Leibniz MMS Days 2026
March 2 - March 4, 2026
Frankfurt (Oder)

Posters

A poster session will take place on March 2, 2026 in the afternoon, comprising the following contributions:

An Accessible Toolbox for Multi-Physics Simulations of Advanced Electronic and Quantum Devices

Patrick Jaap (WIAS Berlin)

We present an open toolbox for multi-physics simulations of advanced electronic and quantum devices with multiscale geometries and coupled effects (electrostatics and elasticity), capturing mesoscale behavior in modern semiconductors. The toolbox, StrainedElectronicDevices.jl, is built on WIAS-PDELib, our ecosystem of Julia PDE solvers. The finite element package supports higher-order elements, anisotropic meshes, complex material tensors, and non-matching periodic boundary coupling. Julia enables reproducibility and access to efficient (non-)linear solvers for large-scale simulations. A high-level interface allows multi-physics setups in a few lines of code, with seamless VTK/ParaView visualization. The toolbox is broadly applicable to advanced electronics and provides an open alternative to proprietary software. We demonstrate its capabilities using strain and electrostatics simulations of a Si/SiGe electron shuttling device (quantum bus).

A Julia-Based Computational Framework for Combined Geothermal and Natural Gas Production in Mature Sandstone Reservoirs

Ernesto Meneses Rioseco (Georg-August-Universität Göttingen, Geoscience Center)

The combined use of geothermal energy extraction and conventional natural gas production represents a promising approach to improve subsurface energy utilization during the energy transition. Mature gas reservoirs offer the potential to supply geothermal heat to nearby communities while extending hydrocarbon recovery through reuse of existing infrastructure. This study investigates combined geothermal-natural gas production in mature sandstone reservoirs, focusing on the Middle Buntsandstein Formation in Northwest Germany. A new computational framework is developed in the Julia programming language, incorporating two-phase flow, gas compressibility, and coupled heat transport. The framework exploits modern Julia features such as automatic differentiation, flexible finite-volume discretization, and multithreading based parallelization to efficiently simulate large-scale reservoir systems. Numerical simulations over a 30-year production period assess the influence of reservoir geometry, stratigraphy, and well configuration on thermal and water breakthrough behavior. Horizontal, anticlinal, and inclined reservoir settings can be evaluated to identify favorable production strategies. Work in progress investigates, if combined geothermal production can enhance cumulative gas recovery compared to conventional scenarios, primarily due to delayed water breakthrough. The framework provides a flexible basis for assessing combined subsurface energy systems while highlighting remaining limitations and directions for future development.

Bayesian optimization for metal additive manufacturing

Dmitry Chernyavsky (IFW Dresden)

Bayesian optimization to avoid valley excitations during long-distance electron shuttling in a Si/SiGe quantum bus

Abel Thayil (WIAS Berlin)

Silicon-germanium quantum dots are a leading platform for scalable quantum computing due to their long spin coherence times and compatibility with existing semiconductor fabrication technologies. A major challenge for scalability is coupling distant qubits, motivating strategies that shuttle electrons over micron-scale distances. Intrinsic alloy disorder in such devices induces fluctuations in the valley splitting-defined as the energy gap between the two lowest electronic states in silicon quantum dots which can lead to a loss of quantum information during shuttling. In this work, we propose a control strategy based on Bayesian optimization (BO) to identify shuttling protocols that maximize preservation of the electron's valley state across a 10 μm channel. As a black-box optimization method, BO does not require prior knowledge of the underlying valley landscape and instead learns from iterative measurements. We further enhance performance using a sequential BO approach that incrementally optimizes longer segments of the shuttling path, leveraging the problem's inherent sequential structure.

Do AI-Generated Biases Shape Human Cognition? Evidence from a Reaction-Time Experiment

Tamara Serrano Romero (Universitat Autònoma de Barcelona, Center for Theoretical Linguistics)

Since recent advances in Artificial Intelligence (AI), new questions have emerged regarding how these tools influence human behavior and cognition. In the case of Large Language Models (LLMsalgorithms trained on vast corpora for language generation their proximity to human linguistic abilities has been increasingly debated. As human communication is goal-oriented, a central concern is whether the information conveyed by AI exhibits bias and whether such bias affects users information processing positively or negatively. Evidence indicates that AI systems can reproduce biases embedded in their training data (Bender et al. 2021), raising the risk that certain statements - potentially non-objective or disrespectful - may become fossilized in human cognition. Our study collects statements about men and women from Google's autocomplete feature and presents them to participants while measuring reaction times to assess if and how AI-related biases influence cognitive processing. This approach examines whether exposure to such information modifies human opinions or behaviors from a gender-based perspective.

Extending the model lid of UA-ICON

Tom Dörffel (IAP Kühlungsborn)

We present recent efforts on extending the upper model lid of the upper-atmosphere extension of the Icosahedral Non-hydrostatic modeling framework. This presentation will highlight the mathematical and numerical challenges and presents high-resolution results of global whole-atmosphere simulations from ground to 250 km.

Funding for Diamond-OA-Journals. With KOALA from passion to profession

Holger Israel (TIB Hannover)

We present recent efforts on extending the upper model lid of the upper-atmosphere extension of the Icosahedral Non-hydrostatic modeling framework. This presentation will highlight the mathematical and numerical challenges and presents high-resolution results of global whole-atmosphere simulations from ground to 250 km.

HybridSolver: flexible, resilient and efficient hybrid solver for multiscale simulations

Denis Korolev (WIAS Berlin)

In this poster, we present the HybridSolver project, which aims to develop a new generation of numerical methods that simultaneously address all scales in composite material modelling by using physics-informed neural networks and conventional numerical techniques. At each scale, an appropriate neural network-based solver or numerical approximation method is selected to best fit the nature of the problem. Coupling into a monolithic computational framework is achieved through materials science scale bridging, and the hybrid numerical solution is obtained by solving a new class of PDE-constrained optimization problems.

Integration of logical tensor networks into LLMs for explainable and efficient reasoning

Janina E. Schütte (WIAS Berlin)

Large language models (LLMs) have found their way into everyday life. However, their underlying mechanism, next-token prediction based on learned probability distributions, does not guarantee logical understanding. This leads to challenges in explainability and reasoning. In contrast, highly efficient logical tensor networks offer a powerful neuro-symbolic approach for representing relational and logical knowledge and for performing reasoning. Despite their advantages, the construction and manipulation of such networks often require domain expertise, limiting their accessibility to practitioners. To address these complementary limitations, we propose an agentic framework that integrates probabilistic LLMs with logical tensor networks, combining intuitive language-based interaction with efficient, logic-driven reasoning.

Machine Learning Applications in Low-Temperature Plasma Research Beyond Supervised Learning

Ihda Chaerony Siffa (INP Greifswald)

Machine learning (ML) has proven to be an indispensable set of tools in many industries and scientific fields, and low-temperature plasma (LTP) science and technology is not an exception. LTP research has enabled various technologies of high societal significance, and here, ML can be of use to improve the modeling and analysis of complex LTP systems. Applications of ML in LTP research have mostly focused on supervised learning, which depends on a large amount of labeled data. In this contribution, we bring forth examples of ML applications across various learning paradigms. First, we demonstrate a supervised learning approach to construct surrogate models for LTP simulations. Second, we introduce physics-informed neural networks (PINNs) utilized to solve complex physical equations in LTP physics, such as the electron Boltzmann equation, in a self-supervised manner. Lastly, we present a preliminary result of reinforcement learning for optimizing equivalent circuits of gas discharges based on electrical measurement data. These examples demonstrate various ML methods beyond the typical supervised learning approach.

MaRDI Knowledge Graph and MathModDB for Mathematical Models (interactive demonstration)

Thomas Koprucki (WIAS Berlin)

The MaRDI Knowledge Graph, developed by the Mathematical Research Data Initiative (MaRDI) within NFDI, is a web-service and infrastructure linking publications, authors, software packages, data sets, and now mathematical models via MathModDB. This knowledge graph is directly connected to MaRDMO, a mathematics-specific extension of the Research Data Management Organiser (RDMO). In this interactive demo, you will explore how 200 mathematical models (with 20,000 semantic statements) are structured and linked in the MaRDI Knowledge Graph. Furthermore, you will learn how to query the graph for models, software, and publications relevant to your research. You will be guided through adding your own models to the graph. Finally, we will discuss how this infrastructure can support your research, and how community-driven contributions will be handled and curated. We will brainstorm applications in model validation, reproducibility, and interdisciplinary collaboration.

Mathematical Modeling of Ethylene Dynamics in Fruit Storage Using Respiration-Based Physical Measurements

Akshay Dagadu Sonawane (ATB Potsdam)

Ethylene plays a central role in regulating fruit ripening, but accurate real-time measurement is limited due to the scarcity and cost of reliable sensors. To address this, a modeling approach was developed to predict ethylene production using fruit respiration rates and CO² dynamics. The model uses O², CO², relative humidity, and temperature as input parameters, providing a physics-based framework to estimate ethylene dynamics in storage and in small-scale sensor boxes without direct ethylene measurement. A small sensor box, acting as a respirometer, was placed inside the storage to capture fruit respiration by flushing storage air. Respiration rates were then correlated with ethylene production rate, and CO² diffusion was linked to ethylene diffusion, enabling the development of a predictive model using mass balance. This approach allows monitoring of ripening behavior and climacteric responses in stored fruit.

Modelling fracture-controlled geothermal reservoirs with application to fault-damage zones in the Upper-Rhine Graben, Germany

Ernesto Meneses Rioseco (Georg-August-Universität Göttingen, Geoscience Center)

The DEKAPALATIN-BERTHA joint research project aims to advance the use of deep geothermal energy from deep-seated reservoirs in the central Upper-Rhine Graben to support the decarbonization of the heating sector. Because the geological targets are fault damage zones, numerical modelling of thermo-hydraulic processes in fault- and fracture-controlled reservoirs is essential for ensuring sustainable and optimized operation of future geothermal doublets. This contribution presents a modelling framework that integrates fracture network geometry, fault-zone petrophysical properties, and coupled fluid heat transport to assess reservoir performance. Applied to representative fault damage zones in the Upper-Rhine Graben, the results highlight the importance of fracture connectivity and permeability contrasts for long-term thermal recovery and operational stability.

Modelling of spatio-temporal ignition patterns in dielectric barrier discharges

Cristian Flores (INP Greifswald)

The spatio-temporal pattern occurring in a dielectric barrier discharge (DBD) is analysed using a combined experimental and modelling approach. The DBD has a spatial 1D lateral geometry and is driven by a high-voltage sinusoidal waveform. Under these conditions, it is observed that filamentary discharges occur in two stages at alternating positions during each half-period. A fluid-Poisson model is used to study the dynamics of the DBD and to validate the estimations of the filament ignition times obtained from a data-driven equivalent electric circuit (EEC) model using variations of the breakdown voltage, the transferred charge, and initial conditions. Once validated, the EEC is used to interpret the experimentally observed filament ignition patterns. The results suggest that a spatial variation in the gap voltage is responsible for the observed effects. This work was funded by the Deutsche Forschungsgemeinschaft (DFG), project number: 535827833.

Modelling Pragmatic Language Production as a Markov Decision Process

Anton Benz (ZAS Berlin)

The DEKAPALATIN-BERTHA joint research project aims to advance the use of deep geothermal energy from deep-seated reservoirs in the central Upper-Rhine Graben to support the decarbonization of the heating sector. Because the geological targets are fault damage zones, numerical modelling of thermo-hydraulic processes in fault- and fracture-controlled reservoirs is essential for ensuring sustainable and optimized operation of future geothermal doublets. This contribution presents a modelling framework that integrates fracture network geometry, fault-zone petrophysical properties, and coupled fluid heat transport to assess reservoir performance. Applied to representative fault damage zones in the Upper-Rhine Graben, the results highlight the importance of fracture connectivity and permeability contrasts for long-term thermal recovery and operational stability.

Numerical Instabilities in Nonlinear Pulse Propagation

Uwe Bandelow (WIAS Berlin)

The propagation of optical pulses along nonlinear and dispersive optical fibers is often described by the Generalized Nonlinear Schrödinger Equation (GNLSE). Split-step methods provide an efficient approach for the numerical solution of GNLSE, because they allow to separate complex problems into simpler, more manageable parts. We analyze numerical stability and provide an estimate for the largest possible integration step h.

Pore-scale simulation of interphase mass transfer during two-phase flow in porous media

Huhao Gao (LIAG Hannover)

Two-phase flow and interphase mass transfer in porous media are important processes for a wide range of scientific and engineering applications, such as geological storage of carbon dioxide and the remediation groundwater contaminants. This study builds a new interphase mass transfer model for the pore-scale direct numerical simulations. The model employs a continuous mass transfer formulation based on the phase field method. We verify the model with the analytical solutions of transport involving advection, reaction and diffusion processes. The model is tested for two-phase flow conditions in a conceptual 2D slit. The applicability of the model is demonstrated in NAPL/water drainage scenarios in a conceptual porous domain, comparing the results in terms of the spatial distribution of the phases and solute concentration.

Quantum applications: Role of strain-engineered Germanium (Ge)

Meera Bishnoi (IHP Frankfurt/Oder)

Germanium (Ge) is an intriguing material because of its high hole mobility, strong spin-orbit coupling, and near-direct bandgap, making it very attractive for next-generation quantum devices. This work explores strain engineering in Ge caused by silicon nitride (SiN) stressors, using finite element method (FEM) simulations in COMSOL to calculate normal and shear strain components resulting from different SiN stressor geometries, such as height, spacing between stressors, and stress levels. The resulting tensor is applied to a 6x6 Bir Pikus Hamiltonian to determine heavy-hole (HH) and light-hole (LH) splitting, which can be adjusted from 36 to 165 meV under 1-4 GPa of stress. The study emphasizes the significant effect of uniaxial and biaxial strains (including shear strain) on valence band ordering. This work shows that SiN stressors offer a precise, scalable, and CMOS-compatible method for tailoring Ge band structure in hole spin qubits and quantum wells.

Recognising research software in performance-based funding: Why research software must count as a scientific output

Anja Eggert (FBN Dummerstorf)

Research software and code are essential components of modern, data-driven science, yet they are often undervalued in performance-based funding schemes (LOM). Without access to software, scientific results are not fully reproducible, transparent, or reusable. High-quality research software increases efficiency, reduces duplication of work, enhances visibility through citation and reuse, and provides key qualifications for early-career researchers. International policies already recognise software as a legitimate research output, including guidance from the DFG, and the European Commission under Horizon Europe, as well as the Leibniz Association. This poster argues that LOM schemes must explicitly include research software as a valued output. Recognising software in evaluation criteria is not an optional add-on, but a prerequisite for reproducible, sustainable, and high-quality science.

Single-pixel imaging for Ge-based metasurface photodetectors

Paul Oleynik (IHP Frankfurt/Oder)

Structuring the absorbing layer of photodetectors as metasurfaces enhances device performance and enables functionalities such as wavelength- and polarization-selectivity for imaging applications. However, even if the fabrication of single pixels can be achieved, the fabrication of pixel arrays presents significant additional challenges, motivating other approaches such as single-pixel imaging. Single-pixel cameras can produce spatially-resolved images with single (or few) pixels by using a digital mirror device to spatially select optical data and direct it towards the photodetector. In conjunction with compressed sensing, this can be used to evaluate device performance at an early stage. Here, we present the current status of our single-pixel setup for multispectral imaging at visible and short-wave-infrared wavelengths. We present first results not only on image acquisition but also on utilizing metasurface-based photodetectors for wavelength-selective imaging.

Techno-economic assessment on wellbore heat loss of the high-temperature aquifer thermal energy storage system in the area of Burgwede, Germany

Dejian Zhou (LIAG Hannover)

High-temperature aquifer thermal energy storage offers a promising solution for balancing fluctuating energy supply and demand. While previous studies have primarily focused on heat and mass transfer within reservoir, wellbore heat loss has received comparatively limited attention. In this study, a coupled wellbore-reservoir model is developed and validated with a real district case to quantify the impacts of wellbore heat loss on overall system performance. Sensitivity analyses and a comprehensive techno-economic assessment are conducted to identify the dominant factors influencing heat loss. The results indicate that continuous heating losses occur in both the reservoir and the wellbore throughout system operation. At a low flow rate of 10 l/s, nearly 90?f the total energy loss originates from the wellbore. At a higher flow rate of 40 l/s, the contribution of wellbore heat loss decreases but still increases over time, rising from ca. 60?n the 1st year to ca. 70?s the operation proceeds. Furthermore, variations in flow rate and injection strategy primarily affect the system levelized cost of heat through changes in the total recovered energy, whereas improvements in wellbore insulation performance mainly reduce costs by decreasing heat loss within the wellbore.

Temperature-dependent performance of a GeSn thermophotovoltaic cell

Intatii Zaitsev (IHP Frankfurt/Oder)

Thermophotovoltaics (TPV) shows promise for direct heat-to-electricity conversion. Unlike solar cells that are powered by the Sun, TPV sources (concentrated sunlight heating, waste heat, power beaming), typically at 1000-2000 K, are closer, which yields radiation with significantly higher power densities (5-60 W/cm² versus ~0.1 W/cm² for solar) but below the bandgaps of many common semiconductors. This has drawn attention toward low-bandgap, CMOS-compatible materials (particularly group-IV semiconductors such as germanium (Ge) and germanium-tin alloys (GeSn). This work presents a comprehensive finite element model (FEM) of a TPV system employing a low-bandgap, direct-gap GeSn photovoltaic cell, accounting for indirect recombination, multilayer generation dynamics, and the thermal equilibrium between emitter and cell. It supports in-depth study of Ge-based materials under varying strain, composition, and temperature, providing insights for the design of future devices.