This paper describes a hybrid quantum-classical computational approach to designing synthesizable deuterated tris(8-hydroxyquinolinato) aluminum (Alq 3 ) emitters with desirable emission quantum efficiency (QE). This multi-pronged approach first uses classical quantum chemistry to create a machine learning dataset, which is then used to construct an Ising Hamiltonian by a factorization-machine-based model to predict the QEs of Alq 3 emitters. Finally, the Ising Hamiltonian is applied to perform simulations using the variational quantum eigensolver (VQE) and quantum approximate optimization algorithm (QAOA) on a quantum device to discover molecules with optimal QE. Moreover, to improve the simulations on the noisy quantum device, we developed the recursive probabilistic variable elimination method, which recursively eliminates qubits depending on the probability that each qubit has a binary value. We demonstrated that the accuracy of VQE and QAOA optimized for a noisy device can be improved from a probability of 0.075 to 0.97.
This paper describes a hybrid quantum-classical computational approach to designing synthesizable deuterated tris(8-hydroxyquinolinato) aluminum (Alq 3 ) emitters with desirable emission quantum efficiency (QE). This multi-pronged approach first uses classical quantum chemistry to create a machine learning dataset, which is then used to construct an Ising Hamiltonian by a factorization-machine-based model to predict the QEs of Alq 3 emitters. Finally, the Ising Hamiltonian is applied to perform simulations using the variational quantum eigensolver (VQE) and quantum approximate optimization algorithm (QAOA) on a quantum device to discover molecules with optimal QE. Moreover, to improve the simulations on the noisy quantum device, we developed the recursive probabilistic variable elimination method, which recursively eliminates qubits depending on the probability that each qubit has a binary value. We demonstrated that the accuracy of VQE and QAOA optimized for a noisy device can be improved from a probability of 0.075 to 0.97.
Abstract This study describes a hybrid quantum-classical computational approach for designing synthesizable deuterated Alq3 emitters possessing desirable emission quantum efficiencies (QEs). This design process has been performed on the tris(8-hydroxyquinolinato) ligands typically bound to aluminum in Alq3 and involves optimization of the hydrogen/deuterium (H/D) ratios in these molecules for the discovery of optimally deuterated high-efficiency OLED emitters. It involves a multi-pronged approach which first utilizes classical quantum chemistry to predict the emission QEs of the Alq3 ligands. These initial results were then used as a machine learning dataset for a factorization-machine-based model which was applied to construct an Ising Hamiltonian to predict emission quantum efficiencies on a classical computer. We show that such a factorization machine-based approach can yield accurate property predictions for all 64 deuterated Alq3 emitters with 13 training values. Moreover, another Ising Hamiltonian could be constructed by including synthetic constraints which could be used to perform optimizations on a quantum simulator and device using the variational quantum eigensolver (VQE) and quantum approximate optimization algorithm (QAOA) to discover a molecule possessing the optimal QE and synthetic cost. We observe that both VQE and QAOA can correctly predict the optimal molecule with greater than 0.95 probability on quantum simulators. These probabilities decrease to 0.83 and 0.075 for simulations with VQE and QAOA, respectively, on a quantum device, while the application of the readout error mitigation technique only improves probabilities to 0.90 and 0.084. We have developed the Recursive Probabilistic Variable Elimination (RPVE) method for simulations on noisy devices that recursively eliminates qubits depending on the probability that each qubit possesses a binary value and have demonstrated that RPVE improves the accuracies of VQE and QAOA for predictions of optimal solutions with probabilities of 0.97 on a quantum device.
The integration of quantum chemistry, machine learning, and optimization calculations is expected to accelerate materials discovery by making large chemical spaces amenable to computational study, a challenging task for classical computers. In this study, we develop a quantum-classical computing scheme involving the computational-basis variational quantum deflation (cVQD) method for calculating the excited states of a general classical Hamiltonian, such as an Ising Hamiltonian. We apply this scheme to the practical use case of generating photochromic diarylethene (DAE) derivatives for photopharmacology applications. Using a dataset of 384 DAE derivatives from quantum chemistry calculation results, we show that machine learning can accurately predict the wavelength of maximum absorbance, λ max , of 4,096 DAE derivatives. After screening over 4,096 molecules using the computing scheme, we identified 5 DAE candidates that have important applications in photopharmacology. In detail, a 12-qubit cVQD calculation provides the ground state and 4 excited states of an Ising Hamiltonian corresponding to DAE candidates possessing large λ max . On a quantum simulator, results are found to be in excellent agreement with those obtained by an exact eigensolver. Utilizing error suppression and mitigation techniques, cVQD on a real quantum device produces results with accuracy comparable to calculations on the simulator. Finally, we show that quantum chemistry calculations for the DAE candidates provides a path to achieving large λ max and oscillator strengths by means of the molecular engineering of DAE derivatives.
New approaches into computational quantum chemistry can be developed through the use of quantum computing. While universal, fault-tolerant quantum computers are still not available, and we want to utilize today's noisy quantum processors. One of their flagship applications is the variational quantum eigensolver (VQE)-an algorithm for calculating the minimum energy of a physical Hamiltonian. In this study, we investigate how various types of errors affect the VQE and how to efficiently use the available resources to produce precise computational results. We utilize a simulator of a noisy quantum device, an exact statevector simulator, and physical quantum hardware to study the VQE algorithm for molecular hydrogen. We find that the optimal method of running the hybrid classical-quantum optimization is to: (i) allow some noise in intermediate energy evaluations, using fewer shots per step and fewer optimization iterations, but ensure a high final readout precision; (ii) emphasize efficient problem encoding and ansatz parametrization; and (iii) run all experiments within a short time-frame, avoiding parameter drift with time. Nevertheless, current publicly available quantum resources are still very noisy and scarce/expensive, and even when using them efficiently, it is quite difficult to perform trustworthy calculations of molecular energies.
Abstract: This study explores the implementation of hybrid quantum-classical algorithms in the realm of drug discovery and molecular simulation, with an emphasis on optimizing molecular interaction prediction. The integration of quantum computing paradigms, particularly the Variational Quantum Eigensolver (VQE), with classical machine learning techniques is evaluated for its efficacy in simulating molecular systems. Key results demonstrate superior accuracy in predicting binding affinities and reaction energetics using the hybrid models compared to traditional methods. The findings offer significant implications for accelerating the drug development pipeline, particularly in early-stage candidate screening. Keywords: Hybrid computing, Quantum simulation, Molecular interaction, Drug discovery, Variational Quantum Eigensolver, Quantum-classical integration
Accurately and efficiently predicting the equilibrium geometries of large molecules remains a central challenge in quantum computational chemistry, even with hybrid quantum–classical algorithms. Two major obstacles hinder progress: the large number of qubits required and the prohibitive cost of conventional nested optimization. In this work, we introduce a co-optimization framework that combines Density Matrix Embedding Theory (DMET) with Variational Quantum Eigensolver (VQE) to address these limitations. This approach substantially reduces the required quantum resources, enabling the treatment of molecular systems significantly larger than previously feasible. We first validate our framework on benchmark systems, such as H 4 and H 2 O 2, before demonstrating its efficacy in determining the equilibrium geometry of glycolic acid (C 2 H 4 O 3 )─a molecule of a size previously considered intractable for quantum geometry optimization. Our results show the method achieves high accuracy while drastically lowering computational cost. This work thus represents a significant step toward practical, scalable quantum simulations, moving beyond the small, proof-of-concept molecules that have historically dominated the field. More broadly, our framework establishes a tangible path toward leveraging quantum advantage for the in silico design of complex catalysts and pharmaceuticals.
Abstract The emergence of noisy intermediate-scale quantum (NISQ) devices has fueled the development of hybrid quantum-classical algorithms for machine learning (ML), promising potential advantages over classical methods. However, the practical efficacy of these algorithms remains uncertain due to challenges like noise susceptibility, trainability issues, and a lack of rigorous benchmarking. This study presents a comprehensive empirical evaluation of two flagship hybrid algorithms the Variational Quantum Eigensolver (VQE) for classification and the Quantum Approximate Optimization Algorithm (QAOA) for combinatorial optimization against state-of-the-art classical baselines under equivalent resource constraints. Through extensive simulations incorporating realistic noise models, we demonstrate that while these quantum approaches achieve competitive performance (e.g., 91.2% accuracy for VQE on a binary classification task), they consistently underperformed optimized classical models like Support Vector Machines (94.8% accuracy) and require orders of magnitude more computational time. Furthermore, we provide empirical validation of critical barriers to scalability, including the barren plateau phenomenon and significant performance degradation under noise, even with advanced error mitigation. Our results indicate that for the tasks studied, current NISQ era algorithms do not yet provide a tangible quantum advantage. This work contributes a rigorous benchmarking framework and a critical, realistic assessment of the field, suggesting that future progress hinges on fundamental algorithmic innovations and hardware improvements rather than incremental refinements of existing paradigms.