UO > Research > Projects & Grants

Projects & Grants

Fixed-node diffusion Monte Carlo as a benchmark method for large noncovalent systems
Project Id18-24321Y
Main solverIng. Matúš Dubecký, PhD.
Period1/2018 - 12/2020
ProviderJuniorský projekt GA ČR
AnotationIn recent years, fixed-node diffusion Monte Carlo method with 1-determinant input wave functions (FNDMC) reached high accuracy in computations of noncovalent interaction energies, comparable to ?gold standard? (CCSD(T) in complete basis set limit). Thanks to its more favourable CPU cost scaling and access to periodicity, FNDMC thus gains popularity as an unprecedented benchmark tool for large noncovalent systems (including periodic). Recent results suggest, that the required reference level (0.1 kcal/mol) is not always accessible, or, the results depend on method parameters that have not yet been taken into account. These problems are now largely ignored, or, the results are overvalued. We plan systematic large-scale screening of FNDMC accuracy for noncovalent interactions, identification and conceptual understanding of the key error sources, method improvements, and applications. In addition to deep physical insights to the FNDMC methodology, the results of the project will lead to rational usage and fine bias control of this method important for large noncovalent systems.
social hub