Machine-learned density functional based quantum chemical computations for ethane: performance of DeepMind 21 on...

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Artificial Intelligence in Science

Background:

• Machine learning (ML) offers a promising approach for quantum chemistry calculations.
• DeepMind 21 (DM21) is a neural network-based functional demonstrating superior performance in Density Functional Theory (DFT).
• There is a need for more research applying AI models like DM21 to quantum science computations.

Purpose of the Study:

• To investigate quantum chemistry algorithmic computations for the ethane molecule (C2H6) using ML.
• To employ the DeepMind 21 neural density functional to compute molecular properties and potential energy surfaces.
• To assess the accuracy of DM21 by comparing its predictions with conventional DFT methods and high-level benchmarks.

Main Methods:

• Utilized the DM21m TensorFlow neural network model to predict exchange-correlation potentials.
• Computed self-consistent field (SCF) energies using the PySCF package with a cc-pVDZ basis set.
• Calculated dipole moment, molecular orbitals (HOMO/LUMO), and long-range interactions for ethane.

Main Results:

• The DM21 functional accurately predicted potential energy surfaces and molecular properties for ethane.
• Results from DM21 showed close agreement with the reference Coupled Cluster Singles Doubles with Perturbation theory (CCSD(T)) benchmark.
• DM21 demonstrated efficacy comparable to conventional DFT methods (B3LYP, PW6B95) and established literature values.

Conclusions:

• The DeepMind 21 functional is suitable for quantum science computations, including potential energy surface generation for the ethane molecule.
• This study marks the first demonstration of deep learning density functionals for quantum chemical computations of ethane.
• The findings highlight the potential of AI-driven functionals for advancing quantum chemistry research.