Konferenzprogramm und Talks

Montag, 24.09.2018, 14:30-15:00 Uhr

Particle-based methods are increasingly used for simulating high-resolution violent fluid effects with splashing and large deformations. In particular, the Fluid-Implicit-Particle (FLIP) method can easily use hundreds of millions of particles and has led to a tremendous increase in the number of particles used in visual effects simulations. Besides the simulation, rendering these gigantic particle sets becomes a severe bottleneck in the production pipeline, where in particular previewing tools for such simulations are important to analyze and control the effects. In my talk I will shed light on some recent developments in FLIP-based fluid simulation and particle rendering. In the first part, I will demonstrate a stable narrow band approach that efficiently reduces the number of particles in a FLIP simulation and leads to significant speed-up time. The second part of my talk is devoted to the rendering of particle-based simulations, demonstrating how to achieve interactive rendering of simulations reaching the giga-particle scale. By making use of an adaptive voxel-grid, bandwidth limitations can be relaxed throughout the entire rendering pipeline and parallel GPU processing can be employed to achieve high performance.

Montag, 24.09.2018, 16:00-16:30 Uhr

Results from a round robin study are presented, in which the issue of systematic errors in molecular simulations was addressed. The same set of simulation tasks was given to independently working user groups that worked with different simulation codes to solve these tasks. Under ideal circumstances, i.e. in the absence of systematic errors, the results from the different groups should agree within their statistical uncertainties. The results from the study show that this is not the case in real life. The study is broad: there were five participating groups, which worked with eight different molecular simulation codes. The tasks were quite simple: the determination of the density and the energy of four pure liquids, i.e. ethane, propane, n-butane, and iso-butane, on a given temperature–pressure grid. Three molecular model types were used, all of them with inter­nal degrees of freedom. The collected data demonstrate that systematic errors are important in molecular simulations. In many cases, the deviations between the results ob­tained by the different groups exceed the statistical uncertainty of the individual results by far. Potential reasons for the systematic simulation errors are plentiful and are briefly discussed. The conclusion is that systematic simulation errors can generally not be avoided once a certain degree of complexity of the simulation is reached. But there are rules of best practice that help mitigating them.

Montag, 24.09.2018, 17:40-18:10 Uhr

In nanoelectronics realistic system sizes for the core functionalities are within the range, which is accessible by high-precision quantum-mechanical numerical methods at the electronic structure level. The building blocks presented here comprise single molecules in photo-switchable tunnel junctions, arrays of extended donor-acceptor polymer chains with hopping-type carrier transfer, and more extended nanowires from elemental group-IV semiconductors with ballistic conduction. In all three cases, however, the operation of the functional element is strongly dependent on interactions with the environment, and the latter may be too large or too complex for explicit modeling with the same high level of precision. Hence, schemes to include influences from the environment are based on a series of approximations, which lead from a fully quantum-mechanical treatment to parameter transfer into classical rate equations. I will present examples, which rely on suitable extensions of the electronic Hamiltonian to include excitations by light and by fields, on the sampling of the phase space under coupling to thermostat and / or barostat, and on embedding into circuit simulations from classical electrical engineering. By carefully choosing the description of the environmental effects and their coupling to the core functionality model can be derived, which – at least to some extent – even have predictive power.

Dienstag, 25.09.2018, 9:00-9:30 Uhr

Solid-state nanopores are single molecule sensors used for the detection and analysis of biomolecules like DNA, RNA and proteins. The idea is as simple as intriguing since the analysis allows single molecule sensitivity and solely relies on measuring the changes in ionic current through the nanopore. First, we will discuss our recent results on DNA translocations. We observe an entropic barrier limited, length dependent translocation rate at 4M LiCl salt concentration and a drift-dominated, length independent translocation rate at 1M KCl salt concentration. These observations are described by a unifying convection-diffusion equation which includes the contribution of an entropic barrier for polymer entry. Based on our understanding of DNA translocation, we introduce DNA carriers that have specific protein binding sites to control protein translocation. Protein detection down to the single protein level is achieved, allowing for identification a single protein species within complex mixtures and multiplexed protein sensing. Since we fully understand the electro-kinetic flows and forces in and around the nanopore control of the translocation process is possible.

Dienstag, 25.09.2018, 10:50-11:20 Uhr

Numerical simulations on large supercomputers have become the tool of choice to predict the non-linear outcome of cosmic structure formation. These calculations connect the simple initial conditions left behind by the Big Bang with the highly non-linear evolved Universe today, and allow tests of central cosmological assumptions, like the existence of a gravitationally dominating collisionless dark matter component. In my talk, I review part of the basic methodology of cosmological simulations, which involve very large particle numbers and aim to address an enormous dynamic range as well as multi-physics complexity. I will also describe current forefront results, which help to understand the formation process of galaxies and its regulation by feedback from supernova explosions and supermassive black holes.

Dienstag, 25.09.2018, 13:30-14:00 Uhr

Solvents are widely used in chemical industry to facilitate a broad variety of applications. It is well-known that they may strongly influence reaction kinetics as well as reaction yield, i.e. reaction equilibria. The key property for describing reaction equilibria is the thermodynamic equilibrium constant  which is directly related to the equilibrium concentrations of reactants and products () and their activity coefficients (). The equilibrium constant  only depends on temperature (and pressure) but neither depends on reactant concentrations nor on solvents. In contrast, is often strongly influenced by solvents and reactant concentrations and that’s why also (and therewith the yield) depends on these properties. Using a thermodynamic model (e.g. PC-SAFT) for determining the activity coefficients of reactants and products allows for predicting the influence of solvents on the equilibrium concentrations once  is known. This will be demonstrated for various chemical and biochemical reactions. Predictions at varying reactant concentrations as well as for different solvents and solvent mixtures will be compared to experimental data. In all cases, the equilibrium concentrations (and therewith yield) were strongly influenced by reactant concentration/solvents and the predictions were in almost quantitative agreement with the experimental data. As solvents influence the reactant activity coefficients, they also do influence reaction kinetics. If the catalyst is not affected by the solvent, this can be again almost quantitatively predicted using a thermodynamic model which will be shown for various example reactions. Thus, using thermodynamic modeling allows for predicting the solvent influence on chemical (and biological) reactions and therewith drastically reduces the experimental effort for selecting the best-suitable solvent for a reaction of interest.  

Dienstag, 25.09.2018, 15:20-15:50 Uhr

Nucleotide excision repair of damaged DNA requires recognition of the DNA lesion and a conformational transition (flipping) of damaged nucleotides from an intra-helical to an extra-helical conformation. It is still unclear how the damage is efficiently identified and looped out from an energetically stable stacked and base paired state without any external energy source. Typically, binding of repair enzymes leads to local and global deformations of the target DNA. For two frequent DNA lesions, oxidation of guanine to form 7,8-dihydro-8-oxo-guanine (8oxoG) and cis-syn-cyclobutane pyrimidine dimer (CPD) formation, we have performed advanced sampling free energy simulations to study the flipping process for globally deformed DNA conformational states. The simulations indicate that global untwisting deformation of the DNA towards the enzyme bound form alone (in the absence or protein) significantly reduces the penalty for damage flipping. The finding offers a mechanistic explanation how binding free energy that is transformed to binding induced DNA deformation facilitates flipping and gives insight into the role of protein-DNA contacts. Further simulation studies on the general structural effects of twist deformations in DNA will also be presented. 

Mittwoch, 26.09.2018, 9:00 - 9:30 Uhr

Intense femtosecond-laser pulses are able to induce ultrafast nonthermal phase transitions in different materials along pathways that are inaccessible under thermodynamic conditions. In order to investigate the motion of atoms in the nonthermal transient state produced by femtosecond laser excitation we performed ab-initio molecular-dynamics (MD) simulations on laser-excited potential energy surfaces using our code CHIVES (Code for Highly excited Valence Electron Systems). We found surprising results, like excitation of squeezed thermal phonons, which constitute the precursor of nonthermal melting as a function of fluence, the presence of  atomic fractional diffusion, as a transient state preceding the formation of a nonthermal liquid, and laser-fluence dependent anisotropic energy redistribution. Also, the possibility of controlling nonthermal melting by pulse shaping was considered. In this talk, a brief overview on some of the results of will be given. In order to extend our method for the study of nucleation and cooperative phenomena, we developed an analytical interatomic potential for laser-excited silicon, which depends on the electronic temperature. With the help of this potential we were able to perform large-scale simulations and study the nonthermal dynamics. In addition, results from a hybrid atomistic-continuum model for the laser-induced nanostructuring of gold surfaces upon excitation with spatially modulated ultrashort UV pulses will be shown. A TTM-MD approach allowed us to perform calculations on experimental spatial and time scales and to achieve a one-to-one comparison with experiments. Finally, the application of a novel nonequilibrium Markov-State-Modeling approach to study the influence of microwave radiation on proteins will be presented.