Skip to main content





Spinodal Decomposition in Binary Polymer Blends: Monte Carlo Simulations and Dynamic Mean Field Theory

Using large scale computer simulations we have investigated the interplay between single chain dynamics and the kinetics of phase separation in a symmetric binary polymer blend. In the framework of a coarse grained lattice model — the bond fluctuation model on a three dimensional lattice — we monitor the growth of concentration fluctuations after a quench from the one phase region into the miscibility gap. Chains of 64 effective segments are simulated in a cell of linear dimension L = 160, i.e., each simulation box contains 256 000 particles. The growth rate of composition fluctuations is averaged over 64 realizations of the temperature quench.The simulation results are compared to dynamic mean field theory without any adjustable parameter. Two theoretical approaches have been investigated: dynamical self-consistent field theory and external potential dynamics. The quantitative comparison between simulation and theory reveals the pronounced influence of the single chain dynamics on the dynamics of collective variables. A Rouse-like single chain dynamics can be incorporated into the dynamical self-consistent field theory via a non-local Onsager coefficient. The external potential dynamics results in Rouse-like dynamics without the need of a non-local Onsager coefficient. Moreover, the latter method is about an order of magnitude computationally faster than the dynamic self-consistent field theory.

E. Reister, M. Müller, K. Binder

Dynamics of Convection and Dynamos in Rotating Spheres

Large scale computer simulations of thermal convection in rapidly rotating spheres have been carried out with and without magnetic fields generated by the dynamo process. Relaxation oscillations and localized convection activity represent coherent phenomena of turbulent convection in the absence of magnetic fields. In the presence of the latter the coherent structures are destroyed and the heat transport is enhanced. A tendency towards saturation of the magnetic energy with increasing Rayleigh number has also been found.

E. Grote, F. H. Busse

Recent Developments in IMD: Interactions for Covalent and Metallic Systems

We describe the recent developments of IMD (ITAP Molecular Dynamics), a general purpose pro gram for classical molecular dynamics simulations on workstations and massively parallel supercomputers. As pair potentials are not entirely suitable for many classes of materials, several further types of interactions with many body forces have been implemented, thereby extending the range of applicability of IMD. IMD now supports, in particular, also EAM (Embedded Atom Method) potentials for the simulation of metals, and Stillinger-Weber and Tersoff potentials for the simulation of covalent systems, such as ceramics and semiconductors.

Erik Bitzek, Franz Gähler, Jutta Hahn, Christopher Kohler, Galib Krdzalic, Johannes Roth, Christoph Rudhart, Gunther Schaaf, Jörg Stadler, Hans-Rainer Trebin

Finite Difference Modelling of Seismic Wave Phenomena within the Earth’s Upper Mantle

The analysis of the propagation of elastic waves (seismic phases) plays an essential role for understanding important questions concerning the structure, composition and evolution of the Earth. Since direct observation of the Earth deeper interior is not possible, elastic waves generated from earthquakes or artificial sources are recorded with seismographs, usually located as array at the Earth’s surface, and are analyzed for seismic phases. These phases carry important information about the velocity structure within the Earth. In most cases the direct derivation of a velocity model of the Earth is not possible. To interpret the recorded seismic data and to understand the propagation of a seismic wave field through the Earth, synthetic seismograms are calculated for an initial starting model. These synthetic seismograms are then compared with the observations and the initial model is modified. Again, synthetic seismograms are calculated, until those and observed seismograms are similar enough to stop this iterative process. In this paper we discuss several applications of synthetic seismograms calculated with finite difference schemes that can significantly improve our knowledge about the elastic fine structure of the Earth’s crust and upper mantle. Finite difference schemes had previously already been shown to be most efficient for high-performance computing on seismological wave propagation tasks [5].

Trond Ryberg, Marc Tittgemeyer, Friedemann Wenzel

Collisional Dynamics of Black Holes and Star Clusters Using Massively Parallel Computing

We study the transport of angular momentum in dense star clusters, such as found in galactic nuclei. Two types of numerical experiments are described: 1) the sinking of two massive black holes in a spherical stellar system; 2) the evolution of rotating, isolated clusters of stars up to core-collapse. In the binary black hole simulations, we find a continuous hardening of the binary even in late stages, where the interaction with the stellar system through gravitational friction is ineffective. Three body interactions with stars and the motion of the massive binary relative to the galactic centre are responsible for this. Simulations of rotating star clusters start off from axisymmetric generalisations of King profiles. Results obtained for two sets of single-mass cluster simulations are presented. These confirm the more rapid evolution of even mildly-rotating clusters. A cluster model with rotational energy comparable to ωCentauri’s reaches core-collapse in less than half the time required for non-rotating clusters. The singularities encountered in Newtonian gravitational dynamics of point mass sources require integration schemes which are not only able to follow the dynamics to a very high accuracy, but also provide a proper regularisation treatment of hard binaries, triples, and stars bound in multiple systems generally. An introduction level discussion of the methods developed to perform these challenging computations is presented as well in the text.

Marc Hemsendorf, Christian Boily, Steinn Sigurdsson, Rainer Spurzem

Three-Dimensional Direct and Inverse Electromagnetic Scattering

The direct and inverse three-dimensional time-harmonic electromagnetic scattering from inhomogeneous media is considered. Volume integral equations are used to describe mathematically the physical problem of electromagnetic scattering from known objects. When solving inverse scattering problems one tries to retrieve information about the unknown scatterer from the knowledge of incident probing waves and measured scattering data. This paper deals with methods to reconstruct the geometry and the material properties of inhomogeneous media from scattering data. The objects considered in this context are isotropic lossy dielectries. The objects are assumed to be nonmagnetic. The inverse scattering problem can be formulated as a nonlinear optimization problem which is solved by means of iterative optimization schemes. Numerical examples demonstrate the efficiency of the proposed methods.

Wolfgang Rieger, André Buchau, Günther Lehner, Wolfgang M. Rucker

Precession Driven Flow in Ellipsoidal Cavities

The simulations described here are concerned with fluid flow in a rotating ellipsoidal container whose rotation axis is precessing. Without the precessing motion of the container, the fluid would simply be entrained by viscous forces at the boundaries until it rotates uniformly like asolid body. Precession disturbs the solid body rotation.

S. Lorenzani, A. Tilgner

The Computation of Highly Exited Hyperbolic 3D-eigenmodes and its Application to Cosmology

The numerical aspect of our project consists of the computation of the eigenvalues and eigenfunctions of the Laplace-Beltrami operator of the hyperbolic 3D-space with constant negative curvature with respect to a compact fundamental cell. The eigenfunctions have to obey certain boundary conditions on the fundamental cell.

Ralf Aurich

Fluid Jet Simulations using Smoothed Particle Hydrodynamics

Our goal is to use Smoothed Particle Hydrodynamics to model the primary breakup of a Diesel jet as it is injected into the cylinder of an engine. We have performed two-dimensional simulations with parameters similar to those in a real Diesel injection process, and have identified some of the physical and numerical effects that have to be taken into account for a realistic simulation. We point out directions for future research.

Erik Schnetter, Stefan Kunze, Roland Speith

Solid State Physics


Spectral Properties of CuO2 Planes in a Cluster Perturbation Approach

We analyze the single-particle properties of the three-band Hubbard model in two dimensions in the half-filled and in the overdoped regime (δ=0.25). By means of a cluster perturbation theory we extend the Green functions obtained from Exact Diagonalization of finite size clusters to an infinite system. The single particle spectra obtained are compared with finite temperature Quantum Monte Carlo calculations (QMC). We find results in excellent agreement with QMC spectra calculated for large clusters (8 x 8) at ßt = 10.

Christopher Dahnken, Robert Eder, Enrico Arrigoni, Werner Hanke

Electronic, Structural and Vibrational Properties of Chalcogenides on Si(001) and Ge(001) Surfaces

The (001) surfaces of Si and Ge do not exist in their structurally ideal form, because they would be highly reactive. These surfaces exhibit (2 x 1) reconstructions, while the adsorption of chalcogenides can restore the (1 x 1) symmetry. In this way the reactivity of the surface can be reduced, the surface is passivated.A Theoretical treatment of the electronic and structural properties of these particular systems by ab-initio methods is possible on conventional workstations. Corresponding results have been published in the last years, while information about the vibrational properties has been lacking. Only the use of more sophisticated architectures like the IBM RS/6000 SP/256 at the Scientific Supercomputing Center Karlsruhe or the CRAY T3E-900/512 at the High-Performance Computing Center Stuttgart makes it possible to calculate surface phonons of such systems from first principles within a reasonable time. Using the so-called density functional perturbation theory, a straight forward parallelization allows to achieve a very high efficiency that is independent of the particular surface. The purpose of this article is to present technical aspects of this parallelization. Especially specific features of the expansion of occuring functions into plane waves will be discussed with regard to the convergence of the results.

Ulrich Freking, Albert Mazur, Johannes Pollmann

Dynamical Properties of the t-J Model

We present a new quantum Monte Carlo method for the determination of the one-particle propagator in the t-J model. The method can be used both at zero doping, where it is free from the notorious sign problem, and at finite doping with holes. For one dimension we show, that a simple slave particle picture is able to describe the overall features of the spectral function at half filling. Additionally we give results at small doping. In two dimensions we observe a dispersion as predicted by self-consistent Born approximation. We observe flat bands at k = (π, 0), and a minimum of the dispersion at k = (π/2, π/2). We further show the existence of string excitations by considering the excitations above the quasiparticle peak at k ≈ (π/2, π/2). As opposed to from the one-dimensional case, the quasiparticle weight is finite in the thermodynamic limit in two dimensions.

Catia Lavalle, Michael Brunner, Fakher F. Assaad, Alejandro Muramatsu

Effects of Three Nucleon-Interactions in A = 4

Within the Refined Resonating Group Model (RRGM) we use variational techniques to investigate n - t and p -3 He scattering. Results for the total and differential cross sections and the analyzing powers for Argonne ν′8 and ν18 with and without additional Urbana IX or Texas-Los Alamos three-nucleon force are discussed. We find large differences between the calculations and the experimental data which arise from an underprediction of the 3P2 and 3P1 phase shifts by more than 10 degrees at Ecm = 5MeV. Therefore these systems might be well suited to study three-nucleon forces.

B. Pfitzinger, H. M. Hofmann

Phase Transitions in Insulating 1D Electron Systems

We study the ground state phase diagram of different extensions of the one-dimensional Hubbard model at half filling by numerically diagonalizing finite systems with the Lanczos and DMRG method. We discuss the different kinds of insulator-insulator phase transitions that may occur in these models and the remaining open questions regarding their phase diagrams. First numerical results obtained on the IBM RS/6000 SP at the Scientific Supercomputing Center (SSC) Karlsruhe are presented.

Ph. Brune, A. P. Kampf

Excited States of Semiconductors and Molecules

The investigation of the electronic structure of materials and of electronic excitations forms a major field in condensed-matter physics. The quantum-mechanical ground state of the electrons in the potential of the atomic nuclei plays an important role for the geometric arrangement of matter, for chemical bonding, structural phases, and for crystal and molecular dynamics. Excitations of the ground state are important for spectral properties, like optical spectra, as well as, for transport and particle scattering problems. Excited states and their spectra are commonly used to characterize materials. Furthermore, they are relevant for a large variety of technological applications, like photovoltaics, photochemistry, dye chemicals, light-emitting devices, etc.

Michael Rohlfing

Jacobi-Davidson Algorithm with Fast Matrix-Vector Multiplikation on Massively Parallel and Vector Supercomputers

The exact diagonalization of very large sparse matrices is a numerical problem common to various fields in science and engineering. We present an advanced eigenvalue alorithm - the so-called Jacobi-Davidson algorithm - in combination with an efficient parallel matrix-vector multiplication. This implementation allows the calculation of several specified eigenvalues with high accuracy on modern supercomputers, such as CRAY T3E and NEC SX-4. Exemplarily the numerical technique is applied to analyze the ground state and spectral properties of the three-quarter filled Peierls-Hubbard Hamiltonian in relation to recent resonant Raman experiments on MX chain [-PtCl-] complexes.

M. Kinateder, G. Wellein, A. Basermann, H. Fehske



Time-dependent Reactive Scattering for Ion-neutral Collisions

Quantum mechanical calculations of reaction probabilities for atom - mole- cule collisions are calculated time-dependently with wave packets (using the Chebychev or split-operator method for propagation). Our aim is to perform the time-dependent treatment of collision dynamics on coupled potential energy surfaces (PESs) for the systems H + H+2, H+ + H2, NeH+2, NeH2, H-3, H3.For the PESs of the electronic ground states of Ne+H+2 and H+HH+2 very accurate ab initio PESs are available. In the case of charge-transfer investigations we need two PESs (H+ + H2, H + H+2). In addition our interest is to investigate intermediately build collision complexes.We calculate “state-to-state” - reaction probabilities. The propagated wave packet will be investigated depending on different starting conditions like collison energy, specifications of the ro-vibrational quantum numbers of the reactant diatom, the influence of total angular momentum and different coupling approximations (with and without Coriolis coupling).The numerical procedure of wave packets can be summarised as follows: discretization of the wave packet on a grid, propagation with the Chebychev-method, kinetic energy and potential energy terms are calculated using “fast fourier transformation” (FFT) and “discrete variable representation” (DVR) methods. A “parallel” version of the code is tested on an IBM-SP2.

Christian Morari, Robert Röhse, Ralph Jaquet

Ground and Exited States of the Hydrogen Negative Ion and Negative Donor Systems Strong Magnetic Fields

The lowest bound states of the hydrogen negative ion and negative donor systems in a homogeneous magnetic field are investigated theoretically via a full configuration-interaction approach with an anisotropie Gaussian basis set. The broad magnetic field regime γ = 8·10-4 - 4·103 is covered. Nonrelativistic electron detachment energies, and transition wavelengths are presented assuming an infinite nuclear mass. The binding mechanisms are discussed in detail. The accuracy for the energies is enhanced significantly compared to previously published data. Details of the computational requirements and procedure are discussed.

Omar-Alexander Al-Hujaj, Peter Schmelcher

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 optimising reaction conditions in chemical industry. Where experiments fail to obtain the needful results, modern quantum chemical approaches can be utilised to investigate chemical systems and predict their properties. This is achallenging task for computational chemists and the necessary calculations, particularly at high levels of theory, are 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. The following chapters give an overview about the research of our group using computational resources of the HLR Stuttgart.

Jan Frunzke, Gernot Frenking

Car-Parrinello Density Functional Calculations of the Bond Rupture Process of Thiolate on Gold in AFM Measurements: Progress and First Results

Recently, direct measurements of the forces related to the rupture of a single thiolate-gold bond have been reported [1]. However, the mechanism of the rupture process itself remains unclear. A special interest is focused on the question whether the bond rupture is taking place at the gold-thiolate interface or if gold itself is extracted from the surface during the rupture of of the bond.As there is a strong evidence for thiol-gold bonds having a covalent character this question is not easily accessible by atomistic simulations based on simple potential models. Additionally the exceptional molecular diffusivity and mobility of metallic gold call for molecular dynamics to investigate this problem. The CPMD [2] code used in this work has been modified and tested to meet these requirements. The first calculation results reported here seem to justify this molecular dynamics approach. Different surface models used for this calculation show that the strong binding of the sulfur headgroup to the gold surface causes substantial rearrangement and distortions that might not be accessible from geometrical optimizations alone. Furthermore first constraint dynamic runs at subsequently increased distances between the surface and the second carbon atom indicate, that the strength of the bond might be sufficient to extract gold ions or clusters from the surface. Starting from a threefold bin ding configuration the molecule could be elevated by 1.8Å with only minimal weakening of two of the three gold-sulfur bonds. The process is accompanied by diffusion activity within the surface. Taking into account the peculiar ability of gold to form stable monoatomic wires, which has recently attracted great attention from both experimental and theoretical groups [3–6], it will also be of great interest to wich point the elevation process might proceed. Additional constraint calculations will be necessary to clarify this question and to get quantitative estimates for the forces measured in AFM.

Daniel Krüger, Roger Rousseau, Dominik Marx, Harald Fuchs, Michele Parrinello

Computational Fluid Dynamics


DNS of Laminar-Turbulent Transition in a 3D Aerodynamics Boundary-Layer Flow

Controlled laminar-turbulent transition of a 3-D boundary-layer flow caused by localized surface nonuniformities is investigated by means of spatial direct numerical simulations (DNS) based on the complete incompressible 3-D Navier-Stokes equations and a combined 6th-order compact Finite-Difference / spanwise Fourier-spectral scheme. The considered laminar spanwise-invariant flat-plate base flow is a model of the boundary-layer flow in the front region of a swept wing, with favorable and ensuing adverse chordwise pressure gradient. The primary downstream growth and nonlinear interaction of vortex-packet modes as well as the appearance of a secondary instability mechanism leading to turbulence is investigated in detail. The effectiveness of a steady, nonlinear, upstream flow deformation of the base flow with respect to significant transition delay is shown.

P. Wassermann, M. Kloker, U. Rist, S. Wagner

High-Performance Computing: Numerical Simulation of the Melt Flow in an Industrial Czochralski Cruzible

High resolution calculations with about two million control volumes were performed on a parallel-vector computer (NEC SX-4 and SX-5) using a finite volume method and a quasi-DNS approach to simulate the fluid flow and heat transfer in an industrial Czochralski melt. The Czochralski process is the most important production technique for high quality silicon single crystals. The convergence of the numerical algorithm was increased by a factor of 5.3 by utilizing the FAS multigrid method on four grid levels. A time-averaging procedure of the velocities and the temperature results in reduction of the data set which was used to analyze the main transport phenomena in the silicon melt. A total performance of 5.46 and 10.89 GFlops on seven processors was reached for the simulation on the SX-4 and the SX-5, respectively. Based on a comparison of the parallel performance on one to fourteen processors with single- and multigrid, the configuration with fourteen processors gives the highest speedup factor of 9.99 and 8.94, respectively.

Sven Enger, Michael Breuer

Analysis of an Elastic Wing in Subsonic Flow Using Direct Numerical Aeroelastic Simulation

This work describes the advancements in developing a computational method for the treatment of solid fluid interaction (SOFIA) to analyze and to improve the aerodynamic performance of elastic wings and rotors by direct numerical aeroelastic simulation. The unsteady fluid flow and the structural dynamics are computed simultaneously in a fully coupled manner. The fluid flow is modeled by the Euler or by the Reynolds-averaged Navier-Stokes equations. The elastic wing is described by a generalized Timoshenko-like beam structure with six degrees of freedom for a material cross-section. Numerical results including comparison to measured data are shown for a rectangular wing model built in the framework of the collaborative research centre SFB401 “Modulation of Flow and Fluid-Structure Interaction at Airplane Wings” of the University of Technology Aachen.

Gerd Britten, Markus Werle, Michael Hesse, Josef Ballmann

Large Eddy Simulation of the Flow over a Matrix of Surface-mounted Cubes

The paper presents Large Eddy Simulation of the flow over a matrix of surface- mounted cubes. The flow with and without heat transfer is investigated. A new boundary model for recirculating flows is proposed and tested in the present configuration.

Jochen Fröhlich, Fabrice Mathey, Wolfgang Rodi

Simulation of Bubbly Gas-Liquid Flows by a Parallel Finite-Difference/Front-Tracking Method

Bubbly gas-liquid flows represent a prototype of dispersed multiphase flow systems, which appear in many natural phenomena and industrial applications. It is of fundamental interest to understand how the dispersed elements in such systems interact with each other and the ambient fluid field, and the collective motion and induced turbulence that arise from these interactions. The special challenges of simulating bubbly flows are to follow the motion of deformable phase boundaries and to accurately account for the stress boundary conditions at the interfaces. A parallel version of a finite difference/front tracking method is used to perform direct numerical simulations of mono- and bidisperse bubble size distributions rising in a stagnant liquid. The Navier-Stokes equations are solved on a fixed, regular, three-dimensional grid, while the interfaces between the gas and the liquid are tracked by two-dimensional surface-fitted moving meshes.

Manfred F. Göz, Bernard Bunner, Martin Sommerfeld, Grétar Tryggvason

Rotary Wing Aerodynamics and Aeroelasticity

A numerical approach for the aeroelastic analysis of helicopter rotors in hover and forward flight is presented. Fluid and structure codes are maintained as separate programs and are time—accurately coupled using TCP/IP socket connections for data exchange throughout run time. This paper includes results of both stand—alone Navier—Stokes hover analyses and Euler computations with fluid—structure coupling applied to both helicopter hover and forward flight.

H. Pomin, A. Altmikus, B. Buchtala, S. Wagner

Unsteady Flow Simulations for Turbomachinery Applications on Dynamic Grids

Preliminary results for the simulation of viscous unsteady transonic flow over oscillating blades in a linear cascade are presented. The experimental setup to be investigated essentially consists of a tip section of a rotor blade of the last stage of a steam turbine that is elastically suspended by cantilevers to allow for oscillations characterized by the first bending mode of the real 3D turbine blade. Time series show Mach number contours for the flow that has been calculated between two blades vibrating with the interblade phase angles 0° and 180°. So far only 2D computations have been performed to further explore and enhance the run-time behaviour of the flow solver ITSM3D on the NEC SX-4 and SX-5 platforms.

Holger Bauer, Jürgen F. Mayer, Heinz Stetter

Testing Turbulence Models by Comparison with DNS Data of Adverse-pressure-gradient Boundary Layer Flow

A favourable/adverse-pressure-gradient boundary layer flow is computed with the zonal grid approach applied to direct numerical simulation (DNS). The computational domain covers a big part of Watmuff’s experimental flow domain [15]. Experimental and DNS results are compared with statistical calculations by a boundary layer code, using the к - ω model and Menter’s SST model [10]. The streamwise development of global quantities are shown as well as the mean streamwise velocity component, Reynolds shear stress and turbulent kinetic energy. Boundary layer calculations with turbulence models are in fair agreement with the DNS data and the experiment. Nevertheless, the prediction of the Reynolds shear stress and the turbulent kinetic energy could be improved.

T. J. Hüttl, G. Deng, R. Friedrich, M. Manhart

Large Eddy Simulation of Subcritical Flow around Sphere

Possible large modeling errors in the case of statistical turbulence models on the one side and small realizable Reynolds numbers in direct numerical simulation on the other side suggest the usage of large-eddy simulation(LES), when the other two approaches fail. This is the case with the flow around a sphere at a subcritical Reynolds number of 50 000, which is here computed using LES in combination with a fully conservative finite-volume method. Two simulations have been carried out and are compared against each other and available experimental data. The most difficult task is to evaluate the performance of the LES model. Several items directly influence the accuracy of the simulation that can’t be investigated separately from each other. The variable arrangement (staggered/colocated) influences the calculation of the velo city gradient, which is related to the model contribution. The discretization scheme acts in the case of second-order central differences like a filter because the interpolation to the cell face corresponds to spatial averaging of variables [8]. The filter width depends upon the local mesh size and only the sharp Fourier cutoff filter removes properly all frequencies of the turbulent spectrum higher than the filter [7]; the natural choice of the unstructured finite-volume code is, however a tophat filter. The filter, the model for the subgrid scales and the discretization scheme can’t be viewed independently, since the errors introduced in individual steps interact and may either partially cancel out or augment each other.

M. Schmid, M. Perić

LES of Turbulent Flows Trough 90°-Pipe Bends on NEC SX-4

Oscillatory phenomena in turbulent flows through 90°-pipe bends are numerically investigated using large-eddy simulations on the NEC SX-4 of the High-Performance Computing Center Stuttgart (HLRS). Since the ratio of the time scales of the oscillatory phenomena to those of the turbulent motion is large, this investigation is very costly with regard to computing time. The bend flows are simulated at different Reynolds numbers and curvature radii. Additionally, the flows through two double 90°-bends are investigated. The computations are performed on block-structured curvilinear meshes with 600,000 to 1,300,000 grid points. The force spectra show two distinct frequencies that relate to vortex shedding and an oscillation of the Dean vortices. The characteristic computing times, the storage capacity requirements and the achieved computer performance are reported. The analysis shows the results in good agreement with experimental data.

Frank Rütten, Matthias Meinke, Wolfgang Schröder

Computations for the European LESFOIL Project

The code LESOCC is applied to perform Large Eddy Simulations of the flow around an airfoil at high Reynolds number and large angle of attack. The flow over a periodic series of hills is considered as a test case to investigate resolution requirements and modelling issues.

Ch. P. Mellen, J. Fröhlich, W. Rodi

Reactive Flows


Correlation Analysis of Permixed Turbulent Flames Using Direct Numerical Simulations

In this work direct numerical simulations of premixed turbulent flames are performed using both, detailed chemical kinetics and detailed transport models. Accounting for all turbulent as well as chemical time scales allows a detailed investigation of the coupling of the chemical kinetics with the turbulent flow field. The results of the DNS are analyzed using a correlation analysis technique which allows to extract information on the chemical kinetics which can then be used to improve submodels for turbulent flame calculations.

Wilhelmina Tsai, Dietmar Schmidt, Ulrich Maas

Adaptive Chemistry Computation to Accelerate Parallel DNS of Turbulent Combustion

Direct numerical simulation (DNS) has become an important tool to study turbulent combustion processes. Especially in the case of using detailed models for chemical reaction kinetics, computation time still severely limits the range of applications accessible by DNS. The computation of the chemical source terms is one of the most time-consuming parts in such simulations. An adaptive evaluation of the chemical source terms can strongly reduce this time without a significant loss in accuracy which is shown for DNS of several premixed and non-premixed reactive flows. A dynamic load-balancing scheme is used to maintain a high efficiency in the parallel adaptive computations.

Marc Lange

The Generation of Dissipative Quasi-Particles near Turing’s Bifurcation in Three-Dimensional Reacting Diffusion Systems

In order to model pattern formation processes in a dc driven semiconductor-gas discharge system on a phenomenological level, we investigate a three-component reaction-diffusion system of 1-activator-2-inhibitor-type. The solutions of this system show localized moving and stationary structures which interact by scattering, annihilation or more complex scenarios. Because of this particle-like behaviour the structures are called dissipative quasi-particles. This work deals with the generation mechanism of dissipative quasi-particles related to Turing’s destabilisation of homogeneous states. Two- and three-dimensional simulations are shown and in the two-dimensional case compared with experimental results.

Andreas W. Liehr, Mathias Bode, Hans-Georg Purwins

Upwind Relaxation Algorithm for Reentry Nonequilibrium Flows

The further development of the Navier-Stokes solver URANUS (Upwind Relaxation Algorithm for Nonequilibrium Flows of the University of Stuttgart) in 1999 will be described.

H.-H. Frühauf, M. Fertig, F. Olawsky, F. Infed, T. Bönisch

Numerical Simulation of the Coupled Dynamic Processes of the Water-Steam Cycle and the Furnace-System

Within the scope of the present project a validation exercise is presented based on a complete set of measurements carried out at a bituminous coal fired utility boiler with an electrical output of 200 MWe. Flue gas concentrations and temperature measurements are available at the furnace exit as well as in the burner near field during both partial and full load operation. Furthermore, temperature measurements are provided at the furnace exit during change of load. The objective of the calculations performed is to investigate the reliability of numerical predictions of both, steady-state and time-dependent processes in full-scale combustion systems. This is a principal requirement for the coupling of the models describing fluid flow in the steam generation system and combustion processes within the furnace.Furthermore, the concept of detailed numerical simulation of the coupled processes in steam power plants is introduced. For a 600 MWe brown coal fired utility boiler, calculations of both, the combustion processes within the furnace and the water-steam cycle are presented.

Alexander Bundschuh, Christoph Sauer, Uwe Schnell, Klaus R. G. Hein

Structural Mechanics


Container Size Dependence of the Velocity Fluctuations in Suspensions of Monodisperse Spheres

We show that a constraint force method can be used in an advantageous fashion to replace the no-slip boundary conditions in a Navier-Stokes simulation of rigid particles suspended in Newtonian fluids. The constraint forces mimic the presence of particles in the fluid. We use the method to study the container-size dependence of the velocity fluctuations in batch sedimentation under periodic boundary conditions and in quadrilateral containers with fixed walls.

Kai Höfler, Esa Kuusela, Christian Manwart, Reinmar Mück, Stefan Schwarzer

Computer Science

Implementing Luby’s Algorithm on the CRAY T3E

We present an implementation of Luby’s algorithm for the calculation of maximal independent sets in graphs on the Cray T3E.

Jürgen Gross, Markus Lohrey

Spatial Partitioning for Parallel Hierarchical Radiosity on Distributed Memory Architectures

This paper presents an efficient, highly scalable parallel implementation of the Hierarchical Radiosity Algorithm. We present a clever mapping of Hierarchical Radiosity to high-dimensional spaces that manifests a locality property, which can greatly reduce communication on parallel distributed memory architectures. We use a very simple dynamic spatial partitioning method to keep the mapping balanced. We describe solutions for some key implementation problems: grouping of elements and links, and data reference locality. Speedup plots give an impression of the scalability of our implementation. On a Cray T3E the speedup curve is almost linear up to 64 processors. This is better than previously published attempts on massively parallel distributed memory computers.

Robert Garmann

Construction of Large Permutation Representations for Matrix Groups on Parallel Supercomputers

In the first [*, High Perf. Comp. in Science and Engin. ‘98, HLRS Stuttgart, Springer-Verlag Berlin, 1998, pp. 430–452] article on this topic the author described an implementation and the construction of the corresponding hash table of the modified Cooperman-Finkelstein-Tselman-York algorithm [Cooperman, Finkelstein, Tselman, and York, Constructing permutation representations for large matrix groups, Proc. of Int. Symp. on Symb. and Algeb. Comp. ISSAC ‘94 (New York), (Oxford), ACM Press, 1994, pp. 134–138] constructing a transitive permutation representation from a given matrix representation of a finite group G over a finite field F and its non-trivial restriction to a given subgroup.The main task of this computation project was then to find the order of a group generated by an 1333-dimensional 11-modular irreducible representation [Lempken, Constructing J4in GL(1333, 11), Comm. Algebra 21 (1993), 4311–3251] of Janko’s largest simple sporadic group J4 [Janko, A new finite simple group of order 86, 775, 571, 046, 077, 562, 880 which possesses M24and the full covering group of M22as subgroups, J. Algebra 42 (1976), 564-596] using parallel supercomputers.The computation required the construction of a 173 067 389-dimensional permutation representation first. Some necessary conditions that had not yet been checked in the first article of this sequel are now proven. The result is used in [Cooperman, Lempken, Michler, and *, A new existence proof of Janko’s simple group J4, Proc. of Euro-Conference “Computational methods for representations of groups and algebras”, Progress in Math., vol. 173, Birkhäuser Verlag, Basel, 1999] to give a new existence proof of J4 ≅ G.To optimize the calculations a 112-dimension GF(2) submodule of the permutation representation was constructed.

Michael Weller

Advances in High-Performance Computing: Multigrid Methods for Partial Differential Equations and its Applications

The program package UG provides a software platform for discretizing and solving partial differential equations. It supports high level numerical methods for unstructured grids on massively parallel computers. Various applications of complex up to real-world problems have been realized, like Navier-Stokes problems with turbulence modeling, combustion problems, two-phase flow, density driven flow and multi-component transport in porous media. Here we report on new developments for a parallel algebraic multigrid solver and applications to an eigenvalue solver, to flow in porous media and to a simulation of Navier-Stokes equations with turbulence modeling.

Peter Bastian, Klaus Johannsen, Stefan Lang, Sandra Nägele, Christian Wieners, Volker Reichenberger, Gabriel Wittum, Christian Wrobel

Recent Advances of SKaMPI

The goal of SKaMPI is the creation of a database containing performance measurements of parallel computers in terms of MPI. This data supports software developers in creating portable and fast programs. To meet this goal port ability is a crucial property of our this benchmark. A large number of platforms to test on and to examine is important for the realization of our vision. Since access to the Stuttgart Cray T3E in October 1998 we were able to improve the code’s portability, to solve memory alignment problems, to compare the quality Cray’s implemented MPI gather algorithm to the one used in IBM’s SP MPI library, and to investigate the blocking behavior of collective MPI operations. Recently we added the parameterization of measurements suites with (even user defined) datatypes.

R. H. Reussner

Porting SPLASH-2 Benchmarks to the T3E

The SPLASH-2 suite of parallel benchmarks is used for numerous studies in computer architecture. Therefore the quantitative and qualitative characteristics of the individual benchmarks (four kerneis and eight applications) are well-known, easing the evaluation of new architectural concepts. The benchmark suite was originally developed for shared-memory systems using the PARMACS macro package for parallelization. We ported three kernels and three applications to a distributed-memory machine, i.e. the Cray T3E, using the Cray SHMEM library. After a short description of the individual programs within the benchmark suite, one kernel and one application are described in detail and experimental results for those benchmarks are given.

Andreas Grävinghoff, Andreas Paul
Weitere Informationen