At the heart of our research efforts is always the design and implementation of efficient computer code. This frequenctly includes the development of low-scaling algorithms, or the highly parallel implementation of existing algorithms for modern supercomputers like the Oak Ridge supercomputer Summit. We are currently contributing to the widely used software packages DIRAC, AMS, and TURBOMOLE. For more details, check out our software overview page.
Short summaries of our research efforts in different areas are given below. For more details, check out our publication list.
We develop hybrid methods that partition a complex chemical system into an active space and an environment. The active space part is described using a high-level method like coupled cluster, full configuration interaction, or the density matrix renormalization group while the environmental degrees of freedom and couplings with the active space are described with methods based on density functional theory, many-body perturbation theory, or approximate density functional theory. To define the active space degrees of freedom, we often rely on localized orbitals. A large part of our research efforts in this direction is dedicated to algorithms in which the active space Hamiltonian is solved on a quantum computer, for instance using the variational quantum eigensolver. Here, we actively collaborate with researchers from Leiden University and the University of Amsterdam.
Recent key publications:
Standard quantum chemistry struggles with high-precision calculations for heavy-element molecules due to strong nuclear electrostatic fields accelerating electrons to near-light speeds. Dirac’s relativistic quantum theory properly describes this regime, enabling a more general quantum chemical framework. Through a long-standing European collaboration, we develop relativistic methods with a focus on accurately treating electron correlation effects. Working with DIRAC program users, we have contributed to diverse topics, including metrology, quantum computing, nuclear quadrupole moments, and main group metal clusters. We primarily focus on actinide chemistry, a field now accessible for accurate quantum modeling. By employing multireference coupled cluster techniques with relativistic Hamiltonians, we achieve chemical accuracy for realistic systems. We collaborate with applied theoretical groups in the US, Germany, and France to explore various aspects of actinide chemistry.
Recent key publications:
We develop highly accurate and cost-efficient computational methods to model spectroscopic processes and ground state properties in complex finite and periodic systems using methods from many-body perturbation theory based on single-particle Green's functions. A central area of our research is the GW approximation and the GW-Bethe-Salpeter equation approach to describe excited states and the random phase approximation for ground state energies. We also work on going beyond these methods with so-called vertex corrections. Combination of these techniques with multi-reference methods via Green's function embedding techniques is currently explored in our group. The main application areass are photochemistry, heterogeneous catalysis, and the investigation of degradation processes in Lithium-ion batteries.
Recent key publications:
We develop methods and algorithms to calculate optical and vibrational spectra within time-dependent density functional theory and beyond. Our group focuses on infrared and resonance-Raman spectroscopy, as well as vibrational circular dichroism that measures the difference in absorption of left- and right-circularly polarized infrared light by chiral molecules, providing insights into their molecular structure and stereochemistry. We also work on techniques within relativistic density functional theory to calculate optical excitations in complicated open-shell molecules containing heavy elements that are of relevance for instance in the rational design of organic light-emitting diodes. To bypass relatively costly first-principle calculations we also develop tight-binding approximations, machine learning models, and graph neural networks to calculate spectroscopic properties.
Recent key publications: