Skip to main content



Earth Sciences

High-Resolution Studies of Transport Processes in the Atmospheric Boundary Layer Using the Synergy of Large Eddy Simulation and Measurements of Advanced Lidar Systems

On poorly resolved and sub-grid scales, turbulent processes may have significant impact on processes that are well represented on the computational grid. Even though the scales of motions of primary interest are generally much larger than those of turbulent motions, the turbulence profoundly affects all scales due to nonlinear interactions.

Tijana Janjić, Volker Wulfmeyer

High Resolution Climate Change Simulation for Central Europe

Two regional climate simulations are performed with a high-resolution limited area model for Central Europe, representing present-day climate conditions and a future climate change scenario according to a doubled global CO2-concentration. Results of a global climate change simulation are used to initialize the regional model and to drive the simulations via time-dependent boundary values. The regional simulations show considerable changes in temperature and precipitation with noticeable geographical and seasonal modifications. For Germany, a significant warming of up to 4K emerges but the simulated tendency of decreasing summer rainfall cannot be approved as significant.

Klaus Keuler, Alexander Block, Eberhard Schaller

Water on Mars

A milestone in the understanding of the Martian hydrological evolution and water cycle is being marked by the Gamma-Ray Spectrometer (GRS) onboard Mars Odyssey ([1], [2], [4]). These gamma rays are supposed to be produced during inelastic collision and/or capture of secondary neutrons which were produced in the soil by cosmic rays. The signals are indicative of the presence of a large amount (at least 35 wt %) of hydrogen below a dry layer in this region. The Neutron Spectrometer [2] and the High Energy Neutron Detector (HEND) [4] measured leakage fluxes of fast, epithermal and thermal neutrons produced in the same way. As hydrogen has a strong ability to slow down the neutrons, depression of neutron fluxes, especially of faster ones, can be interpreted by a large abundance of hydrogen and hence H2O. Both instruments detected a large region around the south pole southward of 60°S (with an area of 107 km2) with strongly reduced epithermal fluxes indicative of large water abundance. In general the upper 15–20 cm of the soil at low and mid latitudes was inferred to be water-poor dust layer underlain by a slightly water-richer regolith [2]. The global map of fast neutrons (indicative of the upper layer) showed a somewhat different picture, particularly those measured by HEND. The area with a large deficit of fluxes was larger at northern high latitudes (107 km2) than at southern latitudes (4 x 106 km2). In all maps the global distribution is not symmetric about the equator and some distinct longitudinal variation is also discernible.

Tetsuya Tokano

Viscosity Stratification and a 3-D Compressible Spherical Shell Model of Mantle Evolution

The viscosity stratification has a strong influence on the thermal evolution of a compressible Earth’s mantle with time-dependent internal heating. The differential equations for infinite Prandtl-number convection are solved using a three-dimensional finite-element spherical-shell method on a computational mesh derived from a regular icosahedron with 1.351.746 or, alternatively, 10.649.730 nodes. We formulate a radial viscosity profile from solid-state physics considerations using the seismic model PREM. New features of this viscosity profile are a high-viscosity transition layer beneath the usual asthenosphere, a second low-viscosity layer below the 660-km endothermic phase boundary and a considerable viscosity increase in the lower 80 % of the lower mantle. To be independent of the special assumptions of this derivation, we vary the level and the form of this profile as well as the other physical parameters in order to study their consequences on the planforms and on the convection mechanism. The effects of the two mineral phase boundaries at 410 and 660 km depth proved to be smaller than effects of the strong variation of viscosity with radius. The latter had more influence on the convective style than all other parameters. Values of our material parameters are time independent and constant in the lateral directions, except for viscosity.The focus of this paper is a variation of non-dimensional numbers as Rayleigh number, Ra, Nusselt number, Nu, the reciprocal value, Ror, of the Urey number, viscosity-level parameter, r n , etc. We explored the parameter range for special solutions. For the wide parameter range −0.5 ≤ r n ≤ +0.3, that includes our preferred viscosity profile, we obtain solutions characterized by reticular connected thin cold sheet-like downwellings. The downwellings are thinner than similar features in previous publications. They bear a resemblance to observed subducting slabs but are purely vertical. We find it remarkable that the downwellings penetrate the high-viscosity transition layer. They remain sheet-like to 1350 km depth. Below this depth they begin to lose definition but their locations are still visible at 1550 km depth. Such thin subducting sheets are notable since the viscosity is Newtonian. On the other hand, it is not surprising there are no transform-like features at the surface of the model. We compute laterally averaged heat flow at the Earth’s surface, the ratio of heat output to radiogenic heat production, Ror, the Rayleigh number and the Nusselt number as a function of time. Nu(2) denotes the temporal average of the Nusselt number of a run for the last 2000 Ma of the evolution, Ra(2) is the temporal average of the Rayleigh number, respectively. For a wide parameter range, we obtain Nu(2) = 0.120 Ra(2)0.295 in this model.

Uwe Walzer, Roland Hendel, John Baumgardner



Collisional Dynamics of Black Holes, Star Clusters and Galactic Nuclei

We use high-precision direct N-body integration to study questions of the thermodynamic behaviour of dense stellar systems. The processes examined include mass segregation and equipartition processes, the study of planetary orbits in dense star clusters, and stellar orbits in galactic nuclei with thick accretion disks.

Emil Khalisi, Chingis Omarov, R. Spurzem, M. Giersz, D. N. C. Lin

Formation and Propagation of Jets Around Compact Objects

The phenomenon of jets is very common among astrophysical objects. Jets appear around supermassive black holes in radio galaxies and quasars, stellar black holes in microquasar or black hole X-ray binaries (BHXBs), around young stellar objects, but also around stellar compact sources as white dwarfs in symbiotic stars and neutron stars in low mass X-ray binaries (LMXBs). In interacting binaries containing compact objects, the secondary — a main-sequence star or red giant — loses matter through a stellar wind or Roche lobe mass overflow. This matter is forced by the gravitational field of the compact primary to form an accretion disc. On the surface of the compact object, explosive nuclear burning causes nova-like outbursts. Through interactions between the accretion disc and the magnetic field of the compact object, matter is accelerated towards the polar regions and ejected as jets. For all objects, the exact mechanism, how jets are launched and the material is accelerated, is not understood very well. A few analytic approaches are developed, but to solve this problem, time-dependant numerical magnetohydrodynamics (MHD) simulations with very high spatial resolution have to be used. Blandford & Payne [BlP82] examined the magnetically forced extraction of energy and angular momentum from the disc and found centrifugally driven outflows, if the angle between the poloidal magnetic field and the disc surface is less than 60°. With increasing distance from the star-disc-system, the toroidal field becomes important, collimating the outflow to jets. Camenzind [Cam90] discovered the first self-consistent model considering all parts of a protostellar star-disc-system.

Matthias Stute, Max Camenzind

Large Scale Simulations of Jets in Dense and Magnetised Environments

We have used the vectorised and parallelised MHD code NIRVANA on the NEC SX-5 in parallel mode to simulate the interaction of jets with a dense environment on a scale of more than 200 jet radii. A maximum performance of 0.75 GFLOP per processor could be reached.

Martin Krause, Max Camenzind

Crack Propagation in Icosahedral Model Quasicrystals

Propagation of mode-I cracks in a three-dimensional model quasicrystal is studied by molecular dynamics simulations. The samples are endowed with an atomically sharp crack and subsequently loaded by linear scaling of the displacement field. The response of the system is then monitored during the simulation. In particular, the crack surface morphology is investigated in dependence of the orientation of the fracture plane. For this purpose, fracture surfaces perpendicular to two- and fivefold axes are compared. For both directions, brittle fracture with rough fracture surfaces is observed.

Christoph Rudhart, Frohmut Rösch, Franz Gähler, Johannes Roth, Hans-Rainer Trebin

Structure and Spectrum of Poly-Porphyrin

We discuss the structural and spectroscopic properties of porphyrin-derived polymers within an ab-initio framework. The polymer is characterized by small effective masses of the relevant electronic bands, accompanied by significant electron-hole interaction. This results in a small fundamental band gap and strong optical absorption in the infrared.

Michael Rohlfing

How Do Droplets Depend on the System Size? Droplet Condensation and Nucleation in Small Simulation Cells

Using large scale grandcanonical Monte Carlo simulations in junction with a multicanonical reweighting scheme we investigate the liquid-vapor transition of a Lennard—Jones fluid. Particular attention is focused on the free energy of droplets and the transition between different system configurations as the system tunnels between the vapor and the liquid state as a function of system size. The results highlight the finite size dependence of droplet properties in the canonical ensemble and free energy barriers along the path from the vapor to the liquid in the grandcanonical ensemble.

P. Virnau, L. González MacDowell, M. Müller, K. Binder

Solid State Physics


Numerical Studies of Collective Effects in Nano-Systems

We have studied quantum effects, structures and phase transitions in Nano-systems. In the following sections an overview is given on the results of our computations on atomic wires, clusters, pore condensates, Bose fluids, elastic properties of model colloids and model colloids in external fields.

M. Dreher, D. Fischer, K. Franzrahe, G. Günther, P. Henseler, J. Hoffmann, W. Strepp, P. Nielaba

Gas-Phase Epitaxy Grown InP(001) Surfaces From Real-Space Finite-Difference Calculations

Density-functional calculations based on finite-difference discretization and multigrid acceleration are used to explore the atomic and spectroscopic properties of P-rich InP(001)(2x1) surfaces grown in gas-phase epitaxy. These surfaces have been reported to consist of a semiconducting monolayer of buckled phosphorus dimers. This apparent violation of the electron counting principle was explained by effects of strong electron correlation. Our calculations show that the (2x1) reconstruction is not at all a clean surface: it is induced by hydrogen adsorbed in an alternating sequence on the buckled P-dimers. Thus, the microscopic structure of the InP growth plane relevant to standard gas-phase epitaxy conditions is resolved and shown to obey the electron counting rule.

W. G. Schmidt, P. H. Hahn, K. Seino, M. Preuß, F. Bechstedt

Amorphous Silica at Surfaces and Interfaces: Simulation Studies

The structure of surfaces and interfaces of silica (SiO2) is investigated by large scale molecular dynamics computer simulations. In the case of a free silica surface, the results of a classical molecular dynamics simulation are compared to those of an ab initio method, the Car—Parrinello molecular dynamics. This comparative study allows to check the accuracy of the model potential that underlies the classical simulation. By means of a pure classical MD, the interface between amorphous and crystalline SiO2 is investigated, and as a third example the structure of a silica melt between walls is studied in equilibrium and under shear. We show that in the latter three examples important structural information such as ring size distributions can be gained from the computer simulation that is not accessible in experiments.

J. Horbach, T. Stühn, C. Mischler, W. Kob, K. Binder

Quantum Monte-Carlo Simulations of Correlated Bosonic and Fermionic Systems

We review recent results of quantum Monte Carlo simulations applied to correlated electronic and bosonic systems. We concentrate on three subjects. 1) Using a recently developed hybrid quantum Monte-Carlo algorithm we investigate the excitation spectra of the one-dimensional t — J model. Our results give strong numerical support for the existence of antiholons, which along with spinons and holons correspond to the elementary excitations of this model. 2) Very recently, it was experimentally demonstrated, that it is possible to attain temperatures low enough, such that degenerate quantum gases can be studied in magneto-optical traps, the most prominent example being Bose-Einstein condensation of alkali atoms. Under the action of a periodic potential created by interfering laser beams, such systems can be brought to a strongly correlated state. We present numerical simulations in one-dimension in order to understand theses new states of matter. 3) Taking the step from one to two and three dimensions poses a formidable numerical challenge. In particular for fermionic models the quantum Monte Carlo method suffers from the so-called sign problem which renders simulations exponentially expensive in CPU time as a function of inverse temperature at lattice size. We show that by considering multi-flavored models this problem is reduced and in some special cases altogether removed.

C. Lavalle, M. Rigol, M. Feldbacher, M. Arikawa, F. F. Assaad, A. Muramatsu

Ab initio Simulation of Clusters: Modeling the Deposition Dynamics and the Catalytic Properties of Pd N on MgO Surface F-Centers

Nano-catalysts are studied in an ab inito framework by solving the Kohn-Sham equations of density functional theory for the supported clusters and a finite zone of the underlying surface. An efficient and accurate numerical parallel implementation of the Kohn-Sham solver using plane waves for the kinetic energy calculations and a real space grid for the potential energy evaluations permits first principle molecular dynamics simulations of the nano-catalyst formation process namely the low-energy deposition of neutral Pd N clusters (N = 2–7 and 13) on a MgO(001) surface with oxygen vacancies (so called F-centers, FC). The main findings of this simulations are a steering effect by an attractive “funnel” due to the polarizing F-center. This results in strong adsorption of the cluster, with one of its atoms pinned atop of the FC confirming that corresponding experiments are performed with supported size-selected nano-clusters and not with larger structures grown by coalescence. Interestingly, the deposited Pd2-Pd6 clusters retain their gasphase geometries, while for N>6 the clusters adopt structures which maximize the contact area with the surface. Furthermore, we show that a large number of NO molecules can adsorbe on the low coordinated sites of the supported Pd clusters. For instance, the Pd4 was able to capture up to 5 NO in our simulations (4 on Pd-Pd bridges and one molecule on top of the tetrahedral cluster). In order to demonstrate the accuracy of our method, we report on an additional study of finite temperature photoelectron spectra for sodium cluster anions.

M. Moseler, B. Huber, H. Häkkinen, U. Landman

Reactive Flows


DNS of Turbulent Premixed CO/H2/Air Flames

A program for the direct numerical simulation (DNS) of reactive flows is presented. In favor of using detailed models for chemical kinetics and molecular transport only spatially two-dimensional simulations are performed. This scientific application code has been used as a benchmark on several platforms, including Intel Itanium 2 systems, IBM’s SMP server pSeries 690, and a Fujitsu-Siemens hpc-Line cluster with Intel Xeon CPUs. The overhead for parallelization on these systems is discussed separately from the single processor performance. Up to 512 processor elements of a Cray T3E have been used and it is shown that even more processors could be used while maintaining a high parallel efficiency. Direct simulations of flames evolving after induced ignition of a premixed CO/H2/air mixture under turbulent conditions have been carried out. Results of this investigation are presented in the application part of the paper.

Marc Lange

Transition from Stationary to Rotating Bound States of Dissipative Solitons

By the example of a cluster of two dissipative solitons, which are well localized solitary solutions of a 3-component reaction-diffusion system in 2-dimensional space, we demonstrate that in dissipative systems a bifurcation of stationary well localized structures to uniform rotating structures is possible. The underlying mechanism is similar to the mechanism of the drift (traveling) bifurcation. For appropriate choice of the path in parameter space of the considered reaction-diffusion system the rotational bifurcation precedes the drift bifurcation. The theoretically predicted velocities are compared to solutions of the reaction-diffusion system.

A. W. Liehr, A. S. Moskalenko, H.-G. Purwins

Computational Fluid Dynamics


Investigation of the Flow Randomization Process in a Transitional Boundary Layer

This work is devoted to the investigation of the late stages of the laminar-turbulent transition process in a flat-plate boundary layer without pressure gradient. Similar to a turbulent boundary layer these stages are dominated by nonlinear flow dynamics and by the occurrence of coherent vortex motions in the boundary layer, which are far from being well understood, up to now. The work consists of two parts: a description of some salient features of the numerical method, and an account of some specific results of a numerical study of the flow randomization process in K-type transition that has been performed conjointly with a wind-tunnel investigation at the Technical University of Berlin.

Daniel Meyer, Ulrich Rist, Markus Kloker

Numerical Simulation of 3D Unsteady Heat Transfer at Strongly Deformed Droplets at High Reynolds Numbers

The dependency of the heat transfer on an initial deformation of droplets has been investigated at high droplet Reynolds numbers. The two-phase flow has been computed with an inhouse 3D DNS program (FS3D) using the Volume-of-Fluid method. For the droplets initial prolate and oblate shapes with an axial approaching flow has been studied. In addition, a spherical shape has been used as reference. The initial droplet Reynolds number for the present study has been Re0 = 660 for all investigated cases. Due to the fact that the steady droplet velocity for the considered droplets has been much lower than the initial velocity of the droplets, the droplet velocity is decreased during the simulation. To gain more knowledge about the influence of deformation on the heat transfer, the time dependent, spatial averaged Nusselt number Nu t and the time and spatial averaged Nusselt number Nu m has been matched by the temperature and velocity field around a deformed droplet. By this comparison the oscillation phase with the largest heat transfer has been observed. The simulations have been performed on the Cray T3E/512-900 at the HLRS with 32 processors. The parallel performance in dependency of the number of processors has been investigated.

Matthias Hase, Bernhard Weigand

3D Simulations of Supersonic Chemically Reacting Flows

The mixing and combustion of hydrogen in a model scramjet (supersonic combustion ramjet) engine is investigated numerically. For an improved mixing a lobed strut injector is employed. It will be shown that due to the chosen shape of the injector strong streamwise vortices are induced which improve the mixing and therefore shorten the necessary combust or length. The hydrogen is injected with Mach 1.4 into a Mach 2 supersonic air flow. Computational grids with 0.43 and 3.2 million volumes have been used for these combustion simulations. Due to the complex physical phenomena in high speed flows extremely fine grids are required in the vicinity of walls. The use of finite-rate chemistry additionally causes long computational times. The chosen chemistry reaction mechanism employs 20 reactions and 9 different species. Thus efficient numerical solvers are required as well as facilities that allow high performance computing. The numerical code is parallelized by domain decomposition using MPI and the simulations shown are performed on a Cray T3E using up to 208 nodes.

Fernando Schneider, Peter Gerlinger, Manfred Aigner

Numerical Investigation of Semi-Turbulent Pipe Flow

Surface deformations of a jet ejected from a straight-pipe atomizer may be due to the turbulent fluctuations in the interior of the pipe. At relatively low Reynolds number (Re) near the transition from laminar to turbulent pipe flow (e.g. Re = 3000 based on the mean velocity and the pipe diameter), an unsteady semi-turbulent state exists which differs from the better known fully developed flow at higher Re. In the present work low- and high-Re jets are investigated experimentally by the DLR group. To identify the influence of the inner injector flow condition to the jet surface phenomena a special injector set-up has been designed. With this setup it was possible to eliminate relative velocity effects between jet and ambient fluid. Using shadowgraphy and a novel image processing approach, wavelengths, amplitudes and undisturbed jet length could be determined. The corresponding pipe flow is simulated numerically using Direct Numerical Simulation (DNS) by the University group. A new second order finite difference scheme in space and time for the incompressible Navier-Stokes equations in cylindrical coordinates is applied. The slope of the axial mean velocity profile near the wall is smaller than that for higher Reynolds number. The turbulent intensities are smaller than those at higher Re. The observed jet surface waves agree well with the computed lengths scales of the turbulent structures within the injector. Varicose instability is observed. Atomization is affected when Re is reduced to the semi-turbulent state(Re=3000), due to the thicker laminar envelope.

Emad Khalifa, Eckart Laurien

Numerical Simulation of Forced Breakup of a Liquid Jet

Even nowadays, the physics of jet breakup is still not well understood and this is the reason why a lot of effort has been put recently into the investigation of this kind of flow. Within this framework, special attention has been focused on the use of excitation sources of defined amplitude and frequency in order to understand their effect on the (possible) jet disintegration process [SCO99, CGO+00]. The lack of analytical models for describing the strong non-linear behavior of this type of flow on the one hand, and the difficulty of accessing experimentally flow features like pressure and/or velocities on the other hand, have made it necessary to develop and use numerical methods for this purpose. In the recent past, numerical methods based on the Navier-Stokes equations with the possibility of handling free-surfaces have been developed and proved to fulfill the above requirements; they also have shown to be able to reproduce qualitatively well the jet-breakup behavior observed in the experiments [AMP00]. In the work presented here, after a brief description of the numerical method used, the results of simulations of the forced breakup of a round jet of ethanol into quiescent air will be presented. Three different types of excitation have been used and their effects on the jet breakup have been analyzed. Special attention has been paid to the capture of droplet generation.

Frank-Olivier Albina, Milovan Perić

The Effect of Impinging Wakes on the Boundary Layer of a Thin-Shaped Turbine Blade

A series of Large Eddy Simulations (LES) of flow in a modern, highly loaded low-pressure turbine cascade, with and without oncoming wakes, has been performed. The computations were carried out using 64 processors on the Hitachi-SR8000 F1. The blade shape is that of newly designed turbine rotors which allow controlling the blade weight. Five different operating conditions were investigated by changing strength and frequency of the incoming wakes. The large number of simulations was made possible by a previous careful tuning of the code.

Vittorio Michelassi, Jan Wissink, Wolfgang Rodi

Numerical High Lift Research II

The project NHLRes is concerned with the simulation of aircraft aerodynamics and thus belongs to the research field of computational fluid dynamics (CFD) for aerospace applications. NHLRes comprises the numerical simulation of the viscous flow around transport aircraft high lift configurations based on the solution of the Reynolds-averaged Navier-Stokes equations. The project NHLRes II, a follow-on activity of the NHLRes project [1], consists of three parts representing a analysis of complex 3D-flow features, wake vortex simulations and an optimization task for selected three-dimensional high lift flow problems.

S. Melber-Wilkending, E. Stumpf, J. Wild, R. Rudnik

Prediction of the Model Deformation of a High Speed Transport Aircraft Type Wing by Direct Aeroelastic Simulation

The aerodynamic performance, maneuverability and flight stability of aircrafts are highly dependent on the deformation of their wings under aerodynamic loads. The accurate prediction of aeroelastic properties, such as aeroelastic equilibrium configurations under cruise conditions, is therefore crucial in early design stages. Due to increasing computer power and further development of numerical methods, direct numerical aeroelastic simulation, in which the governing equations for the fluid and the structure are solved consistently in time, has become feasible [1, 2]. In the collaborative research center SFB 401 “Flow Modulation and Fluid-Structure Interaction at Airplane Wings” at Aachen University the numerical aeroelastic method SOFIA (SOlid Fluid InterAction) for direct numerical aeroelastic simulation is being progressively developed [4, 5]. According to the coupled field approach, three modules can be identified in SOFIA: flow solver, structural solver and a grid deformation tool, which is necessary since the boundaries of the computational grid for the flow solver always have to coincide with the deforming aerodynamic surface of the structure.

C. Braun, A. Boucke, M. Hanke, A. Karavas, J. Ballmann

Rayleigh-Bénard Convection at Large Aspect Ratios

Rayleigh-Bönard convection is simulated in a plane layer with periodic boundary conditions in the horizontal directions. A spectral method allows us to reach Rayleigh numbers up to 107 even for an aspect ratio of 10.

T. Hartlep, A. Tilgner



Quantum Chemical Calculations of Transition Metal Complexes

Transition metal complexes show a wide variety of chemical reactions. To gain insight into the bonding situation of these complexes and the transition states involved in these reactions is not only crucial for understanding the underlying principles, but even more for finding new reaction pathways or optimizing reaction conditions in chemical industry. Where experiments fail to obtain the needed results, modern quantum chemical approaches can be utilized to investigate chemical systems and predict their properties. This is a challenging task for computational chemists and demanding in computational resources. Such calculations have been carried out in order to predict geometries, bond energies, and Lewis basicity of various transition metal complexes. A number of projects are still in progress which are including CCSD(T) and MP2 calculations that have been carried out on the HLR. The following chapters give an overview about the research of our group using computational resources of the HLR Stuttgart.

Stefan Erhardt, Gernot Frenking

Quantum Mechanical Studies of Boron Clustering in Silicon

Boron-interstitial clusters (BICs) are known to be a key problem to controlling diffusion and activation of ultra-shallow boron implants in ULSI silicon device technology. During post-implantation annealing the self-interstitials, which had been created by the radiation damage, mediate fast transient diffusion of boron, during which stable and metastable BICs are formed. The BICs are either electrically inactive or the number of holes they can provide per number of boron atoms is significantly less than one. This causes a significant decrease in the activation rate. Therefore, sophisticated annealing strategies have to be developed to regain isolated boron substitutionals from BICs.

Péter Déak, Ádám Gali, Peter Pichler, Heiner Ryssel

Protonation States of Methionine Aminopeptidase Studied by QM/MM Car-Parrinello Molecular Dynamics Simulations

Methionine aminopeptidases (MetAPs) play a central role for in vivo protein synthesis as they remove the starter methionine from newly synthetized proteins. MetAPs are metaldependent enzymes. It is not clear which metal activates the MetAPs in vivo. For in vitro experiments, cobalt is commonly used because it activates all known MetAPs and the cobaltsubstituted enzymes are usually the most active. [1] Zinc and iron(II) have also been shown to activate some MetAPs. [2] The metalchelating residues in all known MetAPs are two glutamates, two aspartates and one histidine. The geometric arrangement of these residues is practically identical in all MetAP x-ray structures.

Christian D. P. Klein

Molecular Transport Through Single Molecules

Present trends in the miniaturization of electronic devices suggest that ultimately single molecules may be used as electronically active elements in a variety of applications [1, 2]. Recent advances in the manipulation of single molecules now permit to contact an individual molecule between two electrodes (see Fig. 1) and measure its electronic transport properties [3, 4, 5, 6, 7, 8]. Interesting and novel effects, such as negative differential conductance [9], were observed in some of these experiments, which still, by-and-large, beg theoretical explanation. In addition to generic principles of nanoscale physics, e.g. Coulomb blockade [6, 10, 11], the chemistry and geometry of the molecular junction emerge as the fundamental tunable characteristics of molecular junctions [3, 4, 12, 13, 14, 15, 16].

P. Stampfuß, J. Heurich, M. Wegewijs, M. Hettler, J. C. Cuevas, H. Schoeller, W. Wenzel, G. Schön

Computer Science


Towards a Holistic Understanding of the Human Genome by Determination and Integration of Its Sequential and Three-Dimensional Organization

Genomes are one of the major foundations of life due to their role in information storage, process regulation and evolution. However, the sequential and three-dimensional structure of the human genome in the cell nucleus as well as its interplay with and embedding into the cell and organism only arise scarcely from the unknown, despite recent successes e.g. in the linear sequencing efforts and growing evidence for seven genomic organization levels. To achieve a deeper understanding of the human genome the structural, scaling and dynamic properties in the simulation of interphase chromosomes and cell nuclei are determined and combined with the analysis of long-range correlations in completely sequenced genomes as well as the analysis of the chromatin distribution in vivo: This integrative approach reveals that the chromatin fiber is most likely folded according to the Multi-Loop-Subcompartment (MLS) model in which the chromatin fiber bents into 63–126 kbp big loops aggregated to rosettes connected by again 63–126 kbp linkers. The MLS model exhibits fine-structured multi-scaling and predicts correctly the transport of molecules by moderately obstructed/anomalous diffusion. On the basic sequence level, genomes show fine-structured positive long-range correlations, allowing classification and tree construction. This, DNA fragment distributions after carbon ion irradiation and on the highest structural level, the nuclear morphology visualized by histone autofluorescent protein fusions in vivo, agrees again best with the MLS model. Thus, the local, global and dynamic characteristics of cell nuclei are not only tightly inter-connected, but also are integrated holisticly to fulfill the overall function of the genome.

Tobias A. Knoch

Efficient and Object-Oriented Libraries for Particle Simulations

We present two libraries for the parallel computation of particle simulations. One is the object-oriented library sph2000 written in C++, the other is ParaSPH, a library written in C, that supports hybrid architectures (clustered SMPs). They are portable and performant on a variety of parallel architectures with shared and distributed memory. We give details of the object-oriented design of sph2000, the parallelization of ParaSPH for hybrid architectures using MPI and OpenMP and discuss the speedups of the codes. Further, we give three examples of applications based on these libraries, which simulate protoplanetary discs, colliding rubber rings and the injection of diesel into a combustion chamber.

S. Ganzenmüller, M. Hipp, S. Kunze, S. Pinkenburg, M. Ritt, W. Rosenstiel, H. Ruder, C. Schäfer

SKaMPI — Including More Complex Communication Patterns

SKaMPI is now an established benchmark for MPI implementations. In autumn 2002 the development of the “new SKaMPI” has started in three major directions: (i) extension of the benchmark to cover more functions of MPI and a redesign of the benchmark allowing it to be extended more easily (thus matching requests from SKaMPI users); (ii) construction of a collection of important algorithm kernels which are not supported by core MPI collective operations; (iii) development of a simulator which (at least partially) supports the simulation of one MPI implementation while using another MPI implementation (possibly running on a different kind of machine).In the present paper we give an overview of the new SKaMPI and describe fast implementations of MPI_Alltoall and of an algorithm for transposing multidimensional matrices as first instances of SKaMPI-Alg.

Michael Haller, Thomas Worsch

Performance Analysis Using the PARbench Benchmark System

This project is not meant to find a solution to any known problem in the field of science or technique. It’s rather a question of the evaluation of some high performance computer systems used for this purpose. These systems located at the High Performance Computing Center Stuttgart were analyzed using the PARbench benchmark system. Special attention was dedicated to scheduling system within the scope of this analysis. Conclusions on the optimization of the examined computer systems become possible that way.

Andreas Kowarz
Weitere Informationen