# Computational Chemistry Group - Research

The Computational Chemistry Group simulates chemical and biochemical reactions under consideration of environmental effects. We complement experiment and provide answers where experiment alone is not able to do so. We study enzymes, biological receptors, astrochemistry, materials science, and catalysis. Many of the projects make use of the QM/MM method, the combination of quantum mechanics (QM) and empirical force fields (MM).

We develop methods that allow us to investigate these kinds of problems in innovative ways. The program package ChemShellis co-authored by us and we lead the development of the optimization library DL-FIND. Our group extends the capabilities of geometry optimization in systems with many thousand degrees of freedom and developed methods to calculate tunneling processes in large systems as well as free-energy sampling techniques using molecular dynamics.

## Method Development

Tunneling rates in large systems: Instanton theory allows to calculate the rate of tunneling of atoms in chemical reactions. Since the efficiency of tunneling depends on the width of a barrier as well as its height, the optimal tunneling path (shown in red in the Figure on the left) often differs from the dominant classical path (black). | |

Umbrella Integration: a novel analysis method for umbrella sampling simulations. Sampling free-energy changes is becoming more and more important in the simulation of large systems. Umbrella Integration allows to derive the free-energy change as well as its statistical error from umbrella sampling simulations. Umbrella Integration, in conjuction with our geometry optimizer DL-FIND is used in our project umbrella sampling simulations in the Collaborative Research Center 716 . | |

QM/MM method development: QM treats the chemically active center of a system (enzyme, surface, solid state) with highly accurate quantum mechanical methods, and includes the environment using efficient force field methods. We use the modular package ChemShell for QM/MM simulations. However, there is more to QM/MM than just sticking a QM and an MM code together: we implemented a microiterative QM/MM geometry optimization scheme as well as QM/MM free-energy perturbation. | |

Transition-state search, conical intersections, parallel geometry optimization: The PI developed the modular geometry optimization library DL-FIND. It can be interfaced to atomistic simulation codes. We use it in conjunction with ChemShell. This is a collaboration with Daresbury Laboratory, UK. | |

Machine Learning: methods like neural networks and Gaussian process regression are used to build surrogate models for the real potential energy surface. The surrogate models can be evaluated much faster. This is a way to limit the number of required energy evaluations for tasks like rate calculations and geometry optimization. |

## Applications

Astrochemistry: The darkness readily observed between the stars on a clear night sky is far from empty. In fact, a large variety of molecules has been detected during the past century. We contribute with our rate calculations incorporating the effect of atom tunneling down to low temperatures. | |

Signaling of the Epidermal Growth Factor Receptor is induced by ligand binding. This, in turn, induces oligomerization of the receptor. In collaboration with experimentalists (FRET measurements) we investigated the alignment of the receptor on the cell membrane. This induces changes in the ligand binding affinity. | |

Catalytic mechanism of nitrogenase: Nitrogenase is able to break the strongest covalent bond in nature: the N-N triple bond in dinitrogen. We investigated how the enzyme manages this extraordinary task by means of density functional (DFT) calculations. |