Skip to main content
Top

2024 | Book

High Performance Computing in Science and Engineering '22

Transactions of the High Performance Computing Center, Stuttgart (HLRS) 2022

insite
SEARCH

About this book

This book presents the state-of-the-art in supercomputer simulation. It includes the latest findings from leading researchers using systems from the High Performance Computing Center Stuttgart (HLRS) in 2022. The reports cover all fields of computational science and engineering ranging from CFD to computational physics and from chemistry to computer science with a special emphasis on industrially relevant applications. Presenting findings of one of Europe’s leading systems, this volume covers a wide variety of applications that deliver a high level of sustained performance.
The book covers the main methods in high-performance computing. Its outstanding results in achieving the best performance for production codes are of particular interest for both scientists and engineers. The book comes with a wealth of color illustrations and tables of results.

Table of Contents

Frontmatter

Physics

Frontmatter
Simulating Binary Neutron Star Mergers
Abstract
In 2017, the first joint detection of gravitational waves and electromagnetic waves, produced from the merger of a binary neutron star system, inaugurated a new era of multi-messenger astronomy. Due to the strong gravitational fields present in the last stages of the compact binary coalescence, one has to solve Einstein’s field equations for a comprehensive study. For this reason, numerical-relativity simulations are an essential tool to correctly describe and study these compact binary mergers. High-performance computing facilities such as HAWK enable us to perform accurate simulations of binary systems by employing our numerical-relativity code BAM. BAM solves the equations of general relativity together with the equations of general-relativistic hydrodynamics. Within our research project, we use numerical-relativity simulations of binary systems to investigate matter at supranuclear densities, to measure the expansion rate of our Universe, and to calibrate theoretical models for the emitted gravitational and electromagnetic waves.
Tim Dietrich, Parikshit Biswas, Bernd Brügmann, Swami Vivekanandji Chaurasia, Mattia Emma, Francesco Maria Fabbri, Henrique Leonhard Gieg, Maximilian Kölsch, Nina Kunert, Michele Mattei, Anna Neuweiler, Henrik Rose, Peter Tsun Ho Pang, Federico Schianchi, Maximiliano Ujevic Tonino
Microphysical Aspects of Binary Neutron Star Mergers
Abstract
Neutron stars and their mergers challenge our fundamental understanding of dense matter and strong gravity. We present three projects that cover different aspects of these systems, namely the formation of quark matter in neutron star mergers, the determination of the limiting mass to prompt black hole formation and the dynamics of charged particles that occur in magnetised plasmas such as formed around black holes in these events. We also give a brief introduction to the numerical techniques used to model these scenarios and summarise their performance on the HAWK supercomputer system.
Michail Chabanov, Alejandro Cruz-Osorio, Christian Ecker, Claudio Meringolo, Carlo Musolino, Luciano Rezzolla, Samuel Tootle, Konrad Topolski
INTRHYGUE: Simulations of Hyperbolic Binary Black-Hole Mergers
Abstract
INTRHYGUE is a new research effort that aims at developing complete gravitational-wave templates for the observation of intermediate mass-ratio and highly eccentric binary black hole mergers. Accurate templates for this class of mergers are missing although recent LIGO-Virgo observations point to their the possible astrophysical existance. Detections of these black hole mergers will be possible with future ground- and space-based experiments provided that detailed templates will be in place. The project combines high-precision data from \((3+1)\)D numerical relativity simulations with a state-of-art effective-one-body analytical model for generic orbits. Simulations are performed with a novel numerical relativity code with pre-exascale capabilities. The data are employed to inform analytical, suitably resummed, expressions for the gravitational radiation reaction and the merger-ringdown waveforms. The initial focus of the project is the first systematic exploration of mergers from dynamical encounters (hyperbolic orbits). Such numerical relativity simulations have strengthen recent evidence that the signal GW190521 was originated by a dynamical capture of black holes.
Simone Albanesi, Sebastiano Bernuzzi, Boris Daszuta, Rossella Gamba, Alessandro Nagar, Francesco Zappa
Molecular Dynamics Simulations of the Structure of Lipid-Based Nanomaterials
Abstract
Lipid based nanomaterials are being considered as promising nanocarriers for a variety of therapeutic approaches including for example drug delivery, siRNA gene silencing, mRNA vaccines (as those for COVID-19). When the active substances to be delivered are nucleic acids, which are negatively charged in water, the lipid formulation includes typically a ionizable cationic lipid which is believed to interact with the nucleic acid and help protect it from degradation. However only limited information is available about the internal arrangement of the mRNA inside the nanoparticle. Here we report the results of multiscale molecular dynamics simulations used to characterize the structure of a lipid-based nanomaterial composed of messenger RNA and lipid formulations containing ionizable cationic lipids. In particular, the role of pH in the formation of the mRNA-lipids complex has been investigated. The simulations have shown that the interaction between the ionizable cationic lipids and RNA is attractive only at low pH and effectively repulsive at high pH. In addition, the ionizable lipid, when uncharged, also shows a significantly increased flip-flop probability, which may have consequences in the process of fusion of the nanoparticle with the endosomal membrane.
Giovanni Settanni, Friederike Schmid
Correlations, Shapes, and Fragmentations of Ultracold Matter
Abstract
This 2022 report summarizes our activities at the HLRS facilities (Hawk) in the framework of the multiconfigurational time-dependent Hartree for indistinguishable particles (MCTDH-X) high-performance computation project. Our results are a bottom-up investigation into exciting and intriguing many-body physics and phase diagrams obtained via the direct solution of the many-particle Schrödinger equation and its comparison to experiments, and via machine learning approaches. We investigated ultracold quantum gases for Pauli crystal melting, crystallization in a cavity, breakup and fragmentation of a condensate in the rotating frame, machine learning observables from single-shot images of ultracold atomic systems, and, finally Josephson dynamics of fragmented BECs, respectively.
A. U. J. Lode, O. E. Alon, A. Bhowmik, M. Büttner, L. S. Cederbaum, R. Chitra, S. Dutta, D. Jaksch, H. Kessler, C. Lévêque, R. Lin, P. Molignini, L. Papariello, M. C. Tsatsos, J. Xiang
Bulk Features of the Quark Gluon Plasma at Finite Density
Abstract
Pressure, energy density, entropy density and baryon number are the basic bulk features of quark gluon plasma (QGP), the deconfined phase of Quantum Chromodynamics (QCD). Knowledge of these quantities is a prerequisite for the successful interpretation of data from heavy ion collision experiments, and for the assessment of various effective models of quark matter. In this project we calculate the QCD equation of state at finite density in a novel expansion scheme, avoiding the shortcomings of the standard Taylor series in the chemical potential, thus circumventing the sign problem of finite density QCD for chemical potentials relevant for the RHIC Beam Energy Scan. In the first year we extrapolated the thermodynamics to finite baryon density, which was a proof of principle result [1]. In the second year we calculated the extrapolation for the actual experimental setting by enforcing strangeness neutrality. The continuum extrapolated results of the second year appeared in Ref. [2]. We also addressed the algorithmic issue of super-critical slowing down by applying parallel tempering for the quark-less case, where the transition is first order [3].
Szabolcs Borsanyi, Zoltan Fodor, Jana Günther, Sandor D. Katz, Attila Pasztor, Paolo Parotto, Ruben Kara, Claudia Ratti, K. K. Szabó

Molecules, Interfaces, and Solids

Frontmatter
Electro-Catalysis for HO Oxidation and Chlorine Evolution
Abstract
Electrocalaysts facilitate the transformation of renewable electrical energy into chemical feedstocks and fuels. However, our limited understanding of electrocatalysis limits our ability to design the catalysts needed to enable a renewable energy economy. Chlorine production through the chlorine evolution reaction (CER) is one of the few industrially relevant electrocatalytic reactions. Modern analytic methods struggle to probe the electrified solid/liquid interface, which has constrained our ability to develop mechanistic understanding. Ab initio computations do not suffer the same difficulty and can be employed to study the electrocatalytic chemistry occurring at electrified solid/liquid interfaces. A major hurdle with the application of ab initio methods is, however, the large computational cost associated with such simulations. Unlike gas-phase catalysis, accurate treatment of the electrolyte/solid interface requires 100s to 1000s of atoms for minimal simulations, and finite temperature effects must be included to prevent electrolyte freezing. Thus, realistic study of even the simplest electrocatalytic system requires long ab initio quality molecular dynamics simulations. In the first phase of the ECHO project we demonstrated such simulations are feasible using efficient density functional theory (DFT) codes on modern supercomputers and went on to uncover the unexpected role of surface oxidation state in the oxygen evolution reaction (OER) over iridium oxide. In this extension of the ECHO project, we have extend this approach to study the CER over iridium oxides. This system is especially pertinent owing to the scale of chlorine production. It also offers a relative simple system with which to study electrocatalytic selectivity, which proves to be more complex than selectivity in thermal catalysis. These results are highlighted, and, as they were only possible due to the efficient scaling of our approach on Hawk, a discussion about parallelization strategies is also included.
Travis Jones
Organic Functionalization on Solid Surfaces
Abstract
The present report discusses the reactivity and properties of several organic molecules on semiconductor surfaces by means of density functional theory. The adsorption modes and possible reaction paths for the adsorption of a prototype nonalternant aromatic compound azulene on Si(001) are presented. The concept of molecular building block has been addressed with the example of adsorption of cyclooctyne derivative on Si(001). The idea of area-selective atomic layer deposition is being exploited in the light of device miniaturization. New software is developed aiming for efficient automatic reaction network analysis in the field of film growth. And, finally, the scaling of large-scale simulations on the High-Performance Computing Center Stuttgart is addressed.
Badal Mondal, Raza Ullah Khan, Florian Kreuter, Patrick Maue, Sudip Pan, Fabian Pieck, Hendrik Weiske, Ralf Tonner-Zech
Polaron Formation Dynamics in Lithium Niobate from Massively Parallel ab-initio Simulations
Abstract
Polarons influence decisively the performance of lithium niobate for optical applications. In this project, the formation of (defect) bound polarons in lithium niobate is studied by ab-initio molecular dynamics. The calculations show a broad scatter of polaron formation times. Rising temperature increases the share of trajectories with long formation times, which leads to an overall increase of the average formation time with temperature. However, even at elevated temperatures, the average formation time does not exceed the value of 100 femtoseconds, i.e., a value close to the time measured for free, i.e., self-trapped polarons. Analyzing individual trajectories, it is found that the time required for the structural relaxation of the polarons depends sensitively on the excitation of the lithium niobate high-frequency phonon modes and their phase relation.
M. Krenz, A. Bocchini, T. Biktagirov, A. Kozub, S. Badalov, S. Neufeld, I. A. Ruiz Alvarado, U. Gerstmann, W. G. Schmidt
Hyperpolarizabilities of LiNbO, LiTaO and KNbO calculated from First Principles
Abstract
LiNbO\(_3\), LiTaO\(_3\), and KNbO\(_3\) are ferroelectric oxides that are largely employed in technological applications due to their strong nonlinear optical response. In this manuscript, we calculate their second and third order nonlinear susceptibilities from first principles. Two different approaches are employed, based on the perturbative estimate of the optical response in the frequency-domain, and on the time-evolution of the electric polarization, respectively. The two approaches predict second harmonic coefficients for LiNbO\(_3\) which are in excellent agreement with each other. We furthermore calculate second and third order nonlinear susceptibilities of all investigated materials. Quasiparticle effects, accounted for by means of a scissors-shift, blue-shift the calculated spectral features and reduce their intensity. A comparison of the linear and nonlinear optical response suggests that the spectra can be understood by multi-photon adsorption within the fundamental bandgap of the modeled ferroelectric oxides.
Mike N. Pionteck, Felix Bernhardt, Christof Dues, Kevin Eberheim, Christa Fink, Kris Holtgrewe, Florian A. Pfeiffer, Nils A. Schäfer, Leonard M. Verhoff, Ferdinand Ziese, Simone Sanna

Computational Fluid Dynamics

Frontmatter
Study of Data-Driven Prediction of Roughness Skin Friction
Abstract
The potential of developing a machine learning predictive model for the roughness equivalent sand-grain size \(k_s\) is assessed in the present work. The training dataset for the machine learning model is obtained by carrying out Direct Numerical Simulations (DNS) on the artificially generated roughness at friction Reynolds number Re\(_\tau =800\). The generation of the roughness is based on a mathematical algorithm in which the roughness power spectrum (PS) as well as the height probability density function (PDF) can be prescribed. A roughness repository that contains 4000 artificial roughness is constructed with the generation algorithm. 50 roughness out of the repository are used for training the model. The selection of the training roughness samples is made following the active learning (AL) framework, where the AL framework selects roughness samples based on the model uncertainty. The ensemble neural network (ENN) model is employed to quantify the model uncertainty. Excellent performance of the ENN model is observed in the present work. Eventually an averaged prediction error of 8.5% is achieved for the ENN model tested on 4 realistic surfaces.
Jiasheng Yang, Alexander Stroh, Pourya Forooghi
Favorable-Pressure-Gradient Influence on Supersonic Film Cooling with Turbulent Main Flow
Abstract
Cooling of a hot supersonic turbulent boundary-layer flow by wall-parallel blowing of a high-heat-capacity gas through a backward-facing step is an effective method to protect the wall from heat overload. Detailed fundamental investigations in previous work concentrated on the flat-plate film-cooling case with zero streamwise pressure gradient (ZPG). In the application to the nozzle-extension wall of a rocket engine however the pressure decreases significantly, accelerating the flow, and it needs be known whether this has a significant effect on the mixing and heat-transport processes. To this end, a fundamental direct-numerical-simulation study has been performed using largely the same setup as in the ZPG case to elucidate the influence of a ‘favorable’ pressure gradient on the cooling and mixing region for a superheated-steam flow with initial Mach number 3.3, cooled by helium injection with Mach 1.8.
Johannes M. F. Peter, Tobias Gibis, Markus J. Kloker
An Investigation of Information Flux between Turbulent Boundary Layer and Porous Medium
Abstract
The interaction between boundary layer turbulence and a porous layer is the cornerstone to the interface engineering. In this study, the spatial resolved transfer entropy is used to assess the asymmetry of the causal interaction next to a permeable wall. The analysis was based on pore-resolved direct numerical simulation of turbulent channel flow over a cylinder array. The spatial map of transfer entropy reveals the information flux between the porous medium and arbitrary nearby position.
Xu Chu, Wenkang Wang, Bernhard Weigand
Towards DNS of Droplet-Jet Collisions of Immiscible Liquids with FS3D
Abstract
In-air microfluidics became a new method for technical production processes with ultra-high throughput formerly performed in micro channels. Direct Numerical Simulations (DNS) provide a valuable contribution for the fundamental understanding of multiphase flow and later application design. This chapter presents a feasibility study with first DNS results of droplet-jet collisions of immiscible liquids using the in-house software Free Surface 3D (FS3D). Two cases were investigated with a setup comparable to experiments by Baumgartner et al. [1], where a droplet chain of a glycerol solution hits a jet of silicon oil which encapsulates the droplets. The droplets’ shapes present are observed to be more complex than comprehensible from the two-dimensional images from the experiments. Thus, DNS with FS3D can provide additional information like the surface area or the velocity contributions in order to find analytical models of such collision processes in the future. Simulations of such increasingly complex systems require constant improvement of the numerical solver regarding the code’s performance. Thus, the red-black Gauss-Seidel smoother in the multi-grid solver, the iterative red-black scheme to compute the viscous forces as well as the momentum advection method were enhanced with a cache- and memory usage optimization. An overall performance gain of up to \(33\%\) was obtained for a representative test case.
Johanna Potyka, Jonathan Stober, Jonathan Wurst, Matthias Ibach, Jonas Steigerwald, Bernhard Weigand, Kathrin Schulte
On the Effects of Wing-Gust Interactions and Wing Flap Deflections on the HTP Aerodynamics
Abstract
Gust loads on aircraft are critical for the structural wing design. This paper investigates the impact of a critical vertical “1-cos” type gust event and different wing flap deflections on the aerodynamics of the Horizontal Tail Plane (HTP) and thus the pitching moment behavior of the aircraft. The wing flaps are deployed for the purpose of active gust load control and comprise spanwise segmented trailing edge flaps and leading edge flaps. URANS simulations are used for the simulations of two generic aircraft configurations with and without empennage at transonic flow conditions. The simulations reveal a significant impact of the wing-gust interaction and the wing flap deflections on the wing’s downwash angle. The change in the downwash affects the effective angle of attack at the HTP. However, the effects of the wing flap deflections on the HTP loads are less significant than the effect introduced by the reference gust. The deflections of the trailing and leading edge flaps on the wing are shown to be uncritical with regard to the pitching moment behavior of the aircraft. The gust and the wing flap induced maximum pitching moments can be significantly reduced via moderate elevator deflections on the HTP. A first approximation based on parametric 2D studies reveals a reduction by approx. 70% of the latter.
Junaid Ullah, Marco Hillebrand, Maximilian Ehrle, Thorsten Lutz, Wolfgang Heinze, Jochen Wild
Advances in Computational Process Engineering using Lattice Boltzmann Methods on High Performance Computers
Abstract
The present annual report documents the parallelization benefits of Lattice-Boltzmann particulate multiphase flow simulations realized in the open source software library OpenLB and performed on the HoreKa supercomputer. For this purpose, the parallel efficiency and total performance of test cases utilizing vectorization on CPUs as well as GPUs is evaluated for simulations up to 10 billion cells and node counts up to 320 CPU resp. 56 GPU nodes. Selected applications of these capabilities to resolved turbulent large eddy multiphase simulations conducted using Eulerian and Lagrangian approaches are described.
Adrian Kummerländer, Fedor Bukreev, Simon F. R. Berg, Marcio Dorn, Mathias J. Krause
Performance Improvements for Large-Scale Simulations using the Discontinuous Galerkin Framework FLEXI
Abstract
Large-scale simulations pose significant challenges not only to the solver itself but also to the pre- and postprocessing framework. Hence, we present generally applicable improvements to enhance the performance of those tools and thus increase the feasibility of large-scale jobs and convergence studies. To accomplish this, we use a shared memory approach implemented in the Message Passing Interface (MPI) libraries. Additionally, we improve the read and write performance of the flow solver during runtime to minimize the load imposed on the file system. A detailed discussion of the current performance and scaling behavior is given for up to 262144 processes. FLEXI shows excellent scalability for all tested features. We conclude by showing selected applications, where we use the introduced improvements to maximize performance.
Marcel Blind, Patrick Kopper, Daniel Kempf, Marius Kurz, Anna Schwarz, Claus-Dieter Munz, Andrea Beck
Applications of a Discontinuous Galerkin Chimera Method on 3D Flow Problems
Abstract
Within the DGDES project a computational fluid dynamics (CFD) code has been developed using a discontinuous Galerkin (DG) method for the spatial discretization. The DG method is a high-order method in space reducing the amount of cells required for numerical simulations compared to traditional CFD solvers. The aim of the project is to investigate the potential of the DG method for the calculation of the flow phenomena at helicopter rotors. The present study concerns the implementation of a Chimera method and its application to 3D problems dealing with unsteady flows and moving geometries. The results for a test case of the flow past a sphere reveal promising results compared to the reference. Furthermore, a first application to a 3D rotor in hover and in forward fligth are presented. In addition, the parallel performance is investigated and shows the good scalability of the code on the HPE Apollo Hawk system installed at the High Performance Computing Center Stuttgart (HLRS).
Fabian Genuit, Manuel Keßler, Ewald Krämer
Modeling the Gas Liquid Interface of Falling Film Reactors in Fully Developed Flow Regime
Abstract
In falling film reactors, the time scale of reaction is typically faster than the time scale of the mass transfer; hence the overall efficiency of the reactor is limited by the rate of mass transport to the reactive interface, which in turn depends on the effective surface area between the liquid phase and the gas phase. Therefore, the performance of these reactors strongly depends on how well the wavy interface dynamics and their influence on the reactive transport are understood at the most fundamental level. In this work, we focused on the numerical analysis of the wavy interface for alternative liquid distribution strategies with Smoothed Particle Hydrodynamics (SPH). In particular, we investigated the flow development in the entrance region and how it evolves into a fully developed region. We also analyzed the film statistics by extracting the probability density functions of the local film thicknesses in both time and frequency domains. Comparisons between SPH simulations and the available literature confirmed that the deployed numerical solution methodology can capture the global interface dynamics. However, higher residence times must be computed to fully capture the complex local mixing patterns in the fully developed region, which cannot be captured with periodic boundary conditions.
K. V. Muthukumar, M. Okraschevski, N. Bürkle, D. M. A. Bermudez, M. Haber, R. Koch, H.-J. Bauer, C. Ates

Transport and Climate

Frontmatter
Impact of Land-Use Change and User-Tailored Climate Change Information from a High-Resolution Climate Simulation Ensemble
Abstract
The KIT KLIWA ensemble of very high resolution climate simulations has been completed at HLRS in 2021 and is now available as a quasi-transient data set from 1971–2100 providing climate information down to the kilometer scale. A scale in which deep convection can be explicitly resolved by the model and is no longer parameterized. Within the report we analyse the benefit of those high-resolution simulations focusing on application issues and user-ready indices for climate adaptation. We found that climate change information could be resolved on the local scale, which showed dependencies of the projections on major landscapes as well as on land use. Furthermore, we analyse the impact of land use changes in climate models. Sensitivity studies using observed past land use changes show improved representation of the annual and summer temperature in large parts of Europe compared to a simulation with constant land cover. A second experiment with targeted land use changes towards an afforestation with deciduous trees shows that this could reduce peak heat wave temperatures compared to the prevalent coniferous forests.
Hendrik Feldmann, Marie Hundhausen, Regina Kohlhepp, Marcus Breil
Closing the Scale Gap for Resolved-Turbulence Simulations in Meteorology
Abstract
Geophysical flow is generally characterized by huge Reynolds numbers—which limits our ability to directly represent these systems on a computer. Most practical applications such as weather forecasting or climate projection rely on the representation of small-scale processes, one of which is turbulent mixing, by parameterizations. Sometimes, however, an explicit representation is inevitable to further process-level understanding and allow for an informed representation of mixing processes in parameterizations. While an explicit representation of turbulence is not possible across the entire geophysical range of scales, hydrodynamic/Reynolds-number similarity can be exploited to quantitatively extrapolate the behavior at reduced scale to the geophysically relevant limit of large scale separation. We outline here the underlying methodological framework and illustrate the approach by two examples, namely a general formulation of the velocity profiles in Ekman flow and the explicit representation of roughness in a channel flow resembling a boundary layer modeled in a wind tunnel.
Cedrick Ansorge, Jonathan Kostelecky
WRF Simulations to Investigate Processes Across Scales (WRFSCALE)
Abstract
Scientific aspects ranging from boundary layer research and land modification experiments to the investigation of a high-impact weather situation in southwestern Germany were addressed with the Weather Research and Forecasting (WRF) model from the km-scale down to the turbulence-permitting scale. Case study simulations in as different regions as the central United States, the United Arab Emirates and southwestern Germany were performed to investigate the evolution of the convective boundary layer or a severe thunderstorm. The multi-nested WRF setup, driven by the operational analysis of the European Centre for Medium-range Weather Forecasts (ECMWF), high-resolution terrain, and land cover data sets simulated a realistic evolution of the internal turbulent structure of the boundary layer including the transitions between daytime and nighttime conditions. Land modification simulations in the United Arab Emirates demonstrated that plantations as small as 10\(\,\times \,\)10 \(\textrm{km}^2\) could modify the weather pattern in this area in a way that more precipitation reaches the desert.
Oliver Branch, Hans-Stefan Bauer

Bioinformatics

Frontmatter
Data-Driven Multiscale Modeling of Self-Assembly and Hierarchical Structural Formation in Biological Macro-Molecular Systems: Pyruvate Dehydrogenase Complex
Abstract
Macro-molecular self-assembly and hierarchical structural formation are crucial for a variety of systems in nature and technology. Especially biological systems often rely on a specific structural organization to enable their function. Examples are multi-enzyme complexes enabling catalytic activity through structure-based phenomena such as metabolic channeling or the self-assembly of virus capsids necessary for transport of the genetic material and overall infection process. This project attempts to improve understanding and modeling capabilities of such systems by developing a multiscale modeling methodology for self-assembly on the scales of micro-meters and milli-second including a data-driven parameterization approach. As model systems the hepatitis B core antigen (HBcAg) and pyruvate dehydrogenase complex (PDC) are used, which feature a macro-molecular self-assembly crucial in enabling their function.
P. N. Depta, Maksym Dosta, S. Heinrich
Parameter Study of Solvent Systems by Molecular Dynamics Simulations (Project: EnzSim)
Abstract
Characterizing the dependence of the thermophysical properties of complex liquid mixtures on parameters such as composition and temperature is pivotal to the choice of an optimal solvent in process engineering. Therefore, it is indispensable to perform comprehensive parameter studies for the exploration of design space. Molecular simulation is a powerful tool for the prediction of properties under conditions that have not yet been explored experimentally. However, simulation results have to be calibrated with published experimental data. In order to make experimental and simulated data available to a comprehensive analysis, we developed a data management and analysis platform based on the standard data exchange format ThermoML. The practicability of integrating thermophysical data from experiment and simulation was demonstrated for two binary mixtures, methanol-water and glycerol-water, by systematically studying the dependence of densities and diffusion coefficients from water content and temperature. Experimental data was extracted manually from literature. The same parameter space was explored by comprehensive molecular dynamics simulations, whose results were directly transferred to the analysis platform. The usefulness of data integration was illustrated by assessing the transferability of the force fields, which had been developed for pure compounds at a specific temperature to different compositions and temperatures, and by analyzing the excess mixing properties as a measure of non-ideality of methanol-water and glycerol-water mixtures. The core of the data management and analysis platform is the newly developed Python library pyThermoML, which represents metadata, the parameters and the experimentally determined or simulated properties as Python data classes. The feasibility of a seamless data flow from data acquisition to a comprehensive data analysis and publication on Dataverse was demonstrated. Because the Dataverse datasets are in ThermoML format, the data is findable, accessible, interoperable, and reusable (FAIR).
Matthias Gueltig, Jan Range, Benjamin Schmitz, Juergen Pleiss

Miscellaneous Topics

Frontmatter
Simulations of Crystal Growth Using Lattice Boltzmann Formulation
Abstract
Crystallization is widely applied in many industrial applications, in particular for chemical engineering and pharmaceutical products. This project will focus on the modeling of crystal growth with a phase-field model using the lattice Boltzmann method. First, simulations of snowflake growth are carried out under different ambient conditions (humidity and temperature) in the plate-growth regime of the Nakaya diagram. The resulting crystal habits show that the phase-field model correctly captures both the crystal shape and the onset of primary and secondary branching instabilities. After that, the growth rate of a single crystal of S-mandelic acid (S-ma) is computed and validated by comparison with experimental results. Finally, the approach is extended to investigate the impact of forced convection on the crystal habits both on the snowflakes and S-ma crystals. Based on there parametric studies, a modification of the reactor geometry is proposed that should reduce the observed deviations from symmetrical growth for the single S-ma crystal.
Q. Tan, S. A. Hosseini, D. Thévenin
High-Performance Computing as a Key to New Insights into Thermodynamics
Abstract
High-performance computing plays an increasing role in engineering since it allows for the investigation of problems on multiple scales. Outstanding results can be achieved when performing simulations at the molecular level because of their extreme resolution in time and space. Such simulations are not only conducted in natural sciences, but also play an important role in the field of engineering thermodynamics. The present work utilizes molecular dynamics (MD) and Monte Carlo (MC) simulations to gain insight into a variety of different thermodynamic problems. The coalescence of two droplets is addressed as well as the development of a Tang-Toennies potential for the representation of the thermodynamic properties of argon. In addition, a detailed look at the Fisher-Widom line and the Widom line is taken. Finally, coefficients for the coupled transport of heat and mass and vapor-liquid, liquid-liquid and vapor-liquid-liquid equilibria are discussed.
Simon Homes, Ivan Antolović, Robin Fingerhut, Gabriela Guevara-Carrion, Matthias Heinen, Isabel Nitzke, Denis Saric, Jadran Vrabec
Geometric Constrained Scalable Algorithm for PDE-Constrained Shape Optimization
Abstract
In this project, the parallel performance of shape optimization schemes is investigated using varying core counts for several levels of refinement. Weak scalability results are presented for up to 49 152 cores on the national supercomputer HAWK, and the effects on performance related to a high number of cores-per-node is discussed. The algorithm developed during the project is benchmarked in an optimal design problem oriented towards fluid dynamics applications. This work is concerned with situations where it is necessary to retain certain geometric properties, such as volume and barycenter. Therefore, the optimization process solves a PDE-constrained optimization problem that incorporates the imposed geometric constraints in the descent directions. It is done by solving the corresponding minimization problem in appropriate Banach spaces. 2d and 3d results are presented where the domain needs a high number of elements to be properly defined. The system of equations consists of a very large number of degrees of freedom, thus it can only be solved efficiently via the most modern distributed-memory systems, such as HAWK at HLRS.
Jose Pinzon, Martin Siebenborn, Andreas Vogel
First-Principles Study of NV Centers Near Extended Defects
Abstract
We present a density functional theory analysis of the negatively charged nitrogen-vacancy (NV\(^-\)) defect complex located at or in the vicinity of (001) and (111) surfaces as well as 30\(^{\circ }\) and 90\(^{\circ }\) partial glide dislocations in diamond. Formation energies, electronic density of states, geometrical deformations, hyperfine structure and zero-field splitting parameters of NV\(^-\) centers in such distorted environments are analyzed. The formation energies of the NV\(^-\) centers at the dislocation cores are up to 3 eV reduced compared to the bulk of diamond whereas near the surfaces the change of the formation energy is not significant (\(\sim \)0.1 eV). For the 30\(^{\circ }\) partial glide dislocation the lowest energy configuration, where the NV\(^-\) axis is oriented parallel to the dislocation line, exhibits a stable triplet ground state. This lowest energy configuration of NV\(^-\) center at the core of 30\(^{\circ }\) partial glide dislocation has hyperfine constants and ZFS values which deviate by \(3\%\) from the NV\(^-\) bulk values and is thus an interesting candidate for self-assembled NV\(^-\) arrays. These crucial theoretical results required time consuming electronic structure calculations, for which we used the HoreKa computing cluster.
Reyhaneh Ghassemizadeh, Wolfgang Körner, Daniel F. Urban, Christian Elsässer
Lattice Boltzmann Simulation of Flow, Transport, and Reactions in Battery Components
Abstract
Microstructures of battery components largely affect electrochemical properties of the whole battery cell. In this context, especially physical phenomena occurring in their pores play a dominant role. However, experimental studies of these phenomena are hardly feasible. Therefore, the lattice Boltzmann method is applied to provide a detailed insight into the relevant processes at the pore scale. The following topics are covered: (1) electrolyte filling of structurally resolved lithium-ion battery electrodes, (2) impact of gas entrapment on battery performance, (3) development and validation of a new model for multi-phase flow in homogenized porous media, (4) chemical surface reactions, and (5) species transport in dynamically changing microstructures. The results give a first insight into how battery performance can be optimized by adapting structural and physico-chemical properties as well as process parameters.
Martin P. Lautenschlaeger, Julius Weinmiller, Benjamin Kellers, Thomas Jahnke, Timo Danner, Arnulf Latz
Fault Tolerant Molecular-Continuum Flow Simulation
Abstract
Molecular-continuum simulations couple molecular dynamics (MD) and computational fluid dynamics (CFD) simulations in a domain decomposition sense to assess fluid flow, e.g., in process engineering applications, at the nanoscale. Running these simulations on extreme-scale supercomputers, an issue consists in single compute cores or nodes failing due to hardware- or software-sided errors. This imposes a challenge to robustness of numerical simulations and, as such, also to molecular-continuum systems. We introduce a fault tolerance method in our macro-micro-coupling tool (MaMiCo) that has been developed in the past as molecular-continuum simulation software solution. With MaMiCo leveraging ensemble simulations to cope with statistical errors in the MD solutions, we extended the ensemble approach to recognize failing MPI processes and react to these failures. Once a failure is encountered, the affected MD simulations are removed from these MPI processes and relaunched on well-operating MPI process groups. We detail our approach and report scalability results for our approach, achieved on the supercomputer HAWK at HLRS.
Vahid Jafari, Piet Jarmatz, Helene Wittenberg, Amartya Das Sharma, Louis Viot, Felix Maurer, Niklas Wittmer, Philipp Neumann
Metadata
Title
High Performance Computing in Science and Engineering '22
Editors
Wolfgang E. Nagel
Dietmar H. Kröner
Michael M. Resch
Copyright Year
2024
Electronic ISBN
978-3-031-46870-4
Print ISBN
978-3-031-46869-8
DOI
https://doi.org/10.1007/978-3-031-46870-4

Premium Partner