Events
Comphys Seminar
on Monday at 15:20h in the HIT building H 42
Electron-Phonon Coupling Through Curvature in Space
Oliver Furtmaier
Chair of Computational Physics
ETH Zurich
Switzerland
Abstract:
Motivated by the work of Ilario G. et al. we investigate the quantitative and qualitative features of an electron-phonon coupling facilitated via curvature in space. We will present an intuitive approach for the coupling and derive an expression for the metric tensor by systematically comparing the resulting Hamilton operator in curved space with more traditional approaches. In addition, we will show the equations for the dynamics of this metric tensor, analogous to the famous Einstein equations.
on Monday at 15:20h in the HIT building H 42
Numerical modeling of fiber reinforced jammed granular media
Pavel Iliev
Chair of Computational Physics
ETH Zurich
Switzerland
Abstract:
Previous experimental investigations by Aejmelaeus-Lindström et.al. have shown the significant potential of fiber reinforced jammed granular structures that can be free-standing and load-bearing. However, the mechanism that prevents the system from fluidization has not been understood. In order to study this phenomenon we have developed a numerical model that takes into account particles of arbitrary shapes and their interaction with an elastic wire. Our simulations show the constraining and thus force-redirecting effect of the wire on the system.
on Monday at 15:20h in the HIT building H 42
Modeling Clogging and Unclogging in Porous Media
Robin Jäger
Chair of Computational Physics
ETH Zurich
Switzerland
Abstract:
We have extended our previous model for deposition and erosion in porous media with a mechanism that allows the uncapping of channels. Our simulations also show pressure drops as found in previous experiments by Filippo et al. We conduct a detailed comparison of the statistics of the pressure drops and study relevant parameters for this behavior; and further investigate other scenarios where this phenomenon might appear.
on Monday at 15:20h in the HIT building H 42
Learning binary rules in a recurrent neural network by localized reinforcement signals
Damian Berger
Chair of Computational Physics
ETH Zurich
Switzerland
Abstract:
In recent years several studies showed that monoamines diffusing through the extracellular space of the brain can mediate plasticity of neural synapses. Such transmitters have the potential to deliver a feedback signal to a larger part of a neural network. It has already been hypothesized that supervised learning might be represented in the brain by diffusing extracellular monamines. Based on these findings we developed a model with a recurrent neural network and studied if and under which conditions this kind of learning is possible. With this model we find that the learning performance is very sensitive to the spacial extent of the feedback signal.