AUTHOR=Corzo Hector H. , Hillers-Bendtsen Andreas Erbs , Barnes Ashleigh , Zamani Abdulrahman Y. , Pawłowski Filip , Olsen Jeppe , Jørgensen Poul , Mikkelsen Kurt V. , Bykov Dmytro 

TITLE=Coupled cluster theory on modern heterogeneous supercomputers

JOURNAL=Frontiers in Chemistry

VOLUME=Volume 11 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/chemistry/articles/10.3389/fchem.2023.1154526

DOI=10.3389/fchem.2023.1154526

ISSN=2296-2646

ABSTRACT=Basic computational limitations, such as the high polynomial scaling that accompanies the growth of dimensionality,  are inherent in attempts to elucidate complex chemical systems and processes with quantum mechanics.  These limitations are especially notable in ab-initio methodologies designed to capture electron-electron correlation effects and to obtain exact numerical solutions to the non-relativistic many-body  Schrödinger equation.
If more efficient quantum chemical methods are adopted, many of these computational bottlenecks could be overcome with current technologies. As such, the Divide-Expand-Consolidate (DEC) approach for coupled cluster (CC) theory is a suitable approach that exploits both rigorous theoretical chemistry concepts and high-performance computing techniques that establishes a link between performance and accuracy for facilitating reliable predictions of energies and observables in large chemical systems. The DEC scheme is a linear-scaling and massively parallel framework. It is designed as a black-box method with predetermined error thresholds in the correlation energy and quantitative molecular properties. Furthermore, massively parallel implementations of the DEC framework that utilizes modern HPC architectures can provide a fast time-to-solution and can ensure performance and portability that will directly benefit from new hardware developments.  
The main purpose of this work is to provide a general overview of the operational background, concepts, and algorithms that allow the DEC framework to expand the applicability of the ab-initio CC methodologies  to   a large variety of applications in biochemistry, polymer chemistry, and catalysis. Additionally, we present how cluster perturbation theory can be used to obtain excitation energies along with strategies for parallelization.