Day 1 - June, 5th 2017

Session 1, 17:00-18:30

G. Jung, Iterative Reconstruction of Memory Kernels

G. T. Vu, Mean curvature as a guiding field for block copolymer conformation

C. Scherer, Evaluating many-body effects in systemtic coarse graining

Day 2 - June, 6th 2017

Session 2, 10:00-11:30

A. Disterhoft, A short introduction to weak formulations of partial differential equations and galerkin methods

D. Spiller, Spinodal decomposition of polymer-solvent systems: first steps towards treating 2D systems

P. Strasser, Energy-stable numerical schemes for multiscale simulations of polymer-solvent mixtures

Session 3, 14:00-16:00

P. Koss, Estimation of crystal nucleation barriers for colloidal crystals from computer simulations

F. Knoch, Time-dependent nonequilibrium Markov state modeling

S. Wörner, Kinetic properties of liquid crystals from coarse-grained and atomistic molecular dynamics simulations

F. Kössel, A kinetic theory for dynamics of self-propelled magnetic suspensions

Day 3 - June, 7th 2017

Free time suggestion: group trip by ship to Bregenz and hiking on the "Pfänder" for a beautiful view over the Bodensee.

Day 4 - June, 8th 2017

Session 4, 10:00-11:30

T. Pal, All-atom molecular dynamics simulation of ionic liquid films on silica surface

D. Lesnicki, Understanding water relaxation at interfaces

V. R. Ardham, A thermodynamic model to simulate particles at an oil-water interface

Session 5, 14:00-16:00

H. Guzman, Scalable and fast heterogeneous molecular simulation with predictive parallelization schemes

M. Heidari, Chemical potential calculation of dense fluids using hamiltonian adaptive resolution simulation

S. Stalter, Molecular dynamics simulations in hybrid particle-continuum schemes: pitfalls and caveats

Session 6, 18:00-19:00

S. Köhler, Computer simulations at BASF

Day 5 - June, 9th 2017

Session 7, 10:00-11:00

D. Reith, Autonomous cars at Bosch

Session 8, 11:00-12:00

K. Schäfer, Refolding dynamics of molecular constructs after a force quench

S. Jaschonek, Molecular dynamics simulations of phase transition in phospholipid bilayers

Session 9, 18:00-19:00

B. Trefz, IT Consulting

Day 6 - June, 10th 2017

Session 10, 10:00-11:00

T. Kemmer, Efficient nonlocal protein electrostatics in the Julia programming language

F. Padua, File locking and concurrency control for filesystems: understanding real-word applications

Iterative Reconstruction of Memory Kernels

G. Jung, F. Schmid

Institut für Physik, Johannes Gutenberg-Universität Mainz, Germany

In recent years it has become increasingly popular to construct coarse-grained models with Non-Markovian dynamics in order to account for an incomplete separation of time scales. One challenge of a systematic coarse-graining procedure is the extraction of the dynamical properties, namely the memory kernel, from all-atom simulations. In this talk we will present the iterative memory reconstruction (IMR) [1], which is the first iterative method suggested in the field of Non-Markovian modeling. This new extraction technique shows various similarities to the reconstruction of static potentials via inverse Monte Carlo (IMC) [2] or iterative Boltzmann inversion (IBI) [3] . The big advantage compared to previously proposed techniques (e.g. [4, 5]) is the efficient reconstruction of coarsegrained models that have precisely the same dynamical properties as the original finegrained systems.

We demonstrate the performance of the above described methods at the example of the anomalous diffusion of a single colloid. For this system we are able to reconstruct realistic coarse-grained models for timesteps about 200 times larger then the original molecular dynamics simulations.

[1] G. Jung and F. Schmid, submitted to JCTC

[2] A. P. Lyubartsev and A. Laaksonen, Phys. Rev. E 52, 3730 (1995)

[3] D. Reith, M. Pütz, F. Müller-Plathe, Comp. Chem. 24, 1624 (2003)

[4] H. K. Shin, C. Kim, P. Talkner and E. K. Lee, Chem. Phys. 375, 316 (2010)

[5] A. Carof, R. Vuilleumier and B. Rotenberg, J. Chem. Phys. 140, 124103 (2014)

Mean curvature as a guiding field for block copolymer conformation

G. T. Vu, A. Abate, L. Gómez, A. Pezzutti, R. Register, D. Vega, F. Schmid

Experimental data on cylinder-forming block copolymer thin films deposited on curved substrates and free standing membranes show that the local orientation of the pattern are strongly coupled with the geometry in which the system is embedded. This coupling affects defect dynamics and also the shape developed by free-standing membranes. Here we performed self-consistent field theory (SCFT) calculations to determine how equilibrium configurations and the elastic bending constants are affected by a geometric field. The stability of the block copolymer thin film against dewetting due to curvature is also analysed. In good agreement with experiments, it is found that the competition between the strong length scale selectivity imposed by the block copolymer structure and curvature dictates pattern orientation.

Evaluating many-body effects in systematic coarse graining

Christoph Scherer

Particle based coarse-graining (CG) is a systematic way of reduction of the number of degrees of freedom describing a physical system. It involves three steps: choice of the CG degrees of freedom, identification of a merit function which quantifies the difference between the fine- and coarse-grained representations, and determination of the CG potential energy surface (PES). The entire procedure is sensitive to the number and types of basis-functions employed in the CG representations: for example, the incorporation of nonbonded three-body interactions in a coarse-grained water model helps to reproduce both thermodynamic and structural properties [1,2]. In this work, we investigate the effect of extending the basis set to three-body interactions for several organic solvents and formulate clear criteria for when these many-body terms are required. The coarse-graining scheme is implemented in the VOTCA-CSG toolkit [3].

[1] V. Molinero, E. B. Moore, J. Phys. Chem. B, 113, 4008 (2009)

[2] L. Larini, L. Lu, G.A. Voth, J. Chem. Phys. 132, 164107 (2010)

[3] V. Rühle, C. Junghans, A. Lukyanov, K. Kremer, D. Andrienko, J. Chem. Theory. Comput. 5, 3211 (2009)

A Short Introduction to Weak Formulations of Partial Differential Equations and Galerkin Methods

Alexej Disterhoft

Partial differential equations (PDEs) are fundamental to physics and chemistry; often used to describe a wide variety of phenomena. Unfortunately, predominantly only simple examples can be solved analytically. Most PDEs do not even have a solution in the classical sense. The concept of weak solutions lifts some of these problems. For many PDEs, this theoretic framework not only answers the question whether a (unique) solution exists, but also allows the construction of computational methods to calculate approximations. Exemplary applications of this framework, and so-called Galerkin methods based thereon, are presented.

Spinodal Decomposition of Polymer-Solvent Systems: First Steps Towards Treating 2d Systems

Dominic Spiller, Burkhard Dünweg, Maria Lukacova and Paul Strasser

We investigate the dynamics of a polymer system falling out of solution, using two complementary methodological approaches: On the one hand, viscoelastic Cahn-Hilliard (CH) continuum equations are solved numerically on a grid (contribution by Paul Strasser). On the other hand, we use a coupled Lattice Boltzmann (LB) / Molecular Dynamics (MD) scheme to simulate a bead-spring model for a system of chain molecules in a hydrodynamic background, where the solvent quality is taken into account by the strength of the non-bonded monomer-monomer interaction. Since the CH calculations are done in two dimensions, we modify our three-dimensional simulation code in order to study two dimensions as well; this is done via a suitable external confining potential. For such a system, we study and test in detail how the virial equation of state must be evaluated, in order to obtain a suitable parametrization of the continuum free-energy functional. First numerical results on these investigations will be presented.

Energy-stable numerical schemes for multiscale simulations of polymer-solvent mixtures

Paul Strasser

Introducing energy stable numerical schemes to solve a macroscopic equation system aiming at the modeling of the dynamics of polymer-solvent mixtures. Focus is the time discretization of the system, since the space discretization done by a combined Finite Difference / Finite Volume scheme is known to conserve physical quantities. A complementary approach to study the same physical system is realized by simulations of a microscopic model based on a hybrid Lattice Boltzmann / Molecular Dynamics scheme. These latter simulations provide initial conditions for the numerical solution of the macroscopic equations to compare the dynamics of both time scales.

Estimation of crystal nucleation barriers for colloidal crystals from computer simulations

P. Koss

A fluid in equilibrium in a finite volume, with a density exceeding the onset of freezing, may exhibit phase coexistence of a crystal nucleus surrounded by liquid. In classical nucleation theory, the barrier of homogeneous nucleation is given by two contributions, the energy gain of creating a droplet and the energy loss due to surface tension of the newly created interface. Using a computational method suitable for the estimation of the chemical potential of dense fluids we obtain the excess free energy due to the surface of the crystalline nucleus. Our novel analysis method is appropriate for crystal nuclei of all shapes without suffering from ambiguities occurring when one needs a microscopic identification of the crystalline droplet. We report that the nucleation barrier for a soft version of the effective Asakura-Oosawa model[1] is compatible with a spherical shape, and consistent with classical nucleation theory [2].

[1] M. Dijkstra, R. van Roij and R. Evans, Phys. Rev. E 59, 5744-5771 (1999).

[2] A. Statt, P. Virnau, and K. Binder, Phys. Rev. Lett. 114, 026101 (2015).

Time-dependent nonequilibrium Markov state modeling

Fabian Knoch, Ken Schäfer, Gregor Diezemann and Thomas Speck.

A major current challenge in statistical mechanics poses the systematic construction of coarse-grained Markov state models that are dynamically consistent, and moreover might be used for systems driven out of thermal equilibrium. We previously introduced a novel prescription that extends the Markov state modeling approach to systems with dynamics breaking detailed-balance. Here, we introduce the concept of nonequilibrium Markov state modeling to systems driven by general time-dependent periodic as well as nonperiodic protocols. We illustrate our new methodology for both types of driving protocols by giving two examples: alanine dipeptide exposed to an oscillating electric field and the Calix[4]arene-catenane dimer elongated employing a constant-velocity pulling protocol.

Kinetic properties of liquid crystals from coarse-grained and atomistic molecular dynamics simulations

Svenja Wörner, Joseph Rudzinski, Kurt Kremer, and Tristan Bereau

Max-Planck-Institut für Polymerforschung, Mainz, Germany

Liquid crystals display liquid-like behavior while maintaining a long-range crystalline order, giving rise to unique material properties. To understand macroscopic processes, e.g. phase transitions, large systems need to be studied on time scales not accessible by atomistic models. Coarse grained models make these sizes and time scales accessible.

In this work we investigate the liquid crystal mesogen 8AB8, containing a stiff, photoisomerizable azobenzene core with flexible alkyl tails. Mukherjee et al. have previously developed a coarse-grained model of 8AB8, which displays the correct thermodynamic and structural properties. Not only is it able to form a smectic phase, the model also reproduces the transition temperature of the disordered to smectic phase transition of the underlying atomistic model. Reducing the degrees of freedom usually leads to a non-trivial modification of the timescales for different processes sampled by the coarse-grained model. Two well-characterized translocation pathways in the smectic A phase are studied in detail, utilizing a Markov state model framework to systematically assess the differences in the transport kinetics between the coarse-grained and atomistic models. Investigating the precise source of the discrepancies between the two models implicates an approach for reparametrization of the coarse-grained model.

A kinetic theory for dynamics of self-propelled magnetic suspensions

Fabian Rouven Kössel and Sara Jabbari-Farouji

Inspired by novel dynamical patterns in magnetotactic bacteria, we present a minimal kinetic model for dilute suspensions of magnetic self-propelled particles. Our kinetic theory couples a Fokker-Planck equation of active particles in an external magnetic field to a Stokes flow equation. Including the hydrodynamic stress contributions of self- propulsion and magnetic torque in the Stokes flow allows us to investigate, the interplay between internal and external drives on the dynamics and effective viscosity of active suspensions. Through the linear stability analysis and full numerical simulations of our model, we ex- amine the role of the external magnetic field on stability of aligned suspensions and their complex flow patterns.

All-Atom Molecular Dynamics Simulation of Ionic Liquid films on Silica Surface

Tamisra Pal and Michael Vogel

Institut für Festkörperphysik Technische Universität Darmstadt , 64289 Darmstadt , Germany

Room temperature ionic liquid films in confined geometries have been recognized for their significant interfacial properties in electrochemical and electronic devices. Depending on the hydrophobicity of the anions, we chose ionic liquid 1-butyl-3-methylimidazolium cation with hexafluorophosphate ([Bmim][PF6]) and tetrafluoroborate ([Bmim][BF4]) counterparts. Here, the dynamical and structural properties of these ionic liquid confined between amorphous silica slabs have been investigated by all-atom molecular dynamics simulation studies at 300 K. Relative number densities of the ions are calculated near the surface, as well as in the middle of the slit. The more hydrophilic [BF4]- ions tend to stay closer to the slab wall than symmetric [PF6]-, whereas the [Bmim]+ ions always resides in the next layer forming a bi-layered arrangement from the wall. A preferred orientation has been observed for the cations with their methyl groups pointing towards the slab surface and the butyl tail projected inwards. The middle of the slit displays more of a bulk behavior in terms of density and ion diffusivities. We have performed spatially-resolved analysis of the mean square displacement (MSD) and incoherent intermediate scattering function (ISF) to understand the very sluggish and heterogeneous dynamics of these ionic liquids in the vicinity of the silica surface, which need to be considered when designing applications.

Vibrational energy relaxation at water interfaces from ab initio molecular dynamics simulations

Dominika Lesnicki, Marialore Sulpizi

Department of Physics, Johannes Gutenberg University Mainz (JGU Mainz) Staudingerweg 7, 55128 Mainz - Germany

Ab initio molecular dynamics simulations represent a versatile approach to investigate the structure and dynamics of water interfaces, especially in the cases of heterogeneous systems where less computationally demanding techniques, such as force field approaches are not easily transferable. Recently, ab initio simulations have been used to calculate Sum Frequency Generation spectra, also thanks to new developments in the calculations of the response functions, which make use of velocity-velocity correlation functions incorporating the appropriate selection rules [1,2].

We present here a new approach to investigate vibrational energy relaxation at water interfaces from the analysis of ab initio molecular dynamics trajectories. We follow the energy relaxation from a locally excited vibrational state using suitable descriptors based on vibrational density of states.

Our method is applied to different water systems, including bulk water and fluorite/water interfaces at different pH. In the case of the fluorite/water interface at low pH we find that water behaves similarly to bulk water, while, in the case of high pH, instead, the energy relaxation is much slower. A molecular interpretation of the different time scales is provided.

[1] R Khatib, EHG Backus, M Bonn, MJ Perez-Haro, MP Gaigeot, M Sulpizi, Scientific Reports, 6 (2016)

[2] R. Khatib and M Sulpizi, J. Phys. Chem. Letters, 8, 6 (2017)

A thermodynamic model to simulate particles at an oil-water interface

Vikram Reddy Ardham, Frédéric Leroy

Technische Universität Darmstadt, Eduard-Zintl-Institut für Anorganische und Physikalische Chemie, Alarich-Weiss-Strasse 8, 64287 Darmstadt, Germany

Liquid-liquid interfaces offer a suitable environment to facilitate self-assembly of nano-particles driven by the free energy difference between the bulk and interface. Particularly, water-oil interfaces have been exploited to synthesize networks of graphene particles that are highly inter-connected and only up to a few layers thick. To simulate and study these systems, we develop and evaluate a thermodynamic method to generate models that are thermodynamically consistent with a reference system. The models here mean the different interaction potentials that are to be used to simulate particles using molecular simulation methods. We reformulate the wetting coefficient (!)¹ for a particle in a two-phase media in terms of an easily computable work of adhesion² and then use this as reference to generate models with different resolutions (all-atom/coarse-grained). We observe that the wetting coefficient accurately determines the equilibrium position of a particle where only dispersion forces are dominant. We however, observed that reproducing the thermodynamics does not necessarily reproduce the structural properties and must be taken care of separately by choosing appropriate liquid models. Further, we illustrate the applicability of method by simulating a relatively large water-oil interface with graphene particles forming an inter-connected network using a coarse-grain model.

[1] Sumita, M.; Sakata, K.; Asai, S.; Miyasaka, K.; Nakagawa, H. Polym. Bull. 1991, 25 (2), 265-271.

[2] Ardham, V. R.; Deichmann, G.; van der Vegt, N. F. A.; Leroy, F. J. Chem. Phys. 2015, 143 (24).

Scalable and fast heterogeneous molecular simulation with predictive parallelization schemes

Horacio V. Guzman, Kurt Kremer, and Torsten Stuehn

Max Planck Institute for Polymer Research, Ackermannweg 10, 55128 Mainz, Germany Christoph Junghans

Computer, Computational, and Statistical Sciences Division, Los Alamos National Laboratory, Los Alamos, NM 87545, USA

Abstract: Multiscale and inhomogeneous molecular systems are challenging topics in the eld of molecular simulation. In particular, modeling biological systems in the context of multiscale simulations and exploring material properties are driving a permanent development of new simulation methods and optimization algorithms. In computational terms, those methods require parallelization schemes that make a productive use of computational resources for each simulation and from its genesis. Here, we introduce the heterogeneous domain decomposition algorithm which is a combination of an heterogeneity sensitive spatial domain decomposition with an a priori sliding subdomain-walls procedure. The algorithm modeling is presented for dual resolution systems and inhomogeneous binary fluids, in terms of scaling properties as a function of the size of the low-resolution region and the high-to-low resolutions ratio. We also show the algorithm competences, by comparing it to both static domain decomposition algorithms and dynamic load balancing schemes. Mainly, two representative molecular systems have been simulated and compared to the heterogeneous domain decomposition proposed in this work. These two systems comprise an adaptive resolution simulation of a biomolecule solvated in water and a phase separated binary Lennard-Jones fluid.

Chemical Potential Calculation of Dense Fluids using Hamiltonian Adaptive Resolution Simulation

Maziar Heidari, Robinson Cortes-Huerto, Kurt Kremer, Raffaello Potestio

The calculation of chemical potential of fluids is a relevant and challenging problem in computational chemistry and physics. Here, we propose a method which employs the recently multi-scale Hamiltonian Adaptive Resolution Simulation (H-AdResS) method to calculate the chemical potential of dense fluids. In H-AdResS, the simulation domain is subdivided in regions of high and low resolutions, coupled through a hybrid region. Since the dynamics of particles are obtained from a global Hamiltonian, the generated statistical ensembles of the system are well-defined. In our method, the fluid within the high resolution region is coupled with an ideal gas of non-interacting particles, and to enforce a uniform density profile an external force is computed on-the-fly and applied. Then the converged compensation forces are integrated across the hybrid region to obtain the Gibbs free energy difference between the two resolutions. The resulting Gibbs free energy is related to the excess chemical potential of the fluid with respect to the ideal gas. We validated this method by calculating the excess chemical potentials of fluid mixtures.

Molecular dynamics simulations in hybrid particle-continuum schemes: Pitfalls and caveats

Stefanie Stalter, Nehzat Emamy, Leonid Yelash, Mária Lukácová, Peter Virnau

We aim to develop an efficient reduced-order hybrid multiscale method in order to simulate complex fluids. With help of the heterogeneous multiscale method (HMM), that makes use of the scale separation into macro- and microlevels, the method combines the continuum and the molecular description. The missing information on the macro-level is the unknown stress-tensor, evaluated by means of a molecular dynamics (MD) simulation under shear on the micro-level. In this talk I will discuss pitfalls and caveats, which arise for this application of MD simulations. This includes the choice of optimum system sizes, thermostats and realistic shear rates, which can be achieved.

Efficient nonlocal protein electrostatics in the Julia programming language

Thomas Kemmer, Andreas Hildebrandt

Software in the field of computational science is generally exposed to a particularly restrictive set of requirements, mainly owing to the natural complexity of its tasks and the algorithms to solve them. Inevitably, these requirements influence the choice of programming language, not only demanding a certain ease of use - allowing rapid prototyping and, thus, rapid adaption to changes in the experimental or theoretical setup - but also the computational efficiency to solve the tasks within a given set of time or hardware constraints. Computational efficiency becomes of particular importance in the case of multi-scale environments, such as the simulation of electrostatic interactions between biomolecules in highly structured solvents. One major challenge in this context is the treatment of the latter, as strong simplifications of the solvent structure invariably lead to considerable inaccuracies in the estimated potentials and free energies. Here, we present efficient protein electrostatics computations as a showcase example for the cross-platform and open-source Julia programming language, which provides a positive user experience comparable to Python or MATLAB alongside high-performance capabilities known from C and C++. By modeling water in an implicit but nonlocal fashion, we account for correlation of molecular polarization due to the water network around the solute and sustain accuracy without suffering from infeasible runtimes as compared to the explicit case.

File locking and concurrency control for filesystems: understanding real-word applications

Federico Padua

File locking is a form of serialization used by parallel filesystems to prevent data corruption or other bugs in case many programs or many processes access the same files at the same time. This is the case for a single MPI program, in which many processes read and write a single shared file or multiple files during the run. If user-level filesystems aim to replace some key functionalities of kernel level filesystems, then it becomes important to verify whether such a filesystem needs to support any form of concurrency control or none. In case a form of concurrency control is needed, it's also important to asses which one would be the best option.
Our interest in studying concurrent file accesses and file locking is motivated by one section of the POSIX specification:
"This volume of POSIX.1-2008 does not specify behavior of concurrent writes to a file from multiple processes. Applications should use some form of concurrency control." It then becomes fundamental that filesystem designers take into account this possible issue when developing their filesystems: either some form of concurrency control is needed or not.
This mandates a clear understanding and study of real world applications that run in HPC centers.

We hope that the study we are embarking on will open up a discussion and a rethinking of some assumptions. For instance, previous works tried to explore the benefits of optimistic concurrency control in the HPC context. We plan first to assess to what extent ad-hoc file system developers should care about concurrency control or not, to support one design choice over another: this can justify optimistic or pessimistic approaches.

In the case of more general purpose filesystems in HPC clusters, such as GPFS and Lustre, the default to concurrency control is to employ a pessimistic approach using some kind of distributed locking (with different level of granularity). We plan to use the study to provide a solid experimental foundation to see whether this pessimistic approach is needed even in the case of general purpose filesystems to correctly support HPC applications.