Research Interests
Theoretical Chemistry is an interdisciplinary field. It combines Mathematics, Physics and Chemistry to predict and analyze the behavior of matter at the molecular scale. Tools developed by Theoretical Chemists are currently used by Experimentalists to analyze and support their data. Applications range from small molecules (e.g. in Astrochemistry) over medium sized molecules (e.g. in Organometallic Chemistry) up to large molecules (e.g. in Biochemistry and drug design).
Theoretical Chemists develop, test and apply theories, methods and models. My main focus is on development and testing of theories, methods and models. On the one hand I develop theories and functionals that are tested for well-known benchmark systems. On the other hand I develop methods to be used by experimentalists for their respective systems.
Electronic Structure Methods (DMFT, DFT, range-separation, and PQT)
I mainly work on Density-Matrix Functional Theory (DMFT) and Density Functional Theory (DFT). I develop functionals that accurately and efficiently describe bond breaking and bond forming processes. The functionals are benchmarked against the dissociation curve of small test molecules. The goal is to reduce the error along the dissociation curve to the same order of magnitude as is achieved by present day DFT functionals around equilibrium distance.
A popular method to to combine different theories is the range-separation scheme. In a recent paper we were able combine DFT and DMFT by range-separation. Preliminary calculations show promising results for the dissociation curve of a few small test systems.
Recently, I got interested in Wave Function Theory (WFT). I work on Projected Quasiparticle Theory (PQT) that deliberately breaks and then restores symmetries by projection. The resulting wave function is of multi-determinantal character, however, the computational cost is that of a mean-field theory. In particular, I focus on recovering the remaining correlation energy.
Isotope Effects (QM/MM, AKIRA and Isotope Effects)
Quantum Mechanical and Molecular Mechanical coupling schemes (QM/MM) are used to simulate large molecules. I’ve worked on a particular implementation in the past focusing on the coupling of the second derivative matrices (the Hessians). The Hessian is needed to obtain the vibrational frequencies of a system. For large systems, the calculation of the Hessian is time consuming and might be impossible. It is possible to calculate only selected frequencies by an eigenvector following method implemented in AKIRA. I’m currently working on extending AKIRA to interface with CHARMM and fDynamo.
In a next step I will develop a method that allows to calculate isotope effects. The challenge is to identify the frequencies that contribute to the isotope effect. Once identified it will be easy to obtain them with AKIRA interfaced to the QM/MM programs CHARMM and fDynamo. It will be interesting to see how accurate the method will predict isotope effects.
Free Energy Surface (Molecular Dynamics, Meta Dynamics and BEMD)
In order to obtain the free energy surface for a given system statistical effects need to be added to the potential energy surface. A popular method to do this is molecular dynamics. To reduce the computational effort the system is biased in the direction of interest. This method is called Meta-Dynamcis (MD). The computational effort of MD scales exponentially with the number of biased directions.
Bias-exchange Meta-Dynamics (BEMD) reduces the computational effort of MD. A number of MD runs are run parallel. In each of the runs the bias is in a different direction. After some time steps the coordinates of the runs are exchanged at random. This allows for a quick exploration of the free energy surface, discovering new minima, that weren’t imagined before. Ultimately, a subspace of the free energy surface can be reconstructed from the collected data.
Programing (Python, PyQuante and Pyevolve)
While most programs in the field of Theoretical Chemistry are written in Fortran or C/C++ I’m also using Python. This high-level object-oriented programing language allows me to check ideas quickly. The programing is easily done and debugging usually doesn’t take too much time.
In particular for my research on OEP I used the PyQuante package. It features a set of electronic structure methods including HF, MP2 and some density functionals.
Recently, I started using pyevolve to optimize parameters in my density-matrix functionals. Pyevolve is a genetic algorithm that optimizes a given function by changing the parameters due selection, reproduction and mutation.