Integration of Variational Quantum Algorithms into Atomistic Simulation Workflows

In this work, we present the integration of Qiskit Nature’s quantum chemistry solvers into the Atomic Simulation Environment (ASE), enabling hybrid quantum-classical workflows for force-driven atomistic simulations. This coupling allows the use of the Variational Quantum Eigensolver (VQE) and its adaptive variant (ADAPT-VQE) not only for ground-state energy calculations, but also for geometry optimisation, vibrational frequency analysis, strain evaluation, and molecular dynamics, all managed through ASE’s calculator interface. By applying ADAPT-VQE to multi-electron systems such as BeH2, we obtain vibrational and structural properties in close agreement with high-level classical CCSD calculations within the same minimal basis. These results demonstrate that adaptive variational quantum algorithms can deliver stable and chemically meaningful forces within an atomistic modelling workflow, enabling downstream applications such as molecular dynamics and active-learning accelerated simulations.

💡 Research Summary

In this work the authors present a comprehensive integration of Qiskit Nature’s variational quantum chemistry solvers (VQE and its adaptive variant ADAPT‑VQE) into the Atomic Simulation Environment (ASE), thereby extending the traditional atomistic simulation workflow to include quantum‑derived energies, forces, dipoles, vibrational frequencies, strain analyses, and molecular dynamics. The core of the implementation is a custom ASE Calculator that receives an ASE Atoms object, forwards the geometry to a PySCF driver within Qiskit Nature to generate one‑ and two‑electron integrals in a user‑specified Gaussian basis (STO‑3G for all benchmark calculations), maps the resulting fermionic Hamiltonian to qubit operators via the Jordan–Wigner transformation, and then runs either a fixed UCCSD ansatz or an ADAPT‑VQE ansatz on a selected quantum backend (statevector, noisy simulator, or IBMQ runtime). Classical optimization of the variational parameters is performed with standard optimizers such as SLSQP or COBYLA, while expectation values are obtained from the Qiskit Estimator with configurable shot counts and noise models.

For force evaluation the authors employ a central finite‑difference scheme: each atomic coordinate is displaced by a small step, the quantum energy is recomputed, and the derivative is formed. Crucially, all displaced geometries reuse identical electronic‑structure settings and convergence thresholds, which dramatically reduces numerical noise and yields consistent forces across geometry optimizations, vibrational analyses, and dynamics. Dipole moments are extracted from auxiliary operators in Qiskit Nature and converted to Debye, enabling direct use of ASE’s infrared spectroscopy tools.

Benchmark studies on minimal‑basis H₂, BeH₂, and F₂ demonstrate the practical capabilities of the framework. For H₂, geometry optimization with the UCCSD ansatz on the statevector simulator reproduces the exact STO‑3G bond length (0.735 Å) and energy (−30.948 eV) within 10⁻⁵ eV of a classical PySCF CCSD reference, and the harmonic frequency deviates by less than 0.04 %. In the more challenging F₂ case, a naïve fixed‑UCCSD VQE fails to converge reliably, yielding an unphysical vibrational frequency (~77 300 cm⁻¹). By contrast, ADAPT‑VQE builds a compact, system‑specific circuit based on energy‑gradient operator selection, achieving a realistic frequency (~1 673 cm⁻¹) and an energy consistent with STO‑3G CCSD. These results highlight that the dominant source of error in the present demonstrations is the minimal basis set rather than the quantum algorithm itself, and that adaptive ansatz construction is essential for stable forces in multi‑electron systems.

To address the prohibitive cost of evaluating quantum forces at every MD step, the authors integrate the FALCON active‑learning framework. During a Langevin dynamics simulation, a Gaussian‑process surrogate model is trained on‑the‑fly from previously computed quantum forces; the surrogate supplies forces for most steps, while the quantum calculator is queried only when the model’s uncertainty exceeds a preset threshold. This strategy enables tractable, long‑time MD trajectories that retain quantum‑mechanical fidelity where it matters most.

All code is released under an open‑source license at https://github.com/thequantumchemist/ase_quantum_vqe/, allowing users to switch backends, adjust optimizer settings, and extend the calculator to larger basis sets or embedding schemes. Although the current study is limited to STO‑3G calculations on simulated devices, the architecture is compatible with real NISQ hardware and with future extensions such as trans‑correlated Hamiltonians, QM/MM partitioning, or more sophisticated noise‑mitigation techniques.

In summary, the paper demonstrates that variational quantum algorithms—particularly the adaptive ADAPT‑VQE—can be seamlessly embedded into a full atomistic simulation stack, delivering chemically meaningful forces and vibrational properties, and enabling downstream applications such as geometry optimization, strain analysis, and active‑learning‑accelerated molecular dynamics. This work represents a significant step toward practical quantum‑enhanced materials modeling and establishes a reusable platform for further methodological advances.