NOMAD CoE
Massively Parallel Coupled Cluster Theory Calculations for Materials Science

NOMAD CoE researchers from TU Wien and the Fritz Haber Institute have developed novel computer codes to enable massively parallel and highly accurate coupled cluster theory simulations of materials.

One of the great challenges in computational materials science, using ab initio methods and high performance computing, is to reach a good compromise between accuracy and computational cost. The coming era of exascale computing holds the promise to achieve significantly more accurate predictions using high-level electronic structure theories that have previously not been feasible. One such method is coupled cluster theory, which has the potential to predict materials properties with significantly more controllable accuracy compared to state-of-the-art methods.

Recently the research group of Andreas Grüneis at TU Wien has released the open-source software Cc4s, which aims at efficient coupled cluster level calculations of periodic solids and surfaces [1]. The combination of its massive parallelization on thousands of compute cores with its interface to widely-used electronic structure theory packages, makes it a promising benchmark tool that was so far missing in the toolbox of computational materials scientists.

Together, researchers from TU Wien and the Novel Materials Discovery (NOMAD) Laboratory of the Fritz Haber Institute have created a novel interface between Cc4s and the FHI-aims code, which can be used to study atoms, molecules and solids on the coupled cluster levels of theory employing numerically tabulated atom-centered orbitals as implemented in FHI-aims [2]. In an upcoming tutorial in Riga [3], the involved researchers will present the newly developed computational tools and describe its current capabilities as well as its implementation.

To demonstrate the capabilities of Cc4s and its interface to the Vienna ab initio simulation package, Tobias Schäfer and members of the NOMAD Centre of Excellence at TU Wien have computed the adsorption energy of a single water molecule on a graphitic carbon nitride sheet [4]. Graphitic carbon nitride is a promising metal-free water splitting material and the adsorption of a single water molecule – depicted above – forms an important first step in this reaction. The calculated molecular adsorption energy on the level of highly accurate CCSD(T) theory can be used as a reliable benchmark for a wide range of dispersion corrected density functionals [4]. In this particular case, the simulated system contains more than 100 atoms, which necessitates the introduction of embedding schemes. However, thanks to the computational efficiency of Cc4s, ab initio studies of systems containing up to 50–100 atoms can also be performed on modern HPC clusters directly.

Future work within the NOMAD Centre of Excellence will include creating additional interfaces between Cc4s and widely-used electronic structure theory codes as well as adapting Cc4s to improve its performance on the next generation of (pre-)exascale HPC clusters.

 

References:

[1]: Cc4s, https://manuals.cc4s.org/user-manual/

[2]: CC-aims, https://gitlab.com/moerman1/fhi-cc4s

[3]: https://nomad-coe.eu/nomadmidsummerschoolriga

[4]: T. Schäfer, A. Gallo, A. Irmler, F. Hummel, and A. Grüneis, J. Chem. Phys. 155, 244103 (2021). https://doi.org/10.1063/5.0074936