Random Orbital Sampling Matches Systematic Methods in Quantum Simulations

Researchers have demonstrated that a randomized approach to selecting orbitals for quantum simulations can achieve accuracy comparable to systematic, energy-based selection methods. Published in the AAPPS Bulletin, the study introduces RO-VQE, an algorithm derived from the optimized orbital variational quantum eigensolver (OptOrbVQE). By employing a randomized procedure for selecting and optimizing orbitals, the method offers a flexible alternative to deterministic strategies, potentially reducing computational overhead while maintaining high precision in molecular ground-state energy estimations.

The work addresses a significant challenge in quantum chemistry: the limited number of qubits and high susceptibility to noise in near-term quantum devices. These constraints restrict the scalability of algorithms like the variational quantum eigensolver (VQE), which is used to solve the electronic structure problem. While VQE has shown potential for small-scale systems, handling large basis sets and many-electron molecules requires further improvements. The authors propose that randomized active-space selection may offer a practical compromise for the noisy intermediate-scale quantum (NISQ) era.

Traditional strategies for fitting within NISQ hardware often rely on minimal basis sets to reduce the number of spin orbitals, thereby cutting qubit requirements and circuit depth. However, this approach sacrifices diffuse and polarization functions essential for high-quality correlation energies. Expanding the basis set increases the number of required qubits and circuit depth, amplifying error accumulation and noise. Orbital optimization aims to retain the accuracy benefits of larger basis sets while mitigating these computational overheads by focusing on the most relevant orbitals.

RO-VQE modifies the OptOrbVQE algorithm, which previously selected molecular orbitals based on Hartree–Fock energies. In contrast, RO-VQE explores whether randomized orbital selection can achieve comparable accuracy. The algorithm projects the Hamiltonian onto a subspace containing a limited number of spin orbitals using a partial unitary transformation. This transformation is initialized with a randomly generated matrix that is orthonormalized to satisfy unitary constraints. The optimization process involves iteratively minimizing the energy with respect to both the variational parameters of the ansatz circuit and the partial unitary matrix representing the orbital selection.

To evaluate the performance of RO-VQE, the authors conducted simulations on hydrogen chain systems, specifically H2 and H4 molecules. These systems serve as well-established benchmarks for quantum chemistry methods due to their controllable testbed for strong correlation. For the H4 molecule, the study employed a 6-31G split-valence basis set to investigate how expanding beyond minimal basis sets influences the relative performance of MO-based versus random selection. For the H2 molecule, the number of spin orbitals was kept constant at four, while varying the basis sets to assess the algorithm’s robustness.

The simulations were implemented using a combination of publicly available libraries, including Qiskit and PyTorch. The orbital rotation matrix was optimized using the projection method with alternating Barzilai–Borwein steps, while the energy minimization with respect to variational parameters utilized the VQE implementation provided by Qiskit. The results were analyzed in terms of potential energy curves (PECs) and one-electron reduced density matrices (1-RDMs) to provide insight into orbital occupations and electron correlations.

In the H4 molecule simulations, increasing the number of qubits from 6 to 8 consistently resulted in lower total energy values, reflecting improved accuracy in approximating the ground-state wavefunction. For both square and linear geometries, RO-VQE and OptOrbVQE produced closely aligned results at lower qubit counts. However, at 8 qubits, the potential energy curves from both methods became distinguishable, with RO-VQE occasionally yielding slightly lower energies. This suggests that randomly selected orbital subsets can surpass MO-based selection when the active space is sufficiently large.

Analysis of the 1-RDMs revealed distinct differences in orbital occupation and coherence between the two methods. In the square H4 geometry, RO-VQE exhibited a broader distribution of occupations and stronger off-diagonal terms, indicating a more delocalized and correlated electron density. In contrast, OptOrbVQE showed a more localized pattern with smaller off-diagonal values. In the linear H4 geometry, both methods yielded nearly identical 1-RDM profiles, reflecting the highly localized electronic distribution and minimal orbital correlations inherent to the symmetry-constrained system.

For the H2 molecule, both RO-VQE and OptOrbVQE delivered nearly identical performance across different basis sets, including 3-21G, 6-31G, and cc-pVDZ. The energy values decreased systematically as the basis set size increased, with cc-pVDZ producing the lowest energy estimates. The conventional VQE result using the STO-3G basis set served as a baseline, showing a relatively shallow potential energy curve due to the limited flexibility of the minimal basis. Both orbital-optimized methods outperformed this reference, demonstrating the critical role of basis set quality in enhancing accuracy under qubit-limited conditions.

The absence of erratic behavior in RO-VQE, despite its stochastic nature, highlights the method’s robustness. In symmetric and strongly correlated systems, multiple orbital combinations can lead to similarly accurate approximations, reinforcing the reliability of RO-VQE under different orbital configurations. The study suggests that the choice of active space may be less critical than previously assumed, as randomized selection can effectively capture essential low-energy states.

These findings indicate that RO-VQE is a viable and computationally inexpensive alternative for modeling strongly correlated systems. The method offers a practical compromise for NISQ-era quantum computations, providing a flexible option for systems constrained by limited quantum resources. The code for this study is available on GitHub, allowing for further exploration and implementation by the research community.

By demonstrating that randomized orbital sampling can match the performance of systematic strategies, the authors provide a new perspective on active-space selection.

The study was conducted by Yoga A. Darmawan, Angga D. Fauzi, Yanoar P. Sarwono, and Rui-Qin Zhang, and published in the AAPPS Bulletin in September 2025. The research specifically targets the electronic structure problem, which remains computationally challenging due to electron correlation effects that current state-of-the-art classical methods, such as Hartree–Fock and density functional theory, struggle to capture accurately in many-electron systems. Hybrid quantum–classical algorithms, notably the variational quantum eigensolver, are considered more resilient to qubit limitations and noise compared to purely quantum algorithms. However, extending these methods to systems with more electrons and large basis sets requires further development to handle increased molecular complexity effectively.

The theoretical framework of RO-VQE relies on the variational principle, which states that the ground state energy of a system can be approximated by the expectation value of the Hamiltonian with respect to a varied wavefunction. The electronic structure Hamiltonian is formulated in second quantization, involving one- and two-body integrals over a set of basis functions. To map this fermionic Hamiltonian to a qubit Hamiltonian suitable for quantum computation, the authors employed the Jordan-Wigner transformation. This mapping results in a qubit Hamiltonian consisting of tensor products of Pauli operators, which are measured on the quantum computer to determine the expectation values necessary for energy calculation.

The optimization process in RO-VQE is structured as a joint minimization over two distinct parameter sets: the variational parameters of the ansatz circuit and the partial unitary matrix representing the orbital selection. The partial unitary matrix must satisfy an orthogonality condition, ensuring its columns form an orthonormal basis. To solve this complex optimization problem, the algorithm divides the minimization into two sequential subproblems that are solved iteratively. First, the energy is minimized with respect to the partial unitary matrix while keeping the variational parameters fixed. Second, the ansatz parameters are optimized while holding the partial unitary matrix constant. This iterative process continues until a predefined convergence criterion is met, ensuring that both the orbital selection and the wavefunction parameters are jointly optimized for the lowest possible energy estimate.

Initialization of the partial unitary matrix in RO-VQE involves orthonormalizing a randomly generated matrix to ensure it satisfies the unitary constraint. To enhance exploration and prevent convergence to shallow local minima, a small random perturbation is introduced to the updated partial unitary matrix during the optimization iterations. This perturbation is drawn from a normal distribution with a mean of zero and a standard deviation of 0.01.

The choice of hydrogen chain systems, specifically H2 and H4, as testbeds is significant because they are well-established benchmarks for quantum chemistry methods due to their controllable nature and strong correlation effects. The H4 molecule in square and linear geometries provides a rigorous test for the algorithm’s ability to handle different symmetry constraints and correlation structures. The use of the 6-31G split-valence basis set for H4 allows for an investigation into how expanding beyond minimal basis sets influences the relative performance of orbital selection strategies. This controlled experimental design helps establish performance baselines that can guide future implementations on NISQ hardware, where resource constraints are a primary concern.

The projection method with alternating Barzilai–Borwein steps was used to optimize the orbital rotation matrix, while Qiskit’s VQE implementation handled the energy minimization with respect to variational parameters.

The study’s findings suggest that the choice of active space may be less critical than previously assumed, particularly when sufficient qubit resources are available. The ability of randomized orbital selection to match the accuracy of systematic strategies indicates that targeted orbital optimization may not always be necessary for achieving high precision in ground-state energy estimations. As quantum hardware continues to evolve, methods like RO-VQE may offer a practical and flexible alternative for addressing the electronic structure problem in larger and more complex molecular systems.

Subject of Research: Randomized orbital sampling in variational quantum eigensolver algorithms for molecular electronic structure simulations.

Article Title: Improving qubit reduction for molecular simulations with randomized orbital sampling

Article References: Darmawan, Y. A., Fauzi, A. D., Sarwono, Y. P., & Zhang, R.-Q. (2025). Improving qubit reduction for molecular simulations with randomized orbital sampling. AAPPS Bulletin, 35(1), Article 27. https://doi.org/10.1007/s43673-025-00167-5

Image Credits: AI Generated

DOI: 10.1007/s43673-025-00167-5

Keywords: quantum chemistry, variational quantum eigensolver, orbital optimization, molecular simulation, NISQ era, hydrogen chains, qubit reduction, randomized algorithms, electronic structure, quantum computing, Improving, qubit