Q-Chem Webinar 35

Mixed-precision algorithms for coupled-cluster calculations

Pavel Pokhilko Image
Presented by Pavel Pokhilko on

Pavel Pokhilko graduated Lomonosov Moscow State University with honors. After that Pavel moved to the University of Southern California for his Ph. D. under the supervision of Prof. Anna Krylov. During his visit in Q-Chem in Pleasanton, he and Dr. Evgeny Epifanovsky generalized tensor operations in Q-Chem's libtensor library, which allowed the implementation of single-precision options for coupled-cluster methods. Pavel is now in the 4th year of his Ph. D., working on spin-orbit coupling and applying his development on large single molecule magnets.


Abstract

Waiting time for solutions in post-Hartree-Fock methods can be frustrating. To speed up the calculations, we implemented mixed- and single-precision algorithms, available for most methods in Q-Chem's ccman2 module. The errors in energies and properties, introduced by single precision, are negligible for most types of calculations. In the talk, I will show how to use single precision in practice and how to monitor the errors.

Supporting Material