Large Scale Quantum Chemistry with Tensor Processing Units.

Area of Science:

• Computational Chemistry
• Materials Science
• Quantum Mechanics

Background:

• Density Functional Theory (DFT) is a powerful quantum mechanical method for electronic structure calculations.
• Conventional DFT methods exhibit cubic-scaling computational complexity, limiting their application to small systems.
• Accelerating DFT calculations is crucial for advancing scientific discovery in various fields.

Purpose of the Study:

• To demonstrate the use of Google's cloud-based Tensor Processing Units (TPUs) for accelerating and scaling conventional DFT calculations.
• To achieve the largest explicit DFT computation to date.
• To enable accessible and systematic DFT at unprecedented scales.

Main Methods:

• Utilized 512 Google TPU cores for parallel computation.
• Performed conventional cubic-scaling DFT calculations.
• Calculated the electronic structure of a large cluster of water molecules.

Main Results:

• Successfully scaled conventional DFT calculations to unprecedented sizes.
• Executed the largest DFT computation to date, involving 247,848 orbitals and 103,270 electrons.
• Demonstrated the feasibility of large-scale, explicit DFT simulations.

Conclusions:

• Google's TPUs can significantly accelerate and scale conventional DFT.
• This approach enables systematic and accessible DFT studies without system-specific constraints.
• Opens new avenues for large-scale quantum mechanical simulations in chemistry and materials science.