, the solid-state NMR quantum processors operate under controlled environmental conditions, including temperature, pressure, and atmosphere, enabling simulation of material properties under realistic operating conditions.

[0231] In further embodiments, the plurality of NMR quantum processors 430 A-43 ON comprises a combination of liquid-state and solid-state NMR systems. In such hybrid configurations, Classical Control System 410 is configured to assign different types of sub-tasks to different processor types. For example, measurement-intensive sub-tasks may be assigned to liquid-state NMR quantum processors, while environment-specific simulations or material-dependent computations may be assigned to solid-state NMR quantum processors. Measurement data obtained from both types of processors may be aggregated by Measurement Aggregation and Correlation 418 to determine Solution / Output 404.ISUS Ref: 53524-006PCT

[0232] In some embodiments, one or more NMR quantum processors 430A-430N incorporate hyperpolarization techniques to enhance signal strength of Nuclear Spins 432A-432N. Such techniques may include Dynamic Nuclear Polarization (DNP), ParaHydrogen Induced Polarization (PHIP), optical pumping, or spin-exchange optical pumping. Hyperpolarization increases the population difference between nuclear spin states, thereby improving signal-to-noise ratio and enabling more accurate measurement data 424A-424N. In certain implementations, hyperpolarization enables each NMR quantum processor to support a larger number of effective qubits and more complex quantum circuits within coherence time constraints.

[0233] In some embodiments, Classical Control System 410 is configured to assign a plurality of NMR quantum processors 430A-430N to perform redundant computation of a same sub-task. In such embodiments, Measurement Data / Fidelity Metrics 424A-424N from multiple processors are compared, and validated results are determined based on statistical agreement. In some implementations, at least three processors are assigned to each sub-task, and majority voting is used to determine a validated result. In other implementations, results are weighted according to fidelity metrics. In certain embodiments, statistical aggregation includes one or more of outlier rejection, averaging, or Bayesian inference.

[0234] In some embodiments, system 400 further comprises a pool of standby NMR quantum processors in addition to active processors 430A-430N. In such embodiments, Classical Control System 410 continuously monitors fidelity metrics associated with Measurement Data 424A-424N and detects degradation of an active processor when its fidelity metric falls below a predetermined threshold. In response to such detection, the Classical Control System 410 reassigns the sub-task associated with the degraded processor to a standby processor and transitions the degraded processor to a recalibrating state. Processors may be dynamically assigned states selected from active, standby, and recalibrating, and transitions between states may be performed during operation without interrupting overall computation.

[0235] In some embodiments, Classical Control System 410 is configured to perform adaptive measurement strategies based on Measurement Data 424A-424N. In such embodiments, a first set of measurement data comprising expectation values of diagonal observables is obtained from one or more NMR quantum processors 430A-430N. BasedISUS Ref: 53524-006PCTon this data, Classical Control System 410 determines whether off-diagonal observables are expected to carry significant information. When off-diagonal observables are determined to be significant, a second set of measurements is selectively triggered on corresponding processors to obtain additional data for full quantum state reconstruction. When such observables are not significant, the second set of measurements may be

[0236] In some embodiments, Classical Control System 410 is configured to iteratively update sub-task assignments based on intermediate Measurement Data 424A-424N.

[0237] In such embodiments, correlations between sub-tasks may be identified and incorporated into subsequent Sub-Task Assignment / Hamiltonian / Parameters 422A-422N, enabling refinement of quantum operations across iterations. This iterative feedback mechanism may be employed in variational quantum algorithms, optimization problems, and molecular simulations.

[0238] In some embodiments, System 400 is configured to solve combinatorial optimization problems by mapping the computational task into a QUBO formulation, which is decomposed into sub-problems compatible with the qubit capacity of individual NMR quantum processors 430A-430N. In other embodiments, system 400 is configured to perform molecular simulations, including evaluation of electronic structure, correlation effects, or reaction pathways, wherein sub-tasks correspond to molecular fragments or geometrical configurations. In further embodiments, system 400 is configured for machine learning applications, wherein outputs of NMR quantum processors 430A-430N are used as features or kernels within a classical learning framework.

[0239] It will be appreciated that the above-described embodiments may be combined in various ways. For example, hyperpolarization techniques may be applied in both liquid-state and solid-state NMR systems, and adaptive measurement strategies may be used in conjunction with redundant computation and dynamic processor pooling. Accordingly, the scope of the invention is not limited to the specific embodiments described herein, but is defined by the appended claims.ISUS Ref: 53524-006PCT

[0240] The present invention may be a system, a method, and / or a computer program product. The present invention may further comprise a hybrid cl as si cal -quantum computing system. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing one or more processors, one or more classical processors, one or more quantum processors, or a combination thereof, to carry out aspects of the present invention.

[0241] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. In some embodiments, instructions may additionally define control sequences for quantum operations, including gate operations, pulse sequences, and / or measurement protocols applied to qubits. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0242] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. In some embodiments, such computing / processing devices may include quantum processing units (QPUs), quantum simulators, or cloud-accessible quantum hardware platforms. TheISUS Ref: 53524-006PCTnetwork may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0243] Computer readable program instructions for carrying out operations of the present invention may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages and / or quantum programming instructions expressed in quantum programming languages or frameworks (e.g., circuit descriptions, pulse-level instructions, or intermediate representations for quantum execution). The computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, execution may be distributed between classical computing resources and quantum hardware, wherein classical processors perform optimization, preprocessing, or post-processing, and quantum processors perform state preparation, evolution, and / or measurement. In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present invention.ISUS Ref: 53524-006PCT

[0244] Aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. In embodiments involving quantum computation, such blocks may represent quantum circuits, quantum state transformations, measurement stages, or hybrid classical -quantum feedback loops. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer readable program instructions.

[0245] These computer readable program instructions may be provided to a processor of a classical processor, quantum processor, or hybrid processing system to produce a machine, including systems configured to manipulate qubits, quantum states, or quantum registers, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0246] The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device, and / or quantum hardware system, to cause a series of operational steps, including quantum state preparation, application of quantum gates, entanglement operations, and measurement of quantum observables, to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0247] The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In quantum embodiments, blocks may correspond to quantum operations that may beISUS Ref: 53524-006PCTexecuted sequentially, in parallel, or in superposition, and may involve probabilistic or measurement-dependent outcomes. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.

[0248] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0249] The corresponding structures, materials, acts, and equivalents of all means or step plus function elements in the Claims below, including implementations involving quantum hardware, quantum -cl as si cal interfaces, and hybrid computational architectures, are intended to include any structure, material, or act for performing the function in combination with other Claimed elements as specifically Claimed. The description of the present invention has been presented for purposes of illustration and description, but is not intended to be exhaustive or limited to the invention in the form disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the invention. The embodiment was chosen and described in order to best explain the principles of the invention and the practical application, and to enable others of ordinary skill in the art to understand the inventionISUS Ref: 53524-006PCTfor various embodiments with various modifications as are suited to the particular use contemplated.

[0250] As used herein, the term “classical computer” refers to one or more non-quantum computing devices configured to process information represented by classical states, including, for example, binary bits, analog electrical values, digital words, symbols, data structures, or combinations thereof. A classical computer may include, without limitation, one or more central processing units (CPUs), graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), microcontrollers, microprocessors, system-on-chip devices, programmable logic devices, memory devices, storage devices, network interfaces, input / output interfaces, buses, chipsets, firmware, operating systems, virtual machines, containers, servers, workstations, client devices, cloud computing resources, edge computing resources, embedded controllers, or combinations thereof.

[0251] In some embodiments, a classical computer executes classical program instructions, firmware instructions, machine instructions, source code, object code, scripts, control software, optimization routines, preprocessing routines, post-processing routines, scheduling routines, calibration routines, compilation routines, error-processing routines, or any combination thereof. The classical computer may be implemented as a single physical machine, a distributed system, a virtualized computing environment, a cloud-based computing service, an edge device, an embedded controller, a control computer coupled to quantum hardware, or a combination of such devices.

[0252] A classical computer may generate, receive, store, transform, transmit, or analyze data associated with a quantum computation, including circuit descriptions, pulse schedules, gate parameters, qubit mappings, measurement results, expectation values, gradients, objective-function values, calibration data, noise data, error-correction data, error-mitigation data, or other information. A classical computer may be operatively coupled to quantum hardware through one or more wired links, optical links, radiofrequency links, microwave links, network connections, cloud interfaces, application programming interfaces, control buses, readout lines, control lines, cryogenic interfaces, or other communication channels.ISUS Ref: 53524-006PCT

[0253] In some embodiments, a classical computer controls, configures, schedules, compiles for, sends instructions to, receives data from, or otherwise interacts with quantum hardware. Unless expressly stated otherwise, a classical computer may include hardware, software, firmware, or combinations thereof that perform classical computation before, during, or after execution of a quantum operation. A classical computer need not itself include qubits and need not itself perform quantum state evolution in physical qubits.

[0254] As used herein, the term “quantum hardware” refers to physical hardware configured to create, maintain, manipulate, couple, evolve, control, read out, measure, initialize, reset, protect, correct, or otherwise operate on one or more quantum states, qubits, quantum registers, quantum modes, quantum particles, quantum oscillators, quantum annealing states, quantum fields, or other quantum information carriers. Quantum hardware may include, without limitation, one or more quantum processing units (QPUs), qubit devices, qubit arrays, quantum registers, quantum annealers, analog quantum processors, gate-model quantum processors, measurement-based quantum processors, physical quantum simulators, or hybrid quantum processing devices.

[0255] Examples of quantum hardware include superconducting qubit devices, trappedion devices, neutral-atom devices, Rydberg-atom devices, photonic quantum devices, spin-qubit devices, semiconductor quantum-dot devices, nitrogen-vacancy-center devices, topological qubit devices, Majorana-based devices, bosonic-code devices, cavity quantum electrodynamics devices, nuclear magnetic resonance quantum devices, quantum annealing devices, analog quantum simulation devices, or any combination thereof. Quantum hardware may also include physical components associated with such devices, including resonators, cavities, ion traps, optical traps, waveguides, interferometers, beam splitters, phase shifters, lasers, microwave sources, radiofrequency sources, pulse generators, arbitrary waveform generators, digital-to-analog converters, analog-to-digital converters, amplifiers, filters, attenuators, mixers, couplers, circulators, isolators, readout resonators, detectors, photon counters, cryogenic components, dilution refrigerators, vacuum systems, magnetic shielding, thermal links, control electronics, readout electronics, calibration electronics, interconnects, packaging, substrates, and signal lines.ISUS Ref: 53524-006PCT

[0256] In some embodiments, quantum hardware receives control information from a classical computer and performs one or more quantum operations in response thereto. Such quantum operations may include, without limitation, qubit initialization, state preparation, single-qubit gate operations, multi-qubit gate operations, entangling operations, analog evolution, Hamiltonian simulation, adiabatic evolution, quantum annealing, measurement, reset, feed-forward operations, error detection, error mitigation, error correction, dynamical decoupling, calibration, or combinations thereof. Quantum hardware may output measurement data, analog signals, digital signals, bit strings, probability estimates, expectation values, syndrome data, calibration data, or other information to a classical computer or other system component.

[0257] The term “quantum hardware” refers to one or more physical devices and associated control, readout, coupling, environmental, and interface components, and is not limited to any particular qubit modality, quantum computing architecture, level of fault tolerance, number of qubits, gate set, physical implementation, or commercial platform. Quantum hardware may be local to a classical computer, remote from a classical computer, accessed through a network or cloud service, shared among multiple users, or integrated with classical control electronics. Unless expressly stated otherwise, quantum hardware may include currently known and later-developed physical quantum computing or quantum information processing devices capable of supporting the operations described herein.

Claims

ISUS Ref: 53524-006PCTCLAIMSWhat is Claimed is:

1. A method for predicting a chemical property of a molecular system, comprising:performing, using a classical computer, an electronic structure calculation for the molecular system to obtain molecular orbital information corresponding to a plurality of geometrical configurations defined along at least one interatomic coordinate in the molecular system;generating a plurality of representations of the molecular system corresponding to the plurality of geometrical configurations;optimizing, using a quantum hardware, a parameterized quantum state with respect to each of the plurality of representations, thereby obtaining a set of optimized variational parameters;performing, using the quantum hardware, quantum measurements with respect to the plurality of representations based on the set of optimized variational parameters, thereby obtaining measurement data associated with the plurality of geometrical configurations;generating, based on the measurement data, a set of reduced quantum- derived data associated with the plurality of geometrical configurations;transforming, using the classical computer, the reduced quantum-derived data into orbital-basis data, based on the molecular structure information corresponding to the plurality of geometrical configurations of the molecular system;extracting features based on the orbital-basis data, including variation of the features across the plurality of geometrical configurations; andpredicting the chemical property based on the extracted features.

2. The method of Claim 1, wherein the plurality of representations of the molecular system comprises a plurality of Hamiltonians.

3. The method of Claim 2, wherein each Hamiltonian represents an electronic structure of the molecular system at a respective geometrical configuration and having eigenvalues corresponding to energies of the molecular system.ISUS Ref: 53524-006PCT4. The method of Claim 2, wherein each Hamiltonian is represented as a linear combination of Pauli operators and mapped to a qubit representation suitable for execution on quantum hardware.

5. The method of Claim 2, wherein the Hamiltonians encode contributions from electronic degrees of freedom outside a reduced active space, including virtual orbital contributions associated with unoccupied molecular orbitals not explicitly included in the reduced active space, wherein the contributions represent electron correlation effects arising from interactions between occupied orbitals in the reduced active space and the unoccupied molecular orbitals.

6. The method of Claim 2, wherein each Hamiltonian is constructed based on a reduced active-space representation of the molecular system comprising a subset of molecular orbitals including at least one occupied orbital and at least one unoccupied orbital, optionally comprising a two-electron, two-orbital active space.

7. The method of Claim 1, wherein the plurality of geometrical configurations comprises a plurality of interatomic distances along a bond formation or dissociation coordinate.

8. The method of Claim 1, wherein generating the plurality of Hamiltonians comprises:performing electronic structure calculations for each geometrical configuration; andconstructing a reduced active-space Hamiltonian.

9. The method of Claim 8, wherein the electronic structure calculations comprise Density Functional Theory (DFT) calculations configured to determine molecular orbital coefficients and electronic integrals for each geometrical configuration.

10. The method of Claim 8, wherein the reduced active-space Hamiltonian comprises a CASCI(2e,2o) representation.

11. The method of Claim 1, wherein said optimizing the parameterized quantum state comprises:ISUS Ref: 53524-006PCTselecting a representation of the molecular system corresponding to an equilibrium configuration of the molecular system, wherein the equilibrium configuration corresponds to a configuration associated with a minimum energy of the molecular system along at least one interatomic coordinate; and optimizing, using the quantum hardware, a parameterized quantum state with respect to the selected representation to obtain optimized parameters.

12. The method of Claim 11, wherein said using the parameterized quantum state to perform quantum measurements comprises:maintaining the optimized parameters fixed cross the plurality of representations; andfor each of the plurality of representations of the molecular system:preparing a quantum state using the optimized parameters; and performing quantum measurements to obtain measurement data.

13. The method of Claim 11 , wherein the optimized parameters approximate a ground state of the molecular system at the equilibrium configuration, such that maintaining the optimized parameters fixed across the plurality of representations enables extraction of configuration-dependent quantum responses.

14. The method of Claim 11, wherein said optimizing the parameterized quantum state comprises performing a Variational Quantum Eigensolver (VQE) to minimize an expected energy.

15. The method of Claim 14, wherein the quantum hardware comprises an NMR quantum processor, wherein VQE is executed on the NMR quantum processor.

16. The method of Claim 15, wherein the NMR quantum processor performs analog quantum evolution based on nuclear spin interactions, thereby capturing contributions from electronic degrees of freedom beyond a programmed active space.

17. The method of Claim 1, wherein performing quantum measurements across the plurality of geometrical configurations enables amplification of correlationISUS Ref: 53524-006PCTinformation associated with electronic degrees of freedom not explicitly represented in a reduced active space.

18. The method of Claim 1,wherein said generating the set of quantum-derived data comprises:obtaining a calibration set of two or more candidate systems sharing a chemically conserved active-space center with the molecular system;for each system in the calibration set, generating a plurality of representations corresponding to a plurality of geometrical configurations of the system defined along the at least one interatomic coordinate in the system;wherein said predicting the chemical property comprises:calibrating a regression model mapping the feature vectors of the calibration-set systems to corresponding experimentally measured values of chemical properties; andpredicting the chemical property of the molecular system by applying the calibrated regression model to the feature vector of the molecular system.

19. The method of Claim 1, wherein said transforming the quantum-derived data comprises transforming the one or more quantum observables into a molecular orbital basis defined by applying a basis rotation defined by eigenvectors of the corresponding representations of the molecular system.

20. The method of Claim 19, wherein the eigenvectors are derived from electronic structure calculations.

21. The method of Claim 1, wherein said transforming the quantum-derived data integrates:classical electronic structure information; andquantum measurement data.

22. The method of Claim 1 , wherein the extracted features are predictive in a molecular orbital basis but not in a computational qubit basis.ISUS Ref: 53524-006PCT23. The method of Claim 1, wherein said transforming the quantum-derived data is performed separately for each geometrical configuration.

24. The method of Claim 1, wherein said extracting the features is performed based on at least one of: quantum observables and variation of the quantum observables across the plurality of geometrical configurations.

25. The method of Claim 24, wherein the quantum observables comprise at least one of:orbital occupation;coherence;entropy;purity;cumulant;mutual information;diagonal dominance;entanglement;concurrence;negativity;spin correlation;natural orbital occupation numbers;gradient of a quantum observable across the plurality of geometrical configurations;integral of a quantum observable across the plurality of geometrical configurations;extremal values of a quantum observable across the plurality of geometrical configurations;coefficient of variation of a quantum observable across the plurality of geometrical configurations; andcenter of gravity of a quantum observable along the at least one interatomic coordinate.

26. The method of Claim 25, wherein the features comprise a virtual-orbital trace deficit.ISUS Ref: 53524-006PCT27. The method of Claim 24, wherein the features are extracted from an orbital-basis- transformed spatial correlation profile across the plurality of geometrical configurations.

28. The method of Claim 1, wherein the features further comprise statistical descriptors across the plurality of geometrical configurations.

29. The method of Claim 28, wherein the statistical descriptors comprise at least one of: mean, variance, slope, range, or curvature.

30. The method of Claim 1, wherein the features comprise at least one of a combination of values at an equilibrium configuration; and variation across the at least one interatomic coordinate in the molecular system.

31. The method of Claim 1, wherein the features comprise a subset of selected features having a cardinality below a predefined threshold.

32. The method of Claim 31, wherein the selected features are determined based on a statistical relevance metric.

33. The method of Claim 1, wherein generating the plurality of representations comprises generating representations for a plurality of spatially distinct active spaces within the molecular system.

34. The method of Claim 33, wherein features derived from the plurality of spatially distinct active spaces are combined.

35. The method of Claim 1, wherein predicting comprises applying a supervised machine-learning model.

36. The method of Claim 35, wherein the model is trained using compounds sharing a common bonding mechanism, such that the training data is restricted to a consistent chemical bond type.

37. The method of Claim 35, wherein training the model comprises leave-one-out cross-validation.ISUS Ref: 53524-006PCT38. The method of Claim 35, wherein the model is trained using chemistry-specific calibration data and validated using permutation testing that re-executes feature selection.

39. The method of Claim 35, wherein feature selection comprises evaluating combinations of a limited number of features.

40. The method of Claim 39, wherein feature selection comprises evaluating combinations of k features, where k is limited to a predefined small number.

41. The method of Claim 35, wherein the model comprises ridge regression.

42. The method of Claim 35, wherein validation comprises permutation testing, wherein permutation testing comprises repeating feature selection for each permutation to correct for multiple-testing bias.

43. The method of Claim 35, wherein the model is trained using a subset of selected features having a cardinality below a predefined threshold.

44. The method of Claim 1, wherein the chemical property comprises at least one of:binding affinity, enantioselectivity, solid-state material impurity, battery electrolyte material, or battery electrode interaction.

45. The method of Claim 1, wherein the molecular system comprises a drug-target system or a catalytic system.

46. The method of Claim 1, wherein said predicting the chemical property further comprises computing at least one of: a confidence metric associated with the chemical property, an uncertainty estimate, contributing features, measurement fidelity, and a confidence indication associated with an applicability domain.

47. A system for predicting a chemical property of a molecular system, comprising:a classical computer configured to:perform an electronic structure calculation for the molecular system to obtain molecular orbital information corresponding to a plurality of geometrical configurations defined along at least one interatomic coordinate in the molecular system, andISUS Ref: 53524-006PCTgenerate a plurality of representations of the molecular system corresponding to the plurality of geometrical configurations; quantum hardware configured to:optimize a parameterized quantum state with respect to each of the plurality of representations to obtain a set of optimized variational parameters, andperform quantum measurements with respect to the plurality of representations based on the set of optimized variational parameters to obtain measurement data associated with the plurality of geometrical configurations;the classical computer further configured togenerate, based on the measurement data, a set of reduced quantum-derived data associated with the plurality of geometrical configurations,transform the reduced quantum-derived data into orbitalbasis data based on molecular structure information corresponding to the plurality of geometrical configurations of the molecular system,extract features based on the orbital-basis data including variation of the features across the plurality of geometrical configurations, andpredict the chemical property based on the extracted features.

48. A method for predicting a chemical property of a molecular or materials system, the method comprising:providing a calibration set of two or more candidate systems sharing a chemically conserved active-space center, and a target system to be predicted;for each system in the calibration set and for the target system, generating a plurality of representations corresponding to a plurality of geometrical configurations of the system defined along at least one interatomic coordinate, and constructing, from each said representation, an active-space second- quantized Hamiltonian;ISUS Ref: 53524-006PCToptimizing, on an Analog Nuclear Magnetic Resonance (NMR) hardware functioning as a molecular quantum simulator and operating as an open quantum system, a parameterized quantum state with respect to at least one of said representations of each said system, thereby obtaining a set of optimized variational parameters per system;preparing, on said analog NMR hardware, the parameterized quantum state under said optimized variational parameters with respect to each of said plurality of representations of each said system, and performing reduced-density-matrix measurements on said analog NMR hardware, thereby obtaining mixed-state measurement data comprising one- and two-particle reduced density matrices associated with the plurality of geometrical configurations of each said system;transforming said reduced density matrices into a canonical orbital basis derived from electronic-structure information associated with each said system, said transforming comprising rotation by eigenvectors of the active-space Hamiltonian;extracting, from the transformed reduced density matrices of each said system, a feature vector comprising statistics of said transformed reduced density matrices computed across the plurality of geometrical configurations of said system, said statistics including, in particular embodiments, integrals, ranges, coefficients of variation, diagonal -dominance measures, and one- or two-particle cumulant components;calibrating a regression model mapping the feature vectors of the calibrationset systems to corresponding experimentally measured values of the chemical property; andpredicting the chemical property of the target system by application of said calibrated regression model to the feature vector of the target system.

49. A hybrid quantum-classical computing system comprising:a plurality of Nuclear Magnetic Resonance (NMR) quantum processors, each comprising a set of nuclear spins configured as qubits and operable to perform quantum operations; anda classical control system communicatively coupled to the plurality of NMR quantum processors;ISUS Ref: 53524-006PCTwherein the plurality of NMR quantum processors are configured to operate independently and in parallel;wherein the classical control system is configured to:decompose a computational task into a plurality of sub-tasks; assign the sub-tasks to respective NMR quantum processors; coordinate execution of quantum operations across the plurality of NMR quantum processors; andaggregate measurement results obtained from the plurality of NMR quantum processors to determine a solution to the computational task.

50. The system of Claim 49,wherein the classical control system is further configured to:obtain a plurality of representations of a molecular system corresponding to a plurality of geometrical configurations defined along at least one interatomic coordinate in the molecular system; assign to each NMR quantum processor a representation of the molecular system corresponding to a distinct geometrical configuration along the at least one interatomic coordinate such that the plurality of NMR quantum processors evaluate the molecular system across the plurality of geometrical configurations in parallel;coordinate independent execution of parameterized quantum circuits on the respective NMR quantum processors;aggregate measurement results from the plurality of NMR quantum processors to construct a spatial correlation profile across the at least one interatomic coordinate; andapply cross-processor fidelity calibration to normalize measurement data obtained from different NMR quantum processors.

51. The system of Claim 49, wherein each NMR quantum processor comprises between 2 and 15 qubits.

52. The system of Claim 49, wherein the plurality of NMR quantum processors operates at room temperature without cryogenic cooling.ISUS Ref: 53524-006PCT53. The system of Claim 49, wherein quantum operations are implemented using radiofrequency pulse sequences configured to implement parameterized quantum circuits.

54. The system of Claim 49, wherein the plurality of sub-tasks comprises fragments of a larger computational problem, wherein the fragments are processed concurrently across the plurality of NMR quantum processors.

55. The system of Claim 54, wherein fragments correspond to subsets of variables or Hamiltonian terms.

56. The system of Claim 54, wherein the fragments overlap to enforce solution consistency.

57. The system of Claim 49, wherein the classical control system is further configured to:assign a plurality of NMR quantum processors to perform redundant computation of a same sub-task;compare results; anddetermine a validated result based on statistical agreement.

58. The system of Claim 57, wherein at least three processors perform the same subtask.

59. The system of Claim 57, wherein a result is validated based on majority voting.

60. The system of Claim 57, wherein the classical control system is further configured to: assign a plurality of NMR quantum processors to perform redundant computation of a same sub-task; determine a fidelity metric for each NMR quantum processor based on physicality constraints of the measurement data; and determine a validated result by weighting results from the plurality of NMR quantum processors according to their respective fidelity metrics.

61. The system of Claim 49, wherein the statistical aggregation comprises at least one of: outlier rejection, averaging, or Bayesian inference.

62. The system of Claim 49, further comprising:ISUS Ref: 53524-006PCTa pool of pre-initialized NMR quantum processors, in a reference state; wherein in response to detecting an error in a processor from the plurality of NMR quantum processors, the classical control system is configured to replace the processor with a standby processor selected from the pool of pre-initialized NMR quantum processors.

63. The system of Claim 49, further comprising a pool of standby NMR quantum processors, each maintained in a calibrated thermal equilibrium state; wherein the classical control system is configured to: continuously monitor a fidelity metric of each active NMR quantum processor during execution; detect degradation of an active NMR quantum processor when its fidelity metric falls below a predetermined threshold; reassign the sub-task of the degraded processor to a standby processor from the pool; and transition the degraded processor to a recalibration state.

64. The system of Claim 63, wherein each NMR quantum processor is assigned a state from the group consisting of: active, standby, and recalibrating; and wherein the classical control system dynamically transitions processors between states based on their respective fidelity metrics.

65. The system of Claim 64, wherein a processor in the recalibrating state undergoes a re-initialization sequence comprising thermal re-equilibration and reference pulse calibration, and upon successful recalibration is transitioned to standby state.

66. The system of Claim 49, wherein the classical control system is configured to:obtain partial measurement data from a subset of the plurality of NMR quantum processors;determine a confidence level based on statistical agreement; and selectively trigger additional measurements from other processors from the plurality of NMR quantum processors, including full state tomography, when the confidence level falls below a threshold.

67. The system of Claim 49, wherein the classical control system is configured to:ISUS Ref: 53524-006PCTobtain, from each of a subset of the plurality of NMR quantum processors, a first set of measurement data comprising expectation values of diagonal quantum observables;determine, based on the first set of measurement data, whether off-diagonal quantum observables are expected to carry significant information; and selectively trigger, on processors where off-diagonal information is expected to be significant, a second set of measurements comprising off-diagonal quantum observables to obtain a complete density matrix reconstruction.

68. The system of Claim 67,wherein the determination of whether off-diagonal observables carry significant information is based on a degeneracy measure of orbital occupations obtained from the diagonal quantum observables, such that when orbital occupations are near-degenerate, the second set of measurements is triggered;wherein the selective triggering is performed independently for each geometrical configuration; andwherein at geometrical configurations where the diagonal measurements indicate non-degenerate orbital occupations, the quantum -derived data is generated without performing the second set of measurements, thereby reducing total measurement time.

69. The system of Claim 49, wherein the classical control system is configured to iteratively update sub-task assignments to the plurality of NMR quantum processors based on intermediate results.

70. The system of Claim 49, wherein correlations between sub-tasks are identified and incorporated into subsequent quantum operations.

71. The system of Claim 49, wherein the plurality of NMR quantum processors comprises liquid-state NMR systems.

72. The system of Claim 71, wherein the liquid-state NMR systems comprise ensembles of molecules in solution providing ensemble-averaged measurement signals.ISUS Ref: 53524-006PCT73. The system of Claim 71, wherein the liquid-state NMR systems comprise a solution containing a macroscopic ensemble of molecules, each molecule providing a corresponding quantum processor, such that quantum operations are performed simultaneously across the ensemble and measurement signals are obtained as ensemble-averaged observables representative of the collective quantum state.

74. The system of Claim 71, wherein quantum measurements are obtained without projective collapse using ensemble averaging.

75. The system of Claim 71, wherein qubits are implemented using nuclear spins in heteronuclear molecular systems.

76. The system of Claim 71, wherein quantum gate operations are implemented using radiofrequency pulse sequences.

77. The system of Claim 49, wherein the plurality of NMR quantum processors comprises solid-state NMR systems.

78. The system of Claim 78, wherein the solid-state NMR systems comprise nuclear spins in representative material samples.

79. The system of Claim 78, wherein the solid-state NMR systems are configured to apply a magic angle spinning (MAS) mechanism.

80. The system of Claim 79, wherein the MAS mechanism is configured to rotate a sample containing the nuclear spin qubits at predetermined Magic Angle of approximately 54.74 degrees relative to an applied magnetic field, thereby reducing anisotropic spin interactions and enhancing coherence and spectral resolution for quantum operations.

81. The system of Claim 79, wherein the magic angle spinning is performed at a frequency in the range of 10-70 kHz.

82. The system of Claim 79, wherein magic angle spinning increases coherence time of nuclear spin qubits.ISUS Ref: 53524-006PCT83. The system of Claim 77, wherein the solid-state NMR systems operate under controlled temperature and pressure conditions corresponding to target material environments.

84. The system of Claim 77, wherein the solid-state NMR systems comprise representative materials selected based on similarity to a target system being modeled.

85. The system of Claim 49, wherein the plurality of NMR quantum processors comprises both liquid-state and solid-state NMR systems.

86. The system of Claim 85, wherein:liquid-state processors perform measurement-intensive tasks; and solid-state processors perform environment-specific simulations.

87. The system of Claim 85, wherein results from liquid-state and solid-state processors are combined to determine the solution.

88. The system of Claim 49, further comprising hyperpolarization.

89. The system of Claim 49, wherein the processors operate at physiological temperature.

90. A method for performing quantum-assisted computation using a distributed Nuclear Magnetic Resonance (NMR) quantum computing system, comprising:decomposing, by a classical control system, a computational task into a plurality of sub-tasks;assigning, by the classical control system, the sub -tasks to respective NMR quantum processors from a plurality of NMR quantum processors, each comprising a set of nuclear spins configured as qubits;executing, by the plurality of NMR quantum processors operating independently and in parallel, quantum operations corresponding to the assigned sub-tasks;coordinating, by the classical control system, execution of the quantum operations across the plurality of NMR quantum processors;ISUS Ref: 53524-006PCTobtaining measurement results from the plurality of NMR quantum processors; andaggregating, by the classical control system, the measurement results to determine a solution to the computational task.