Institute of Solid State Physics


Master Projects

Studying the impact of dynamic disorder on charge transport in MOFs combining machine-learned force fields with the calculation of electronic couplings using DFT

Charge transport in organic semiconductors is crucially impacted by the relative arrangement of neighboring molecules, as it determines the electronic coupling between the molecules and, thus, carrier hopping rates and crystal electron effective masses. As we have shown recently, there is a natural tendency in organic semiconductors to minimize the said electronic coupling, as this also reduces the exchange repulsion between the molecules. Here, metal-organic frameworks (MOFs) and covalent organic frameworks (COFs) offer the possibility to use their 3D network structure for positioning semiconducting molecules such that this problem can be overcome. Consequently, by a careful framework design, the semiconducting building blocks can be arranged in a strong-coupling configuration. This is comparably straightforward to show for the static case. At room temperature, the thermally triggered displacement amplitudes of the involved building blocks can, however, become relatively large. This calls for an improved description of the situation including dynamic disorder. During the present thesis, for systematically varied network structures this will be studied by analyzing the time-dependent couplings along molecular dynamics trajectories calculated using highly accurate machine-learned force fields.

COMPENSATION: 440€ per month / for at least 6 months

CONTACT: Egbert Zojer (Egbert.Zojer@tugraz.at)

Machine-Learning Physical Equations from kMC Simulations

Machine-Learning Physical Equations from kMC Simulations

Keywords: machine learning, kinetic Monte Carlo, extracting physics, theory/experiment collaboration

In times where it becomes increasingly easy to generate large amount of data from computer simulations, analyzing these results becomes increasingly challenging. Especially when multiple variables impact the outcome, manual inspection and interpretation is almost impossible. Instead automated machine-learning algorithms, like the Sure Independent Screening and Sparsifying Operator SISSO, have been developed.

Put shortly, SISSO takes various variables as so-called “primary features” as input and combines them via mathematical operations (addition, multiplication, etc.) into so-called descriptors that resemble physically meaningful equations. It then screens these descriptors for correlation with the outcome of the simulation, selecting the best-fitting formulas for further processing. In recent works, it has been shown that, given the data is correctly curated, SISSO is able to correctly exctract (already known) physical relationships.

The goal of this thesis is to apply SISSO to a physical meaningful system where the underlying physics are not (yet) fully understood. In particular, we aim to extract effective transition rates between organic polymorphs and their relationship to the deposition rate, temperature, and the energetics of the individual building blocks. The extracted relations will later serve (a) as the basis to design organic molecules with an improved growth behavior, and (b) to apply “optimal control theory” to predict the ideal experimental condition to create selected structures. The main challenge of this work is to perform the kMC simulations and to deal with the “noise” of the simulation output, that occurs inherently due to the stochastic nature of kMC.


We offer:


  • A productive group with a friendly atmosphere.

  • Excellent training in computational science, organic electronics, and semiconductor physics.

  • Visiting an international conference

  • Usually a publication at the end of the thesis.

  • Compensation: 440 € Forschungsbeihilfe (6 months)


Contact: Oliver Hofmann (o.hofmann@tugraz.at)

Computational Prediction of Optimal Growth Conditions of Organic Thin Films

Organic thin films play an important role in many applications, ranging from OLED-TVs and mobile phone displays to mechanically flexible photovoltaic cells. A particular challenge for these materials, however, is that they can assume many different structures (polymorphs), with substantially different properties. Which polymorph forms in a particular experiment is strongly dependent on the processing conditions, which are often optimized in a long, tedious and incredibly boring process by trial and error.

Computational material science excels at predicting which polymorph would be most interesting to prepare. However, so far, it has been mostly unable to guide experiment by naming which processing conditions would lead to the desired structure – in fact, being able to predict experimental conditions leading to kinetically trapped structure is probably one of the most sought-after goals in computational material predictions.

Recently, we developed an extensive framework based on density functional theory, machine learning, and kinetic Monte Carlo simulations that provide the necessary overview over polymorph energies, properties, and energy barriers that is required to reach this goal. In a proof-of-principle implementation, we were also able to show that promising experimental “recipes” (i.e., changes of temperatures and pressure during the deposition) can be obtained.

The goal of this master thesis is to expand and apply the present implementation to a realistic system and to predict an “recipe” that is to be verified experimentally by collaborators in Jena, Graz and/or Brno.

We offer:


  • A productive group with a friendly atmosphere.

  • Excellent training in computational science, organic electronics, and semiconductor physics.

  • Visiting an international conference

  • Usually a publication at the end of the thesis.

  • Compensation: 440 € Forschungsbeihilfe (6 months)


Understanding thermal transport in OSCs and MOFs using highly accurate machine-learned force-fields

Thermal transport is of distinct relevance for most technological applications of materials. This is particularly true for organic semiconductors (OSCs) and metal-organic frameworks (MOFs). In the course of the advertised theses (one for OSCs and one for MOFs) we will use our recently developed strategy for calculating thermal transport properties of complex materials employing non-equilibrium molecular dynamics (NEMD) with highly accurate, system-sepcifically parametrized force fields. The said force fields will be obtained combining on-the-fly machine learning approaches based on periodic DFT codes with highly efficient moment-tensor potentials. The NEMD simulations analyze thermal transport in real space. As an alternative approach, the machine-learned potentials will also be used for accurately calculating phonon properties, which - when combined with the Boltzmann Transport Equation - allow analyzing thermal transport in reciprocal space.
COMPENSATION: 440€ per month / for at least 6 months
CONTACT: Egbert Zojer (Egbert.Zojer@tugraz.at)

Proof of surface crystallisation for anthraquinone

The preparation of anthraquinone thin films is performed by solution processing. The first step is the preparation of a dissolution of the molecule anthraquinone with a solvent (like tetrahydrofuran, toluene, chloroform, a.o.) and the second step is the deposition of the dissolution at a substrate surface by drop casting, dip-coating or spin coating. After evaporation of the solvents crystals of anthraquinone remain at the substrates surface. Two different mechanisms are assumed for the crystallization process at the substrates surface. On one hand the nucleation of crystals happens in the solution, crystals grow in the solution due to supersaturation and finally crystals remain at the substrate surface after the evaporation of the substrate. On the other hand, nucleation of the crystals happens at the substrate surface and the crystal nuclei grow und the expense of the molecules in the solution. The thesis should separate these two effects of surface crystallization on basis of anthraquinone thin films.

COMPENSATION: 440€ per month / 6 months
CONTACT: Roland Resel (roland.resel@tugraz.at)

Physics-Informed Deep Learning for Reactive Gas Transport in Paper

Simulating the transport of gasses such as organic volatile molecules through porous media is challenging when the complex geometry of real systems has to be considered, especially when transport is accompanied by sorption and/or chemical reactions. This complexity in geometry challenges traditional numerical solvers in their ability to capture the intricate details of the system in an accurate yet feasible manner.

The transport through paper is influenced by a combination of material properties, including the polarity of the organic compound, its diffusivity in the porous and solid phases of the paper, and the paper’s micro-structural features such as porosity and tortuosity. Physics-informed neural networks (PINNs) can successfully predict such material parameters and the subsequent transport through paper (DOI: 10.1007/s11242-022-01864-7).

Based on experimental migration data, this PINN approach shall be extended to two- and three-dimensional systems by incorporating microstructural information obtained via micro-CT measurements of paper into the framework of PINNs domain.

Contact: Karin Zojer

Preparation of a photonic single crystal

The first part of the work is the preparation of a photonic single crystal by 3D printing technique. 3-dimensional stacking of piles will provide a periodic lattice with periodicity in the range of 1 µm, a characteristic length comparable with the wavelength of visible light. In a subsequent step en experimental set-up have to be built to perform diffraction experiments using visible light. Geometries of classical X-ray diffraction techniques will be used using white light (Laue diffraction) or monochromatic light (single crystal diffraction by using a four circle goniometer).
additional information:
https://www.osapublishing.org/ao/abstract.cfm?uri=ao-51-28-6732

COMPENSATION: 440€ per month / 6 months
CONTACT: Roland Resel (roland.resel@tugraz.at)

      >> mehr >>

The molecule HATCN is a strong electron acceptor that is commercially used in OLEDs to modify the property of metal substrates. Adsorbed on silver, this molecule shows unusual, fascinating physics. At low coverage, the molecule forms honeycomb patterns, which can be exploited as epitaxial growth template. When the coverage is increased, however, the first monolayer rearranges. This drastically changes the material properties, in particular the system’s work function.

At present only very little is known about the geometric and electronic structure of the rearranged, high- coverage phase. This is now at the focus of a joint efforts including the groups of Prof. Resel (TU Graz), and Prof. Fritz (University Jena), which will perform x-ray and low energy electron diffraction experiments and characterize the system via optical spectroscopy. The aim of this thesis is to provide complimentary computational insight to these experiments. Density-functional theory calculations will be performed in order to predict possible geometric structures, characterize the optical and vibrational properties, and understand the driving force that leads to the observed phase transition.

Contact:
Oliver Hofmann
Email: o.hofmann@tugraz.at
Tel: 0316 873 8465
http://www.if.tugraz.at/web.php?85

 

 


Impressum, Datenschutzerklärung