NIC Symposium 2018

22 – 23 February 2018 Forschungszentrum Jülich, Germany

Programme

Bus Schedule

Poster Abstracts

Participants

Programme

Thursday, 22 February 2018

8:30 Transfer from Jülich

9:00 Registration

9:30 Welcome Address by S. M. Schmidt, Member of the Board of Directors, Forschungszentrum Jülich

9:45 Th. Lippert, Forschungszentrum Jülich Supercomputer Evolution at JSC

10:30 Coffee

11:00 T. Korzec, Bergische Universität Wuppertal How Strong are the Strong Interactions?

11:45 C. Urbach, Universität Bonn Hadron-Hadron Interactions from Lattice QCD

12:30 Group Photograph

12:45 Lunch

14:00 L. Pastewka, Universität Freiburg Structure and Mechanical Properties of Cu|Au and Cu|Ni Nanolaminates

14:45 Ch. Scheurer, Technische Universität München Computational Modelling of Solid State Battery Materials

15:30 Coffee

16:00 T. D. Kühne, Universität Paderborn Linear-Scaling Self-Consistent Field Theory Based Molecular Dynamics: Application to C60-Buckyballs Colliding with Graphite

16:45 H.-P. Hsu, Max-Planck-Institut für Polymerforschung, Mainz Effect of Entanglements on Non-Equilibrium Polymer Dynamics

17:30 Poster Session and Reception

19:00 Transfer to Jülich

Friday, 23 February 2018

8:30 Transfer from Jülich

9:00 M. Breit, Universität Frankfurt Numerical and High Performance Computing Methods for Multiscale Computational Neuroscience

9:45 M. B. Poulsen, Niels Bohr Institute, University of Copenhagen Southern Ocean Eddy Compensation Examined with a High-Resolution Ocean Model

10:30 Coffee

11:00 J. P. Mellado, Max-Planck-Institut für Meteorologie, Hamburg The Challenge of Small-Scale Turbulence in Planetary Boundary Layers

11:45 F. K. Röpke, Universität Heidelberg Towards Multidimensional Hydrodynamic Simulations of Stars

12:30 Lunch

14:00 C. Burstedde, Universität Bonn Scalable Algorithms for Adaptive Mesh Refinement: Extension to General Element Types and Application to Fluid Dynamics

14:45 A. Fiolitakis, Deutsches Zentrum für Luft- und Raumfahrt, Stuttgart Numerical Investigation of Reacting Flows in Gas Turbine Model Combustors

15:30 Coffee

16:00 S. Trebst, Universität Köln Exact Results for the Many-Electron Problem: Competing Orders in a Nearly Antiferromagnetic Metal

16:45 Th. Voigtmann, Deutsches Zentrum für Luft- und Raumfahrt, Köln Active Brownian Particle Dynamics at High Densities

17:30 End of NIC Symposium

17:45 Transfers to Jülich and to Düren Train Station

Bus Schedule

Bus is free of charge for participants.

Date Departure Meeting Point Destination Time 22 February 08:30 Bus stop „Neues Rathaus“ (in front of Forschungszentrum Jülich, "Stadtverwaltung") Auditorium

19:00 Forschungszentrum Jülich, Auditorium Jülich City, Bus stop „Neues Rathaus“ 23 February 08:30 Bus stop „Neues Rathaus“ (in front of Forschungszentrum Jülich, "Stadtverwaltung") Auditorium

17:45 Forschungszentrum Jülich, Auditorium Jülich City, Bus stop „Neues Rathaus“ Düren Train Station

Hotel Kaiserhof

Stadthotel Bus Stop "Neues Rathaus" Hotel am Hexenturm

Train/Tram Station Rurtalbahn (RTB) "Jülich Bahnhof"

Poster Abstracts

The number in brackets in front of the title is the number of the movable wall where to place the poster for the poster session.

Elementary Particle Physics

[E 1] Dynamical Charm Effects on the QCD Static Potential

S. Cali, F. Knechtli, T. Korzec, H. Panagopoulos, R. Sommer

We evaluate, in the continuum limit, the effects of a dynamical charm quark on the static potential. The size of these effects is calculated through a comparison between quenched QCD and QCD (푁푓 = 2), with two heavy degenerate quarks of mass 푀 = 푀푐, where 푀푐 is the mass of a charm quark. As applications, we also determine the charm loop effects on other related observables that can be extracted from the force between two static color sources, like the strong coupling in the 훼푞푞-scheme and its Renormalization Group 훽푞푞-function.

[E 2] Calculating the Proton Radius Using Lattice QCD

J. Green, N. Hasan, S. Meinel, M. Engelhardt, S. Krieg, J. Negele, A. Pochinsky, S. Syritsyn

The proton charge radius can be measured in experiments, either from the electron-proton scattering cross section or from hydrogen spectroscopy. However, a recent very precise measurement based on spectroscopy of muonic hydrogen is in significant disagreement with earlier results using both methods; this is known as the proton radius puzzle. If confirmed, this would be evidence for a violation of lepton universality, a well-tested principle in particle physics. A first-principles calculation of the proton radius based on lattice quantum chromodynamics could help to understand this puzzle. Typically, this is done by first computing the electric form factor at several values of the momentum transfer, and then performing a fit to determine the charge radius from the slope at zero momentum transfer. We present new methods for directly computing the radius and compare them with an approach based on fitting, in a calculation using one lattice ensemble with the quark masses set to their physical values.

[E 3] Axion Mass from Lattice QCD

J. N. Guenther

We determine the topological susceptibility and the thermodynamical equation of state for temperatures relevant for the axion production in the early Universe. We use lattice QCD with dynamical fermions. We point out several difficulties in these calculations and address them by introducing novel techniques. Assuming the standard QCD axion scenario and no topological defects we obtain a lower bound on the axion mass. [E 4] Investigation of Theories beyond the Standard Model

S. Ali, G. Bergner, H. Gerber, P. Giudice, I. Montvay, G. Münster, S. Piemonte, P. Scior

The fundamental constituents of matter and the forces between them are splendidly described by the Standard Model of elementary particle physics. Despite its great success, it will be superseded by more comprehensive theories that are able to include phenomena, which are not covered by the Standard Model. Among the attempts in this direction are supersymmetry and Technicolor models. We report about our non-perturbative investigations of the characteristic properties of such models by numerical simulations on high-performance computers.

[E 5] The Nature of the QCD Thermal Transition as a Function of Quark Flavours and Masses

F. Cuteri, O. Philipsen, A. Sciarra

The fundamental theory of the strong interactions is Quantum Chromodynamics (QCD). At high temperatures it predicts a transition from a gas of hadrons (the known nucleons as well as various mesons) to a quark gluon plasma, where the boundaries around the constituents of hadronic matter are lost. The order of this transition depends crucially on the number of dynamical quark flavours and their mass. Understanding this dependence completely is important to constrain the limit of massless quarks as well as the theory at finite baryon number, both of which cannot be simulated directly.

[E 6] The Leading Order Hadronic Contribution to the Anomalous Magnetic Moment of the Muon

B. Toth

The anomalous magnetic moment of the muon is one of the most precisely measured physical quantities. Comparing its experimental value to the theoretical prediction of the Standard Model (SM) provides a stringent test of SM, and a possible disagreement can indicate new physics. The dominant source of uncertainty in the recent theoretical determinations is the hadronic loop corrections arising from the hadronic vacuum polarization (HVP). Due to the large coupling constant of the strong interaction at low energies, the HVP contribution is only accessible by non-perturbative methods. Here we apply lattice QCD to address these hadronic contributions non-perturbatively. Our calculations include all contributions from the u, d, s, and c quarks, directly at the physical values of their masses, in their quark-connected and quark-disconnected configurations. We find that our value for the leading order hadronic contribution to the anomalous magnetic moment of the muon, within its combined error of 2.7%, is compatible with the current experimental results.

[E 7] Nucleon Sigma Terms, Corrections to Dashen's Theorem and the Light Quark Mass Ratio from Lattice QCD

L. Varnhorst

We present a QCD calculation of the u, d, and s scalar quark contents of nucleons based on 47 lattice ensembles with 푁푓 = 2 + 1 dynamical sea quarks, 5 lattice spacings down to 0.054 fm, lattice sizes up to 6 fm, and pion masses down to 120 MeV. Using the Feynman- Hellmann theorem, we obtain 푓푢푑, 푁 = 0.0405(40)(35) and 푓푠, N = 0.113(45)(40), which translates into 휎휋, N = 38(3)(3) MeV, 휎푠, N = 105(41)(37) MeV for the sigma terms, where the first errors are statistical and the second errors are systematic. Using isospin relations, we also compute the individual up and down quark contents of the proton and neutron. Based on the 푁푓 = 2 + 1 QCD simulations to which QED effects have been added in a partially quenched setup we determine the corrections to Dashen’s theorem and the individual up and down quark masses. For the parameter which quantifies violations to Dashen’s theorem, we obtain 휀 = 0.73(2)(5)(17), where the third error is an estimate of the QED quenching error. For the light quark masses we obtain, 푚푢 = 2.27(6)(5)(4) MeV and 푚푑 = 4.67(6)(5)(4) MeV in the modified minimal subtraction scheme at 2 GeV and the isospin breaking ratios 푚 푢 = 0.485(11)(8)(14), 푅 = 38.2(1.1)(0.8)(1.4), and 푄 = 23.4(0.4)(0.3)(0.4). Our results ⁄푚푑 exclude the 푚푢 = 0 solution to the strong CP problem by more than 24 standard deviations.

[E 8] Extended Investigation of the 12 Flavor Beta Function

Z. Fodor, K. Holland, J. Kuti, S. Mondal , D. Nogradi, C. H. Wong

We report on continued studies of the non-perturbative beta function in SU(3) gauge theory with 12 fundamental massless flavors, a model which remains controversial as to the existence of an infrared fixed point. We extend our previous work (Phys. Rev. D94 (2016) no.9, 091501) to stronger gauge couplings and test conformality for this model.

Atomic/Nuclear Physics

[AK 1] The Strong Interaction at Neutron Rich Extremes

K. Hebeler, J. Holt, S. König, A. Schwenk, J. Simonis, R. Stroberg, I. Tews

We present results of recent calculations of medium-mass nuclei based on modern nuclear interactions derived within chiral effective field theory, constraints for the equation of state of dense matter including applications to the structure of neutron stars, as well as novel studies of three- and four neutron resonances.

[AK 2] Relativistic Magneto-Hydrodynamics for Heavy Ion Collisions

G. Inghirami, M. Bleicher

We present an overview of why and how ideal relativistic magneto-hydrodynamics has been implemented in the ECHO-QGP code, with the aim to describe in a self-consistent way the dynamical evolution of the Quark Gluon Plasma in the presence of strong magnetic fields, like those generated by the fast-moving charges of the nuclei in heavy ion collisions. We briefly illustrate the adopted formalism and we present some preliminary results.

[AK 3] Light Neutron-Rich Hypernuclei from the IT-NCSM

R. Wirth, R. Roth

We present an ab initio method suitable for the theoretical description of hypernuclei in the p shell, the hypernuclear importance-truncated no-core shell model. Using two- and three- baryon interactions from chiral effective field theory and a similarity renormalization group transformation consistently up to the three-baryon level to improve the model-space convergence, we investigate ground- and excited-state systematics in the lambda-helium and lambda-lithium isotopic chains. We compute neutron separation energies and find indications that the neutron drip lines for the nuclear and hypernuclear chains are identical. We also discuss the hyperon puzzle in neutron star physics, providing a natural explanation for the appearance of strong repulsive hyperon-nucleon-nucleon interactions that delay the appearance of hypernuclei in the interior of neutron stars.

Materials Science

[MAT 1] Microstructure Formation of Metallic Nanoglasses and Their Mechanical Properties

O. Adjaoud, C. Kalcher, K. Albe

Metallic glasses combine unique mechanical properties, such as high strength and hardness but usually exhibit negligible global plasticity under uniaxial tension: If loaded in a state of uniaxial stress, metallic glasses fail catastrophically along a single or a small number of shear bands. One promising way to improve the plasticity is to use metallic nanoglasses which are fully amorphous materials with microstructural features, as an analogue of nanocrystalline materials. We performed molecular dynamics simulations to investigate the microstructure formation of metallic nanoglasses and the influence of their microstructure on the mechanical deformation [1,2,3]. The results reveal that during cold compaction most of the glassy spheres deform by homogeneous plastic flow and in some glassy spheres strain localization occurs in a shear band which traverses the whole glassy sphere [2]. Moreover, the porosity is closed if hydrostatic pressures exceed 4 GPa. The resulting nanoglasses are composed of glassy regions connected by glass-glass interfaces. Shear strain analysis during uniaxial tensile deformation shows that theses interfaces promote shear band nucleation and prevent strain localization [3,4]. In addition, we show that it is possible to reinforce the nanoglass without compromising its ductility by manipulating the glass-glass interfaces in a similar way to grain boundary strengthening in crystalline materials [3].

[1] O. Adjaoud et al., Acta Mat. 113, 284-292 (2016). [2] O. Adjaoud, et al., Acta Mater. 322-330 (2018) 322-330. [3] C. Kalcher et al., Scr. Mater. 141 (2017) 115-119. [4] C. Kalcher et al., Acta Mater. 141 (2017) 251-260.

[MAT 2] Modelling of Materials for Si-Compatible Microelectronics

J. Dabrowski, G. Kissinger, G. Lippert, G. Lupina, M. Lukosius, P. Sana, T. Schroeder

Ab initio density functional theory (DFT) is an established method to model the behavior of materials at the atomic scale. At the IHP (Innovations for High Performance Microelectronics), we use it to investigate materials systems that are of interest to the most popular and cost- efficient technology, by which electronics is made today: the silicon technology. Here we report on the results obtained for various materials: (a) for strictly 2D atomic sheets (graphene), (b) for heteroepitaxial layers (oxides and nitrides), their surfaces, and the interfaces between these films, and (c) for bulk crystals (defects in silicon). The graphene sheets are intended as components of chemical sensors, optical modulators, and high-speed and high-power transistors. The chemical reactions and diffusion processes governing the nucleation and growth of graphene on perfect (flat and stepped) and defected surfaces of germanium films were simulated, and the mechanisms responsible for the observed growth modes were elucidated. The Sc oxide and nitride films constitute the topmost part of a heterostructure on which GaN diodes, lasers, and high-power transistors can be assembled. The simulations provided insight into the intermixing of oxygen and nitrogen. The substrate on which all these films and other device structures are grown, is crystalline silicon. For numerous application it is critical that the substrate getters (collects and binds) the impurities that are unintentionally introduced by the technological process. The formation of oxygen precipitates used as the gettering centers is associated with the presence of missing atoms (vacancies) in the Si bulk. We studied the process of vacancy clustering and oxidation, we extrapolated the clustering results to infinite separation between the defects, and we discussed the implications also for the interpretation of deep level transient spectroscopy (DLTS) or for the strategy to perform numerically expensive defect calculations (as done with hybrid potentials), among others.

[MAT 3] A Theoretical View on the Current Transport in ReRAM Devices

C. Funck, R. Waser, S. Menzel

The exponential law of miniaturization known as Moore’s law has reached the nanometre range. Due to this miniaturization established FLASH memory techniques are running into physical limitations, which prevents further downscaling. This limits a further increase in the data storage density. New concepts are required to overcome the problems of downscaling and energy consumption. One of the most promising concepts are redox-based resistive switching memories, which show excellent properties in all of these fields. These devices consist of a thin ionically conducting oxide layer. The resistance in these devices can be switched between a low resistive state (LRS) and a high resistive state (HRS). The resistance is a result of a reordering of oxygen vacancies, in a filamentary like region, which is formed in a previous electroforming step. The total nature of the resistive switching process, however, is still not completely understood. The information in these memory cells is stored in the resistance of these cells, consequently, the electrical conduction is one of the key parameters. The small dimensions of the conduction path in these devices require an atomistic approach to understand the electronic conduction mechanism. Using quantum mechanical simulations based on density functional theory and the non-equilibrium Green’s function method, we developed a theory for the transport in for the model system Nb:SrTiO3/SrTiO3-x/Pt. While the data could be fitted using Schottky emission theory, we could disprove Schottky emission as the dominating transport mechanism. Our theory highlights the reason for misleadingly understanding the conduction mechanism as Schottky emission. From our experimental and theoretical results, we show that the conduction mechanism is described by thermally assisted tunnelling into the conduction band. Schottky emission over the barrier contributes only little to the overall current. The thermal excitation that is required in order to tunnel into the conduction band is the main reason for the misinterpretation of the conduction mechanism. The derived theory from the models system of SrTiO3/Pt has already successfully applied to systems based on TiO2.

[MAT 4] First Principles Calculations on Incorporation of Cu and Na in β-In2S3

E. Ghorbani, K. Albe

β-In2S3 is an n-type semiconductor considered as a Cd-free buffer layer for the Cu(In,Ga)Se2 (CIGS) based thin film solar cells. A fundamental characteristic of In2S3 is the existence of four vacant tetrahedral sites per unit cell, which makes it a vulnerable host against third elements diffusing out from the absorber (CIGS) and/or front contact (ZnO). According to various experimental reports, when indium sulfide is deposited on CIGS, they form an intermixed absorber-buffer interface with a high concentration of Cu and Na. So, the question which arises here is whether the CIGS/In2S3 interface is stable or we would expect to have Na- containing and Cu-containing secondary phases at interface region. From a theoretical point of view, the stability of the absorber/buffer interface can be investigated either by defect thermodynamics calculations or through conducting direct interface calculations. Our point defects calculations show that there is a massive driving force for the occurrence of side reactions leading to the formation of Cu-containing and Na-containing interfacial secondary phases. We conclude that independent of the interface orientation, the absorber/buffer interface is thermodynamically unstable against the diffusion of Na and Cu into the In2S3. Moreover, direct interface calculations indicate large lattice distortions and massive structural relaxations in the vicinity of the pn-junction.

[MAT 5] Effect of Al- and Ta- Substitution on Elastic Properties of Li7La3Zr2O12 for Solid State Batteries

J. F. Nonemacher, C. Hüter, H. Zheng, J. Malzbender, M. Finsterbusch, R. Spatschek, M. Krüger

The current study investigates LLZO in its pure, tetragonal phase and its stabilized cubic phase by substitution with different amounts of Ta and Al. The experimental results obtained via nano-indentation testing and theoretical predictions based on ab initio calculations using the VASP software for DFT are compared.

[MAT 6] Atomistic Simulations of Nuclear Waste Materials

Y. Ji, H. Si, P. M. Kowalski

The world-wide strategy for nuclear waste management is to dispose the highly radioactive waste in a deep geological disposal and it is planned to be accomplished within a few decades. In order to make the nuclear waste safe, the immobilization of radionuclides in durable solid forms such as glass or ceramic materials is considered as a solution for high level nuclear waste. We apply ab initio and force-field atomistic modeling methods in order to understand the atomic scale-driven properties and long-term behavior of such nuclear materials under final disposal conditions. We present results of investigation of radiation damage processes in monazite-type ceramics (LnPO4). With computer-aided simulations we have delivered information on the atoms threshold displacement energies and probabilities of defects formation for a series of monazites [1]. In particular we discuss the temperature dependence of these parameters. We also will discuss the molecular dynamics simulations of the critical amortization dose and compare the results with the experimental data. Among other materials properties we will show the results of simulations of the elastic properties and the thermal conductivities and discuss the incorporation of actinides into different waste hosts and secondary phases [2]. Last, but not least, we will emphasize the importance of joint atomistic modeling and experimental effort for full characterization of the properties of challenging nuclear materials.

[1] Ji, Y., Kowalski, P. M., Neumeier, S., Deismann, G. "Atomistic modeling and experimental studies of radiation damage in monazite-type LaPO4 ceramics", Nuclear Instruments and Methods in Physics Research B. 393, 54 (2017). [2] Ji, Y., Beridze, G., Li, Y., & Kowalski, P. M. "Large Scale Simulations of Nuclear Waste Materials", Energia Procedia, 127, 416 (2017).

[MAT 7] Amorphous-Crystalline Interfaces in Silicon Heterojunction Solar Cells via Ab Initio Calculations

F. Li, P. Czaja, U. Aeberhard, S. Giusepponi, M. Gusso, M. Celino

Heterojunction solar cells with hydrogenated amorphous passivation layers hold the efficiency record among silicon solar cells. The exact configuration of the interface region, especially the density and energy of defects, but also the band offsets and doping induced band bending plays a crucial role in transport and recombination across the interface and, in consequence, for the overall device performance. In our contribution, we investigate the electronic structure and transport at amorphous-crystalline silicon interfaces by means of first principle calculations. Both a-Si:H/c-Si and a-SiOx/c-Si interfaces are created using ab initio molecular dynamics, and the electronic structure is analyzed with density functional theory. Special attention is given to the characterization of localized defect states at the interface. Ab initio electronic transport from maximally-localized wannier functions is investigated on the level of ballistic transmission based on the electronic structure using the NEGF formalism, in order to assess the role of subgap states in the charge transfer across the hetero-interface. The transmission function of the interface region is compared to the mobility gap as defined by the spread of the wave functions.

[MAT 8] The Electrostatic Origin of Polar Hydrophobicity: A Unique Property of Perfluorinated Carbon Materials

L. Mayrhofer, G. Moras, M. Moseler

Fluorinated carbon materials are of high importance in many fields of science and technology including chemistry, biochemistry and materials science. Fluorination is often used to enhance the hydrophobicity of carbon materials. Probably the most prominent example is PTFE, better known as Teflon. The poor wetting properties of perfluorinated carbon compounds can be traced back to the very weak interaction with water at the molecular level. On a first glance this might come as a surprise: F is the most electronegative element, thus C-F bonds are highly polar and commonly polar surfaces are associated with hydrophilic behavior. This seeming contradiction is known as polar hydrophobicity [1]. Using fluorinated diamond surfaces as model systems we show that the polar hydrophobicity of perfluorinated materials is caused by the high packing density of C-F dipoles and that it is a very unique phenomenon peculiar to this class of materials [2]. Although our conclusions were derived from a combination of ab initio DFT calculations and classical molecular dynamics simulations, the polar hydrophobicity can also be understood in terms of a simple electrostatic model based on point charges. Moreover, we will show under which conditions the polar C-F bonds can indeed act as reasonably strong H-bond acceptors and hence can lead to hydrophilic behavior of partially fluorinated carbons. The concepts derived for fluorinated diamond surfaces can be easily transferred to fluorinated molecules.

[1] J. C. Biffinger, H. W. Kim, and S. G. DiMagno, ChemBioChem 5, 622-627 (2004). [2] L. Mayrhofer, G. Moras, N. Mulakaluri, S. Rajagopalan, P. A. Stevens, and M. Moseler, J. Am. Chem. Soc. 138, 4018–4028 (2016).

[MAT 9] Ionic Liquids at Interfaces

N. Vučemilović-Alagić, R. Stepić, D. Berger, C. R. Wick, D. M. Smith, A.-S. Smith

In the last decades it has become a common practice to use ionic liquids films in the context of catalysis. A particular advantage of these systems is their low vapor pressure and powerful solvation. Understanding the solvent effects both in the context of interface wetting and chemical reactions is vitally important for the technological applications of ILs, however, the understanding of these processes is still not satisfactory. To address this problem, we employ atomistic molecular dynamics (MD) simulations and investigate the behavior of the archetypical imidazolium-based IL [C2Mim][NTf2] at interfaces (i.e. a hydroxylated sapphire surface and vacuum). We establish a refined additive force field model for [C2Mim][NTf2], which reproduces the experimentally observed x-ray reflectivity, surface tension and diffusion coefficients with reasonable accuracy. Based on our validated new force field, we are able to investigate the structuring of the IL and the changes of its transport properties as a function of the distance from the interface of choice. [MAT 10] Thermodynamic Aspects for Sr-Related Degradation Issues in SOFCs

X. Yin, R. Spatschek

Sr is a reactive element in (La,Sr)(Co,Fe)O3-δ (LSCF) cathode and has significant influence on several degradation issues of SOFCs performance. Thermodynamic calculations are performed to have further understanding of the two Sr-related degradation issues: Cr- poisoning of the LSCF cathode and volatile Sr species formation.

Condensed Matter

[KM 1] Hybrid Organic Magnetic Metal Interfaces

N. Atodiresei, V. Caciuc, S. Blügel

The density functional theory provides a framework with predictive power that can be used to describe organic-metal hybrid systems in a realistic manner. In this respect, ab initio studies elucidate how the subtle interplay between the electrostatic, the weak van der Waals and the strong chemical interactions determine the geometric, electronic and magnetic structure of hybrid organic-metal interfaces. More precisely, the interaction between the π-like electronic cloud of organic materials with the magnetic states of a metal influences the (i) spin- polarization, (ii) magnetic exchange coupling, (iii) magnetic moments and (iv) their orientation at the hybrid interfaces. In this poster we will briefly summarize how first-principles calculations (i) provide the basic insights needed to interpret surface-science experiments and (ii) are a key tool to design novel materials with tailored properties that can be integrated in carbon-based spintronic devices.

[1] N. Atodiresei, J. Brede, P. Lazić, V. Caciuc, G. Hoffmann, R. Wiesendanger, S. Blügel, Phys. Rev. Lett. 105, 066601 (2010). [2] J. Brede, N. Atodiresei, S. Kuck, P. Lazić, V. Caciuc, Y. Morikawa, G. Hoffmann, S. Blügel, R. Wiesendanger, Phys. Rev. Lett. 105, 047204 (2010). [3] N. Atodiresei, V. Caciuc, P. Lazić, S. Blügel, Phys. Rev. B. 84, 172402 (2011). [4] R. Decker, J. Brede, N. Atodiresei, V. Caciuc, S. Blügel, R. Wiesendanger, Phys. Rev. B 87, 041403(R) (2013). [5] M. Callsen, V. Caciuc, N. Kiselev, N. Atodiresei, S. Blügel, Phys. Rev. Lett. 111, 106805 (2013). [6] J. Brede, N. Atodiresei, V. Caciuc, M. Bazarnik, A. Al-Zubi, S. Blügel, R. Wiesendanger, Nature Nanotechnology 9, 1018 (2014). [7] R. Friedrich, V. Caciuc, N. Kiselev, N. Atodiresei, S. Blügel, Phys. Rev. B 91, 115432 (2015). [8] R. Friedrich, V. Caciuc, N. Atodiresei, S. Blügel, Phys. Rev. B 92, 195407 (2015). [9] R. Friedrich, V. Caciuc, N. Atodiresei, S. Blügel, Phys. Rev. B 93, 220406(R) (2016). [10] K. V. Raman, A. M. Kamerbeek, A. Mukherjee, N. Atodiresei, T. Sen, P. Lazic, V. Caciuc, R. Michel, D. Stalke, S. K. Mandal, S. Blügel, M. Münzenberg, J. S. Moodera, Nature 493, 509 (2013). [11] B. Warner, F. El Hallak, N. Atodiresei, Ph. Seibt, H. Prüser, V. Caciuc, M. Waters, A. J. Fisher, S. Blügel, J. van Slageren, C. F. Hirjibehedin, Nat. Comm. 7, 12785 (2016). [12] F. Huttmann, D. Klar, N. Atodiresei, C. Schmitz-Antoniak, A. Smekhova, A. J. Martínez- Galera, V. Caciuc, G. Bihlmayer, S. Blügel, T. Michely, H. Wende, Phys. Rev. B 95, 075427 (2017). [13] F. Huttmann, N. Schleheck, N. Atodiresei, T. Michely, J. Am. Chem. Soc. 139, 9895 (2017).

[KM 2] Orbital Magnetism of 3d Adatoms Deposited on the Pt(111) Surface

S. Brinker, M. dos Santos Dias, S. Lounis

The classical definition of the orbital magnetic moment using the ground state charge current is well-defined for finite systems, while for periodic systems the modern theory of orbital magnetization applies. Quantum-mechanically, the orbital moment is due to the spin-orbit coupling (SOC) tying spin and orbital degrees of freedom. In this work we consider the intermediate case: small magnetic nanostructures deposited on surfaces from first-principles. As an application we consider transition metal adatoms deposited on the Pt(111) surface, and their ground state charge current distributions. We find two contributions to the orbital magnetic moment a local contribution corresponding to the swirling charge currents around each atom and a non-local contribution related to the net currents flowing between atoms. We show that the non-local contribution, which was neglected in previous works, is an important part of the total orbital magnetic moment, and can even exceed the local contribution. Our findings can be relevant to describe magnetometry measurements of adatoms and dimers with SP-STM [1] and for future experiments with nano-diamonds [2].

This work was supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (ERC Consolidator Grant No. 681405 DYNASORE). We acknowledge the computational resources provided by the Jülich Supercomputing Centre.

[1] A. A. Khajetoorians et al., Nature Physics 8, 497 (2012). [2] L. Rondin et al., Rep. Prog. Phys. 77, 056503 (2014.)

[KM 3] Hybrid Molecular-Based Interfaces from First Principles

V. Caciuc, N. Atodiresei, S. Blügel

The hybrid systems consisting of physisorbed or chemisorbed molecules on surfaces are a key component in the technologically emerging molecular electronics and molecular spintronics fields. From a theoretical point of view, a major challenge in the density functional theory (DFT) is to properly describe the non-local correlation effects responsible for the van der Waals (vdW) interactions that are crucial for bonding in the physisorbed molecular systems. Therefore, we explored how to mitigate this DFT limitation via a first-principles non-local correlation functional (vdW-DF) or a semi-empirical approach and applied them to investigate a prototypical non-aromatic molecule as cyclooctatetraene (C8H8) on several metal surfaces [1] and the clean and intercalated graphene on Ir(111) [2-4]. In particular, our ab initio vdW-DF calculations performed for naphthalene (C10H8) on graphene/Ir(111) demonstrated that the intercalated atomic layers with an electropositive or electronegative character can modulate the strength of the vdW interactions between a π-conjugated molecule and the n- or p-doped graphene on Ir(111) [4]. Furthermore, on this basis, we also investigated how the physisorbed and chemisorbed molecules can be used to tune the strength and spin texture of the Rashba spin-split surface states of the BiAg2/Ag(111) surface alloy [5, 6].

[1] H. Harutyunyan, M. Callsen, T. Allmers, V. Caciuc, S. Blügel, N. Atodiresei and D. Wegner, Chem. Comm. 49, 5993 (2013). [2] C. Busse, P. Lazić, R. Djemour, J. Coraux, T. Gerber, N. Atodiresei, V. Caciuc, R. Brako, A. T. N'Diaye, S. Blügel, J. Zegenhagen and T. Michely, Phys. Rev. Lett. 107, 036101 (2011). [3] W. Jolie, F. Craes, M. Petrovíc, N. Atodiresei, V. Caciuc, S. Blügel, M. Kralj, T. Michely and C. Busse, Phys. Rev. B 89, 155345 (2014). [4] F. Huttmann, A. J. Martínez-Galera, V. Caciuc, N. Atodiresei, S. Schumacher, S. Standop, I. Hamada, T. O. Wehling, S, Blügel and T. Michely, Phys. Rev. Lett. 115, 236101 (2015). [5] R. Friedrich, V. Caciuc, G. Bihlmayer, N. Atodiresei and S. Blügel, New J. Phys. 19, 043017 (2017). [6] R. Friedrich, V. Caciuc, B. Zimmermann, G. Bihlmayer, N. Atodiresei and S. Blügel, Phys. Rev. B 96, 085403 (2017). [7] M. Paβens, V. Caciuc, N. Atodiresei, M. Feuerbacher, M. Moors, R. E. Dunin-Borkowski, S. Blügel, R. Waser and S. Karthäuser, Nature Comm. 8, 15367 (2017).

[KM 4] Towards Understanding Photomigration: Insights from Atomistic Simulations of Azopolymer Films Explicitly Including Light-Induced Isomerization Dynamics

M. Böckmann, N. L. Doltsinis

The light-induced surface modification of a thin film of poly-(disperse orange-3- methylmethacrylate) is investigated computationally using atomistic molecular dynamics simulations specifically tailored to include photoisomerization dynamics. For a model surface consisting of a periodic pattern of alternating irradiated and dark spots, it is shown that repeated photoisomerization in the irradiated areas initially leads to a local temperature increase and a raised surface profile accompanied by a migration of molecules away from the bright spots. After switching off the light source and letting the system cool down, this leads to an inversion of the surface profile, i.e., dips in the bright spots and bumps in the dark spots. To separate the effect of photoisomerization from the pure heating effect, a second simulation is performed in which no photoisomerization is allowed to occur in the bright spots, but the equivalent amount of energy is introduced there locally in the form of heat. This also leads to a raised surface in these areas; however, no outward migration of molecules is observed and the surface pattern practically vanishes when the system is subsequently cooled back to room temperature. Furthermore, a variety of different effects including the rate of heat dissipation, the direction of linearly polarized light, and the periodicity of the light pattern is investigated systematically in a series of simulations.

[KM 5] High-Performance Functional Renormalization Group Calculations with Truncated Unities

J. Ehrlich, T. Reckling, G. Schober, J. Lichtenstein, D. Sánchez de la Peña , C. Honerkamp

Functional renormalization group calculations allow an unbiased investigation of the electronic system of solids but are computationally challenging as the interaction vertex scales with the cube of the number of momenta or frequencies. We discuss how the insertion of truncated unities of a form-factor basis represents an advantageous approximation that reduces the vertex to scale only linearly with the number of momenta and quadratically with the number of form-factors. Moreover, we show that this method allows a highly efficient parallelization with OpenMP as well as MPI and compare the results with previous conventional functional renormalization group studies for the 2D Hubbard model.

[KM 6] Quantitative Convergence of Functional Renormalization Group and Dynamical Cluster Approximations in the Two-Dimensional Hubbard Model

C. Honerkamp

The two-dimensional Hubbard model on the square lattice is a standard model for correlated electrons. Despite the simplicity of its Hamiltonian, there are no exact solutions for relevant parameter regimes and often quantitative predictions are missing. Here we compare numerical results for effective interactions and susceptibilities obtained with two completely different methods, the dynamical cluster approximation (DCA) and the functional renormalization group (fRG) for the half-filled fully nested band. We show that, qualitatively, both methods describe the physics. In addition, in particular the susceptibility data agrees quite well on the quantitative level if one takes into the frequency-dependence of the interactions and self-energy effects also in the fRG. This convergence of the two methods represents an important check that increases the predictive power for other physical situations and models.

[KM 7] Real-Time Broadening of Nonequilibrium Density Profiles and the Role of the Specific Initial-State Realization

R. Steinigeweg, F. Jin, D. Schmidtke, H. De Raedt, K. Michielsen, J. Gemmer

The real-time broadening of density profiles starting from nonequilibrium states is at the center of transport in condensed-matter systems and dynamics in ultracold atomic gases. Initial profiles close to equilibrium are expected to evolve according to the linear response, e.g., as given by the current correlator evaluated exactly at equilibrium. Significantly off equilibrium, the linear response is expected to break down and even a description in terms of canonical ensembles is questionable. We unveil that single pure states with density profiles of maximum amplitude yield a broadening in perfect agreement with the linear response, if the structure of these states involves randomness in terms of decoherent off-diagonal density- matrix elements. While these states allow for spin diffusion in the XXZ spin-1/2 chain at large exchange anisotropies, coherences yield entirely different behavior.

[KM 8] Universality of Defect-Skyrmion Interaction Profiles

I. Lima Fernandes, J. Bouaziz, S. Blügel, S. Lounis

Magnetic skyrmions are non-collinear spin textures with particle-like properties, which are currently under intense scrutiny due to the rich science and technological potential for future information and communication devices. However, incorporating them as possible bits of information hinges on their interaction with material inhomogeneities ubiquitous to any device since their nucleation, motion and velocity are heavily affected in contrast to phenomenol- ogically-based predictions. Thus, a detailed knowledge of defects-skyrmion interactions is desirable but unknown, which limits the realization of such topological magnetic entities as possible bits of information. Using JURECA, we performed complex simulations using the full- potential relativistic Korringa-Kohn-Rostoker (KKR) Green function method to map, for the first-time from fully ab initio, the energy-landscape of single magnetic skyrmions interacting with single-atom impurities, establishing a generic shape as function of the defect's electron filling. Depending on their chemical nature, foreign 3d and 4d transition metal adatoms or surface-implanted defects can either repel or pin skyrmions generated in a PdFe bilayer deposited on Ir(111) surface. With a careful analysis of the hybridization of the electronic states, the mechanism behind the expulsion and pinning can be related to the degree of filling of bonding and anti-bonding electronic states inherent to the proximity of the non-collinear magnetic structure. The universality of the interaction profile may give guidance for the ultimate design of devices generating, controlling and guiding skyrmions.

Funding provided by the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (ERC-consolidator grant 681405 - DYNASORE).

[KM 9] Lattice Dynamcis and the Coupled Crystal Field – Phonon Excitations in CeAuAl3

B. Liu, P. Cermak, O. Sobolev, C. Franz, C. Pfleiderer, A. Schneidewind

Cerium intermetallic compounds of CeTX3 type (T = transition metal d element and X = Si, Ge or Al) have attracted considerable attention, due to the discovery of many interesting physical phenomena such as magnetic properties, heavy-fermion behavior and unconventional superconductivity. In most of these systems, the crystal electronic field (CEF) excitations and phonons can be studied independently since they are considered to decouple with each other. However, strong magneto-elastic (MEL) coupling could result in the formation of a bound state of CEF and phonons that shows up in the magnetic cross section as described by a vibronic model for cubic CeAl2 [1] and recently for tetragonal CeCuAl3 [2]. In order to study if such interesting feature exists or not in CeAuAl3 system, neutron scattering experiment has been performed on a single crystal sample in the paramagnetic phase. "Mixed-mode" excitations (anticrossing of phonons and CEF) have been directly observed. Considering the one-ion magneto-elastic interaction [3], which corresponds to the direct coupling between the deformations of the lattice and the 4f shell and can be constructed by group theory, an analytical expression for the mixed-mode excitations have been obtained and the calculated spectrum agrees well with experimental data. In this work, the second order magneto-elastic interaction as well as the linear rotational interaction is neglected.

[1] P. Thalmeier and P. Fulde, PRL 49 1588 (1982). [2] D.T. Androja et al, PRL 108 216402 (2012). [3] E. Callen, H. B. Callen, Phys. Rev. 139, A455 (1965).

[KM 10] Structures and Phases in (Nano-) Systems in Confined Geometry

M. Beck, B. Gast, S. Haun, M. Matt, F. Müller, M. Pütz, M. Ring, R. W. U. Schmid, U. Siems, U. Steinwandel, T. Vater, P. Nielaba

Computer simulations by members of the NIC-project HMZ07 are reviewed. Structural and transport properties have been investigated in atomic wires by Molecular Dynamics and electronic transport methods and in model colloids by Brownian Dynamics. Spinodal decomposition and domain growth of liquid-vapour systems and galaxy magnetism have been studied by hydrodynamics simulations, and nucleation phenomena in lattice systems by rare event methods.

[KM 11] Tuning the Optical Spectra of Carbon Nanotubes

M. Rohlfing, T. Deilmann, M. Drüppel, Y. Ma

The optical spectra of nanostructured systems can be manipulated (e.g., red-shifted) when the environment is changed, e.g. by physisorbed material, even without chemical modification. The spectrum of a carbon nanotube (CNT) can be red-shifted by several 10 meV by environmental polarizability, e.g. from another CNT at touching distance (see the figure for peak shifts in the visible spectrum of several (N,0) CNTs), or by the admixture of charge- transfer configurations. Similar effects are observed in layered materials, like graphene/ graphite of boron nitride. Spectra can also be modified by atomic adsorbates, like oxygen or hydrogen atoms on the surface of a CNT. We discuss all these effects in terms of the Bethe- Salpeter equation (BSE) for electron-hole pair states on the basis of a preceding GW calculation.

[1] M. Drüppel, T. Deilmann, P. Krüger, and M. Rohlfing, Nat. Comm. 8, 2117 (2017). [2] T. Deilmann and M. Rohlfing, Nano Lett. 17, 6833 (2017). [3] T. Deilmann, M. Drüppel, and M. Rohlfing, Phys. Rev. Lett. 116, 196804 (2016).

[KM 12] TITAN: A Code and its Applications for Time-dependent Transport and Angular Momentum in Nanostructures

J. R. Suckert, F. Souza Mendes Guimarães, J. Chico, S. Lounis

Spintronics has been a highly studied topic for decades, with already successful applications in technological devices and other promising routes for future implementations. In this work, we explore the magnetization dynamics using a multi-orbital tight-binding approach including the spin-orbit interaction by calculating the full magnetic susceptibility in a linear response formalism. This quantity is mapped into analytical expressions obtained from a phenomenological model to obtain all the relevant parameters - in particular, the damping constant (also known as Gilbert parameter). We use JURECA and JUQUEEN super- computers to investigate typical magnetic bulk systems - Fe, Ni and Co - and compare the different methods available in the literature, establishing their range of validity.

[KM 13] Development of Electron Transport Calculation Code and Its Application to Molecular Junctions

S. Tsukamoto, V. Caciuc, N. Atodiresei, S. Blügel

The first-principles electron transport calculation methods used currently are categorized into two: One is based on the nonequilibrium Green function, and the other uses the wave function matching (WFM) technique. The latter discussed in this presentation computes scattering wave functions (SWFs) of a transition region between a pair of semi-infinite electrodes, which provide the direct pictures of transport processes. The SWF at each end of the transition region is represented as a linear combination of the generalized Bloch waves of the corresponding electrode. A set of generalized Bloch waves is composed of not only propagating waves but also evanescent waves. Sørensen et al. proposed to exclude rapidly decaying evanescent waves under the assumption that the rapidly decaying evanescent components of a SWF are negligible inside the electrodes. This contributes to avoiding expensive and unstable computation for the rapidly decaying evanescent waves. However, the transition region has to be extended by attaching extra layers to both ends, resulting in increase of computational cost for SWFs. In addition, the translational invariance of transmission/reflection probability is not preserved when moving the matching planes between the electrode and transition regions. We ascribe the lack of the translational invariance to using pseudoinverses in calculating transmission/reflection coefficients, which accompany the exclusion of rapidly decaying evanescent waves. We introduce a formulation to calculate transmission/reflection probability without explicit use of the coefficients. Consequently, the sum of the transmission and reflection probabilities exactly agrees with the number of transport channels, and the accuracy is largely improved. The translational invariance of the transmission/reflection probability is also preserved. The accuracy of the present method based on the WFM technique is comparable to those of the methods based on the nonequilibrium Green function. In addition to the method development, we present a couple of realistic electron transport calculations for molecular junctions. One of them demonstrates the possibility of tuning electron transport properties of a molecular junction by systematically changing the chemical properties of molecular anchoring groups. When B (N) atoms are substituted for the C atoms in the anchoring groups, the HOMO-(LUMO-)derived electron transmission peak appears at around the Fermi energy in the transmission spectra. The other one demonstrates the possibility of improving transmission probability of a poly- peptide molecular junction by replacing the side chains. By replacing a neutral side chain with an electronegative one, an electronic state bridging the polypeptide molecule appears at around the Fermi energy to form a transport channel through the molecular junction, resulting in large increase of the transmission probability.

Computational Soft Matter Science

[POLY 1] DPD with Energy Conservation Simulation of Thermophoretic Particle

F. Alidadi Soleymani, D. Fedosov, M. Ripoll, G. Gompper

The self-propelled particle converts environmental energy into the directed motion. Examples range from chemotactic cells and bacteria to artificial micro-swimmers which are widely studied due to their applications in drug delivery and micro/nanomachines in a fluid. The main physical mechanism of propulsion is an inhomogeneous field e.g. a flexible magnetic filament under an applied magnetic field or a self-propelled particle in an inhomogeneous concentration (diffusiophoresis phenomenon) or temperature field (thermophoresis phenomenon). Janus particles are colloidal particles with the inhomogeneous surface feature which can form the field gradient. The Janus particle with a metallic cap absorbs more energy from an external source which can be the heat source (laser beam) or magnetic field. Energy absorption increases the temperature of one cap and the temperature gradient is imposed mainly at poles. We investigate the behavior of the thermophoretic Janus colloid in its temperature gradient by the dissipative particle dynamics method with energy conservation (DPDE). The simulation results show how local fluid-colloid interactions and the temperature gradient near the colloid’s surface control the swimming velocity.

[POLY 2] Tissue competition: The Role of Cross Adhesion

T. Büscher, N. Ganai, J. Elgeti

Cells grow and divide, which implies a change in volume. In physical terms, the conjugate force to a change in volume is a pressure. Thus, in order to grow, cells must exert mechanical pressure on the neighbouring tissue. In turn, mechanical stress influences growth. Indeed, experiments on the growth of a cancer cell line under pressure display a reduction in proliferation due to mechanical pressure. This effect leads to a mechanical contribution when tissues compete for space. The tissue with higher homeostatic pressure, i.e. the pressure at which cell division and death balance, overwhelms the weaker one. We expand these works to include different adhesion properties. We find that the cross adhesion between the two tissues plays a crucial role in the dynamics of the competition. Besides one overwhelming the other, we observe a variety of states in which the two tissues coexist, ranging from flatly separated tissues to a bi-continuous structure. Interestingly, adhesion also plays a crucial role in cancer development: Cancer cells usually express less adhesion proteins.

[POLY 3] Excess Entropy Scaling Applied to Coarse-Grained Polymer Models

G. Garcia Rondina

Multiscale methods make it possible to simulate molecular systems in order to understand phenomena happening at various length and time scales. One very successful method is coarse-grained modeling, in which two or more atoms are grouped together and treated as a single interaction site. This greatly reduces the number of degrees of freedom of a system, making it possible to simulate longer time scales and larger length scales. However, the reduction of degrees of freedom comes at a cost. The force fields derived for coarse-grained models are often too soft and the dynamics of the models are greatly accelerated in comparison to atomistic or experiments. This means that coarse-grained derived diffusion coefficients, for instance, exceed their atomistic or experimental counterparts by one to two orders of magnitude, and viscosities are too low by the same amount. The question addressed in this project is whether the artificially accelerated dynamics of coarse-grained models can be quantified in terms of the variation of excess entropy between molecular models at different resolutions. We approach this question by investigating how the relations between excess entropy and dynamical properties behave for coarse-grained models. Extensive simulations of bead- spring model polymers at three different resolutions have been performed for a large number of state points. From these simulations, transport coefficients and entropy data were calculated and their correlations were analyzed with the aid of scaling entropy rules, such as those proposed by Rosenfeld [1] and Dzugutov [2]. The scaling of the excess entropy provides a quantitative analysis of the acceleration of the dynamics of the coarse-grained models.

[1] Y. Rosenfeld, Phys. Rev. A, 1977, 15, 2545-2549. [2] M. Dzugutov, Nature, 1996, 381, 137-139.

[POLY 4] Blood Flow Regulates Platelet-Polymer Aggregation

M. Hoore, D. A. Fedosov, G. Gompper

Platelets form aggregates as they adhere to the stretched von Willebrand factors (VWFs) at high shear rates. Their aggregation is critically dependent on shear rate and dissolves reversibly at low shear rates. In blood flow, red blood cells (RBCs) keep away from the vessel walls and leave a RBC free layer (RBC-FL) to which platelets and VWFs marginate. The formed aggregates in the RBC-FL gain significant hydrodynamic lift force and penetrate the RBC-core. Expectations from the experimental evidence on the shear-activated VWF-platelet interaction imply that these demarginated aggregates have to dissociate in the center of the vessel where the shear rate is low enough. Mesoscopic hydrodynamic simulations of the blood flow including RBCs, platelets, and VWFs, support this prediction. The regulation of undesired aggregates is beneficial for the vasculature, prohibiting undesired spontaneous thrombosis and imminent blockage or stroke, and can be altered if the affinity of the platelets to VWFs changes, such as in von Willebrand disease type 2B.

[POLY 5] High-Throughput Screening of Drug-Membrane Thermodynamics

R. Menichetti, K. H. Kanekal, K. Kremer, T. Bereau

In an in silico search for structure-property relationships aiming at inverse molecular design, the extensive computational resources required by molecular dynamics simulations performed with the atomistic detail seriously hamper the possibility of spanning large regions of chemical compound space. This issue can be efficiently addressed by relying on coarse- grained models, which provide a means to mitigate the computational expense while still capturing the relevant physical properties of the system under investigation. In this work, we introduce a high-throughput screening of compound space by means of coarse-grained molecular dynamics simulations of the Martini force field, and apply it to the determination of the insertion thermodynamics of an ensemble of small molecules in a phospholipid bilayer environment. We employ a novel importance sampling technique in coarse-grained compound space, akin to the more traditional sampling of conformational space used in computational physics. The study allows us to identify simple relationships between bulk properties and key features characterizing the thermodynamic stability of a compound in a bilayer, for different lipid compositions. Moreover, we show that Martini, as a transferable coarse-grained model, yields a significant reduction in the size of chemical space. This makes our results representative of the transmembrane behavior of a set of more than 500,000 chemically different small molecules.

[POLY 6] Mesoscopic Simulations of Electrokinetic Phenomena

N. Rivas, W. David, K. Yung, I. Pagonabarraga, J. Harting

Mesoscopic simulations of electrokinetic phenomena can provide valuable information in parameter ranges where theoretical understanding is poor, experiments are difficult, and atomistic simulations too expensive. This poster presents a lattice-Boltzmann-based framework to simulate binary electrolyte solutions with charged particles. The scaling performance is shown for different configurations and specific algorithms. Furthermore, exemplary cases of the capabilities of the algorithm are presented including neutral and charged drops deformation, particle electrophoresis, and particle-coated droplets.

Computational Plasma Physics

[PLA 1] Toward Enhancement of Betatron Radiation Flux (For JuSPARC Facility)

Z. Chitgar, A. Sobatta, M. Büscher, P. Gibbon, A. Lehrach

A short and bright X-ray source is of great importance in different branches of science to study the atomic structure down to sub-nanometer range and ultrashort time scale events. Synchrotrons and free electron lasers are the common sources of X-rays, both of which have limited accessibility because of their huge scale and costs. These limitations could potentially be overcome by laser-driven betatron radiation as a novel femtosecond X-ray source. In this project, preparatory simulations are performed to investigate the feasibility of achieving a high X-ray flux and energy using a TW-class laser soon to be installed as part of the JuSPARC facility. Two different injection schemes to enhance the beam current in plasma-based laser- electron acceleration are studied: ionization injection and self-injection. For ionization injection method a cluster-gas jet target is proposed to be applied. A double-pulse scheme is also applied to optimize self-injection.

[PLA 2] Numerical Studies towards a Plasma-Driven Free-Electron Laser

M. Kirchen, S. Jalas, L. T. Campbell, B. W. J. McNeil, A. R. Maier

In a plasma accelerator, a laser or particle beam excites a plasma wave that can sustain orders of magnitude higher electric fields than the RF-cavities of conventional accelerators. These compact devices have the potential to drive next-generation free-electron lasers (FEL) of reduced size and cost. However, the highly sensitive FEL mechanism imposes strict requirements on the electron beam properties. So far, the quality of plasma accelerated beams cannot compete with that of conventional accelerators. Successfully driving the self- amplifying process demands the development of methods with increased control of the beam’s phase space - in combination with novel FEL designs that leverage the unique features of the plasma accelerated beams. Demonstrating a plasma-driven FEL therefore heavily relies on theoretical studies and requires numerical models that correctly represent the physics involved under the restriction of limited computational resources. Here, we present our recent work towards a plasma-driven FEL, based on the particle-in-cell (PIC) code FBPIC, and the non-averaging 3D FEL code PUFFIN.

[PLA 3] Radiation-Dominated Plasma Dynamics in the Interaction of Super-Intense Circularly Polarized Pulses with Thick Plasma Targets

T. Liseykina, A. Macchi, S. Popruzhenko

We present a self-consistent theory describing the interaction of superintense (with intensities about 1023 W/cm2 and higher) circularly polarized electromagnetic pulses with thick dense plasma target. In contrast to the case of thin plasma slabs, thick targets are known to provide an interaction regime with extremely high radiation losses, because a high plasma surface density prevents longitudinal acceleration of electrons. Relying on our previous work [1] and on the model suggested in 1975 by Zeldovich [2], we developed an analytic theory which predicts that the combined effect of the radiation reaction force and of the charge separation in the plasma leads to the self-limited (although high) radiation losses, making a classical picture of radiation fairly accurate up to the intensity 1025 W/cm2, which is currently at the edge of capabilities expected from the Extreme Light Infrastructure pillars [3]. We expect that at these ultrahigh but nevertheless foreseen in the near future intensities the radiation losses saturated at a roughly 50% energy conversion from the laser pulse to high-energy photons. As a result of this conversion, a quasi-static magnetic field with the magnitude up to 30 gigagauss can be generated. Our theoretical predictions are supported by results of relativistic 3D particle-in-cell simulations.

[1] T. V. Liseykina, S. V. Popruzhenko and A. Macchi, New J. Phys. 18, 072001 (2016). [2] Y. B. Zeldovitch, Usp. Fiz. Nauk 115, 161 (1975). [3] Extreme Light Infrastructure, https://eli-laser.eu.

[PLA 4] Electron-Injection Techniques in Plasma-Wakefield Accelerators for Driving Free-Electron Lasers

A. M. de la Ossa, J. Osterhoff, E. Svystun, L. D. Amorim, A. Aschikhin, R. Aßmann, Á. F. Pousa, R. Fonseca, A. Helm, A. Knetsch, L. di Lucchio, J. L. Martins, T. Mehrling, P. Niknejadi, B. Sheeran, L. O. Silva, J. Vieira, M. Vranic, V. Wacker, M. Weikum

Plasma-based accelerators (PBAs) are seen as the leading technology for future compact radiation sources. However, despite remarkable experimental and theoretical achievements, the realization of a PBA matching the demanding requirements on beam quality and stability for applications such as Free-Electron-Lasers (FELs) remains a challenge. In the context of our project HHH23 on the supercomputer system JUQUEEN at Forschungszentrum Jülich we investigate a number of regimes and methods for reliable production of high quality beams from PBAs to identify the most promising plasma wakefield accelerator design to drive FELs. In this poster we highlight the most prominent research results on production, quality- preserving acceleration and transport of FEL-capable electron beams in a PBAs. Computational Chemistry

[CH 1] Simulation of Electron Transport through Graphene-Molecule Junctions

D. Weckbecker, P. B. Coto, M. Thoss

While most experiments on single-molecule junctions have employed metal electrodes, recent works demonstrate that graphene has a number of advantages over metallic leads [1-2]. In this contribution, we investigate, from a theoretical perspective, charge transport in molecular junctions with molecular bridges covalently bonded to the zigzag or armchair edges of graphene leads. Specifically, we analyze the possibility to use a proton transfer reaction as novel mechanism for switching the conductance of a molecular junction. Our simulations are performed using state-of-the-art DFT and non-equilibrium Green's function methods, which are implemented in the well-known TRANSIESTA code [3], which is highly optimized to make efficient use of many cores on parallel computers employing MPI- and OpenMP-based parallelization [3]. We find that an intramolecular proton transfer can change the conductance of the graphene- molecule junction significantly [4]. While a proton transfer parallel to the transport direction could be used to realize a molecular switch or diode, a junction with a perpendicular proton transfer can resemble a transistor.

[1] K. Ullmann et al., Nano Lett. 15, 3512 (2015). [2] C. Jia et al., Science 352, 1443 (2016). [3] N. Papior et al., Comput. Phys. Commun. 212, 8 (2017). [4] D. Weckbecker et al., Nano Lett. 17, 3341 (2017).

Computational Biology and Biophysics

[BIO 1] Free Energy Profile and the Dimer-To-Monomer Equilibrium of the Phospholipase PlaF From Pseudomonas Aeruginosa

S. Schott-Verdugo, S. Ahmad, R. Batra-Safferling, J. Granzin, F. Kovacic, K.-E. Jaeger, H. Gohlke

P. aeruginosa is a Gram-negative bacterium, which was recently classified by the WHO as a pathogen with critical priority for research and development of new antibiotics due to its pathogenicity and resistance to conventional treatments [1]. PlaF, a recently identified phospholipase A and integral inner membrane protein, has been shown to be a relevant virulence factor, making it an appealing study target [2]. Interestingly, the crystal structure and crosslinking experiments reveal a dimeric configuration, even when it shows activity as a monomer. Here, we studied the dynamics of the dimeric and monomeric PlaF configurations, the energetics of dimerization and discuss the dimer-monomer equilibrium in vivo. Using OPM predictions [3] of the protein orientation in the membrane, two possible configurations of PlaF were obtained: I) the catalytic tunnel is parallel to the membrane plane in the dimer and II) the catalytic tunnel is perpendicular to the membrane in the monomer. The different orientations suggest that they might affect substrate access to PlaF. Microsecond molecular dynamics (MD) simulations were performed with the dimer, the monomer in a dimer orientation (“split”), and the oriented monomer (“tilted”). The results show that both, the PlaF dimer and the “tilted” monomer, remained in their initial configuration during the MD simulation, while the “split” monomer adopts a “tilted” configuration. For studying the free energy profile of the dimer-monomer equilibrium in PlaF and the transition between the “split” and “tilted” configurations, we performed umbrella sampling simulations together with the Weighted Histogram Analysis Method (WHAM). The calculated dimerization constant from the profiles reveal that at the expected concentrations in P. aeruginosa, PlaF stays mainly as a monomer. Additionally, it explains the increased dimer configuration under overexpressing condition and suggests possible implications for changes in the concentration of the enzyme in vivo.

[1] World Health Organization (WHO). Retrieved from http://who.int/medicines/publications/WHO-PPL-Short_Summary_25Feb-ET_NM_WHO.pdf, 2017. [2] Kovacic, F. HHU PhD Thesis, 2010. [3] Lomize, M.A., et al. Nucleic Acids Res, 2012. 40(Database issue):D370-6.

[BIO 2] Resolving the Mechanism of Secondary Active Glutamate Transport

C. Alleva, K. Kovalev, V. Gordeliy, C. Fahlke, J.-P. Machtens

Excitatory amino acids transporters (EAATs) terminate synaptic transmission by taking up glutamate from the synaptic cleft into glial cells. EAATs are secondary active transporters: substrate uptake is coupled to the co-transport of 3 Na+ and 1 H+, and to the counter- transport of 1 K+. This stoichiometry ensures low extracellular glutamate concentrations with up to 106-fold concentrations gradients across the membrane. Available X-ray crystal structures have shed light on major conformational changes involved in EAAT transport. However, the conformational changes caused by ligand binding and the mechanisms of coupled transport remained elusive. Here we use all-atom molecular dynamics simulations, fluorescence spectroscopy, and x-ray crystallography of the archeal EAAT homolog GltPh to investigate the coupling between Na+ and substrate binding, as well as extracellular gate opening. Free energy calculations reveal that the apo transporter assumes preferentially a closed state, while Na+ binding locks the transporter in an open state, thereby making the glutamate binding pocket accessible to the substrate. We obtained a high-resolution x-ray crystal structure of Na+-only bound wild-type GltPh, which displays rearrangement of a number of residues induced by sodium binding. Two residues in particular, Arg397 and Met311, assume a central role in signal transduction from Na+ binding to gate opening. Mutation of these residues to smaller side chains modifies Na+/gate coupling by altering the equilibrium distribution between open and closed states of the extracellular gate. Comparison between our new structure and a previous sodium-bound structure of the Arg397Ala mutant further corroborates the coupling between Arg397 and gate opening. Stopped-flow fluorescence experiments on purified GltPh protein reveal that mutating Met311 to alanine affects both the kinetics and the KD for Na+ as well as the binding affinity of the substrate. In summary, our results provide a dynamic view on EAAT function, where a cooperative regulation of gate dynamics by Na+ ions and glutamate ensures stoichiometric Na+/substrate co-transport.

[BIO 3] Lipid transport by the ABC Transporter MDR3

M. Bonus, M. Prescher, L. Schmitt, H. Gohlke

Hepatocytes regulate the directional transport of endogenous substances and xenobiotics through a variety of membrane transporters in the sinusoidal and canalicular membrane [1]. Among other important exporters, the canalicular membrane contains the ATP binding cassette (ABC) transporter multidrug resistance protein 1 (MDR1, ABCB1, P-glycoprotein) and the homologous multidrug resistance protein 3 (MDR3, ABCB4). Despite their high sequence similarity of 86 %, MDR1 and MDR3 differ greatly regarding tissue expression and function [2]. MDR1 is ubiquitously expressed and transports a variety of structurally unrelated compounds including drugs, lipids, and peptides. In contrast, MDR3 is almost exclusively found in the canalicular membrane of hepatocytes and is a floppase specific for phosphatidylcholine lipids [3]. However, the structural basis for this divergent specificity has remained elusive so far as most residues in the drug-binding cavity of MDR1 are identical in MDR3. Based on molecular dynamics (MD) simulations and configurational free energy calculations, we provide evidence that the mechanism of substrate translocation in MDR3 may be different from the classical cavity-mediated transport in MDR1. Initial MD simulations indicated that MDR3 does not use a dedicated lipid binding cavity for the translocation of phosphatidylcholine lipids. We therefore hypothesized that MDR3 facilitates phosphatidylcholine translocation via a conspicuous arrangement of non-conserved, hydrophilic amino acids on the surface of transmembrane helix 1 (TMH1). Subsequent configurational free energy calculations revealed that phosphatidylcholine flip-flop along this pathway is ~7 kcal/mol more favorable and, hence, approximately five orders of magnitude faster than spontaneous transversal diffusion. This finding is in excellent agreement with experimentally determined flip-flop rates. The predictions based on our MD simulations shall be validated by in vitro studies with respect to the proposed mechanism. In particular, we aim to transform MDR1 into an MDR3- like lipid floppase by site-directed mutagenesis. Eventually, these results could contribute to a better general understanding of the structure-function relationship of ABC transporters.

[1] M. Nicolaou et al., J Pathol 2012, 226, 300-315. [2] J. J. Smit et al., Lab Invest 1994, 71, 638-649 [3] A. van Helvoort et al., Cell 1996, 87, 507-517.

[BIO 4] 3D Reconstruction of Nerve Fibers in the Human, the Monkey and the Rodent Brain

O. Bücker, A. Müller, A. Lührs, S. Münzing, S. Köhnen, P. Schlömer, M. Schober, N. Schubert, D. Schmitz, M. Axer

3D-Polarized Light Imaging (3D-PLI) is a neuroimaging technique that has opened up new avenues to study the complex architecture of nerve fibers in post mortem brains. It allows reconstructing the three-dimensional pathways of nerve fibers with a resolution of a few micrometers by means of birefringence measurements of the brain tissue. The main goals of the project JINM11 is to analyze these birefringence measurements (images) of thousands of unstained histological brain sections obtained from different species (human, monkey, rodents) and to realign (register) serial section images into coherent volumes to recover the original 3D brain shapes. Apparently, the project is purely data driven.

In the context of JINM11, a complex image analysis and processing workflow had to be developed: - image and signal processing techniques (e.g., segmentation, stitching, blind source separation) applied to each brain section - linear and non-linear registration techniques to generate 3D brain models - volume data analysis (e.g., orientation distribution functions, structure tensor analysis) - 3D visualization

Considering the pure size of the targeted data set (Terabyte range), it is evident that a high- performance computing solution is required to produce the unique datasets in a reasonable time frame. JURECA is the supercomputer and UNICORE the framework of choice to carry out the data analysis in an efficient and automatized way. At the poster we will present the current analysis workflow as well as new reconstruction results obtained from a vervet monkey brain.

[BIO 5] Full Dynamic Brain Simulation Using GATE in a High-Performance Computer

L. Caldeira, S. Lalitha, M. Lenz, R. Deepu, W. Klijn, C. Lerche, N.J. Shah, U. Pietrzyk

Modern PET scanners have increased sensitivity due to improved hardware capabilities [1]. The increased sensitivity enables new applications, for instance dynamic PET. Dynamic PET allows tracking of changes in the activity distribution in the course of time, in contrast to static imaging in conventional PET. A number of methods is being developed for dynamic PET imaging. Geant4 Application for Tomographic Emission (GATE) [2] is an open-source simulation tool to study scanner system design and methods, but there is still the need for simulated datasets that can be shared among researchers. This work aims at creating such datasets and a framework that can easily simulate different brain activity distributions in a reasonable time. In this work we report on the comparison of the current CPU based version of GATE and a GPU version that generates phantom scatter information after our previous code changes. Furthermore, we compare the CPU and GPU version of GATE on the number of true, scatter, and random coincidence events and provide an overview for the full runtime of the simulation. Additionally, we compare the runtime of the CPU and GPU simulations. The CPU simulations were performed with GATE v7.2. A full dynamic simulation of an hour was split in 3600 one- second frames. The differences between CPU and GPU were consistently under 2% across all time points. The runtime of the simulation is highly dependent on the configuration used. For the run times values presented, we used only Coincidences as output. Increasing the output by including Singles and Hits will increase the total run time. Unfortunately, the new code changes decrease the performance of the simulation (from 7:02:22 to 8:48:02). In HPC, the so-called energy to solution is an important characteristic for an application: The aim is to complete the task with the least amount of energy (money). The GPU GATE simulation has a time to solution of 20 minutes when running in isolation on a GPU node, but this setup is wasting resources and energy. For the optimal use of resources in JURECA, 48 GATE simulations should be simultaneously started, but the time to solution will be longer. However, the time to solution does not scale linearly, that is, it is shorter than 48 times 20 minutes (16 hours). Thus, for optimal resource usage, multiple concurrent GATE instances, each simulating a single frame of one second, were started on each node: 24 for the CPU simulations and 48 (with hyperthreading) for the GPU simulations. For short simulations of one single frame the GPU version is preferable as it is faster than the same simulation in CPU. To conclude, a full dynamic simulation of one hour can be simulated in a few days, that is, in 1000 node hours. We expect that performance improvements can still be made as the focus so far has been in validation of the code changes. Our developments will be made available to the GATE community. In the future, we are planning to simulate different dynamic patterns.

[1] Slomka, P. J. et al. "Recent advances and future progress in PET instrumentation." Seminars in nuclear medicine.l.(2016). [2] Jan, S. et al. "GATE V6: a major enhancement of the GATE simulation platform enabling modelling of CT and radiotherapy. "Physics in medicine and biology (2011).

[BIO 6] Collective Behaviour of Motile Cells

O. Duman, G. Gompper, J. Elgeti

Motile biological tissues show different phases of motility, generally described as nonmotile, jammed, and fluid-like phases [1-5]. The different phases support particular functions. In wound healing, cells move collectively to close a wound. In tumors, cells evolve into a motile phenotype, during cancer metastasis. In in vitro experiments, cell-cell contacts mature, leading to a glass-like arrest of the previously fluid-like tissue [3]. For motile tissues, we distinguish generally an ordered and jammed solid-like phase and a disordered and migratory fluid-like phase. We develop a novel computational model of tissues which includes the finite extend of single cells. The model reproduces the solid-to-fluid and the reverse transition as a function of physical parameters. In particular, we find that cell-cell adhesion favours a solid- like phase, while motility and deformability lead to more fluid like structures.

[1] N. Podewitz, F. Jülicher, G. Gompper, J. Elgeti, New J. Phys. 18 083020 (2016). [2] N. Podewitz, M. Delarue, J. Elgeti, EPL 109 58005 (2015). [3] S. Garcia, E. Hannezo, J. Elgeti, J.-F. Joanny, P. Silberzan, N. Gov, PNAS 112 15314 (2015). [4] M. Basan, J. Elgeti, E. Hannezo, W.-J. Rappel, H. Levine, PNAS 110 2452 (2012). [5] J. Ranft, M. Basan, J. Elgeti, J.-F. Joanny, J. Prost, F. Jülicher, PNAS 107 20863 (2010).

[BIO 7] Molecular Determinants of Glutamine Synthetase Deactivation by Tyrosine Nitration

B. Frieg, B. Görg, N. Homeyer, D. Häussinger, H. Gohlke

Tyrosine nitration is a covalent post-translational protein modification mediated by reactive nitrogen species [1]. Levels of nitrated tyrosine residues have been established as a biomarker for nitroxidative stress. Moreover, tyrosine nitration in proteins has been linked to structural and functional changes and is suggested to play a crucial pathophysiological role in human diseases [2, 3]. Human glutamine synthetase (GS) is highly sensitive to the nitration of Tyr336, causing GS deactivation [4, 5], which has been linked to hyperammonemia and cerebral ammonia intoxication [6]. However, the molecular mechanism how Tyr336 nitration deactivates GS has remained elusive. Here, we provide evidence that suggests that nitrated Tyr336 in its deprotonated form hampers adenosine triphosphate (ATP) binding to GS. By means of unbiased molecular dynamics simulations as well as binding and configurational free energy computations, we observed that, first, Tyr336 nitration weakens the direct interaction with ATP, and, second, Tyr336 nitration introduces structural and energetic barriers along the ATP binding path. Both results indicate a reduced binding affinity of ATP if Tyr336 is nitrated. Furthermore, we computed a marked decrease in the pKa of nitrated Tyr336 in its protein environment, suggesting that only the negatively charged variant is relevant for GS deactivation under physiological conditions. The suggested pH sensitivity of GS function may be of clinical importance, as a reduced GS activity leads to hyperammonemic conditions, which, in turn, may then completely abolish GS activity.

[1] R. Radi, Acc. Chem. Res. 2013, 46, 550-559. [2] L. Castro, V. Demicheli, V. Tortora, R. Radi, Free Radic. Res. 2011, 45, 37-52. [3] R. Radi, Proc. Natl. Acad. Sci. U. S. A. 2004, 101, 4003-4008. [4] B. Görg, M. Wettstein, S. Metzger, F. Schliess, D. Häussinger, Hepatology 2005, 41, 1065- 1073. [5] B. Görg, N. Foster, R. Reinehr, H.J. Bidmon, A. Hongen, D. Häussinger, F. Schliess, Hepatology 2003, 37, 334-342. [6] N. Qvartskhava, P.A. Lang, B. Görg, V.I. Pozdeev, M.P. Ortiz, K.S. Lang, H.J. Bidmon, E. Lang, C.B. Leibrock, D. Herebian, et al., Proc. Natl. Acad. Sci. U. S. A. 2015, 112, 5521-5526.

[BIO 8] Oligomerization of the GPCR TGR5

C. G. W. Gertzen, L. Wäschenbach, V. Keitel, C. Seidel, D. Häussinger, H. Gohlke

G-protein coupled receptors (GPCRs) are among the most important drug targets in the human body as they regulate a multitude of functions from eye-sight through blood pressure to intestinal motility. It has been shown that GPCRs can homo- and hetero-oligomerize in the endoplasmic reticulum and on the cell surface [1]. This oligomerization can impact the activity and membrane localization of GPCRs, hence allowing the targeting of GPCR di- and oligomers as a novel pharmacological tool. This way, influencing the activity of the GPCR TGR5 could prove beneficial. TGR5 is involved in the regulation of blood-glucose levels, metabolism, and inflammation but has also been found to be overexpressed in cholangiocarcinoma cells [2], making it a highly interesting drug target. Yet, to address the dimerization of TGR5 with small molecules in a sophisticated approach, knowledge of the protein-protein interfaces is necessary. Here, we present an integrated modeling approach using molecular dynamics (MD) simulations, umbrella sampling, and high precision MFIS- FRET to identify and evaluate the possible dimerization interfaces of TGR5. Based on X-ray structures of known GPCR dimers, we could show that the primary dimerization interface of TGR5 involves transmembrane helix 1 (TM1) and helix 8, and that TGR5 forms higher order oligomers using other interfaces [3]. Further energetic evaluation of TGR5 dimerization pathways via umbrella sampling and Molecular Mechanics Poisson-Boltzmann Surface Area calculations showed that another dimerization interface of TGR5 could involve TM4 and TM5. Thus, TGR5 could form rows of higher order oligomers involving both the TM1-helix 8 and TM4-TM5 interface.

[1] C. Hiller, J. Kühhorn, P. Gmeiner, J. Med. Chem. 2013, 56, 6542-6559. [2] V. Keitel, R. Reinehr, M. Reich, A. Sommerfeld, K. Cupisti, W. T. Knoefel, D. Häussinger, Z. Gastroenterol. 2012, 50, P5_24. [3] A. Greife, S. Felekyan, Q. Ma, C. G. W. Gertzen, L. Spomer, M. Dimura, T. O. Peulen, C. Wöhler, D. Häussinger, H. Gohlke, V. Keitel, C. A. M. Seidel, Sci. Rep. 2016, 6, 36792.

[BIO 9] Conformational Dynamics of the Unbound Lipase-Specific Foldase Lif Studied by MD Simulations and Fluorescence Spectroscopy

N. Verma, J. Kubiak, P. Dollinger, F. Kovacic, C. Seidel, K.-E. Jaeger, H. Gohlke

Intrinsically disordered proteins (IDP) may adopt an astronomical number of conformations [1]. The lipase-specific foldase (Lif) is such an example of a highly flexible steric chaperone with intrinsically disordered regions in the unbound state. Lipase A (LipA), produced by the gram-negative bacterium Pseudomonas aeruginosa, requires the assistance of the inner membrane-bound chaperone Lif to achieve its enzymatically active conformation. The crystal structure of the Burkholderia glumae LipA-Lif complex has been solved, in which Lif forms an α-helical scaffold embracing LipA in a headphone-like structure. At present, the exact mechanism of action of Lif is still unknown [2,3]. Due to the intrinsic disorder of Lif in its unbound state, the characterization of typical conformations by means of molecular dynamics (MD) simulations is challenging. To quantitatively characterize the conformational dynamics of Lif, we performed MD simulations, starting from the open, headphone-like structure of Lif to generate a conformational ensemble, and refined the resulting ensemble against high precision Förster Resonance Energy Transfer (hpFRET) measurements by using the maximum entropy method. We identified 35 Lif conformations ranging from a partially closed to an entirely closed state. These structures provide the basis for a detailed atomistic model of Lif dynamics in the unbound state.

[1] H. A. Leung, O. Bignucolo, R. Aregger, S. A. Dames, A. Mazur, S. Bernèche, S. Grzesiek, J. Chem. Theory Comput. 2015, 12, 383-394. [2] K. Pauwels, A. Lustig, L. Wyns, J. Tommassen, S. N. Savvides, P. V. Gelder, Nat. Struct. Mol. Biol. 2006, 13. [3] F. Rosenau, J. Tommassen, K. E. Jaeger, ChemBioChem 2004, 152-161.

[BIO 10] The Mechanics of Vesicle Blebbing

S. Hillringhaus, G. Gompper, D. A. Fedosov

A broad range of in silico models, including liquid and viscoelastic drop models, has been introduced for simulating the complex mechanical properties of different cell types. These models are used to understand and quantify experimental measurements. In this work, we employ a coarse-grained cell model in two and three dimensions which incorporates the membrane properties similar to the RBC-model and an elastic inner mesh to include the cytoskeletal properties. The model is formulated in the framework of the dissipative particle dynamics simulation method. It is used to investigate cell-blebbing, which is observed in synthetic vesicles. Cell-blebbing describes the dissociation of the membrane from the inner network, in this case as result of inner stress. We analyze the influence of different parameters on the blebbing process and show that the occurrence of blebbing is a result of the instability of the connection between membrane and actin-network.

[BIO 11] Mining Aggregation Patterns in MCMC Simulations of the Tau-Protein Derived Peptide PHF6

R. Jing, O. Zimmermann

Markov Chain Monte Carlo (MCMC) Simulation enables simulation of biomolecular processes which act on timescales of seconds or longer and are currently out of reach for Molecular Dynamics (MD) simulations. MCMC therefore enables novel insights into protein folding and peptide aggregation in atomic detail. As the individual configurational updates in MCMC can be large existing tools for analyzing MD trajectories cannot directly be applied. Here we report on the results of a large scale peptide aggregation simulation and the tools we developed for the analysis. The data are from a replica exchange simulation containing 200 chains of the peptide PHF6, an aggregation promoting part of the tau protein which itself is involved in several neurodegenerative diseases such as Alzheimer or Parkinson’s. The simulation was generated by our High Performance MCMC software ProFASi. With 32 replicas per temperature, 32 temperatures and 10000 stored configurations per replica, the simulation is comprised of 106 configurations to be analyzed. Using our in-house parallel analysis framework MCmine, as well as downstream statistical analysis and visualization using Python scripts, we describe the aggregation process of PHF6 in great detail, focusing on cluster initiation, cluster growth, and the formation of multi layered clusters. In particular we show how different properties contribute to the binding preferences that underlie the formation of the highly ordered fibrillar aggregates observed for PHF6.

[BIO 12] Anion Conduction by a Calcium-Activated Lipid Scramblase of the TMEM16 Family

A. Y. Kostritskii, J.-P. Machtens

The lipid bilayer is a perfect electrical insulator and provides the basis for molecular information processing at the cell membrane. Transport proteins establish electrochemical potential gradients across the membrane, providing the energy for passive ion flux through channels. Electrical signaling in cells is mediated by the opening and closing of ion channels in response to electrical or chemical stimuli and the consequent transmembrane voltage changes. Furthermore, active lipid transporters selectively redistribute lipid molecules between the inner and outer leaflets of the cell membrane. Upon activation, scramblases dissipate this transbilayer asymmetric distribution of phospholipids, thereby generating chemical signals at the cell surface. The TMEM16 family (also known as anoctamins) represents a class of membrane proteins found in virtually all eukaryotic cells. In humans, ten closely related TMEM16 isoforms exist with a functional specialization into either calcium-activated chloride channels, phospholipid scramblases, or both. The common denominator of TMEM16 proteins is their activation by intracellular calcium ions and positive voltage. The physiological functions of these proteins range from brain and nervous information processing, vision, motor coordination, transepithelial transport to cell volume regulation, and programmed cell death, with implications in various genetic diseases. The recent determination of an X-ray crystal structure of the fungal homolog nhTMEM16 provided first structural insights into this protein family. The dimeric structure suggests an unusually hydrophilic protein cavity, the so-called subunit cavity, facing the hydrophobic bilayer as a potential site of catalysis for lipid scrambling. However, the molecular details of anion conduction and lipid translocation are still elusive. In particular, it is unclear how these two distinct functions are accommodated in the protein, and if anions and lipids share a common transport pathway. nhTMEM16 operates as both ion channel and scramblase; the crystal structure thus enables us to investigate both TMEM16 functions by molecular simulations. With a combination of all-atom MD simulations and electrophysiological experiments, we aim at deciphering the molecular basis of selective ion conduction and lipid transport by TMEM16 proteins. We here performed all-atom molecular dynamics simulations to investigate the interactions of nhTMEM16 with the lipid bilayer and permeant ions. Using Computational Electrophysiology to apply a sustained transmembrane voltage, we have been able to directly simulate selective anion permeation through nhTMEM16 and thereby define the anion conduction pathway. We demonstrate that lipid head groups partly line the hydrophilic subunit cavity, where anion permeation occurs. Our simulations provide a mechanistic framework for understanding the dual functionality of TMEM16 proteins for subsequent experimental validation.

[BIO 13] Finite-Difference Time-Domain Simulations Assisting to Reconstruct the Brain’s Nerve Fiber Architecture by 3D Polarized Light Imaging

M. Menzel, M. Axer, H. De Raedt, K. Michielsen

The neuroimaging technique 3D-Polarized Light Imaging (3D-PLI) reconstructs the brain’s nerve fiber architecture with micrometer resolution [1-4]: Unstained histological brain sections are placed in a polarimeter that measures the birefringence (optical anisotropy) of the nerve fibers, revealing their spatial orientations. The anisotropic structure of the nerve fibers that is responsible for the birefringence also leads to diattenuation of brain tissue, i. e. the intensity of light that is transmitted through the tissue depends on the direction of polarization relative to the fiber orientation. As an extension to 3D-PLI, Diattenuation Imaging (DI) can be used to provide additional information about the underlying fiber structure [5]. In order to better understand the physical processes behind 3D-PLI and DI, and to improve the reliability and accuracy of the reconstructed fiber orientations, the propagation of the polarized light wave through the brain tissue was simulated by means of a massively parallel 3D Maxwell Solver [6]. The Maxwell Solver is based on a Finite-Difference Time-Domain (FDTD) algorithm which computes the electromagnetic field components by discretizing space and time and approximating Maxwell's equations by finite differences [7-8]. The simulations were performed on the supercomputer JUQUEEN [9]. Here, we demonstrate that FDTD simulations are a valuable tool for modeling 3D-PLI and DI measurements: The simulations do not only help to explain experimental observations, but also to improve the nerve fiber reconstruction. A combined measurement of 3D-PLI and Two-Photon Fluorescence Microscopy [10] revealed that the average transmitted light intensity (transmittance) in 3D-PLI measurements decreases with increasing inclination angle of the fibers. With FDTD simulations, we could show that the inclination dependence of the transmittance is caused by light scattering due to the small numerical aperture of the imaging system and that the transmittance can be used to distinguish e.g. horizontal crossing from vertical fiber structures. Combined 3D-PLI and DI measurements have shown that brain tissue exhibits two different types of diattenuation: For some brain regions, the transmitted light intensity becomes maximal (minimal) when the polarization of light is oriented parallel to the fiber structures, referred to as D+ (D-) effect [5]. Over time, the D- effect decreases and the correlation between diattenuation and birefringence increases. With FDTD simulations, we could show that anisotropic light scattering leads to an inclination-dependent diattenuation (D- for flat, D+ for steep fibers), which decreases over time. Moreover, we could demonstrate that the diattenuation also depends on tissue properties like the distribution of fiber radii and orientations. This makes DI a promising imaging technique revealing additional tissue properties.

[1] M. Axer, K. Amunts, D. Grässel, C. Palm, J. Dammers, H. Axer, U. Pietrzyk, and K. Zilles. A novel approach to the human connectome: Ultra-high resolution mapping of fiber tracts in the brain. NeuroImage, 54(2):1091-1101, 2011. doi:10.1016/j.neuroimage.2010.08.075. [2] M. Axer, D. Grässel, M. Kleiner, J. Dammers, T. Dickscheid, J. Reckfort, T. Hütz, B. Eiben, U. Pietrzyk, K. Zilles, and K. Amunts. High-resolution fiber tract reconstruction in the human brain by means of three-dimensional polarized light imaging. Frontiers in Neuroinformatics, 5(34):1-13, 2011. doi:10.3389/fninf.2011.00034. [3] M. Dohmen, M. Menzel, H. Wiese, J. Reckfort, F. Hanke, U. Pietrzyk, K. Zilles, K. Amunts, and M. Axer. Understanding fiber mixture by simulation in 3D Polarized Light Imaging. NeuroImage, 111:464-475,2015. doi:10.1016/j.neuroimage.2015.02.020. [4] M. Menzel, K. Michielsen, H. De Raedt, J. Reckfort, K. Amunts, and M. Axer. A Jones matrix formalism for simulating three-dimensional polarized light imaging of brain tissue. Journal of the Royal Society Interface, 12:20150734, 2015. doi:10.1098/rsif.2015.0734. [5] M. Menzel, J. Reckfort, D. Weigand, H. Köse, K. Amunts, and M. Axer. Diattenuation of brain tissue and its impact on 3D polarized light imaging. Biomedical Optics Express, 8(7):3163-3197, 2017. doi:10.1364/BOE.8.003163. [6] M. Menzel, M. Axer, H. De Raedt, and K. Michielsen. Finite-Difference Time-Domain Simulations for Three-dimensional Polarized Light Imaging. In K. Amunts, L. Grandinetti, T. Lippert, and N. Petkov, Eds., Brain-Inspired Computing, BrainComp 2015, Lecture Notes in Computer Sciences, Vol. 10087, Chp. 6. Springer International Publishing, Cham, 2016. doi:10.1007/978-3-319-50862-7_6. [7] A. Taflove and S. C. Hagness. Computational Electrodynamics: The Finite-Difference Time-Domain Method. Artech House, MA USA, 3rd Edn., 2005. [8] H. De Raedt. Advances in unconditionally stable techniques. In A. Taflove and S. C. Hagness, Eds., Computational Electrodynamics: The Finite-Difference Time-Domain Method, Chp. 18. Artech House, MA USA, 3rd Edn., 2005. [9] M. Stephan and J. Docter. JUQUEEN: IBM Blue Gene/Q Supercomputer System at the Jülich Supercomputing Centre. Journal of large-scale research facilities, 1, A1, 2015. doi:10.17815/jlsrf-1-18. [10] I. Costantini, M. Menzel, L. Silvestri, N. Schubert, M. Axer, K. Amunts, and F. S. Pavone. Polarized Light Imaging and Two-Photon Fluorescence Microscopy correlative approach for 3D reconstruction of the orientation of myelinated fibers. Optics in the Life Sicences Congress, OSA Technical Digest (online), paper BrW4B.5, 2017. doi:10.1364/BRAIN.2017.BrW4B.5.

[BIO 14] Disinhibition and Inhibition of HCN2 Channel Function by Ligand Binding to the Cyclic Nucleotide Binding Domain

C. Pfleger, M. Bonus, B. Frieg, M. Otte, A. Schweinitz, M. Kondapuram, T. Leypold, J. Kusch, K. Benndorf, H. Gohlke

Hyperpolarization-activated cyclic nucleotide-modulated (HCN) channels mediate electrical pacemaking activity in specialized cells of the heart and neurons of the brain [1]. The activity of HCN channels is controlled by two stimuli: (I) membrane hyperpolarization and (II) the binding of cyclic nucleotides (cNMP). However, the mechanism underlying the activation of HCN channels is presently only poorly understood. Here, we address the central question how changes in the conformational dynamics and energetics of the homotetrameric HCN2 upon cNMP binding relate to the ligand-dependent channel gating. Because cNMPs regulate a variety of targets, we further aimed to identify cNMP analogues selective for HCN2. As model systems we used the full-length structure, and the isolated tetrameric and monomeric cytosolic domain of HCN2. The full length HCN2 was used in molecular dynamics simulations for studying the influence of cNMPs on a network of interactions linking the transmembrane domain (TMD) and the cyclic nucleotide binding domain (CNBD). The relevance of these interactions on the HCN channel opening has been shown in a mutation experiment for a homologous channel [2]. As to the isolated CNBD, we used rigidity analyses [3,4] for identifying residues in HCN2 involved in the allosteric transmission upon cNMP binding. Our results depict pathways connecting (I) the CNBDs of each subunit and (II) the CNBD with the C-linker region of HCN2. Upon mutation, three out of four selected pathway residues show an altered mechanism of HCN2 activation in electrophysiological experiments. To identify HCN2-selective cNMP analogues, we first predicted the binding mode of known N6-substituted cNMP derivatives, which guided a subsequent virtual screening experiment. From this, we identified a set of cNMP analogues, which will be tested for selective binding on HCN2 channels.

[1] C. Wahl-Schott, M. Biel, Cell. Mol. Life Sci. 2009, 66, 470-494. [2] C. Nava et al., Nat Genet. 2014, 46, 640-645. [3] C. Pfleger, P. C. Rathi, H. Gohlke, et al., J. Chem. Inf. Model. 2013, 53, 1007-1015. [4] C. Pfleger, H. Gohlke, et al., J. Chem. Theory Comput. 2017, 13, 6343-6357.

[BIO 15] Changes in Brain Structure in Parkinson's Disease

P. Pieperhoff, M. Südmeyer, L. Dinkelbach, C. Hartmann, S. Ferrea, A. S. Moldovan, S. Diaz, K. Zilles, A. Schnitzler, K. Amunts

Background and Objectives: Neuropathological changes in sporadic Parkinson's Disease (sPD) have been found in the substantia nigra, brain stem nuclei, the cerebral cortex and basal ganglia. Braak et al. [1] postulated a six-stage model for the propagation of sPD related pathology in the brain based on the distribution of the protein α-Synuclein (AS), which is contained in Lewy-bodies in postmortem brains. Deviating findings in another neuropathological study [2], however, challenged the hypothesis of an association between the AS distribution and the symptomatology of sPD. The present longitudinal study aimed to quantify individual, local volume changes in in vivo MR-images of sPD patients and healthy control subjects in order to separate structural changes in sPD from those of normal ageing, and to characterize specific pathophysiological changes. In addition, associations of these changes with progredient motor and cognitive impairments were examined.

Methods: T1-weighted magnetic resonance (MR) images of 38 sPD patients (m/f 24/14, age 53 ± 13 years) and 27 control subjects (m/f 12/15, age 60 ± 10 years) were acquired on a Siemens 3T Trio scanner. Each subject was scanned about 8±4 times and over periods of 3.9 ± 2.4 years (maximum: 8.8 years). In addition, the UPDRS-III and MMSE were recorded at each time point. Individual series of MR images were analysed by Deformation-based morphometry [3]: The different follow-up images of each subject were non-linearly registered with his initial MR image, in order to measure the local volume changes relative to the first time point. The registration algorithm had to detect subtle structural changes in the range of a tenth of the voxel resolution, and at the same time suppress noise artefacts. To this end, an optimization algorithm was applied which aimed to minimize a dissimilarity metric, and which was constrained by modelling an elastic deformation. This yielded a system of partial differential equations which had to be solved in each iteration of the optimization by a multigrid solver. The spatial transformation resulting from the image registration was defined by a vector field, from which voxel-wise volume differences between each pair of registered images were derived. Next, the JuBrain atlas [4] was transformed onto each subject's initial MR image in order to compute the volume changes of specific neuroanatomical regions. Group differences in region-based volume changes and associations with motor and cognitive scores were analysed.

Results and Conclusions: The MR morphometry revealed significantly accelerated volume decreases in sPD patients as compared to controls bilaterally in the extrastriate visual cortex and fusiform gyrus (FG1-4), temporal, and inferior parietal lobule. Additional changes involved the left insula and Globus pallidus, and the right amygdala and Ncl. basalis Meynert. The pattern of accelerated volume decreases of the occipital and temporal lobe, as well as the amygdala in brains of sPD patients appears to coincide with the occurrence of AS- pathology in stages IV-VI, as described in [1]. Thus, longitudinal DBM enables the representation of structural change in individual brains, which may be relevant for both, neuroscience research and also for diagnosis and therapy monitoring. Further analyses will examine the associations between structural changes of individual symptoms and additional putative influencing factors.

Acknowledgements: The authors gratefully acknowledge the computing time granted by the JARA-HPC Vergabegremium on the supercomputer JURECA at Forschungszentrum Jülich.

[1] Braak, H., et al., Mov. Disord. 21, (2006): p. 2042-2051. [2] Parkkinen, L., Pirttilä, Tuula, and I. Alafuzoff, Acta Neuropathol, 115, (2008): p. 399-407. [3] Pieperhoff, P., et al., Neuroimage. 43, (2008): p. 269-287. [4] Amunts, K. and K. Zilles, Neuron. 88, (2015): p. 1086-107.

[BIO 16] Biomolecular Coevolution and Its Applications: Going from Structure Prediction Toward Signaling, Epistasis, and Function

A. Schug

On the molecular level, life is orchestrated through an interplay of many biomolecules. For many of them modern sequencing techniques provide us with an unprecedented wealth of genomic sequences, a 'Biological Big Data'. Novel statistical tools based on statistical physics such as Direct Coupling Analysis (DCA) take advantage of this explosive growth of sequential databases and trace residue co-evolution to infer secondary and tertiary contacts for proteins [1] and RNAs [2]. These contacts can be exploited as spatial constraints in structure prediction methods leading to excellent quality predictions [1,2,3]. Going beyond anecdotal cases of a few protein families, we have applied our methods to a systematic large-scale study of nearly 2000 PFAM protein families of homo-oligomeric proteins [4]. Also, we can infer mutational landscapes by capturing epistatic couplings between residues and can assess the dependence of mutational effects on the sequence context where they appear [5].

[1] Weigt M. et al., Proc Nat Acad Sci USA (2009) 106, 67-72; F. Morcos et al., Proc Nat Acad Sci USA (2011) 108, E1293-E1301 [2] E. De Leonardis et al., Nucl Acids Res (2015), gkv932. [3] Schug A. et al., Proc Nat Acad Sci USA (2009) 106, 22124-22129; Dago A. et al., Proc Nat Acad Sci USA (2012), 109: E1733-42. [4] G. Uguzzoni et al., Proc Nat Acad Sci USA (2017). [5] M. Figliuzzi et al., Mol. Bio. Evol. (2016), 33:268-280, msv211.

[BIO 17] NEST - The Neural Simulation Tool

D. Terhorst, J. Pronold

The highly scaling framework for spiking neural network simulations offers an efficient and model independent platform to neuroscientists. The rich set of features can be freely combined and users can customize and extend the simulator with neuron models, synapses, and other modules. The neural network dynamics and function of cerebral cortex rely on interactions on multiple scales, from the spiking activity of individual neurons to brain-wide interactions between areas. This poster shows the different applications of the NEST simulations and as a prime example presents simulations of the macaque brain at full cellular and synaptic resolution, and investigations of the relationship between its network structure and its resting-state dynamics.

[BIO 18] Collective Cell Behavior - a Cell-Based Parallelization Approach for a Phase Field Active Polar Gel Model

S. Praetorius, D. Wenzel, A. Voigt

We consider a continuum model for collective cell movement. Each cell is modeled by a phase field active polar gel model and the cells interact via steric interactions. We provide a finite element implementation with a parallel efficiency of at least 0.5 in the number of cells. This is achieved by considering each cell on a different processor and various improvements to reduce the communication overhead to deal with the cell-cell interactions. We describe implementation details and demonstrate results for up to 768 cells.

Miscellaneous Engeneering

[SE 1] Cable Fire Simulation

T. Hehnen, L. Arnold

In fire safety engineering, design fires are often used as relatively simple means to assess the effectiveness of smoke management strategies. However, their rigid, prescribed nature makes them difficult to use for the assessment of material release in case of fire. A different approach is the simulation of fire propagation, based on thermal degradation (pyrolysis) of the combustible material and the environmental conditions near the fire seat. Here efforts are presented to create material parameter sets that make it possible to estimate fire propagation on solid combustible materials. Simulations have been performed utilizing the Fire Dynamics Simulator (FDS).

[SE 2] From Elastic Wave Simulation to Ultrasonic Wavefield Imaging and Inversion

L. T. Nguyen, E. H. Saenger

Towards converting the power the advanced computation and simulation tools and modern computing machines offer into routine engineering practices, we investigate, deploy, and further develop ultrasound imaging methods based on the use of such tools and machines. The ComLabgo project at the International Geothermal Centre in Bochum aims to establish a computer-based facility and imaging tools for characterizing of natural materials (rocks and drilled core samples) as well as engineered materials (e.g. concrete and pipework). Based on an in-house rotated staggered finite difference program (Heidimod) and a public- domain spectral element program (SPECFEM3D), we numerically investigate the time reversal principle of elastic wave propagation. The time reversed wavefield is simulated by re- emitting the time reversed signals at locations where the signals are recorded. The simulated time reversed wavefield is then the vital element in the imaging steps within implementations of the wavefield self-focusing time reverse modeling (TRM) and the forward-reverse wavefield cross-correlation in the manner of the reverse time migration (RTM). One other approach is full waveform inversion (FWI) of the ultrasound measurement, which is intrinsically a model based data fitting approach. In FWI, the nonlinear data misfit function is most often minimized by a linearized optimization approach due to the high dimensionality of the inverse problem. In the context of seismic imaging at various scales, optimization strategies used for FWI have been extensively developed to deal with non-uniqueness and slow convergence rates. Due to the high dimensionality and algorithmic complexity of FWI, the inversion is most effectively executed on a high-performance computer and managed by a sophisticated workflow program. In this poster presentation, we demonstrate the efficiency of the full ultrasonic wavefield imaging methods for nondestructive material evaluation. We present a combined workflow that is potential for increasing the detectability and resolution of an ultrasound screening in complex engineered structures. Also, a computation demanding 3D model based ultrasound mapping for inspection of pipework, which allows for localization and sizing of possibly multiple pipe defects to be achieved in one imaging step, is presented. As for nondestructive testing, in many cases where 2D and small 3D models fit, massive parallelization over the computational domain and ultrasound shots combined with efficient imaging workflow management may allow near real-time wavefield imaging to be carried out. This capability, if can be achieved, will encourage the practice of model based imaging in nondestructive testing routines.

[SE 3] Size, Shape and Spectral Analysis of Nanoparticles by Analytical Ultracentrifu- gation Using UltraScan

M. Uttinger, J. Walter, W. Peukert

Analytical Ultracentrifugation (AUC) is a powerful and versatile tool to determine the size, shape and density of macromolecules and nanoparticles (NPs) directly in solution based on their sedimentation properties. Even complex mixtures can be studied as the measured signals of individual species are not superimposed. Data evaluation is achieved using the sophisticated software package UltraScan3 which makes use of direct boundary models and HPC. The sedimentation profiles are fitted to adaptive space-time finite element solutions of the Lamm equation [1] considering the sedimentation and diffusion transport in the AUC [2]. AUC equipped with a multiwavelength extinction detector (MWL-AUC) detects wavelengths reaching from 230 to 1100 nm simultaneously instead of just one. In our contribution the possibilities of direct boundary modelling will be presented. For proteins and quantum dots we could perform MWL experiments, in such way revealing the unique spectral properties of individual species and sizes [3, 4]. Moreover, first studies covering non-ideal sedimentation will be presented. When studying concentrated systems in AUC, their sedimentation behavior is affected by hydrodynamic and thermodynamic non- ideality [5]. High data resolution is of major importance in order to resolve the non-ideality effects in the sedimentation boundaries. Consequently, data analysis requires expensive numerical computations. As a comprehensive, multi-platform software package, UltraScan3 allows to analyze such AUC experiments. The time for data analysis is significantly reduced by HPC as the computing time scales linearly with the number of cores [6]. The extension of the existing numerical methods towards the incorporation of non-ideality by means of a concentration dependency of the sedimentation and diffusion coefficients will be outlined in our poster. AUC in combination with HPC is a powerful technique allowing for a direct correlation of size, shape, optical properties and non-ideality phenomena of NPs and macromolecules. This is highly relevant for a variety of new applications where multidimensional particle properties are in the focus.

[1] W. Cao and B. Demeler, Biophys. J., 2005, 89, 1589. [2] E. Brookes, W. Cao and B. Demeler, Eur. Biophys. J., 2010, 39, 405. [3] E. Karabudak, E. Brookes, et al., Angew. Chem. Int. Ed. 2016, 55, 11770. [4] J. Pearson, J. Walter, et al., Anal. Chem. 2018, 90, 1280. [5] A. Solovyova, P. Schuck, et al., Biophys. J., 2001, 81, 1868. [6] E. H. Brookes, R. V. Boppana, B. Demeler, SC 2006 Conference, Proceedings of the ACM/IEEE, Nov. 2006, 2006.

[SE 4] Light Scattering by Irregular Particles Much Larger Than the Wavelength

Y. Grynko, J. Foerstner

The knowledge of optical properties of natural particles like atmospheric aerosols, powder constituents, cosmic dust, etc., is needed in the field of optical remote sensing. Such particles demonstrate great variety of sizes, shapes and morphologies. If the size parameter X = πd/λ, where d is the particle size and λ is the wavelength, exceeds several tens the problem becomes multi-scale and accurate simulations of light scattering by realistic geometries require significant computer resources. The only efficient approach here is to apply geometrical optics (GO) approximation. This implies, however, ignoring small-scale surface roughness and, correspondingly, wave effects. We apply computer modeling to study light scattering properties of compact irregular particles in the wide range of size parameters from X = 10 to 200. To solve the light scattering problem we apply the Discontinuous Galerkin Time Domain method [1]. It allows optimal spatial discretization based on unstructured meshing and excellent parallel scalability that is critical for large-scale simulations. With such systematic variation of sizes we are able to track qualitative changes in the angular dependencies of intensity and linear polarization degree of scattered light. Interestingly, all main linear polarization features are preserved in the entire size range. At X = 150-200 we approach the geometrical optics regime where we can apply ray trajectory analysis. This gives us understanding of the light scattering mechanisms that work at smaller sizes.

[1] Hesthaven, J. S., and Warburton, T., 2002: Nodal High-Order Methods on Unstructured Grids: I. Time-Domain Solution of Maxwell's Equations. J. Comp. Phys. 181, 186-221.

Fluid Mechanics

[ST 1] Investigation of Cavitation Erosion with a Density-Based CFD-Method on a Hydrofoil with Circular Leading Edge

R. Skoda, M. Blume, P. Limbach

A hyperbolic flow solver together with the statistics of vapor dynamics and detection of void collapses is utilized for the cavitation erosion prediction on the circular leading edge (CLE) hydrofoil in dependence on cavitation number. Good agreement is achieved with experimental results regarding erosion sensitive wall zones and relative flow aggressiveness. Erosion is one of the most serious consequences of cavitation in hydraulic machinery. The most aggressive type regarding cavitation erosion is cloud cavitation, which occurs on the CLE-hydrofoil for the chosen range of cavitation numbers σ between 2.0 and 2.5 at Reynolds number 130000. Due to the similarity of the CLE-profile and pump blades, it is an adequate test case for pump flow. A CFD-method is chosen that is able to capture and resolve the relevant physical phenomena of cavitation erosion such as shock waves and compressibility effects. Thus, a density-based, compressible flow solver with explicit time integration is utilized. The flux function was developed for cavitating flow [1]. Cavitation is modelled with an equilibrium model via a barotropic equation of state [2], realized as look-up table for efficiency. No explicit turbulence model is applied. In order to investigate cavitation erosion, the collapses near the hydrofoil wall have been identified and analyzed statistically as in [3], yielding a load collective, i.e. cumulative collapse rate ccr vs. collapse pressure pcoll. In order to obtain local information about erosion sensitive wall zones, the procedure is applied to axial segments of length Δx = 10 mm above the hydrofoil: Only near-wall collapses above a threshold value of 200 bar are taken into account to yield local information about erosion sensitive wall zones. The erosion sensitive wall zones found by local evaluation of collapses agree qualitatively with the found erosion patterns for the three analyzed cavitation numbers σ: with increasing cavitation number erosion occurs further upstream and is concentrated to a smaller area. Flow aggressiveness can be defined as area under the ccr curve. For both experiment and simulation the highest load is observed for σ = 2.3. In the experiments [4, 5] a correlation has been found between vapor dynamics and erosion sensitive wall zones: CCD images, in which vapor is seen white and liquid black, have been analyzed regarding mean and RMS of pixel color. High gradients of average values and high RMS values correspond to erosion sensitive wall zones. For the simulation, the same comparison is made for iso surfaces of vapor volume fraction (with value 0.05), for which a binary evaluation is made. The vapor dynamics and erosion sensitve wall zones also show good agreement. Both the RMS of the void fraction field (vapor dynamics) as well as the local number of collapses are good indicators for erosion sensitive wall zones. The former can even be used with implicit CFD methods. The latter is needed in order to assess flow aggressiveness.

[1] Schnerr G. H., Sezal I. H., Schmidt S. J., Physics of Fluids 20 040703 (2008). [2] Iben U., System Analysis Modelling Simulation 42 (9), (2002). [3] Mihatsch M. S., Schmidt S. J., Thalhamer M., Adams N. A., Skoda R., Iben U., WIMRC 3rd International Cavitation Forum, Warwick (2011). [4] Bachert B., PhD Thesis, TU Darmstadt (2004). [5] Dular M., Bachert B., Stoffel B., Sirok B., WEAR 257 (11) : 1176-1184 (2004).

[ST 2] Multi-Physics, Multi-Scale Simulations Using CIAO

M. Bode, D. Denker, J. H. Göbbert, D. Goeb, J. Boschung, F. Hennig, A. Attili, H. Pitsch

Many applications in energy science feature concurrent and complex physical phenomena, which interact over a wide range of length and time scales. Full-scale simulations of these multi-physics and multi-scale problems are very challenging and resolving all scales in direct numerical simulations (DNS) is computationally not possible with current supercomputers. One way to address this limitation is to resolve only the large scales and model small-scale effects as successfully demonstrated with large-eddy simulations (LES) for turbulent flows. Due to the lack of reliable models for many physical phenomena such as interface dynamics or particle interactions in multiphase flows, the applicability of LES to complex multi-physics problems is still limited. At the Institute for Combustion Technology (ITV), RWTH Aachen University, the CIAO code is developed for predictive simulations of industrial applications. The code allows to run DNS and LES with high-order numerics, complex geometries and boundary conditions, and multi-physics phenomena in one framework and is thus well-suited for small-scale model development. Over the last two years, five different physical phenomena have been isolated, have been separately addressed with petascale DNS performed with CIAO, and small-scale models are currently under development: First, the effects of a large density ratio and a large change of vorticity at the outer edge of turbulent jets on turbulent mixing and entrainment was studied (JPHC09). Second, the interfacial processes in liquid/gas configuration under incompressible and compressible conditions was simulated. Different ligament formation processes were identified (JHPC18). Third, several temporal jets with non-premixed combustion were computed. Especially, the mechanism of extinction was addressed (JHPC22). Next, coal combustion was investigated by interacting particle simulations, which focus on the effect of particle inertia on the transport, heat exchange, and devolatilization of particles (JHPC48). Finally, reacting spray simulations were performed and ignition models were improved (JHPC49). A summary of the simulation results of this project is given on this poster. M. Bode received the JARA Excellent Junior award for his work as part of this project and wants to acknowledge the Jülich Aachen Research Alliance (JARA) for this honor and the ongoing support.

[ST 3] Instabilities and Turbulence in Magnetohydrodynamic Duct Flows

D. Krasnov, V. Bandaru, L. Bühler, T. Boeck

Liquid metal flows in the presence of a uniform magnetic field experience electromagnetic induction. Eddy currents and associated Lorentz force density modify the flow and give rise to thin electromagnetic boundary layers on the walls of the channel or duct. The transition to turbulence in magnetohydrodynamic sidewall jets is studied in direct numerical simulations. These jets occur in a duct flow when the Hartmann walls perpendicular to the magnetic field are electrically conducting. In addition, the modification of the magnetic field and the turbulence in a duct with insulating walls is investigated for the case when the magnetic diffusion time is not small compared to the flow time scale.

[ST 4] Direct Numerical Simulation of Liquid-Gas-Solid Flows

S. Bogner, J. Harting, U. Rüde

Three-phase flows involving suspended particles are common in nature and technical engineering. We present simulation results of bubble-particle interaction obtained from a free surface lattice Boltzmann model. The model is a direct numerical simulation technique, that is, all of the relevant length scales are resolved. Typically the smallest involved length scale is the particle size, while the gas bubbles are orders of magnitudes larger. To cope with the computational complexity, we rely on the massively parallel lattice Boltzmann framework waLBerla, that is member of Jülich’s High-Q Club. This allows the simulation of fully resolved particle beds consisting of hundred thousands of suspended particles in interaction with gas bubbles in a containing liquid.

[ST 5] External Intermittency in a Turbulent Shear Flow with Variable Fluid Properties

Y. Brahami, E. Varea, L. Danaila, F. Hunger, M. Gauding

Turbulent flows encountered in engineering and environmental applications are often free shear flows, where the variation of thermo-physical properties (like viscosity and/or molecular diffusivity) can be large due to variations of temperature or species composition. A characteristic feature of free turbulent shear flows is a thin distinct interface, that divides the flow in a fully developed turbulent core region, and a non-turbulent outer region, being associated with negligible vorticity fluctuations. At the so-called turbulent/non-turbulent interface (hereafter T/NT), strong fluctuations are created and transported through the mechanism of entrainment into the core of the shear flow. By using highly resolved direct numerical simulations of turbulent shear flows, it is shown that the T/NT-interface plays an important role for turbulent mixing. Conditional statistics with respect to the interface distance reveal that effects of variable fluid properties prevail in the vicinity of the T/TN-interface and are less dominant in the fully developed turbulent core region of the flow. This result is important to understand the inter-scale coupling between the large, non-universal structures and the small scales, where fluctuations are destroyed by viscosity or molecular diffusivity.

[ST 6] Generation of a Database with Detailed Numerical Simulation of Mixed-Mode Combustion

T. Zirwes, F. Zhang, J. A. Denev, P. Habisreuther, H. Bockhorn, D. Trimis

It is a common feature in practical combustion systems that fuel and oxidizer are not fully mixed. These partially premixed conditions are either deliberately generated, such as in stratified engines, or induced by instabilities, such as in aircraft engines or gas turbines. Mixed-mode combustion prevails in these cases, where combustion regimes range from fully non-premixed to premixed flame mode. Changes in the flame structure as it transitions through these modes has only recently started to be investigated. More crucially, the numerical modeling of such flame-regime transitions presents a challenge, as most of the existing combustion modeling concepts are designed specifically for either fully premixed or fully non-premixed combustion. This work presents results of a quasi direct numerical simulation (DNS) for a well- documented, mixed-mode combustion configuration of laboratory scale: the Sandia/Sydney flame FJ200-5GP-Lr75-57, which is operated with an inhomogeneously mixed methane/air flame, having a Reynold number of 28000 in the core region of the unburnt fuel jet. The objective is to provide a comprehensive database for the development of novel combustion modeling approaches, which can be applied to the numerical simulation of combustion processes with such flame-regime transition. The simulation has been made by using an in- house developed solver within the OpenFOAM framework. A computational grid with 150 million cells and a smallest resolution of 10 µm is used, employing detailed molecular diffusion and a complex reaction mechanism consisting of 19 species. The inflow condition with inhomogeneous mixture is generated by a previous DNS of only the mixing process within the burner nozzle. The simulation has been carried out on 12.000 processor cores from the Hazel Hen cluster maintained by the High-Performance Computing Center (HLRS) in Stuttgart, therewith consuming approx. 10 million core hours. The simulation results show very good agreement with measured data with respect to time mean and rms values of the chemical scalars like temperature and species mass fractions. Compared to the existing experimental data, the DNS database additionally includes: - statistical values for the flow velocities - temporal evolution of selected variables for 2000 monitor points with a time resolution of 0.2 µs - temporal evolution of selected variables on 2D slices passing through the centerline axis for 500 time steps with a sampling rate of 50 µs - volume data with all variables computed for the complete computational domain for 33 time steps These lead to an amount of totally 10 TB data, which, after evaluating and publishing of the simulation results, will be made publicly available for download. Such an extensive database with detailed numerical simulation of combustion process for mixed-mode combustion has not been presented before. Potentially, the database will generate a great interest for the combustion community which is working on modeling mixed-mode combustion concepts.

[ST 7] Liquid Water Transport in Stochastic Material of Gas Diffusion Layers of Polymer Electrolyte Fuel Cells

D. Froning, J. Yu, U. Reimer, W. Lehnert

In polymer electrolyte fuel cells of the type PEFC, DMFC and HT-PEFC, the gas diffusion layer (GDL) connects the electrodes with the feeding channels of the bipolar plate. To ensure efficient operation of fuel cells it is required that the electrodes are sufficiently supplied by fluid fuels from the channels. Also, reaction products must be transported away from the electrodes. The GDL also has to provide electric contact to the bipolar plates but its major task is the mass transport of these fluids. The GDL is typically composed of materials based on carbon fibers, e.g., paper, woven and non-woven textiles. In low temperature fuel cells, e. g., PEFC, the presence of liquid water in the GDL can substantially affect the efficiency of the fuel cell. Transport of liquid water in the micro structure of GDL is simulated by means of the Lattice Boltzmann method. Geometric structure data of GDL are available for paper type [1] and non- woven fleece type GDL [2], both real data and also data created by a stochastic geometry model. The artificial geometry is stochastic equivalent to the real 3D structure of the GDL. Transport simulations on paper-type GDL structures ran on the JURECA hardware of the Jülich Supercomputing Centre, grant JIEK30. The inhomogeneity of liquid water at the GDL/channel interface is analyzed with stochastic methods. This is concerning local characteristics of droplets exiting the GDL beneath the air channel of the flow field [3] and also statistical information on the distribution of the droplets entering the channel [4]. In particular, the apparent contact angles at the GDL surface show variations caused by the irregular microstructure of the stochastic model of paper-type GDL. These characteristics can potentially bridge the gap between multiple scales by providing essential information on droplets entering the channel to be used by cell and stack scale simulations.

[1] R. Thiedmann, F. Fleischer, C. Hartnig, W. Lehnert, V. Schmidt. Stochastic 3D Modeling of the GDL Structure in PEMFCs Based on Thin Section Detection, JECS 155 (4), B391-B399 (2008). [2] G. Gaiselmann, R. Thiedmann, I. Manke, W. Lehnert, V. Schmidt. Stochastic 3D modeling of fiber-based materials, Comput. Mater. Sci. 59, 75-86 (2012). [3] J. Yu, D. Froning, U. Reimer, W., Lehnert. Apparent contact angles of liquid water droplet breaking through a gas diffusion layer of PEMFC, Int. J. Hydrogen Energy, accepted. [4] J. Yu, D. Froning, U. Reimer, W., Lehnert. Liquid water breakthrough point distances on gas diffusion layer of polymer electrolyte membrane fuel cell, J. Power Sources, submitted. [ST 8] Using Highly-Resolved Direct Numerical Simulations to Analyze the Universality of Small-Scale Turbulence

M. Gauding, J. H. Goebbert, L. Danaila, E. Varea

The universality of a passive scalar advected in homogeneous isotropic turbulence is studied by scale-by-scale budget equations for higher order moments. Based on an analytical development in the dissipative range, a scaling for higher order structure functions is proposed. A similarity scale analysis is used to show the validity of the proposed scaling in the dissipative range and the inertial range. The analysis is based on highly resolved direct numerical simulations (DNS) with different Reynolds numbers. To this end, a comprehensive DNS data base of turbulence has been created. To resolve all relevant scales of turbulence the grid size is as high as 68 billion grid points. This data base allows a consistent analysis of small-scale turbulence and scaling laws of turbulent flows.

[ST 9] Numerical Investigation of Turbulent Combustion Noise Using a Hybrid LES/CAA Approach

S. Herff, K. Pausch, M. Meinke, W. Schröder

The acoustic response of round turbulent flames is investigated by a hybrid large-eddy simulation / computational aeroacoustics (LES/CAA) with respect to the burner geometry and inflow turbulence distribution. The first study is conducted to extend the findings for a rectangular slot burner studied by Schlimpert et al. to round burners. The results show that the round flame is more stable than the rectangular slot flame due to the additional expansion direction and the initial curvature. Consequently, smaller flame front perturbations and less flame pockets are generated at the flame tip reducing the noise emission in this region. Compared to the round burner, the additional length scale of the rectangular slot burner leads to a medium frequency region in the heat release spectrum with a stronger roll-off behavior and a lower peak frequency in the acoustic spectrum. In the second study, the hybrid approach is used to investigate the impact of the inflow turbulence on the flame’s acoustic emission. It is evident that various turbulence distributions lead to different turbulent kinetic energy spectra in the low frequency region of the cold jet. These differences also result in changes of the flame motion and significantly influence the acoustic spectra in the low frequency regime. Furthermore, a non-reactive LES computation including the burner’s plenum revealed significant differences in the turbulent kinetic energy spectrum compared to the LES results of the reduced domain.

[ST 10] Turbulent Superstructures in Rayleigh-Bénard Convection

A. Pandey, J. D. Scheel, J. Schumacher

Turbulent Rayleigh-Bénard convection displays a large-scale order in form of rolls and cells on lengths larger than the layer height once the fluctuations of temperature and velocity are removed. These turbulent superstructures are reminiscent of the patterns close to the onset of convection. They are analyzed by numerical simulations of turbulent convection in fluids at different Prandtl number ranging from 0.005 to 70 and for Rayleigh numbers up to 107. For each case, we identify characteristic scales and times that separate the fast, small-scale turbulent fluctuations from the gradually changing large-scale superstructures. The typical scales of large-scale patterns, which change with Prandtl and Rayleigh number, are also found to correlate with the boundary layer dynamics that generates clusters of thermal plumes at the top and bottom. Our analysis suggests a scale separation and thus the existence of a simplified description of turbulent superstructures in geo- and astrophysical settings. [ST 11] Transient Simulation of Centrifugal Pumps with OpenFOAM

A. Pesch, N. Casimir, R. Skoda

For the transport of solid loaded fluids, e.g. wastewater or sludge, pumps with few blades (e.g. 1 or 2 blades) are used in order to prevent the pump hydraulics from clogging. The machines are designed with focus on high reliability while appropriate methods to improve their energy efficiency are not available today. Conventional centrifugal pumps are optimized for operations near the design-point and provide poor efficiency with differing operating conditions like partial load and overload. In both these cases, the flow is dominated by transient phenomena. State of the art simulation methods with statistical URANS turbulence models overpredict integral values such as pump head and efficiency. The project aims to enhance CAE methods for fluid-dynamic and vibration analysis based on the OpenFOAM software. This will be tackled by improving the handling of unsteady boundary conditions as well as the usage of scale resolving turbulence modeling (Scale-Adaptive Simulation, SAS [1]). In the first part of the project, detailed validation studies were conducted to assess the quality of state of the art URANS simulations of a single-blade pump and a centrifugal pump with low specific speed using OpenFOAM. An implicit pressure-based solver with dynamic mesh handling and a merged PISO-SIMPLE-algorithm (pimpleDyMFoam) was used. The coupling between rotor and stator was utilized via a General-Grid-Interface (GGI) [2]. Turbulence was modelled with the k-omega-SST approach [3]. A grid convergence study was conducted to assure grid independent simulation results. The results, including integral values such as pump head and power as well as the transient pressure trace at different positions in the volute and the internal flow field were compared to in-house and external experimental data [4] as well as commercial simulation software results. The observation was made that unphysical pressure oscillations occur at the rotor-stator- interface caused by a mass defect. A way to tackle these pressure oscillations at the rotor- stator interface in OpenFOAM is presented as well.

[1] F. R. Menter and Y. Egorov: "The Scale-Adaptive Simulation Method for Unsteady Turbulent Flow Predictions. Part 1: Theory and Model Description" Flow, Turbulence and Combustion, vol. 85, no. 1, pp. 113-138, 2010. [2] H. Jasak and M. Beaudoin: "Development of a Generalized Grid Interface for Turbomachinery simulations with OpenFOAM" Open Source CFD International Conference 2008, 2008. [3] F. R. Menter and T. Esch: "Elements of Industrial Heat Transfer Prediction" 16th Brazilian Congress of Mechanical Engineering, 2001. [4] S. Meschkat: Experimentelle Untersuchung der Auswirkungen instationärer Rotor-Stator- Wechselwirkungen auf das Betriebsverhalten einer Spiralgehäusepumpe, Dissertation, Technische Universität Darmstadt, 2004.

[ST 12] Model Development for Meteorological Applications

V. Heuveline, M. Baumann, P. Zaspel, S. Gawlok, C. Song, P. Gerstner, N. Schween

Numerical simulations based on the Finite Element method (FEM) are conducted for three different areas of Computational Fluid Dynamics. First, heat transfer in a vertical cylinder annulus filled with a dielectric fluid under applied temperature gradient and voltage between inner and outer wall is considered. The situation can be modeled by means of the thermal electro hydrodynamical Boussinesq equations (TEHD) which are based on the standard Boussinesq approximation for natural convection, augmented by DEP force and Gauss law for describing the electric field inside the fluid as a function of temperature. We performed simulations for different temperature gradients, leading to varying classical and electric Rayleigh numbers in the order of 104. The numerical solutions show emergence and transition of a varying number of azimuthally to axially oriented vortices, where the latter are present in the final state of the dynamic system. Due to these vortices, radial heat transfer is enhanced. Another application in this study is about a method comparison for cyclone-cyclone interaction (CCI). The weather forecast of the motion and evolution of the interaction of tropical cyclones is a very challenging task, not only the underlying physics is very complicated to model, but also the computing time needs to be realistic in order to make the prediction possible. We compare therefore two models: the compressible Navier-Stokes model and the Low-Mach model. The Low-Mach model can be considered as an extension of the compressible Navier- Stokes equations based on the Low-Mach number approximation. The central question is about how large the time step can be applied for the Low-Mach model comparing to the compressible model, such that we can reduce the computational time by a notable factor and maintain a comparable quality of numerical results. We consider in this study the tropical cyclones which have several 100km of diameter, it leads to a computational domain of 4000km × 4000km × 13km. The comparison is analyzed with respect to the numerical and physical scenarios, e.g. the computational costs and the cyclone trajectory. A reduction of the computational time by a factor 15 is achieved by using the Low-Mach model. Uncertainty Quantification (UQ) receives considerable attention in the last years owning to the improved quality and reliability of numerical simulations for physical problems. In this project, we consider the Spectral-Stochastic Finite Element Method (SSFEM) to model the uncertain inputs in a blood pump device. The main difficulty is that the stochastic Galerkin method introduces a coupled system on the discrete level crossing all stochastic modes, the standard linear solvers and preconditioners are no more efficient. Therefore, we use a Multilevel preconditioner to precondition the global matrix, this technique exploits the hierarchical structure of the Chaos Polynomials, a satisfactory convergence behavior can be achieved. For the blood pump device, we utilize the Food and Drug Administration (FDA) blood pump. Three uncertain parameters are taken into account: inflow boundary condition, dynamic viscosity and rotational speed. The deterministic problem is modeled with the variational multiscale method (VMS). We developed and tested a scalable solver for linear systems that arise when discretizing multi-physics systems by means of the Finite Element method for each of the three applications. This solver is based on a two-level Schur complement approach that decouples the solution process for the overall system into subsequent solutions of simpler, elliptic problems which are well suited for being solved with highly scalable methods, such as the Algebraic Multigrid Method. All simulations are performed by means of the multi-purpose FEM package HiFlow3.

[ST 13] Simulation of Oil Jets for Piston Cooling Applications

L. Wendling, V. Karyofyli, M. Frings, A. Hopf, S. Elgeti, M. Behr

Inside the next generation of Internal Combustion Engines (ICE), pistons will require new cooling techniques to maintain good durability. The most promising technique actively investigated consist in spraying lubricating oil underneath the piston to remove the excess heat. This cooling medium was chosen for its high cooling performance and availability in the engine. As an alternative to traditional test benches, we are developing numerical simulations for the design phase with the objective of providing detailed insights into the complex interactions between oil, air and the moving piston. In jet simulations, it is necessary to capture accurately the large deformations of the interface including the splitting and merging behaviour of the oil droplets. One method suitable for those conditions is the level set method which implicitly keeps track of the interface position. Furthermore using this method, the interface can be explicitly assessed which allows an accurate account of the capillary effects. For each Newtonian fluids, air and oil, the motion is governed by the incompressible Navier-Stokes equations. The resulting two-phase flow problem is solved with a staggered approach and is discretised using space-time finite element method which allows implicit mesh deformation necessary to model the piston movement. Simulations results for jet atomisation will be presented. Details of the oil spreading were obtained with different geometries: a flat surface and a real piston geometry. Using snapshots from the experiment we successfully validated our model for different jet regimes, ranging from laminar to atomised jets. Like in the experiment, the inflow mass flow rate and the temperature influence the jet behaviour. We further extended our model by modelling a moving piston which includes a cooling gallery to improve the cooling efficiency. The piston movement is achieved by means of elastic mesh deformation and allowed to study the complex interface interactions resulting from the rising oil jet and the moving. The oil filling ratio inside the cooling gallery is recorded throughout the simulation and will be correlated in the near future with the heat transfer coefficient across the surface of the cooling gallery. This correlation can then be used as a guideline to design the next generation of pistons with better cooling performance.

Earth and Environment

[MET 1] Scale-Dependence of Atmosphere – Surface Coupling Through Similarity Theory

C. Ansorge

Surface-layer similarity theory is derived under the assumptions of horizontal heterogeneity and stationarity, but it is used in ever more complex and heterogeneous configurations. This work independently assesses the scale-dependence of surface-layer similarity theory by obtaining values of the surface friction from wind and velocity fields within the surface layer. Therefore, a classical set of Monin-Obukhov similarity functions is inverted analytically; this allows to investigate the joint convergence of individual samples of the wall-friction and its estimate from similarity theory as a function of the spatial and temporal averaging scale. Spatially and temporally resolved data originate from a direct numerical simulation of turbulent Ekman flow, an idealized representation of the atmospheric boundary layer. A systematic underestimation of variance in surface fluxes estimated through similarity theory versus the actual flux is found and quantified as a function of height, averaging time and length scale. A sufficient convergence of the local friction estimate by similarity theory to the actual local surface friction is only obtained for averaging box sizes of several thousand wall units, and only for data filtered along both horizontal dimensions; the three-dimensional structure of the turbulence elements appears to limit the convergence of data filtered along any of the single dimensions time, the streamwise or the spanwise direction.

[MET 2] Full-Waveform Inversion of Seismic Data

N. Athanasopoulos

Living in a time where natural resources are scarce and precious, the number of underground constructions is increasing, and the storage of waste and other material in the earth is necessary, it is important to find accurate ways of mapping geological structures in the Earth’s interior. Most seismic imaging methods analyse the arrival time of recorded seismic signals. Full- waveform tomography (FWT) is a cutting-edge inverse method that accounts for the full- seismic waveform recorded over a broad range of frequencies and apertures. It iteratively retrieves multiparameter models of the subsurface by solving the full wave equations. It allows for a mapping of structures on spatial scales down to less than the seismic wavelength, hence providing a tremendous improvement of resolution compared to traveltime tomography which is based on ray-theory. The main part of the FWT consists of seismic modeling which is a straightforward method, i.e., synthetic seismic data are computed by the numerical solution of the acoustic or elastic wave equation. Modeling is essential for simulations of wave propagation and the computation of data sets based on model assumptions of the subsurface. In contrast, the FWT is an inverse method, whose aim is to recover the underlying model using an observed seismic data set. We apply the elastic multi-parameter full-waveform Inversion (FWI) for shallow seismic applications using our software IFOS ((Inversion of Full Observed Seismograms). These cases are extremely important for geotechnical and hydrogeophysical characterisation, since they allow to reconstruct essential information of the shallow subsurface. Specifically, the aim lies on reconstructing the P- and S-wave velocities of the subsurface along with density, allowing the coupling with various other parameters for safer constructions and better resource management. The main information comes from Rayleigh waves since they exhibit a high signal to noise ratio in field data recordings and they are easily exited by using hammer blows. As Part of the BMBF founded Wave-Project we developed a new object orientated modelling code for 2D and 3D (visco)acoustic and (visco)elastic wave equations. In this new code the spatial derivatives in the wave equation are solved via matrix-vector products. With this formulation it becomes possible to implement the HPC library “LAMA". The main advantage is that we can run the modelling code hardware-independent. "LAMA" facilitates the development of fast and scalable software for distributed HPC systems currently including multicore CPUs, Nvidia GPUs and Intel Xeon Phi accelerators. Besides, it allows a seamless integration of future architectures while offering an independent formulation of algorithms, which makes rewriting code for new hardware obsolete. We performed weak and strong scaling benchmarks of the 3D elastic implementation on JURECA to test the efficiency on CPUs and GPUs.

[MET 3] On the Predictability of Exceptional Error Events in Wind and Solar Power Forecasting – An Ultra Large Ensemble Approach

J. Berndt, H. Elbern

Exceptional error events in wind and solar power forecasting impose a major obstacle to today's economic and reliable power supply. The predictability of such error events is fundamentally restricted by the underlying weather forecast, resting on limitations of state-of- the-art numerical weather prediction systems. These forecasts must be furbished with likelihood, implying the operation of model ensembles. To this end, the standard sizes of meteorological ensembles are increased from O(10) to an ultra large ensemble size of O(1000) members to accomplish an improved approximation of the probability density function. The computational complexity is far beyond the capacities of traditional massively parallel platforms and calls for the dedication of JUQUEEN. The increased ensemble size favours the application of Sequential Monte Carlo Methods (Particle Filter), a fully nonlinear data assimilation technique, while imposing the challenge of growing computational expenses of a resampling step within the particle filter algorithm. For this purpose, a novel approach of an ensemble control system has been developed on JUQUEEN (ESIAS-met), which realizes a parallel execution of the ensemble within a single executable. Performance measurements demonstrate strong scalability of the system with up to 4096 ensemble members utilizing 262,144 cores. The ESIAS-met system is further applied to investigate the benefit of an increased ensemble size on the predictability of recent exceptional error events. The analysis reveals, that despite the large ensemble size, the forecast error is only represented by single outliers. As an approach to identify imminent forecast errors, higher order moments prove to provide a robust measure of the proper direction of forecast error and to assess their likelihood. It is shown, that at least O(100) of ensemble members are needed to resolve the higher order moments sufficiently well. Hence, the results achieved in this work yield important potential for future warning capabilities of exceptional error events.

[MET 4] Observability and Predictability of Volcanic Ash Transports

A. C. Lange, P. Franke, H. Elbern

Volcanic ash emissions come with large uncertainties due to unknown emission parameters that results in uncertain volcanic ash concentrations in the atmosphere. This issue is addressed in terms of observability and predictability analysis. Observability analysis identify beneficial observations and may suggest areas and times at which observations are most valuable. Predictability analysis provide uncertainty estimation of the emission parameters as well as of the volcanic ash forecast. Both analysis are constraint by SEVIRI observations of vertically integrated column mass loadings of volcanic ash. One challenge is the analysis of vertically resolved emission profiles from the vertically integrated observations, which make the use of an ensemble of model runs inevitable. The observability analysis is performed with an ensemble of nine 4-D var data assimilation runs with differing emission heights using the EURopean Air pollution Dispersion – Inverse Model (EURAD-IM). This ensemble was applied to the Eyjafjallajökull eruption 2010 in Iceland. Observation-constraint areas in the analyzed volcanic ash concentration field were identified. In an identical twin experiment the new ensemble EURAD-IM was integrated in the Ensemble for Stochastic Integration of Atmospheric Simulations (ESIAS) system in order to analyze the predictability of volcanic ash forecasts. Here, an ensemble was generated in which in each ensemble member only a small amount of volcanic ash at distinct times and height was emitted. These emission packages were optimally combined in a Nelder-Mead minimization algorithm before the variance of the ensemble was optimized by a resampling step. The identical twin experiment shows the ability of the ESIAS system to provide reliable estimates of vertically resolved emission profiles and its uncertainty under ideal conditions, which are a long assimilation window length and a high wind speed. The use of the high dimensional chemistry transport model EURAD-IM (up to O(107)) in an ensemble framework with O(100) ensemble members requires high computational power. Additionally, the 4-D var data assimilation method consists of an iterative computation of the EURAD-IM and its adjoint and, thus, needs high computational power, too.

[MET 5] European Extreme Events Simulations with the Fully Coupled TerrSysMP

C. Furusho, S. Kollet, K. Goergen, J. Keune, K. Kulkarni

20 year (1989-2008) time series of fully-coupled groundwater-to-atmosphere simulation results using the Terrestrial Systems Modeling Platform (TerrSysMP) over the EURO- CORDEX domain at 12km resolution is presented. The focus of this study is placed on the added value of 3D groundwater dynamics on land-surface and atmosphere feedback processes to reproduce extreme events as heat waves and intense precipitation. In previous studies, the capacity of current available regional climate models (RCMs) of reproducing heat waves in the past two decades has been examined and compared to observational reference data. In this study, we interrogate, whether the 3D representation of groundwater dynamics potentially improves simulated spatial variation of soil moisture and therefore better captures the feedbacks between land-surface and atmosphere at the climate time scale. The present long-term simulations allow further investigation on the capacity of the model to represent extremes and climatological annual and seasonal mean values. TerrSysMP simulations are compared to the outputs of other EURO-CORDEX models, focusing particularly on climate indices to evaluate the added value of including groundwater dynamics to the simulation of extreme events. In addition to the typical atmospheric and land surface variables simulated by RCMs, the fully coupled model provides the climatology of groundwater depth and ground water recharge, improving our understanding of past evolution of the water cycle at the continental scale.

[MET 6] Convection Permitting WRF Climate Simulations: Precipitation Statistics and Impact of Land Surface Properties

S. Knist, K. Goergen, C. Simmer

High-resolution regional climate models with a detailed representation of heterogeneous land surface properties, and an explicit treatment of deep convection can lead to an improved simulation of meteorological processes and the climate system at the meso-gamma scale. In this presentation we focus on (1) precipitation statistics and (2) the impact of land surface properties on land-atmosphere exchange fluxes and meteorology. With topic (1) we address the following questions: How well can precipitation observations be reproduced with current models and what is the added value of high-resolution runs? Will the precipitation intensity distribution change in the future? Results from the WRF RCM, driven both with 10 years current climate ERA-Interim and 3x10 years MPI-ESM-LR climate change simulations, all at a 3km convection-permitting spatial resolution for central Europe, are analyzed focusing on precipitation statistics. Evaluation simulations from both resolutions are compared and evaluated against sub-daily synop station data over three regions with a moderate, low mountain and high mountain topography. An added value in the 3km simulation is found especially at the sub-daily scale in the reproduction of intensity, diurnal cycle and spatial extent of precipitation. A positive precipitation bias found for both resolutions is more dominant in the 12km simulation with too much light precipitation. Precipitation clearly differs between both simulations depending on the season with largest differences over mountainous regions and during summer months with high convective activity. Furthermore, we examine changes in precipitation intensity distributions and extreme precipitation indices based on MPI-ESM-LR RCP4.5 driven control and scenario time slices, again for areas with different topography. With topic (2) we address the following question: What is the impact of the spatial scales of the land-use patterns, soil moisture and orography on convection-permitting RCM simulations in terms of atmospheric patterns and domain wide averages. We perform five WRF simulations each with the same atmospheric setup at 3km resolution but different combinations of coarsely resolved (12km) land use and soil type, initial soil moisture and orography for summer 2003 over central Europe. A coarser-resolved orography alters the large-scale flow pattern, which results, e.g., in a weaker Föhn and in enhanced locally generated convective precipitation, that peaks earlier in the afternoon compared to simulations with highly-resolved land-use patterns. The coarser-resolved land-use mainly affects overall percentages rather than the heterogeneity of the atmosphere. Soil moisture initial conditions on the contrary have a stronger relative impact than land use heterogeneity changes. In general, the differences caused by coarse land surface patterns in 3km runs are much smaller than differences with a 3km vs a 12km atmospheric model resolution.

[MET 7] The Role of Mesoscale Eddies in Energy Transfers and Tracer Uptake in the Ocean

A. Griesel, C. Eden

Mesoscale eddies play an important role in the oceanic uptake of tracers, such as CO2, and they are an important player in the energy transfers over different scales in the ocean. Oceanic eddies derive their energy primarily from baroclinic instability processes and are currently parameterized using different eddy diffusivities in down-gradient parameterizations. CO2 uptake in coarse resolution models where eddies are parameterized with isopycnal and thickness diffusivities is usually very sensitive to increases in Southern Ocean wind stress, while sensitivity is thought to be reduced when eddies are resolved. While spatial variability of eddy diffusivities has been investigated before, the time variability and response to changes in wind energy input has not received much attention. We use here a global realistic high-resolution ocean model, parallelized to run on 2000 cores, to quantify tracer uptake and eddy diffusivities from both eddy tracer fluxes and Lagrangian particle trajectories, and their variation with time.

[MET 8] Soil Moisture Assimilation into TerrSysMP at Different Scales

H.-J. Hendricks Franssen, W. Kurtz, B. Naz, S. Gebler, K. Görgen, S. Kollet

Soil moisture is a key driver for water and energy exchange at the land surface and plays an essential role for water management, food production, flood forecasting, or climate projections. Soil moisture can be measured at different spatial and temporal scales: point scale with in situ sensor networks (e.g., TDR, FDR); intermediate-scale with for example cosmic-ray probes and large-scale with remote sensing (e.g., SMOS, SMAP). We investigated the value of soil moisture measurements at different scales for improving the characterization of system states and water and energy fluxes in terrestrial systems. This investigation was carried out using: (i) the Earth system model TerrSysMP, which includes the subsurface, land surface and atmosphere (Shrestha et al., 2014); (ii) soil moisture measurement data, either at the point scale (Rollesbroich, TERENO), the field scale (TERENO-catchment Rur-Eifel) or the continental scale (the merged remote sensing product CCI from ESA); and (iii) the Parallel Data Assimilation Framework PDAF (Nerger & Hiller, 2013, Kurtz et al., 2016), which is coupled to TerrSysMP. TerrSysMP-PDAF is a highly efficient framework which shows a good scalability at JUQUEEN for more than 50,000 processors. This is needed for solving the governing equations at a high spatial and temporal resolution and in ensemble mode. Results show that the TerrSysMP-PDAF framework improves soil moisture predictions for hillslope to continental scale models using data from different soil moisture monitoring techniques.

[MET 9] High-Performance Computing in Basin Modeling: Simulating Mechanical Compaction Through Vertical Effective Stress Using Level Sets

S. McGovern, S. J. Kollet, C. M. Buerger, R. L. Schwede, O. G. Podlaha

In the context of sedimentary basins, we present a model for the simulation of the movement of a geological formation (layers) during the evolution of the basin through sedimentation and compaction processes. Assuming a single phase saturated porous medium for the sedimentary layers, the model focuses on the tracking of the layer interfaces, through the use of the level set method, as sedimentationdrives fluid-flow and reduction of pore space by compaction. On the assumption of Terzaghi's effective stress concept, the coupling of the pore fluid pressure to the motion of interfaces in 1-D is presented in McGovern, et al. (2017) [1]. The current work extends the spatial domain to 3-D, though we maintain the assumption of vertical effective stress to drive the compaction. The idealized geological evolution is conceptualized as the motion of interfaces between rock layers, whose paths are determined by the magnitude of a speed function in the direction normal to the evolving layer interface. The speeds normal to the interface are dependent on the change in porosity, determined through an effective stress-based compaction law, such as the exponential Athy's law. Provided with the speeds normal to the interface, the level set method uses an advection equation to evolve a potential function, whose zero level set defines the interface. Thus, the moving layer geometry influences the pore pressure distribution which couples back to the interface speeds. The flexible construction of the speed function allows extension, in the future, to other terms to represent different physical processes, analogous to how the compaction rule represents material deformation. The 3-D model is implemented using the generic finite element method framework Deal II, which provides tools, building on p4est and interfacing to PETSc, for the massively parallel distributed solution to the model equations [2]. Experiments are being run on the Jülich Supercomputing Centre’s Jureca cluster [3].

[1] McGovern, et al. (2017). Novel basin modelling concept for simulating deformation from mechanical compaction using level sets. Computational Geosciences, SI:ECMOR XV, 1-14. [2] Bangerth, et al. (2011). Algorithms and data structures for massively parallel generic adaptive finite element codes. ACM Transactions on Mathematical Software (TOMS), 38(2):14. [3] Jülich Supercomputing Centre (2016). JURECA: General-purpose supercomputer at Jülich Supercomputing Centre. Journal of large-scale research facilities, 2, A62. doi:10.17815/jlsrf-2- 121.

[MET 10] The Challenge of Small-Scale Turbulence in Planetary Boundary Layers

J. P. Mellado, K. Fodor, A. Haghshenas, M. Karimi, B. Schulz

Planetary boundary layers (PBLs) are important in climatology-modulating the fluxes between atmosphere, land and ocean-, and in meteorology - influencing weather conditions -, but key properties remain poorly understood, largely because the PBL is turbulent, and understanding and characterizing the multi-scale nature of turbulence remains challenging. High- performance computing and direct numerical simulations are decisively contributing to advance our understanding of PBL properties.

[MET 11] Towards Improved High-Resolution Hydrologic Simulations over Europe Using HPC Based TerrSysMP-PDAF

B. S. Naz, S. Kollet, H.-J. Hendricks Franssen, W. Kurtz, C. Montzka, W. Sharples, K. Goergen, J. Keune, A. Springer

A Parallel Data Assimilation Framework (PDAF) integrated with the Community Land Model, version 3.5 (CLM3.5) in the Terrestrial System Modeling Platform (TerrSysMP-PDAF) was used to improve continental-scale hydrologic estimates of soil moisture, surface runoff, discharge and total water storage. The model was forced with the high-resolution reanalysis COSMO-REA6 from Hans-Ertel Centre for Weather Research (HErZ) and implemented at a spatial resolution of approximately 3km over Europe for a time period of 2000 – 2006. Using this modeling framework, remotely sensed ESA CCI soil moisture (CCI-SM) daily data were assimilated into TerrSysMP-PDAF. The impact of remotely sensed soil moisture data on improving continental-scale hydrologic estimates was analyzed through comparisons with independent observations including ESA CCI-SM, E-RUN runoff, GRDC river discharge and total water storage from GRACE satellite. Cross-validation with independent CCI-SM observations show that estimates of soil moisture improved, particularly in the summer and autumn seasons. The assimilation experiment also showed overall improvements in runoff particularly during peak runoff. The results demonstrate the potential of assimilating satellite soil moisture observations to improve high-resolution hydrologic model simulations at the continental scale, which is useful for water resources assessment and monitoring.

[MET 12] Feedbacks in the General Circulation Model ICON-A for Explicit and Parametrised Convection across Resolutions

M. H. Retsch, T. Mauritsen, B. Stevens

A recently developed General Circulation Model (GCM), ICON-A, is used to conduct experiments similar to ‘aquaControl’ and ‘aqua4K’ of the Cloud Feedback Model Intercomparison Project (CFMIP) at different resolutions and at two different setups of the model. One setup uses the default convection parametrisation and the other setup switches the convection parametrisation off, which is called explicit convection. The resolutions range from 2525 to 5km globally. ICON-A decreases its Equilibrium Climate Sensitivity (ECS) for resolutions higher than 631km in both setups. With explicit convection the longwave (LW) feedback is more negative, resulting in a lower ECS, whereas the shortwave (SW) feedbacks are more positive. The reasons for the different behaviour of the two setups are traced back mainly to more negative LW clear-sky and cloud feedbacks in the tropics and a less positive SW feedback from less increase of total cloud water in the high latitudes for explicit convection.

[MET 13] Towards an Ensemble Data Assimilation System of the Coupled Atmosphere- Land-Surface-Subsurface System

B. Schalge, H.-J. Hendricks-Franssen, S. Kollet, C. Simmer

Our hypothesis is that fully coupled atmosphere, land-surface and subsurface models can benefit from data cross-compartmental assimilation, which means that measurement data taken in a certain terrestrial compartment not only update the states of the same compartment, but also states of other compartments. For example, soil moisture data would not only subsurface states but also states of the lower atmosphere. In order to set up a working ensemble based data assimilation system it is necessary to investigate the sensitivity of fluxes and states of different compartments with respect to states and parameters of (other) compartments. This provides additional insights in the potential of weakly or fully coupled data assimilation to further improve characterization of states in the atmosphere-land surface-subsurface system. In this study we present preliminary results from 16 members of a 32-member ensemble where for 8 members soil properties and for 8 other ensemble members vegetation properties were varied. The impact of the variation of these soil and vegetation properties on different states and fluxes of the fully coupled model system were analyzed. The Terrestrial Systems Modeling Platform (TerrSysMP) is used to realize the simulation for a catchment which mimics the Neckar river catchment located in SW-Germany. The spatial resolution of the atmospheric compartment of TerrSysMP is 2.8km while the land-surface and subsurface run at 800m resolution. Even at this relatively low resolution substantial computing recourses are needed to realize the one year time series we aimed for. We found that without having different atmospheric forcings to consider the differences between the members were relatively small for many states and fluxes, except for Transpiration, sensible heat flux and convective precipitation in the summer. Once the analysis with the full ensemble is complete it can be decided if feedbacks are still strong enough to warrant developing a coupled data assimilation system. It is found that the impact of uncertain vegetation states on land-atmosphere exchange fluxes is larger than the role of uncertain soil properties. In addition, in summer the role of uncertain soil and vegetation properties is larger than in winter. Furthermore, feedbacks from the surface and subsurface may be more impactful if larger domains, such as continental scales are considered.

[MET 14] Terrestrial Systems Numerical Simulations in the Framework of Jülich Research on Exascale Cluster Architectures (JURECA)

P. Shrestha, M. Sulis, S. Poll, C. Simmer, S. J. Kollet

This study presents an overview of numerical simulations carried out using a novel integrated hydrological modeling platform (TerrSysMP) in the framework of Jülich Research on Exascale Cluster Architectures (JURECA). The simulations are performed over the North-Rhine Westphalia domain located in West Germany using different model configurations. A broad range of scientific issues spanning from the detailed analysis of the predictive skills of the modeling platform to scaling behavior of both atmospheric and land surface processes are presented.

[MET 15] Isoprene Secondary Organic Aerosol

S. Stadtler, T. Kühn, S. Schröder, D. Taraborrelli, H. Kokkola, M. Schultz

Within the framework of the global chemistry climate model ECHAM-HAMMOZ a novel explicit coupling between the sectional aerosol model HAM-SALSA and the chemistry model MOZ was established to form isoprene derived secondary organic aerosol (iSOA). Isoprene oxidation in the chemistry model MOZ is described by a semi-explicit scheme consisting of 147 reactions, embedded in a detailed atmospheric chemical mechanism with a total of 779 reactions. Low volatile compounds (LVOC) produced during isoprene photooxidation are identified and explicitly partitioned by HAM-SALSA. A group contribution method was used to estimate their evaporation enthalpies and corresponding saturation vapor pressures, which are used by HAM-SALSA to calculate the saturation concentration of each LVOC. With this method, every single precursor is tracked in terms of condensation and evaporation in each aerosol size bin. This approach leads to the identification of ISOP(OOH)2 as a main contributor to iSOA formation. Further, reactive uptake of isoprene epoxidiols (IEPOX) and isoprene derived glyoxal were included as iSOA sources. The parameterization of IEPOX reactive uptake includes a dependency on aerosol pH value. This model framework connecting semi-explicit isoprene oxidation with explicit treatment of aerosol tracers leads to a global, annual isoprene SOA yield of 16% relative to the primary oxidation of isoprene by OH, NO3, and ozone. With 445 Tg (392 TgC) isoprene emitted, an iSOA source of 148 Tg (61 TgC) is simulated. The major part of iSOA in ECHAM-HAMMOZ is produced by IEPOX (24.4 TgC) and ISOP(OOH)2 (28.3 TgC). The main sink process is particle wet deposition which removes 143 Tg (59 TgC). The iSOA burden reaches 1.6 Tg (0.7 TgC) in the year 2012.

[MET 16] Element Transport in Aqueous Fluids in the Interior of the Earth: New Insights from Ab Initio Molecular Dynamics Simulations

J. Stefanski, S. Jahn

Aqueous fluids are important agents in metasomatic, magmatic and hydrothermal processes in the interior and the surface of the Earth. The excellent solvent properties of water at ambient conditions depends on its extremely high dielectric constant [9]. The dielectric constant of aqueous fluids in the Earth’s crust increases with pressure but decreases much stronger with temperature [7]. Nevertheless, these fluids will never be pure water and carry different metal cations, which form solute species. The capacity of an aqueous fluid to host a certain element or to dissolve a certain amount of a phase component depends on the chemical potential of this solute species [1]. Thus, knowledge of the interplay between the concentration of molecular species in aqueous fluids and their thermodynamic properties is required to interpret, understand and eventually predict the behavior of fluids in various geological environments. Besides the lack of thermodynamic and speciation data for complex fluids at high temperatures, T, and pressures, P, the existing thermodynamic models have clear limitations [2, 3]. The combination of ab initio molecular dynamics simulations with advanced sampling methods such as constrained dynamics or metadynamics allows to predict thermodynamic properties of solute species without any empirical assumptions [4]. In the framework of the HPO15 project, we investigate the transport of beryllium in fluorine- rich late magmatic fluids. In this study, we are able to reproduce experimental stability constants of Be-F complexes at room temperature and interpolate our data in pressure- temperature space to gain new thermodynamic data. From this data, we conclude that at high fluoride activity BeF2 (aq) and BeF3-(aq) are the majority species. With the same approach, we started to compute stability constants of rare earth element species in highly saline solutions as they occur, e.g., in the Earth’s deep crust [8, 5]. Moreover, by using the metadynamics approach we reproduce the acid dissociation constant (pKa) of hydrofluoric acid at room T and make predictions for higher P/T conditions. These results allow us to constrain the availability of fluoride ligands in different geological environments, e.g., in subductions zones at relevant pH conditions [4].

[1] Anderson G. (2009) Thermodynamics of Natural Systems. 2nd ed., Cambridge University Press, Cambridge. [2] Dolejš D. (2013) Thermodynamics of Aqueous Species at High Temperatures and Pres- sures: Equations of State and Transport Theory. Rev. Mineral. Geochem. 76, 35–79. [3] Driesner T. (2013) The Molecular-Scale Fundament of Geothermal Fluid Thermodynamics. Rev. Mineral. Geochem. 76, 5–33. [4] Galvez M. E., Connolly J. A. D. and Manning C. E. (2016) Implications for metal and vola- tile cycles from the pH of subduction zone fluids. Nature 539, 420–424. [5] Keppler H. (2017) Fluids and trace element transport in subduction zones. Am. Mineral. 102, 5–20. [6] Marx D. and Hutter J. (2012) Ab Initio Molecular Dynamics: Basic Theory and Advanced Methods., Cambridge University Press. [7] Sverjensky D. A., Harrison B. and Azzolini D. (2014) Water in the deep Earth: The dielec- tric constant and the solubilities of quartz and corundum to 60 kb and 1200 °C. Geochim. Cosmochim. Acta 129, 125–145. [8] Tsay A., Zajacz Z. and Sanchez-Valle C. (2014) Efficient mobilization and fractionation of rare-earth elements by aqueous fluids upon slab dehydration. Earth Planet. Sci. Lett. 398, 101–112. [9] Wagner W. and Pruß A. (2002) The IAPWS Formulation 1995 for the Thermodynamic Properties of Ordinary Water Substance for General and Scientific Use. J. Phys. Chem. Ref. Data 31, 387–535.

[MET 17] Performance Analysis and Simulations for the Project "High Definition Clouds and Precipitation for advancing Climate Prediction (HD(CP)²)"

C. I. Meyer, O. Stein, C. Nam, S. Kneifel, S. Crewell, O. Sourdeval, J. Kretzschmar, J. Quass, L. Hoffmann

The High Definition Clouds and Precipitation for advancing Climate Prediction (HD(CP)² ) project is an initiative coordinated by BMBF to improve our understanding of cloud and precipitation events, to increase the quality and number of available observations, and to evaluate and improve modeling capabilities. The major aim of HD(CP)² is to enable the newly developed ICOsahedral Nonhydrostatic (ICON) climate and numerical weather prediction model to run on very high resolutions to be able to overcome a suite of parameterisation challenges. By synthesizing the observational and modeling approaches HD(CP)² advances both basic understanding of the role of clouds and precipitation processes in climate, as well as how to better represent clouds and precipitation in lower resolution global circulation models. The computing time of this project is used to execute simulations requested by the scientific community of the HD(CP)² project. Here, we will highlight a few examples:

1) The response of clouds to anthropogenic forcings: The response of clouds to anthropogenic forcings, such as increased carbon dioxide concentrations, is the main cause for diversity among climate models simulating global climate change. Recent studies have shown that a considerable portion of the climate sensitivity, approximately one third, is due to rapid adjustments of clouds. To tackle this problem two simulations of May 2, 2013 were performed, where the CO2 was quadrupled in one simulation and the change in cloud properties are evaluated. 2) Cloud microphysics: In a simulation for November 24, 2015, a widespread Atlantic frontal system can be seen which passed Germany from the North-West caused by a low pressure system over Island. The descending ice clouds associated with the warm front showed sublimating ice and snow on the cloud bottom which transformed over the day into light drizzle. During the second half of the day the precipitation on the ground transformed from rain to snowfall. Overall, this winter case includes a multitude of relevant cold cloud microphysical processes. Intensive observations at the JOYCE and TRIPEX campaigns (Bonn, Karlsruhe, Köln) exist for this day which makes it a "golden case" to evaluate the ICON performance for cold cloud processes and winter precipitation. 3) Cloud response to aerosols: The cloud response to aerosol is highly uncertain, yet a very important climate forcing mechanism. It is in particular unclear how cloud microphysical processes and cloud dynamics respond to perturbations in cloud condensation nuclei (CCN) perturbations. The response strongly depends on the way cloud- and precipitation microphysical processes are implemented in the model.

[MET 18] Physics of Rotating Convection in Planetary and Stellar Interiors

S. Stellmach, U. Hansen

Turbulent rotating convection occurs in many geo- and astrophysical bodies, but it is still far from being well understood. In this poster contribution, we present recent findings obtained using the JUQUEEN system. Our simulations reveal that tiny viscous boundary layers, so- called Ekman layers, are much more important in rapidly rotating convection than previously thought. We further show that the convective heat transfer is accurately predicted by asymptotic reduced models of rotationally constrained convection. We also discuss evidence for upscale kinetic energy transport generating large-scale, coherent structures in the turbulent regime of rotating convection. Finally, we briefly consider the geo- and astrophysically relevant case of convective systems in which the fluid parcels in the deeper parts of the convective region get compressed significantly by the weight of the overlying fluid. We show that in the presence of rotation, such compressibility effects can drive alternating jets similar to those observed on Jupiter and other giant planets.

[MET 19] Formaldehyde Oxidation in Presence of Clouds

B. Franco, D. Taraborrelli, A. Kiendler-Scharr, A. Wahner

Formic acid plays key roles in atmospheric aqueous-phase chemistry. It facilitates cloud droplet activation and significantly contributes to cloud and rainwater acidity. Current knowledge of its sources is deficient as state-of-the-art models severely underestimate its atmospheric burden. Accommodating either a large direct emission or photochemical production in models to match the observed concentrations would imply a profound revision of the reactive carbon oxidation schemes and suggests that relevant key processes are yet to be identified. Here, we show how the kinetically limited dehydration of methanediol in cloud droplets leads to a large and pervasive gas-phase production of formic acid. The magnitude of this newly identified source is a sextuple of the known sources and allows to reconcile model predictions with remote-sensing measurements. The efficient conversion of formaldehyde to formic acid significantly lowers the carbon monoxide yield from methane oxidation which agrees with isotope enabled inversion studies. This also bears implications for top-down estimates of formaldehyde and isoprene emissions. The increased atmospheric burden of formic acid decreases the cloud and rainwater pH by ~0.2 and ~0.5 in remote and isoprene-dominated environments, respectively. [MET 20] More Homogeneous European Wind Conditions Under Strong Climate Change

J. Wohland, J. Weber, D. Witthaut

Avoiding dangerous climate change in line with the Paris Agreement requires massive modifications of our energy infrastructure. Energy related emissions have to reach zero within a couple of decades posing an unprecedented challenge to energy system operation [Rogelj et al., 2015]. These strong emission reductions require usage of renewable energies, most prominently wind and solar, that are governed by the weather. However, it is thinkable that the Paris targets are not met despite substantial additions of renewable generators. To ensure robustness of the mitigation pathway, we therefore explore the vulnerability of wind-driven power systems to strong climate change (rcp 8.5) and find increasing homogeneity of the wind resource that will partly impede international balancing of electricity.

Astrophysics

[A 1] Planet Formation with High-Resolution Hydrodynamic Simulations

H. Klahr, H. Baehr, A. Schreiber, N. Manger

Planet formation occurs in the rotating disks around young stars, on many different lengthscales and timescales. To model the processes which may lead to planetary bodies, we use hydrodynamic simulations to resolve the gravitational collapse on various physical domains. To adequately capture the necessary physical effects, we require the high-resolution and high-performance resources provided by the Jülich Supercomputing Centre. We present large-scale simulations of self-gravitating disks which form Jupiter-sized planets in the cool outer disk and smaller-scale simulations triggered gravitational collapse of solid material which are concentrated through disk instabilities such as the streaming instability and the vertical shear instability. With these simulations we aim to constrain the planetesimal distribution which is a crucial parameter to understanding the formation of planets in the Universe.

[A 2] Progenitor Systems of Thermonuclear Supernovae

W. Hillebrandt, M. Bulla, M. Kromer, R. Pakmor, F. K. Röpke, M. Schrauth, S. A. Sim

Despite the importance of Type Ia supernovae for modern astrophysics, their detailed explosion mechanism is not fully understood. We present recent findings from numerical models in the context of observed diversity and discuss how these models can help to shed light on the explosion mechanism and to identify their progenitor systems. It is shown that the models have considerable predictive power and they can aid the identification of sources of systematic errors when used as distance indicators for cosmology.

[A 3] Dynamical Evolution of Dense Star Clusters with and without Central Black Holes

R. Spurzem, T. Panamarev, P. Berczik, A. Just, M. A. Sedda, P. Assmann

A large fraction of galaxies show evidence of supermassive black holes (henceforth SMBH) residing in their center. They are typically embedded in nuclear star clusters (NSC); if resolution allows to observe the NSCs, they are among the densest clusters known. Their size is similar to galactic globular clusters, but they are much heavier and brighter. Their interaction with the central massive black hole is through gravitational (Newtonian and Post- Newtonian) gravitational forces and through tidal disruption of stars which get too close to the black hole. Both processes provide key information for current and future observations in gravitational wave and electromagnetic bands. We are modelling the relaxation driven evolution of a nuclear star cluster together with its growing massive central black hole, the current model has been advanced to some 5 Gyrs and is going on. We currently follow the first generation of older stars in our model with one million bodies and already see some evidence how the paradox of old stars is resolved, they still form a cusp in the center, and for the first time we get a comprehensive and consistent picture of the evolution and distribution of all objects in the galactic center: binaries, compact and exotic objects (white dwarfs, neutron stars and black holes), evolving stars with tidal disruption depending on stellar parameters. Our ongoing simulation in this project is the only one currently with such a high resolution, performance and particle number, and it is only possible using the massive GPU acceleration available.

Computer Science and Numerical Mathematics

[INF 1] Performance Evolution of FLEUR

U. Alekseeva, G. Michalicek, D. Wortmann

The advances in performance optimisation of the all-electron DFT code FLEUR [1] are presented. The improved implementation includes increased modularity of the code, reduced I/O, hybrid MPI/OpenMP parallelisation, additional interfaces to external libraries providing performance portability and provides a significantly improved users experience. This new version has been developed within the European Center of Excellence MaX [2]. Due to the performance boost of the current version of the FLEUR code simulations with unit cells of more than 1000 atoms are now feasible. On our poster we will discuss these changes, the basic algorithms, and the scaling of the code on different architectures.

[1] http://www.flapw.de. [2] http://www.max-center.eu.

[INF 2] Examples of Massively Parallel Non-Numerical Algorithms

M. Axtmann, S. Lamm

We present our work on algorithm engineering for massively parallel discrete problems. In particular, we investigate distributed memory parallel sorting algorithms, as well as distributed memory graph generators. All of our algorithms are both communication-efficient and provide a very good scaling behavior. First, we present our sorting algorithms that scale to the largest available machines and are robust with respect to input size, duplicate keys, and distribution of the input elements. The main outcome are four robust sorting algorithms that cover the entire range of possible input sizes: For very small inputs, we propose a simple algorithm that trades of work for latency. For small inputs, we devise new low overhead mechanisms to make hypercube quicksort robust with respect to duplicate keys and skewed input distributions. Finally, for medium and large sized inputs, we present our robust and very scalable multi-level generalization of sample sort. Next, we present a set of massively parallel graph generators for a variety of commonly used network models. Network generators serve as a tool for building scalable algorithms on a petabyte scale by providing synthetic instances with controllable parameters. In particular, we target the classic Erdos-Renyi model and various spatial network models including Random Geometric Graphs, Random Delaunay Graphs and Random Hyperbolic Graphs. By making use of pseudorandomization and divide-and-conquer schemes, all of our generators follow a communication-free paradigm, i.e. they require no communication. In addition to performance guarantees derived by asymptotic analysis, we present an extensive experimental evaluation of our algorithms. In our sorting benchmark, we compared 7 algorithms on 10 input distributions with varying input sizes over 9 orders of magnitude and used up to 262 144 cores. For all but the largest input sizes, we are the first to perform experiments on such large machines at all and our algorithms significantly outperform the ones one would conventionally have considered. Our graph generators do not only rival the sequential performance of current state-of-the-art generators, but also provide very good weak- and strong-scaling behavior. Due to our communication-free paradigm, we are able to generate instances of up to 243 vertices and 247 edges in less than 22 minutes on 32 768 cores. This is comparable to instances generated in the Graph500 benchmark and thus allows new graph families to be used on an unprecedented scale. [INF 3] Large Scale Simulations of Continuum Models Using Parallel Geometric Multigrid Methods

S. Reiter, A. Vogel, G. Wittum

Two simulations of continuum models on the JUQUEEN supercomputer are presented: permeation of substances into the human skin as an application from computational biology and density driven subsurface flow from the field of computational geosciences. Employing geometric multigrid methods and carefully designed refinement and distribution strategies, we demonstrate the applicability and efficiency of our simulation approaches in massively parallel computations on up to 262,144 processing entities.

[INF 4] FE2TI: Computational Scale Bridging for Dual-Phase Steels

A. Klawonn, S. Köhler, M. Lanser, O. Rheinbach, M. Uran

Advanced High Strength Steels (AHSS) provide a good combination of both strength and formability and are therefore applied extensively in the automotive industry, especially in the crash relevant parts of the vehicle. Dual-phase (DP) steel is an example for such AHSS which is widely employed. The excellent macroscopic behavior of this steel is a result of the inherent micro-heterogeneity and complex interactions between the ferritic and martensitic phases in the microstructure. Thus, considering the microscale is indispensable for realistic simulations. We present our computational scale bridging software package FE2TI, which is based on the FE2 homogenization approach. The FE2 approach splits the computations into a comparably small macroscopic problem, which does not consider the microscopic structure of the dual- phase steel, and many microscopic problems on representative volume elements (RVEs). The latter ones are independent of each other and can be solved in parallel. In FE2TI we use highly scalable nonlinear FETI-DP approaches to solve the RVEs. On this poster, we summarize our efforts of the last four years, including scalability to the complete JUQUEEN for model problems and for a production setup, mechanical results for large production runs using unstructured RVEs and J2-elasto-plasticity models, a comparison of different RVEs, and finally a recently added parallelization approach for the linearized macroscopic problem using BoomerAMG. The latter modification leads to an improved overall weak scalability of FE2TI and enables the simulation of even larger macroscopic problems.

[INF 5] Towards Automatic Generation of Energy-Aware Efficient Geometric Multigrid Solvers

L. Claus, A. Grebhahn, S. Kronawitter, S. Kuckuk, H. Rittich, C. Schmitt

Many problems in computational science and engineering require the numerical solution of large, sparse linear systems of equations that arise from the discretization of partial differential equations. Multigrid is known to be one of the most efficient methods for this purpose. However, the specific multigrid algorithm and its implementation depend on the underlying problem and hardware. Project ExaStencils' goal is a compiler and code generation framework capable of generating automatically highly parallel and efficient geometric multigrid solvers. Given an abstract description of the problem and the target hardware, the approach is to choose from a family of options the most suitable composition of variants. We have already been able to showcase a full compilation flow for the investigated class of elliptic partial differential equations for real, as of yet homogeneous, supercomputers such as JUQUEEN. We provide details of our approach as well as an overview of the different components of our compiler framework. [INF 6] Superconducting Flux Qubits Compared to Ideal Two-Level Systems as Building Blocks for Quantum Annealers

M. Nocon, D. Willsch, F. Jin, H. De Raedt, K. Michielsen

For quantum computers, two theoretical models are nowadays considered to be the most important: the gate-based quantum computer and the quantum annealer. Gate-based quantum computers are based on computational gates just like classical computers, but have potentially more computational power due to the algebra behind quantum theory. A quantum annealer works fundamentally different: First the system is prepared in a known ground state of an initial Hamiltonian, then this Hamiltonian is adiabatically transformed into the final Hamiltonian whose ground state corresponds to the solution of a given problem, usually taken from the class of optimization problems. Quantum annealing works well in theory if the qubits can be modeled as two-level systems. However, in real devices, the qubits are not based on perfect two-level systems, but on a two- dimensional subspace of a larger system. This makes approximations in analytic calculations unavoidable. With a simulation utilizing the Suzuki-Trotter product-formula approach to solve the time- dependent Schrödinger equation, the time-evolution of the full state of such a device based on superconducting flux qubits is investigated.

[INF 7] A Multigrid Accelerated Eigensolver for the Hermitian Dirac Operator in Lattice QCD

A. Frommer, K. Kahl, M. Rottmann, A. Strebel

Computing physical observables in Lattice QCD is in one way or the other oftentimes connected to eigenvalues of the Hermitian Dirac Operator 푄. Examples are the pion and eta- meson correlators, which need small eigenvalues of 푄 to improve the signal-to-noise ratio. In this work we present a Davidson type eigensolver for the Hermitian Dirac Operator. As a preconditioner we used our successful DD훼-AMG multigrid method. Within this framework, we incorporated several modifications specifically tailored to the Hermitian Dirac Operator and the multigrid method. We employed a strategy which introduces a synergy between the preconditioner and the Davidson method, allowing for a significantly better eigenvalue scaling. We show numerical results showing the impact of these modifications and compare with the common and state-of-the-art eigensolvers PARPACK and PRIMME.

[INF 8] Error Analysis of Gate-Based Quantum Computers with Transmon Qubits

D. Willsch, M. Nocon, F. Jin, H. De Raedt, K. Michielsen

Over the past decades, tremendous effort has gone into building a universal quantum computer. In theory, such a device can solve certain problems such as factoring exponentially faster than digital computers. The leading technological prototypes are based on superconducting circuits and contain about 10-50 qubits. Controlling these fragile systems requires an enormous amount of precision, posing a difficult challenge for the experimentalists. We study such quantum systems in detail by solving the time-dependent Schrödinger equation for a generic model Hamiltonian. For this purpose, we have developed efficient product-formula algorithms that are tailored to key features of the model Hamiltonian. This allows us to simulate each individual voltage pulse that is used in experiments to realize a certain quantum gate, as dictated by the computational model of a quantum computer. By optimizing the pulse parameters, we find that even in the ideal case, the best pulses still contain undesirable errors in the realization of the intended quantum gate. The common gate metrics measured and reported in experiments or computed in theory are shown to provide insufficient practical information about the significance of these errors.

Scientific Big Data Analysis

[SDBA 1] Supporting Cytoarchitectonic Mapping on Histological Brain Sections using Transfer-Learning with Convolutional Neural Networks

C. Schiffer, H. Spitzer, K. Kiwitz, K. Amunts, T. Dickscheid

Cytoarchitectonic mapping of cortical areas is a key aspect in creating a multimodal brain atlas. Currently used semi-automatic methods to detect cortical boundaries are precise, but insufficient to handle the steadily increasing quantity of histological brain sections. This motivates the development of an automated approach. To this end, a Convolutional Neural Network (CNN) was trained, which can automatically segment 13 cortical areas across different brains. We try to improve the accuracy of the above CNN by focusing on just one specific area in a few, spatially close sections. Knowing that spatially close sections share a similar texture and geometric structure, we simplify the objective in this way and expect to achieve better results.

Participants

Name First name Institution Poster Adjaoud Omar TU Darmstadt MAT 1 Ahmad Sabahuddin Universität Düsseldorf BIO 1 Alekseeva Uliana Forschungszentrum Jülich INF 1 Ali Sajid Universität Münster Alidadi Soleymani Fatemeh Forschungszentrum Jülich POLY 1 Alleva Claudia Forschungszentrum Jülich BIO 2 Ansorge Cedrick Universität Köln MET 1 Arnold Lukas Forschungszentrum Jülich SE 1 Athanasopoulos Nikolaos Karlsruher Institut für Technologie MET 2 Atodiresei Nicolae Forschungszentrum Jülich KM 1 Attig Norbert Forschungszentrum Jülich Axer Markus Forschungszentrum Jülich Axtmann Michael Karlsruher Institut für Technologie INF 2 Baehr Hans MPI für Astronomie A 1 Berndt Jonas Forschungszentrum Jülich MET 3 Bihlmayer Gustav Forschungszentrum Jülich Binder Kurt Universität Mainz Blume Martin Universität Bochum ST 1 Böckmann Marcus Universität Münster Bode Mathis RWTH Aachen University ST 2 Boeck Thomas TU Ilmenau ST 3 Bogner Simon Forschungszentrum Jülich ST 4 Bonus Michele Universität Düsseldorf BIO 3 Brahami Yacine CORIA ST 5 Breit Markus Universität Frankfurt INF 3 Brinker Sascha Forschungszentrum Jülich KM 2 Bücker Oliver Forschungszentrum Jülich BIO 4 Burstedde Carsten Universität Bonn Büscher Tobias Forschungszentrum Jülich POLY 2 Caciuc Vasile Forschungszentrum Jülich KM 3 Caldeira Liliana Forschungszentrum Jülich BIO 5 Cali Salvatore Universität Wuppertal E 1 Chitgar Zahra Forschungszentrum Jülich PLA 1 Dabrowski Jarek IHP MAT 2 Denev Jordan Karlsruher Institut für Technologie ST 6 Diaz Sandra Forschungszentrum Jülich Dickscheid Timo Forschungszentrum Jülich Name First name Institution Poster Doltsinis Nikos Universität Münster KM 4 dos Santos Dias Manuel Forschungszentrum Jülich Ehrlich Jannis Forschungszentrum Jülich KM 5 Elbern Hendrik Forschungszentrum Jülich Elgeti Jens Forschungszentrum Jülich BIO 6 Fehling Marc Forschungszentrum Jülich Fiolitakis Andreas Deutsches Zentr. f. Luft- und Raumfahrt Fischer Thomas Helmholtz Centre for Environm. Res. Fodor Zoltan Universität Wuppertal Franke Philipp Forschungszentrum Jülich MET 4 Frieg Benedikt Forschungszentrum Jülich BIO 7 Frings Wolfgang Forschungszentrum Jülich Frommer Andreas Universität Wuppertal Froning Dieter Forschungszentrum Jülich ST 7 Funck Carsten RWTH Aachen University MAT 3 Furusho Carina Forschungszentrum Jülich MET 5 Garcia Rondina Gustavo TU Darmstadt POLY 3 Gauding Michael CORIA ST 8 Gerstner Erwin Universität Duisburg-Essen Gertzen Christoph G. W. Forschungszentrum Jülich BIO 8 Ghorbani Elaheh TU Darmstadt MAT 4 Gibbon Paul Forschungszentrum Jülich Goergen Klaus Forschungszentrum Jülich MET 6 Gohlke Holger Universität Düsseldorf BIO 9 Grebhahn Alexander Universität Passau Green Jeremy DESY E 2 Gregory Eric B. Universität Wuppertal Griesel Alexa Universität Hamburg MET 7 Grotendorst Johannes Forschungszentrum Jülich Grynko Yevgen Universität Paderborn SE 4 Guenther Jana N. Universität Regensburg E 3 Haas Sarah Forschungszentrum Jülich Hebeler Kai TU Darmstadt AK 1 Hendricks-Franssen Harrie-Jan Forschungszentrum Jülich MET 8 Herff Sohel RWTH Aachen University ST 9 Hillringhaus Sebastian Forschungszentrum Jülich BIO 10 Höfler-Thierfeldt Sabine Forschungszentrum Jülich Honerkamp Carsten RWTH Aachen University KM 6 Hoore Masoud Forschungszentrum Jülich POLY 4 Hsu Hsiao-Ping MPI für Polymerforschung Name First name Institution Poster Hüter Claas Forschungszentrum Jülich MAT 5 Huysegoms Marcel Forschungszentrum Jülich Inghirami Gabriele Universität Frankfurt AK 2 Ippoliti Emiliano Forschungszentrum Jülich Jahn Sandro Universität Köln Jalas Soeren Universität Hamburg Janetzko Florian Forschungszentrum Jülich Ji Yaqi Forschungszentrum Jülich MAT 6 Jin Fengping Forschungszentrum Jülich KM 7 Jing Runyu Forschungszentrum Jülich BIO 11 Jones Robert Forschungszentrum Jülich Kamps Martina Forschungszentrum Jülich Kirchen Manuel Universität Hamburg PLA 2 Kiwitz Kai Forschungszentrum Jülich Klawonn Axel Universität Köln INF 4 Knechtli Francesco Universität Wuppertal Koch Erik Forschungszentrum Jülich Kollet Stefan Forschungszentrum Jülich Korzec Tomasz Universität Wuppertal Kremer Manfred Krieg Stefan Forschungszentrum Jülich Kronawitter Stefan Universität Passau INF 5 Kühne Thomas Universität Paderborn Lima Fernandes Imara Forschungszentrum Jülich KM 8 Lamm Sebastian Karlsruher Institut für Technologie Lanser Martin Universität Köln Li Feng Forschungszentrum Jülich MAT 7 Lippert Thomas Forschungszentrum Jülich Liseykina Tatyana Universität Rostock PLA 3 Liu Benqiong Forschungszentrum Jülich KM 9 Lounis Samir Forschungszentrum Jülich Machtens Jan-Philipp Forschungszentrum Jülich BIO 12 Mayrhofer Leonhard Fraunhofer IWM MAT 8 McGovern Sean Forschungszentrum Jülich MET 9 Meinke Jan Forschungszentrum Jülich Mellado Juan Pedro MPI für Meteorologie MET 10 Menichetti Roberto MPI für Polymerforschung POLY 5 Menzel Miriam Forschungszentrum Jülich BIO 13 Michielsen Kristel Forschungszentrum Jülich Müller Marcus Universität Göttingen Name First name Institution Poster Münster Gernot Universität Münster E 4 Müser Martin Universität des Saarlandes Naz Bibi Sarwat Forschungszentrum Jülich MET 11 Nguyen Luan Bochum Univ. of Appl. Sciences SE 2 Nielaba Peter Universität Konstanz KM 10 Nocon Madita Forschungszentrum Jülich INF 6 Orth Boris Forschungszentrum Jülich Otte Philipp Forschungszentrum Jülich Pandey Ambrish TU Ilmenau ST 10 Pastewka Lars Universität Freiburg Pesch Andreas Universität Bochum ST 11 Pfleger Christopher Universität Düsseldorf BIO 14 Philipsen Owe Universität Frankfurt E 5 Pieperhoff Peter Forschungszentrum Jülich BIO 15 Pietrzyk Uwe Forschungszentrum Jülich Pitsch Heinz RWTH Aachen University Poulsen Mads Bruun Universität Copenhagen Pronold Jari Forschungszentrum Jülich Retsch Matthias Heinz MPI für Meteorologie MET 12 Rheinbach Oliver TU Freiberg Rivas Nicolas Forschungszentrum Jülich POLY 6 Rohlfing Michael Universität Münster KM 11 Röpke Friedrich Universität Heidelberg A 2 Rosanka Simon Forschungszentrum Jülich Rossetti Giulia Forschungszentrum Jülich Schalge Bernd Universität Bonn MET 13 Scheurer Christoph TU München Schiffer Christian Forschungszentrum Jülich SBDA 1 Schiller Arwed Universität Leipzig Schröder Sabine Forschungszentrum Jülich Schug Alexander Forschungszentrum Jülich BIO 16 Schween Nils Universität Heidelberg ST 12 Shrestha Prabhakar Universität Bonn MET 14 Souza M. Guimaraes Filipe Forschungszentrum Jülich KM 12 Spatschek Robert Forschungszentrum Jülich Spitzer Hannah Forschungszentrum Jülich Spurzem Rainer Universität Heidelberg A 3 Stadtler Scarlet Forschungszentrum Jülich MET 15 Stefanski Johannes Universität Köln MET 16 Stein Olaf Forschungszentrum Jülich MET 17 Name First name Institution Poster Stellmach Stephan Universität Münster MET 18 Strebel Artur Universität Wuppertal INF 7 Suarez Estela Forschungszentrum Jülich Svystun Elena DESY PLA 4 Taraborrelli Domenico Forschungszentrum Jülich MET 19 Terhorst Dennis Forschungszentrum Jülich BIO 17 Toth Balint Universität Wuppertal E 6 Trautmann Alexander Forschungszentrum Jülich Trebst SImon Universität Köln Tsukamoto Shigeru Forschungszentrum Jülich KM 13 Urbach Carsten Universität Bonn Uttinger Maximilian Universität Erlangen-Nürnberg SE 3 Varnhorst Lukas Universität Wuppertal E 7 Vinayak Ashish D. Universität Wuppertal Vogel Annika Forschungszentrum Jülich Voigt Axel TU Dresden BIO 18 Voigtmann Thomas Deutsches Zentr. f. Luft- und Raumfahrt Weckbecker Dominik Universität Erlangen-Nürnberg CH 1 Wendling Loïc RWTH Aachen University ST 13 Wenschuh Ulrich Deutsche Post AG Wick Christian Universität Erlangen-Nürnberg MAT 9 Wiesen Sven Forschungszentrum Jülich Willsch Dennis Forschungszentrum Jülich INF 8 Wirth Roland TU Darmstadt AK 3 Wittig Hartmut Universität Mainz Wohland Jan Forschungszentrum Jülich MET 20 Wolf Dietrich Universität Duisburg-Essen Wong Chik Him Universität Wuppertal E 8 Yin Xiaoyan Forschungszentrum Jülich MAT 10 Zimmermann Olav Forschungszentrum Jülich

This list has been drawn up for the participant's personal information. It may not be passed on to third parties or used for any other purpose.