CP2K program package

CP2K is an opens source electronic structure code for atomistic simulation of a broad range of systems including liquids, molecules and materials. CP2K supports different levels of theory, DFT, TDDFT, MP2, RPA, semi-emperical methods, classical force fields and QM/MM. CP2K provides also different simulation modes such molecular dynamics, metadynamics, Monte Carlo and many more.

Read more

I contributed to different functionalities of CP2K including the QM/MM and force field modules, electrostatic-potential fitting, DFT and integral evaluation and GW. I have developed and implemented the following functionalities:

Image charge augmented QM/MM (IC-QM/MM)

The IC-QM/MM scheme is designed for simulation of molecules at metallic surfaces. The QM part is described by DFT, whereas the interaction between the metal atoms is modeled by classical force fields. The interaction between both subsystems is described at the MM level of theory accounting for the polarization of metal and adsorbate by an image charge approach. The computational overhead compared to a standard QM/MM calculation is negligible. Details of theory and implementation can be found in J. Chem. Theory Comput., 9, 5086 (2013).

See IC-QM/MM tutorial.

Restrained electrostatic potential fitting (RESP) for periodic systems

Electrostatic potential fitting is a popular approach to obtain atomic charges for, e.g., force field or QM/MM molecular dynamics. I implemented a periodic version of RESP fitting in CP2K, as well as, different methods to sample the fitting points, for details of the implementation and application to surface systems, see Phys. Chem. Chem. Phys., 17 , 14307-14316 (2015). CP2K features also a slight variant of periodic RESP fitting, the so-called REPEAT method.

See RESP tutorial

Local resolution-of-the-identity scheme for DFT (LRIGPW)

DFT calculations in CP2K are based on the Gaussian and plane waves (GPW) method. The computational most expensive part in the GPW method are the grid-based calculation of the KS density and KS matrix elements. I implemented a local resolution-of-the identity (LRI) approach for the GPW approach (LRIGPW), which reduces the computational cost for these operations. More precisely, the prefactor for the grid operations is reduced, while the linear scaling of the GPW approach is retained. Details of theory and implementation can be found in J. Chem. Theory Comput., 13, 2202 (2017).

See LRIGPW tutorial

Solid harmonic Gaussian (SHG) integrals

The traditional Obara-Saika scheme for the evaluation of Gaussian integrals can be computationally demanding for high quantum number and highly contracted basis sets. I developed a computationally more efficient integral scheme based on solid harmonic Gaussian (SHG) functions. These integrals are used, for example, in LRIGPW. For details see J. Chem. Phys., 146, 034105 (2017).

Low-scaling GW

I contributed to the low-scaling implementation in CP2K. The GW calculations performed in J. Phys. Chem. Lett., 9, 306 (2018) are one of the largest reported so far.

FHI-aims program package

FHI-aims is an all-electron electronic-structure package for the simulation of molecular systems and materials. FHI-aims supports DFT and different many-body theories, e.g., MP2, RPA or GW.

Read more

I contributed to the many-body methods in FHI-aims, most notably to the GW implementation. I implemented the contour deformation approach to enable accurate calculations of core-level excitations, see J. Chem. Theory Comput. 14, 9 (2018) . I also implemented different levels of partial self-consistency, see also Front. Chem. 7, 377 (2019).