Quantum Graph Neural Network Models for Materials Search.

Area of Science:

• Quantum computing
• Materials science
• Computational chemistry

Background:

• Classical graph neural networks (GNNs) are increasingly used in materials research.
• Predicting molecular properties like energy gaps is crucial for discovering new materials.
• Quantum computing offers new paradigms for complex simulations.

Purpose of the Study:

• Introduce a novel quantum graph neural network (QGNN) model.
• Evaluate QGNN performance in predicting molecular properties.
• Compare QGNNs with classical GNNs for materials research.

Main Methods:

• Developed a QGNN model inspired by classical GNNs.
• Utilized the equivariantly diagonalizable unitary quantum graph circuit (EDU-QGC) framework.
• Applied QGNNs to predict the energy gap of small organic molecules.

Main Results:

• QGNNs achieved lower test loss than classical models with similar trainable variables.
• QGNNs demonstrated faster convergence during training.
• The EDU-QGC framework enabled discrete link features and minimized quantum circuit embedding.

Conclusions:

• QGNNs represent a powerful new tool for predicting chemical and physical properties of molecules and materials.
• The proposed QGNN model offers advantages in accuracy and training efficiency over classical approaches.
• This work provides a foundation for further development of quantum machine learning in materials science.