Machine learning delivers strong predictive performance in scientific and engineering tasks when high-fidelity data are abundant. Yet, real-world models seldom quantify aleatoric (data) and epistemic (model) uncertainties, leading to overfitting on noisy inputs. In addition, collecting adequate high-fidelity data is often expensive or infeasible, whereas low-fidelity data are more accessible but less reliable. To address these challenges, this thesis proposes a general multi-fidelity Bayesian learning framework that enables trustworthy uncertainty disentanglement, and extends its application to constitutive modeling and design of recycled composite materials.

The thesis begins with an introduction to regression from a Bayesian perspective, establishing the connection between deterministic and probabilistic treatments, and covering models from linear to kernel and deep neural network regressions. This chapter provides the theoretical foundation for the remainder of the thesis by formalizing inference techniques, uncertainty quantification strategies, and model evaluation metrics.

Two core machine learning methods are developed to tackle the challenges of uncertainty estimation and multi-fidelity data fusion. First, a cooperative training scheme is proposed that combines a variance estimation neural network with a Bayesian mean neural network, enabling explicit disentanglement of aleatoric and epistemic uncertainties while improving mean prediction. The approach demonstrates strong scalability and generality across tasks and network architectures. Second, a practical multi-fidelity Bayesian learning framework is introduced, which fuses low- and high-fidelity data via a deterministic model, a transfer-learning module, and a Bayesian residual learner. This architecture balances expressiveness and computational efficiency, yielding robust predictions in both data-scarce and data-rich regimes.

To support data-driven mechanics, the proposed framework is extended to generalized constitutive modeling of history-dependent materials. A hierarchical learning scheme is developed that spans from single-fidelity deterministic networks to multi-fidelity Bayesian recurrent neural networks. This addresses two major limitations in data-driven modeling: the reliance on large, clean datasets and the lack of interpretability in neural network predictions.

Finally, the methodology is applied to the sustainable design of recycled composite polymers, where uncertainty arises from microstructural variability and the use of compatibilizers. Leveraging the cooperative training framework, both aleatoric and epistemic uncertainties are quantified, and a novel polymer design is optimized to achieve better expected performance and lower data variation.

Together, this thesis presents a unified and scalable framework for Bayesian learning under uncertainty and multi-fidelity conditions, with broad applicability in scientific computing, materials modeling, and sustainable engineering. The thesis concludes by summarizing key findings, discussing current limitations, and outlining future research directions. ... Read more

Heart failure remains a leading cause of morbidity and mortality worldwide, highlighting the need for reliable tools to assess cardiac function. Myocardial oxygenation is one of the most direct indicators of tissue health, yet current methods lack compact, implantable solutions for continuous monitoring. This work presents an implantable optical sensor that exploits the ultraviolet-excited fluorescence of NADH as a marker of oxygenation. To overcome the limited penetration of ultraviolet light, near-infrared photons are externally delivered and converted into ultraviolet emission by lanthanide-based upconverting nanoparticles(UCNPs), enabling localized excitation without implanted power sources. A Fabry–Perot filter was incorporated to suppress blue emission that overlaps with NADH fluorescence while maintaining high ultraviolet transmittance. The filter design was optimized through multilayer simulations, and deposition conditions were tuned to improve film quality. Upconverting nanoparticles were drop-cast onto the filter surface, and material characterization confirmed the presence of significant nanoparticle coverage. An optical testing platform was further established using both a xenon-based source and a laser diode, which enabled validation of up-conversion performance and filter function. Collectively, these results demonstrate the feasibility of a compact, externally powered light emitter for implantable cardiac oxygen monitoring and establish a foundation for future development of minimally invasive biosensors. ... Read more

In this dissertation the formation of precipitates in steel is studied.
In the first chapter some theoretical background is presented to the topic.
Then in the second chapter an analytical approximation to the classical nucleation and growth model is presented.
In the third chapter a reference-free modified embedded atom method (RF-MEAM) potential is constructed for Fe, suitable for ferrite and austenite.
In chapter four the formation of Nb-C and Ti-C clusters in ferrite is studied, here formation paths with a monotone decrease in Gibbs energy are found. Moreover, the resulting clusters remain stable up to 1100K.
In the fifth chapter other group IV and V transition metal-carbon systems in ferrite are studied in small cells using density functional theory. Showing similar energy patterns as previously found for Nb and Ti. Furthermore the transformation from a metal-carbon cluster to a carbide precipitate is studied, where the matrix strain is found to be the driving force behind the transformation.
in the sixth and last chapter some recommendations are presented. ... Read more

Deep learning has revolutionized scientific and engineering applications by enabling fast, data-driven predictions and optimizations. In materials science, however, its impact is limited by complex structure–property relationships, sparse high-quality data, and the need to respect fundamental physical laws. Overcoming these challenges calls for machine learning frameworks that learn effectively from limited data while producing results that are physically meaningful, interpretable, and actionable.
This thesis develops physics-guided machine learning approaches to accelerate the design, modeling, and interpretation of materials across scales and material classes. It introduces neural network architectures for metamaterial design that learn directly from data while remaining physically consistent. These models retrieve structural designs that achieve specified target properties instantly, enabling rapid exploration of the structure–property landscape and supporting scenarios with multiple design goals more efficiently than traditional optimization methods. For material modeling, this work addresses the limitation that stress fields are not directly accessible in experiments. The proposed frameworks instead learn constitutive laws from measurable quantities such as displacements and forces, while preserving essential physical principles such as thermodynamic consistency. This work further demonstrates how data-driven approaches reveal previously unknown structure–property relationships, such as the role of dangling atomic masses in the thermal transport of crystalline nanoporous materials. Finally, it introduces a chemistry-constrained generative framework that proposes synthesizable, diverse, and novel molecular structures using limited data, while providing interpretable representations of the molecular generation.
Together, these contributions establish that physics-guided machine learning can complement and extend traditional materials science by delivering reliable, interpretable, and generalizable solutions to longstanding challenges in design and modeling. ... Read more

Metallic materials exhibit structural and performance anisotropy at various scales, including the crystal structure, microstructure, and bulk levels. The anisotropy influences how stress and strain distribute within the material. The localized stress concentration is closely related to deformation and several failure mechanisms. A systematic exploration of the anisotropy can deepen our understanding of stress concentration and material failure, and guide the design of high-performance materials tailored for specific applications. Additionally, this work supports sustainable development by enhancing the service life and recyclability of metallic materials—an increasingly critical priority in materials science..... ... Read more

Extreme Ultraviolet (EUV) lithography plays a crucial role in the semiconductor industry, enabling the shrinkage of transistor sizes and sustaining Moore’s law. However, the high cost of EUV light limits the number of available photons for high-volume wafer manufacturing. To maximize the utilization of each incoming photon, metal-oxide resist (MOR) has emerged as a promising candidate to replace conventional chemicallyamplified resist due to its higher absorption coefficient when exposed to EUV light. An open-source Monte Carlo based simulator is used in this study to model electron scattering within the photoresist materials. When EUV light strikes the photoresist, initial high-energy photoelectrons are generated, triggering a series of scattering events that produce a cascade of secondary electrons (SEs). These SEs possess energies capable of altering the chemistry of resist materials, leading to pattern formation in the following processes. In this study, we propose a novel method of applying an electric field to the resist layer to enhance pattern performance under a fixed EUV dose. Simulation results demonstrate that this approach creates an anisotropic electron blur extended in the 𝑧 direction perpendicular to the resist surface) without compromising much the resolution in the 𝑥 and 𝑦 directions (parallel to the resist surface). Additionally, an increase in SE yield is observed. The optimal electric field strength, identified as -400 MV/m for MOR, results in an 11.93% increase in 𝑧-direction blur and a 3.41% increase in SE yield per absorbed photon. Moreover, the asymmetry of 𝑧-direction blur counteracts the EUV light absorption near the surface and contributes to more chemical conversions deeper in the resist. ... Read more

Neuromorphic computing, a novel computing configuration inspired by the brain, aims to perform calculations based on physical neurons and synapses, attracting significant attention in recent years. Resistive random access memory (RRAM) shows great potential in this field, demonstrating high operation speed, nanoscale scalability, long retention time, non-volatile performance, and a simple structure.

Despite the promising performance of RRAM, a high forming voltage potentially hinders the widespread application of the device. This thesis aims to diminish and eliminate the forming voltage. To achieve this, different metals were inserted between the insulator layer and the bottom electrode of the RRAM, serving as an interface metal layer. The interface metal was expected to introduce oxygen vacancies to the insulator, thereby decreasing the forming voltage. Advanced nanofabrication processes were employed in the cleanroom, and a related recipe was developed. The influence of layer thickness and device area was also studied to gain a comprehensive understanding. Among all the samples, Ru-based devices were observed to be forming-free.

Data analysis methods were applied to model the data, with the random forest method found to be the most suitable, achieving an accuracy of 82.4%. The model was verified by measurements of 10 nm Ru-based devices. Feature importance was then calculated to interpret the model. The four most important features determining the forming voltage are the thickness, standard electrode potential, area, and work function of the interface metal. This work adopts a new approach to eliminating the forming voltage, not only providing a forming-free device but also offering a guideline for future research on forming voltage. ... Read more

With high global ambitions for sustainability and circularity and increasing energy prices, the recycling industry keeps innovating to keep up with political goals and customer demand. Reducing the energy consumption in the recycling process reduces environmental impact and reduces financial costs for the customer. A energy intensive process step in recycling is oven drying the plastics to an acceptable moisture content. The energy cost can be reduced by using microwave irradiation as heat source. This research investigates the moisture content and the intrinsic viscosity (IV) of polyethylene terephthalate (PET) during and after the drying process. The measurements of the PET dried in an experimental setup employing microwave irradiation as heat source is compared to the drying process in a oven at 80, 100 and 150 ◦C. The IV of the virgin PET pellets did not decrease significantly after drying in the experimental setup, but it did decrease after drying in the oven at 150 ◦C. The moisture content decreases more rapidly in the oven at all temperatures than in the experimental setup. Scaling up the experimental setup should increase the drying rate. No indications of unfeasibility of microwave assisted drying are found. Keywords: PET (polyethylene terephthalate), Circular economy, Plastic recycling, Intrinsic viscosity (IV), Moisture content
... Read more

Neural networks have made significant progress in domains like image recognition and natural language processing. However, they encounter the challenge of catastrophic forgetting in continual learning tasks, where they sequentially learn from distinct datasets. Learning a new task can lead to forgetting important information from previous tasks, resulting in decreased performance on those earlier tasks. This issue is further intensified in dynamic scenarios where the task sequence varies unpredictably. To address this problem, architectural methods have been developed to modify a neural network's structure, creating or adapting subnetworks to retain task-specific knowledge and mitigate catastrophic forgetting. However, these solutions can lead to network saturation, where the accumulation of task-specific adaptations hampers the network's ability to learn new tasks. This research aims to address the problem of network saturation by developing innovative methods that enable neural networks to maintain high performance across both existing and new tasks in continual learning scenarios. Eventually, the new model improved its learning ability on new tasks in the presence of an allowable forgetting, while demonstrating better overall learning ability. ... Read more

Deep learning models have made enormous strides over the past decade. However, they still have some disadvantages when dealing with changing data streams. One of these flaws is the phenomenon called catastrophic forgetting. It occurs when a model learns multiple tasks sequentially, having access only to the data of the current task. However, this scenario has strong implications for real-world machine learning and engineering problems where new information is introduced into the system over time. Continual learning is a subfield of deep learning that aims to work in this scenario. Therefore, this thesis presents a general continual learning paradigm to tackle the catastrophic forgetting issue in deep learning models, regardless of architecture.

Following ideas from the neuroscience literature, we create task-specific regions in the network, i.e. subnetworks, to encode information there. Thus, some parameters are responsible for solving this task, which mitigates forgetting compared to conventional training where the trainable parameters are simultaneously assigned to all tasks. A proper subnetwork should be then selected by the algorithm to make a prediction or information about the correct subnetwork must be given by the user. The subnetworks can share some connections to transfer knowledge between each other and facilitate future learning.

In the first part of the thesis, we describe the proposed methodology: task-specific subnetwork creation during training and the proper subnetwork selection during inference stages. We examine different subnetwork prediction strategies outlining their advantages and disadvantages. We validate the proposed algorithms on a series of well-known image datasets in computer vision in classification and semantic segmentation tasks. The proposed solution significantly outperforms current state-of-the-art methods by 10-20\% of accuracy.

The second part of the thesis illustrates the benefits of cooperative learning via continual learning in physical sciences and solid mechanic examples. We demonstrate that by sharing parameters, the following subnetwork can be trained either with lower prediction error, requiring fewer training data points, or both, compared to conventional training with one network per task. Importantly, the model does not forget any of the acquired knowledge since once a parameter is assigned to a subnetwork, it is not changed when training new tasks. We would like to highlight the potential importance of further development of continual learning methods in engineering to improve the generalization capabilities of the models.

The thesis concludes by discussing the main results and findings. We also outline the main limitations of the work and directions for improvement. Further development of continual learning models will lead to more advanced artificial intelligence systems that should contribute to solving a wider range of problems. ... Read more

The lack of a decent solid-state ionic conductor has hindered the large-scale application of solid-state batteries, which are considered to be the potential game changer for energy transition. The recently reported K doping CsPbF3 material system has shed light on this problem. This material possesses high ionic conductivity and a wide electrochemical stability window at the same time, making it a highly promising candidate for the next-generation fluoride ion solid-state battery. In order to have a clearer understanding of the structural information of this material and to find out what contributes to the outstanding properties it demonstrates, this thesis project uses Density functional theory(DFT) to calculate its ground state properties. Meanwhile, to better understand its local structure, the Nuclear magnetic resonance(NMR) parameters for this material are also calculated using DFT and analyzed in detail. Results generated from the calculations suggest that the coulombic interaction can be utilized to explain the structural deformation upon doping K into the CsPbF3 system. Additionally, the analysis of the optimized cell structure indicates a tendency for the material system to go through a cubic to tetragonal phase transition, which reproduces the trend observed experimentally and offers a potential explanation for the driving force behind it. Further investigation using Nudged elastic band calculations(NEB) also reveals a relatively low energy barrier for vacancies to diffuse in the crystal structure, which provides insight into the high ionic conductivity of this material. The findings manifested in this thesis project could potentially offer improvement directions for the K-doped CsPbF3 system and contribute to the development of other solid-state ionic conductors. ... Read more

Snow is a natural hazard to human life and infrastructure. This motivates current research efforts to understand the granular material. The material point method models snow as a continuum. Application length scales range from the microstructural level to full scale avalanches. This conventional numerical method relies on solely spatially local information to make local updates. The recent graph neural network machine learning model is shown to include both local and global information in making local updates. This model’s promising attribute motivates its use to replace the conventional snow simulation method. However, it is uncertain if current graph neural network applications to learn physical simulations truly learn the underlying physics. This work is inspired by the finite element community's patch-test proposed in the 1960s. This insight is used to reimagine the means a graph neural network model is evaluated. Through this novel evaluation choice, may the model be investigated on the core properties of numerical methods. Further, a state-of-the-art graph neural network model is improved to utilize unnormalized features and targets in making stable predictions. Future research recommends these machine learning models in this application make architecture design choices such that the core properties of conventional numerical methods are met. ... Read more

Asphalt concrete is one of the most widely used materials in modern road construction. Predicting its functional properties is crucial in the design of new asphalt concrete mixtures. However, current prediction models are limited in accuracy and applicability due to the complex nature of asphalt concrete properties. This thesis researches the use of machine learning algorithms to greatly improve upon existing prediction models. The input is limited to standardized test results in line with Dutch regulations, the output focusses on functional design parameters including stiffness, fatigue resistance, water sensitivity and resistance to permanent deformations. The performance of several machine learning algorithms and the effects of different regression methods are compared. Furthermore, a solution is found for the inverse problem, which allows for greater flexibility when using the models to design new asphalt concrete mixtures. The results show that machine learning algorithms outperform traditional models on accuracy while simplifying the model input parameters. Machine learning algorithms were also able to predict a greater range of output parameters, most of which with a high accuracy. The analysed possibility of modelling asphalt concrete mixtures directly from their desired functional properties is shown to be promising. The proposed machine learning models and their inverse problem counterparts have the potential to greatly improve the accuracy and practical usability of the prediction of asphalt concrete properties, ultimately leading to better mixture design and more durable roadways. ... Read more

Quantum computing has gained a lot of interest from researchers and industry due to its great potential to solve some complex problems in various fields. One of the biggest challenges is developing hardware suitable for the extremely low operation temperatures required by quantum computers. Specifically, the wiring material of a quantum computer must provide good electrical conductivity while keeping the thermal load to a minimum. Delft Circuits’ cryogenic flexible cable design made from metalized Polyimide (PI) with Silver (Ag) thin films as conductors offers a promising solution for quantum computer i/o systems. This thesis aims to investigate the effect of fabrication and design parameters on the electrical and thermal properties of silver thin films, utilizing the Residual Resistivity Ratio (RRR) as an indicator to assess the thermal load of cryogenic flexible cables. The RRR is calculated by taking the ratio of electrical resistivity measured at room temperature and extremely low temperatures of liquid He (4.2 K). The key parameters for the experimental investigation include oven heat treatment, lamination, silver purity, film thickness, aging and storage. The RRR behavior of samples with varying fabrication and design characteristics is explored through cryogenic measurements. Subsequently, the RRR results are analyzed and supported by material characterization via SEM and XRD. The results show that heat treatment results in higher RRR values due to grain growth and lower grain boundary density, leading to lower electrical resistivity at low temperatures. Heat treatment parameters including temperature, pressure, duration, ambiance, and cooling rate play a significant role in the resulting microstructure of the silver film, and consequently, in their RRR values. Furthermore, the study reveals that the lower purity level leads to decreased RRR values of the silver film due to higher electron scattering caused by impurities. A linear relationship is found between the film thickness and the RRR behavior of silver thin films. Lastly, aging and storage do not result in a significant change in the RRR values of heat-treated silver thin films. This thesis provides a deeper understanding on the influence of fabrication and design parameters on the low-temperature resistivity of silver thin films. It highlights the key role of these parameters in tailoring the microstructure of silver thin films to achieve desired material properties. ... Read more

Master thesis about machine learning in materials science ... Read more

High entropy alloys (HEAs) are new potential materials for hydrogen storage applications which could help with the transition towards sustainable energy sources. The rate of hydrogen kinetics is one of the material properties that is important for the storage and other hydrogen related applications. One of the limiting factors on the rate of hydrogen kinetics is hydrogen diffusivity. Thus far, there has been no reports on hydrogen diffusivity for HEAs in relation to hydrogenation kinetics.

Therefore, this project investigates hydrogen diffusivity in equimolar TiVZrHfNb and the influence of hydrogen concentration on hydrogen diffusivity to gain better understanding of hydrogenation kinetics. The selected HEA has been found to absorb the highest amount of hydrogen (2.5 H/M) among other HEAs.



The investigation was done by a computational approach using ab initio molecular dynamics. BCC and face-centered cubic (FCC) supercells with different hydrogen concentrations (H/M = 0.2, 0.8, 1.4, 2, 2.4) were simulated at a temperature range of 773 – 973 K. At the same time, experimental electrochemical hydrogen charging using chronoamperometry and cyclic voltammetry was performed in order to compare computational and experimental values of hydrogen diffusivity.



The electrochemical hydrogen charging did not result in hydrogen absorption, most probably due to the passivation of the sample surface.

From the simulation results, the values of activation energy and pre-exponential factor were estimated to be in the range of 0.26 – 0.48 eV and 0.73 – 2.95 x 10-7 m2/s, respectively. Hydrogen diffusivity was found to be higher in BCC than in FCC. In BCC the hydrogen diffusivity slowly decreases linearly with increasing H/M. In the case of FCC, the hydrogen diffusivity was found to be the highest at 2.4 H/M while at 2 H/M the diffusivity was the lowest. The analysis of hydrogen occupation at 2 H/M shows that most of the hydrogen atoms are trapped inside tetrahedral sites. It is possible that the hydrogen occupation in the tetrahedral sites results in the optimum hydrogen distribution where the repulsive interaction between hydrogen atoms is the lowest. At concentrations above 2 H/M, an additional repulsive force between hydrogen atoms seems to contribute to the increase of hydrogen diffusivity. ... Read more

Interest in transition metal surfaces has grown in fields such as catalysts and semiconductors. Models of initial adsorption and growth of adsorbates are used to verify if certain structures are formed. Structural complexity and diversity of overlayers often determines this growth. Understanding the difference between a clean single crystal transition metal surface and adsorbate covered surfaces at the atomic scale are a first step. Characterization at atomic scale however can be challenging and help can come from electronic structure theory. Much is known about the overlayers of oxygen on a Ru(0001) surface, yet little research has been conducted in understanding the interaction between phosphorus and Ru(0001). In this work the adsorption mechanism of phosphorus on a ruthenium (0001) is studied. Density functional theory (DFT) calculations were performed to find optimal adsorption sites on a Ru(0001) surface for P. The four high coordination positions were then used to form distinct combinations upon stacking layers. There it became evident that in the regime of first layers, the adhesion and stacking of adsorbate layers is governed by the Ru(0001) stacking order and highest probable coordination sites on the surface. Upon stacking more layers, the adsorbate layer follows a pattern of ’on top’ stacking with alternating layer heights. The bond distances and lack of electron transfer show a tendency to form P-P bi-layers. Binding enegies within the P-P bilayer (intra-bilayer) are stronger than binding energies between successive bi-layers (inter-bilayers). Latter one being dominated by VanderWaals forces. A minimal analytical sum of binding interactions is proposed to the surface and bilayer energies showing an accurate description of the DFT results. Bilayer formation and weak inter-bilayer interactions indicate the possibility of formation of two-dimensional phosphorene structures. At last as a point for future work possible supercells are constructed an tabulated that could accommodate adhesion of phosphorene (oxide) structures with minimal strain. ... Read more

Fossil fuels have been the primary source of rising energy requirements for humankind. However, the extensive use of fossil fuels has led to an increase in Earth's surface temperature. To tackle rising energy demands and the increase in Earth's surface temperature, various organizations like Inter-governmental Panel for Climate Change and the European Environmental Agency have suggested the use of renewable energy as an alternative energy supply. E.g., the use of hydrogen as an alternative fuel in transportation will reduce greenhouse gas emissions. Besides, converting CO2 and N2 to fuels and industrial feedstock like CO or NH¬3 can curb the Earth's increasing surface temperature. As a result of this, in this thesis, the catalysts for the synthesis of hydrogen from water-splitting (hydrogen evolution reaction- HER), conversion of CO2 to CO via carbon dioxide reduction reaction (CO2RR), and reduction of N2 from air to NH3 (Nitrogen reduction reaction-N2RR) are studied. The conventional catalysts used for these reactions are Pt for HER, Cu for CO2RR Cu, and Ru for N2RR. Although these catalysts are active and exhibit a high yield of products, they have some disadvantages, such as the long-term availability and cost of Pt and Ru. On the other hand, Cu suffers from the low selectivity for the conversion from CO2 to CO. To overcome these disadvantages; scientists have developed a new kind of catalyst with a higher specific activity, known as the Single-Atom Catalysts (SAC). The SACs use fewer precious elements than the conventional bulk catalysts without compromising the activity. The use of binding energy (EB) as a descriptor for the reactions mentioned has been proven in the literature. Therefore, EB is used in this thesis to predict novel SACs through high-throughput DFT calculations using 3-N doped graphene as the substrate. The descriptor for HER is the EB of H atom, for CO2RR is EB of CO, and that of N2RR is EB of N on the respective catalyst surfaces. These calculated binding energies are compared against the descriptor EB on the conventional catalysts to obtain the novel SACs. With EB as the descriptor, the candidate catalysts for HER are B, Cr, Mn, Fe, Co, Ni, Ge, Ru, In, Sb, La, and Pb. The candidate catalysts for N2RR are Ru, Mo, and Cr. The candidate catalysts for CO2RR are Mg, Al, Ca, Zn and Se,. In addition to this, the charge dissipation of the adsorbent species on the SAC and the effect of atomic size on the EB is studied. It was seen that the computational predictions go hand in hand with the predictions of existing experiments for HER and CO2RR. ... Read more

This thesis demonstrates the feasibility of Extreme-ultraviolet (XUV) high-harmonic generation from structured silica, and was performed at the Advanced Research Centre for Nanolithography (ARCNL). The project focuses on High-harmonic generation (HHG) from condensed matter, and further explores the possibility of high-harmonic generation from micro- and nano-structured silica. HHG from solids was discovered less than a decade ago, and it is expected to be a new source for coherent ultrafast pulses, showing potential in many applications such as HHG spectroscopy, imaging and photonic devices. By generating high-harmonics in structured solids, the capability to control HHG properties by engineering the topology of the surface on solids has been demonstrated. Previous research has demonstrated the control of HHG in the visible light regime by generating from structured semiconductors. In this project, we aim to generate high-harmonics in the XUV regime, and control the properties of XUV light. Our work shows the potential of using structured dielectric materials as new XUV optics, and applications on HHG high-resolution lens-less imaging. ... Read more

Because of its extraordinary material properties, like its high melting point and thermal stress resistance, low erosion and swelling rate, and high radiation damage resistance, highly deformed pure tungsten has been chosen as the plasma facing surface material for the ITER reactor divertor. The study of tungsten’s recrystallization behavior and damage response during operation conditions is thus important because the divertor will have to withstand high heat fluxes and temperatures during service which induce recrystallization. This phenomena alters the microstructure of the material, inducing degradation in its properties, like loss in mechanical strength and embrittlement making it prone to large plastic deformation, surface roughening, crack networks formation and propagation. Understanding this behavior under the ITER reactor operation circumstances is paramount for the success of the reactor.

The aim of this project was to test regimes simulating steady state operation and high frequency and temperature transient pulses called ELMs (Edge Localized Modes) striking the divertor. Recent research has shown that the degradation and behavior of tungsten under these transient conditions does not consistently follow the expected parameters characterized in the literature. According to it, recrystallization, grain growth, and crack formation seem to be suppressed by the plasma loading under these regimes, thus a new understanding of the material behavior for these circumstances must be developed. To do this, ITER grade tungsten samples were subjected to a hydrogen plasma beam at DIFFER’s Magnum-PSI with temperatures at the strike point ranging from ~1000 to ~1500 °C and high frequency pulses that increased the surface temperature by ~200 to ~300 °C above the steady state temperature. The surface thermal shock response to the plasma pulses was characterized by means of infrared and pyrometer readings at the samples’ surface during exposure. Temperature and power density calculations were correlated with identified damage morphologies on the targets and a damage map for the experiments was elaborated, which showed that the most severe damage (cracks and crack networks) begin to appear in the range of the measured recrystallization temperature of the samples, which was lower than expected.

Using Vickers hardness, the recovery and recrystallization kinetics of the material were characterized by means of logarithmic decay and a modified version of JMAK recrystallization kinetics that includes an incubation time for the onset of recrystallization. Recrystallization kinetics were found to accelerate as the hydrogen exposure progresses, thus yielding lower effective activation energies for recrystallization when comparing furnace one hour exposures, plasma one hour exposures, and plasma four hour exposures. This pointed to the presence of hydrogen actively reducing the activation energy for self-diffusion. Simulations of hydrogen diffusion were performed to test this hypothesis, and even though the total concentration is low, given that the high experimental temperature does not permit trapping of the hydrogen, the diffusing atoms may still play a role in accelerating the recrystallization kinetics.

Based on the results of this research, it is proposed that interstitially diffusing hydrogen segregating to voids or grain boundaries is modifying the behavior of surrounding tungsten crystal lattice. Specifically, the mobility of the grain boundaries may be increasing because the hydrogen’s presence would be promoting the creation of ledges in the grain boundary resulting in an overall free energy reduction for grain boundary diffusion. This localized defect formation would require a lower concentration of hydrogen than that required for solute drag or other suppression mechanisms, and might be mechanism behind the behavior observed in this work’s experiments. ... Read more