08 Fakultät Mathematik und Physik
Permanent URI for this collectionhttps://elib.uni-stuttgart.de/handle/11682/9
Browse
48 results
Search Results
Item Open Access Collective variables in data-centric neural network training(2023) Nikolaou, KonstantinNeural Networks have become beneficial tools for physics research. While they provide a powerful tool for data-driven modeling, their success is accompanied by a lack of interpretability. This thesis aims to add transparency to the opaque nature of NNs by means of collective variables, a concept well-known in the field of statistical physics. Three collective variables are introduced that emerge from the interactions between neurons and data. These observables enable one to capture holistic behavior of the network and are used to conduct an analysis of neural network training, focusing on data. Through the investigations, the collective variables are applied to selections from a novel sampling method: Random Network Distillation (RND). Besides studying collective variables, the investigation of Random Network Distillation as a data selection method composes the second part of this thesis. The method is analyzed and optimized with respect to its components, aiming to understand and improve the data selection process. It is shown that RND can be used to select data sets that are beneficial for neural network training, giving rise to its application in fields like active learning. The collective variables are leveraged to further investigate the selection method and its effect on neural network training, revealing previously unknown properties of RND-selected data sets. The potential of the collective variables is demonstrated and discussed from a data-centric perspective. They are shown to be discriminative towards the information content of data and give rise to novel insights into the nature of neural network training. In addition to fundamental research on neural networks, the collective variables offer several potential applications including the identification of adversarial attacks and facilitating neural architecture search.Item Open Access A brief review of capillary number and its use in capillary desaturation curves(2022) Guo, Hu; Song, Kaoping; Hilfer, R.Capillary number, understood as the ratio of viscous force to capillary force, is one of the most important parameters in enhanced oil recovery (EOR). It continues to attract the interest of scientists and engineers, because the nature and quantification of macroscopic capillary forces remain controversial. At least 41 different capillary numbers have been collected here from the literature. The ratio of viscous and capillary force enters crucially into capillary desaturation experiments. Although the ratio is length scale dependent, not all definitions of capillary number depend on length scale, indicating potential inconsistencies between various applications and publications. Recently, new numbers have appeared and the subject continues to be actively discussed. Therefore, a short review seems appropriate and pertinent.Item Open Access Machine learning-driven investigation of the structure and dynamics of the BMIM-BF4 room temperature ionic liquid(2024) Zills, Fabian; Schäfer, Moritz René; Tovey, Samuel; Kästner, Johannes; Holm, ChristianRoom-temperature ionic liquids are an exciting group of materials with the potential to revolutionize energy storage. Due to their chemical structure and means of interaction, they are challenging to study computationally. Classical descriptions of their inter- and intra-molecular interactions require time intensive parametrization of force-fields which is prone to assumptions. While ab initio molecular dynamics approaches can capture all necessary interactions, they are too slow to achieve the time and length scales required. In this work, we take a step towards addressing these challenges by applying state-of-the-art machine-learned potentials to the simulation of 1-butyl-3-methylimidazolium tetrafluoroborate. We demonstrate a learning-on-the-fly procedure to train machine-learned potentials from single-point density functional theory calculations before performing production molecular dynamics simulations. Obtained structural and dynamical properties are in good agreement with computational and experimental references. Furthermore, our results show that hybrid machine-learned potentials can contribute to an improved prediction accuracy by mitigating the inherent shortsightedness of the models. Given that room-temperature ionic liquids necessitate long simulations to address their slow dynamics, achieving an optimal balance between accuracy and computational cost becomes imperative. To facilitate further investigation of these materials, we have made our IPSuite-based training and simulation workflow publicly accessible, enabling easy replication or adaptation to similar systems.Item Open Access Tuning the properties and microstructuring of ionic liquid mixtures at surfaces through atomistic modeling(2021) Kobayashi, Takeshi; Fyta, Maria (Prof. Dr.)Item Open Access Percolativity of porous media(2022) Hilfer, Rudolf; Hauskrecht, J.Connectivity and connectedness are nonadditive geometric functionals on the set of pore scale structures. They determine transport of mass, volume or momentum in porous media, because without connectivity there cannot be transport. Percolativity of porous media is introduced here as a geometric descriptor of connectivity, that can be computed from the pore scale and persists to the macroscale through a suitable upscaling limit. It is a measure that combines local percolation probabilities with a probability density of ratios of eigenvalues of the tensor of local percolating directions. Percolativity enters directly into generalized effective medium approximations. Predictions from these generalized effective medium approximations are found to be compatible with apparently anisotropic Archie correlations observed in experiment.Item Open Access Modeling the translocation of DNA structures through nanopores(2021) Szuttor, Kai; Holm, Christian (Prof. Dr.)Item Open Access On extremal domains and codomains for convolution of distributions and fractional calculus(2022) Kleiner, T.; Hilfer, R.It is proved that the class of c-closed distribution spaces contains extremal domains and codomains to make convolution of distributions a well-defined bilinear mapping. The distribution spaces are systematically endowed with topologies and bornologies that make convolution hypocontinuous whenever defined. Largest modules and smallest algebras for convolution semigroups are constructed along the same lines. The fact that extremal domains and codomains for convolution exist within this class of spaces is fundamentally related to quantale theory. The quantale theoretic residual formed from two c-closed spaces is characterized as the largest c-closed subspace of the corresponding space of convolutors. The theory is applied to obtain maximal distributional domains for fractional integrals and derivatives, for fractional Laplacians, Riesz potentials and for the Hilbert transform. Further, maximal joint domains for families of these operators are obtained such that their composition laws are preserved.Item Open Access Water structuring induces nonuniversal hydration repulsion between polar surfaces : quantitative comparison between molecular simulations, theory, and experiments(2024) Schlaich, Alexander; Daldrop, Jan O.; Kowalik, Bartosz; Kanduč, Matej; Schneck, Emanuel; Netz, Roland R.Polar surfaces in water typically repel each other at close separations, even if they are charge-neutral. This so-called hydration repulsion balances the van der Waals attraction and gives rise to a stable nanometric water layer between the polar surfaces. The resulting hydration water layer is crucial for the properties of concentrated suspensions of lipid membranes and hydrophilic particles in biology and technology, but its origin is unclear. It has been suggested that surface-induced molecular water structuring is responsible for the hydration repulsion, but a quantitative proof of this water-structuring hypothesis is missing. To gain an understanding of the mechanism causing hydration repulsion, we perform molecular simulations of different planar polar surfaces in water. Our simulated hydration forces between phospholipid bilayers agree perfectly with experiments, validating the simulation model and methods. For the comparison with theory, it is important to split the simulated total surface interaction force into a direct contribution from surface-surface molecular interactions and an indirect water-mediated contribution. We find the indirect hydration force and the structural water-ordering profiles from the simulations to be in perfect agreement with the predictions from theoretical models that account for the surface-induced water ordering, which strongly supports the water-structuring hypothesis for the hydration force. However, the comparison between the simulations for polar surfaces with different headgroup architectures reveals significantly different decay lengths of the indirect water-mediated hydration-force, which for laterally homogeneous water structuring would imply different bulk-water properties. We conclude that laterally inhomogeneous water ordering, induced by laterally inhomogeneous surface structures, shapes the hydration repulsion between polar surfaces in a decisive manner. Thus, the indirect water-mediated part of the hydration repulsion is caused by surface-induced water structuring but is surface-specific and thus nonuniversal.Item Open Access PDADMAC/PSS oligoelectrolyte multilayers : internal structure and hydration properties at early growth stages from atomistic simulations(2020) Sánchez, Pedro A.; Vögele, Martin; Smiatek, Jens; Qiao, Baofu; Sega, Marcello; Holm, ChristianWe analyze the internal structure and hydration properties of poly(diallyl dimethyl ammonium chloride)/poly(styrene sulfonate sodium salt) oligoelectrolyte multilayers at early stages of their layer-by-layer growth process. Our study is based on large-scale molecular dynamics simulations with atomistic resolution that we presented recently [Sánchez et al., Soft Matter 2019, 15, 9437], in which we produced the first four deposition cycles of a multilayer obtained by alternate exposure of a flat silica substrate to aqueous electrolyte solutions of such polymers at 0.1M of NaCl. In contrast to any previous work, here we perform a local structural analysis that allows us to determine the dependence of the multilayer properties on the distance to the substrate. We prove that the large accumulation of water and ions next to the substrate observed in previous overall measurements actually decreases the degree of intrinsic charge compensation, but this remains as the main mechanism within the interface region. We show that the range of influence of the substrate reaches approximately 3 nm, whereas the structure of the outer region is rather independent from the position. This detailed characterization is essential for the development of accurate mesoscale models able to reach length and time scales of technological interest.Item Open Access Permeability estimation of regular porous structures : a benchmark for comparison of methods(2021) Wagner, Arndt; Eggenweiler, Elissa; Weinhardt, Felix; Trivedi, Zubin; Krach, David; Lohrmann, Christoph; Jain, Kartik; Karadimitriou, Nikolaos; Bringedal, Carina; Voland, Paul; Holm, Christian; Class, Holger; Steeb, Holger; Rybak, IrynaThe intrinsic permeability is a crucial parameter to characterise and quantify fluid flow through porous media. However, this parameter is typically uncertain, even if the geometry of the pore structure is available. In this paper, we perform a comparative study of experimental, semi-analytical and numerical methods to calculate the permeability of a regular porous structure. In particular, we use the Kozeny-Carman relation, different homogenisation approaches (3D, 2D, very thin porous media and pseudo 2D/3D), pore-scale simulations (lattice Boltzmann method, Smoothed Particle Hydrodynamics and finite-element method) and pore-scale experiments (microfluidics). A conceptual design of a periodic porous structure with regularly positioned solid cylinders is set up as a benchmark problem and treated with all considered methods. The results are discussed with regard to the individual strengths and limitations of the used methods. The applicable homogenisation approaches as well as all considered pore-scale models prove their ability to predict the permeability of the benchmark problem. The underestimation obtained by the microfluidic experiments is analysed in detail using the lattice Boltzmann method, which makes it possible to quantify the influence of experimental setup restrictions.