Highly Efficient, Linear-Scaling Seminumerical Exact-Exchange Method for Graphic Processing Units.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Accurate calculation of molecular electronic structure is crucial for understanding chemical reactions and material properties.
• Previous methods, like the CPU-based Laqua, faced scalability challenges for large systems.

Purpose of the Study:

• To present sn-LinK, a novel graphic processing unit (GPU)-accelerated seminumerical exact-exchange method.
• To improve the efficiency and scalability of quantum chemical calculations for large molecules.

Main Methods:

• Development of the sn-LinK method, leveraging GPU acceleration.
• Implementation of advanced integral bounds and high-accuracy numerical integration grids.
• Assessment of accuracy using established test sets and performance analysis on large molecules (62-1347 atoms).

Main Results:

• sn-LinK achieves errors significantly below 1 mEh for the smallest integration grid.
• Demonstrated outstanding performance, especially for large basis sets (e.g., polarized quadruple-zeta with diffuse functions).
• Efficient scaling for large molecular systems.

Conclusions:

• sn-LinK offers a highly efficient and accurate approach for quantum chemical calculations.
• The GPU acceleration and advanced techniques enable reliable simulations of large molecules.
• This method advances the capabilities of computational chemistry for complex systems.