【Featured in Journal】Towards Improved Quantum Machine Learning for Molecular Force Fields
- 田中拓哉
- Dec 25, 2025
- 1 min read
A collaborative study between Quemix and NGK Insulators has been featured in Physical Review A, highlighting advances in quantum machine learning for molecular force field prediction.
In this work, the team addressed the computational bottleneck in materials simulations by proposing an Equivariant Quantum Neural Network (QNN) that inherently reflects molecular symmetries. The study identified architectural limitations in earlier models and introduced a revised design that incorporates element-specific and distance-aware physical features.
Simulations using up to 16 qubits demonstrated improved force prediction accuracy and reduced overfitting compared to previous quantum models. Meanwhile, challenges such as stable energy prediction and data scaling behavior were also surfaced.
This result provides an important milestone toward the practical application of quantum computing in computational materials science, paving the way for hybrid quantum–classical approaches and refined encoding strategies in future research.
ニュースリリースはこちら
Comments