Atomic orbital-based SOS-MP2 with tensor hypercontraction. I. GPU-based tensor construction and exploiting sparsity.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Theoretical Chemistry

Background:

• Second-order Møller-Plesset perturbation theory (MP2) is crucial for accurate electronic structure calculations.
• Standard MP2 methods scale poorly with system size (quartic scaling).
• Tensor hypercontraction (THC) offers a promising route to reduce computational cost.

Purpose of the Study:

• To develop and implement a computationally efficient THC-based SOS-MP2 (scaled opposite spin second-order Møller-Plesset perturbation theory) method.
• To reduce the formal scaling of SOS-MP2 calculations from quartic to cubic.
• To achieve practical efficiency gains through sparsity and GPU acceleration.

Main Methods:

• Implementation of the tensor hypercontracted (THC) approximation within the scaled opposite spin (SOS) second-order Møller-Plesset perturbation theory (SOS-MP2) framework.
• Exploitation of sparsity in atomic orbital representations.
• Utilization of graphical processing units (GPUs) for accelerating integral construction and matrix multiplication.

Main Results:

• The developed THC-SOS-MP2 method achieves a practical scaling of N(2.6) for calculations involving up to 1600 basis functions.
• GPU acceleration significantly speeds up integral construction and matrix multiplication.
• Correlation energy errors are less than 0.5 kcal/mol compared to density-fitting-SOS-MP2 for systems up to 162 atoms.

Conclusions:

• The GPU-accelerated atomic orbital-based THC-SOS-MP2 method provides a significant speedup for electronic structure calculations.
• This method offers a balance between accuracy and computational efficiency for large molecular systems.
• The approach paves the way for more accurate and feasible calculations in computational chemistry.