My research can be subdivided in three major themes that are briefly discussed below.
1. Subsystem Methods based on Density Functional Theory
In this programme we develop and apply subsystem approaches that connect various quantum and classical mechanical methods. Using density functional theory (DFT) as unifying theory, we work on methods in which local electronic properties are calculated with accurate coupled cluster methods while the electronic structure of the environment is treated using a DFT or approximate DFT approach. This not only speeds up calculations by orders of magnitudes but also keeps a local picture that facilitates the transfer of results to model hamiltonian approaches. This enables multiscale approaches to the modeling of complex systems for a wide range of research fields; examples are enzymatic catalysis, virtual screening of pharmaceutically active compouds and light harvesting, in which respectively reaction rates, binding affinities, and excitonic couplings are input for models used at higher length and time scales.
2. Reducing the time-to-solution of computational models
In this line of research I work closely with the in-house SCM company and computer scientists to improve upon the speed and numerical accuracy of DFT-based modeling. We have recently extended our python-based scripting environment for multiscale computational workflows to automatically exploit parallelism. This enables use of the relatively cheap resources available in computational grids.
To shorten the time spent in indivdual calculations we work on tensor contraction algorithms to scale up to the thousands of processor cores available in the next generation supercomputers. This is combined with the development of algorithms suitable for GPU-accelaration as they are found in both supercomputers as well as in desktop computers.
An important research line in this field concerns the investigation of quantum chemistry algorithms that can be run on quantum computers. Here we consider composite methods in which an otherwise intractable electron correlation problem can be solved using a quantum computer, while the another part of the calculation is done using a conventional computer.
3. Development and application of relativistic computational chemistry techniques
Standard quantum chemistry techniques are not suited for high precision calculations on molecules containing heavy elements because the strong electrostatic potential in the vicinity of the nuclei accelerates the electrons to velocities that approach the speed of light. The Dirac relativistic quantum theory of Dirac describes this high-velocity regime properly and can thus be used to develop a more general quantum chemical framework. In a long-standing European collaboration we work on development of relativistic methods with a focus on methods that allow for accurate treatment of electron correlation effects. In collaboration with users of our DIRAC program system I thereby worked on a wide range of topics (metrology, quantum computing, nuclear quadrupole moments, main group metal clusters) for which the use of highly accurate electronic structure methods is crucial. In my own group I mostly focused on applications to actinide chemistry, a field that has only just become accessible for accurate quantumchemical modeling. We are now able to achieve chemical accuracy by use of multireference coupled cluster techniques based on a relativistic Hamiltonian. With our development of efficient algorithms and advances in computer technology it is nowadays possible to study realistic systems and we collaborate with applied theoretical groups (in the US, Germany and France) to study various aspects of actinide chemistry.
