Eldorado Collection:
http://hdl.handle.net/2003/39779
2024-03-29T10:09:22ZUltrafast coherent lattice dynamics coupled to spins in the van der Waals antiferromagnet FePS3
http://hdl.handle.net/2003/42400
Title: Ultrafast coherent lattice dynamics coupled to spins in the van der Waals antiferromagnet FePS3
Authors: Mertens, Fabian
Abstract: 2D materials, like the antiferromagnetic van der Waals semiconductors FePS3 studied
in this work, open up new possibilities for technological applications due to the
unique interaction of their magnetization with electronic, optical, and mechanical
properties. Furthermore, they provide the potential to study magnetism and
magnetization dynamics in reduced dimensions. Up do date, the coherent control of
the magnetization of these materials has barely been studied. Our research addresses
this gap by using ultrashort light pulses. In this context, time-resolved studies can
give an insight into the evolution of the light-induced dynamics, which essentially
require a dedicated experimental setup.
In this thesis, we present a comprehensive study on the development and application
of a table-top laser setup designed for magneto-optical pump-probe experiments and
adaptable for the investigation of microscopic samples. The system employs two
optical parametric amplifiers, with a tunable photon-energy range of 0.5 eV - 3.5 eV
for both the pump and the probe beam. Remarkable is the high pump amplitude
modulation rate at 50 % of the laser repetition rate, realized via the integration of
an electro-optical modulator, blocking every second pump pulse. Combined with a
high-frequency digitizer, performing single pulse detection, our system can achieve a
high sensitivity, down to 50 µdeg of the probe polarization rotation. The setup can
apply magnetic fields of up to ±9 T, and voltages in the kV regime while providing
a temperature control between 4 K-420 K.
The functionality of the setup’s systems is demonstrated by performing static Kerrrotation
and ultrafast demagnetization measurements in a cobalt single crystal as a
function of the most important experimental parameters.
The major part of this thesis is dedicated to our studies on a coherent optical
lattice mode of terahertz frequency triggered by femtosecond laser pulses in the
antiferromagnetic van der Waals semiconductor FePS3 . This specific 3.2 THz phonon
mode shows a close relation to the antiferromagnetic order, as it vanishes above the
Néel temperature and hybridizes with a magnon mode in the presence of a magnetic
field. We investigate it as a function of sample temperature, probe polarization,
excitation photon energy and externally applied magnetic fields. The resonant
excitation of a crystal-field split electronic ..-.. transition efÏciently pumps the
displacive excitation process of the mode, while the magnetic linear dichroism is
identified as the magneto-optical effect, which reflects the phonon mode in the probe
rotation. By applying magnetic fields of up to 9 T we can generate and observe
the coherent hybridized phonon-magnon mode, thus exploiting the hybridization to
excite coherent spin-dynamics. Furthermore, we investigate the coherent phonons
in the bulk form of FePS3 and in an exfoliated flake with a thickness of 380 nm.2023-01-01T00:00:00ZModulation of the transient magnetization of an EuO/Co bilayer by controlled optical excitation
http://hdl.handle.net/2003/42246
Title: Modulation of the transient magnetization of an EuO/Co bilayer by controlled optical excitation
Authors: Mönkebüscher, David
Abstract: Der ferromagnetische Halbleiter Europiummonoxid (EuO) gilt als vielversprechender Kandidat für neuartige spintronische Anwendungen, da er ein großes magnetisches Moment und starke magneto-optische Effekte mit isolierenden Eigenschaften vereint. Obwohl EuO mit T_C = 69 K die höchste Curie-Temperatur unter den Europiumchalkogeniden aufweist, ist sie für kommerzielle Anwendungen zu niedrig. Viele Ansätze zur Erhöhung von T_C, wie zum Beispiel die Dotierung mit Gd-Ionen oder epitaktische Verformung, wurden bereits erfolgreich untersucht. Jedoch basieren sie alle auf einer Veränderung der Stöchiometrie und Leitfähigkeit des Seltenerdoxids. Das Ausnutzen des Proximity-Effektes könnte eine alternative Herangehensweise für das starke Erhöhen der magnetischen Ordnungstemperatur von EuO darstellen, die gleichzeitig dessen intrinsischen Eigenschaften bewahrt. Dieser Effekt beruht auf der Kopplung an einen Ferromagneten mit hoher Curie-Temperatur und ist in der Literatur für ähnliche System bereits demonstriert worden.
In dieser Arbeit wird ein EuO/Co-Zweischichtsystem dünner Filme mittels des statischen und zeitaufgelösten magneto-optischen Kerr-Effekts (MOKE) untersucht, um einen Nachweis für eine erhöhte Curie-Temperatur von EuO aufgrund der Nähe zum Übergangsmetall Co zu finden. Des Weiteren wird der Einfluss von Co auf die Spindynamik von EuO untersucht. Statische Messungen der Hysterese der EuO/Co-Probe zeigen eine antiferromagnetische Kopplung zwischen den beiden ferromagnetischen Schichten. Aufgrund der Überlagerung des Signals beider Schichten übersteigt die Co-Hysterese einen möglichen Restbeitrag von EuO bei erhöhten Temperaturen. Zeitaufgelöste MOKE-Messungen zeigen eine transiente Verstärkung der EuO-Magnetisierung, die auch dann auftritt, wenn selektiv nur das Übergangsmetall angeregt wird. Dieses Verhalten wird auf die Erzeugung eines superdiffusven Spinstroms von Majoritätselektronen bei der Entmagnetisierung der Co-Schicht zurückgeführt. Der Spinstrom breitet sich in Richtung der EuO-Schicht aus, um deren 5d-Zustände zu besetzen, was zu einer ähnlichen Magnetisierungsverstärkung wie bei einer direkten Photoanregung des Seltenerdoxids führt. Die Beiträge beider Schichten zur transienten Spindynamik zeigen entgegengesetze Vorzeichen. Daher bietet die EuO/Co-Probe ein System, in dem die transiente Kerr-Rotation durch Variation externer Parameter wie der Probentemperatur, des angelegten Magnetfelds und der Pumpstrahlfluenz beeinflusst werden kann. Durch eine starke Anregung der Co-Schicht wird ihre Magnetisierung signifikant verringert, wodurch die Hysterese der EuO-Schicht bei transienten Hysteresemessungen zugänglich wird. Sie ist auch noch bei einer Temperatur von 300 K zu beobachten, was auf eine starke Erhöhung der magnetischen Ordnungstemperatur von EuO, bedingt durch die Nähe zu Co, hindeutet.; The ferromagnetic semiconductor europium monoxide (EuO) is a promising candidate for new spintronic applications due to its large magnetic moment and strong magneto-optical effects combined with its insulating properties. Although EuO has the highest Curie temperature among the europium chalcogenides with T_C = 69 K, it still requires excessive cooling in real applications. Many approaches to increase its T_C have been successfully studied, such as doping with Gd ions or epitaxial straining, which inevitably change the stoichiometry and conductivity of the rare earth oxide. An alternative pathway to greatly increase the Curie temperature of EuO while preserving its conductivity and stoichiometry could be based on the magnetic proximity effect. This effect relies on the coupling to a high T_C ferromagnet, and has been demonstrated in the literature for similar systems.
In this thesis, a EuO/Co bilayer of thin films is studied using the static and time-resolved magneto-optical Kerr effect (MOKE) to find an evidence for an elevated EuO T_C due to the proximity to the transition metal Co. Furthermore, the influence of Co on the spin dynamics of EuO is investigated.
Static measurements of the EuO/Co bilayer hysteresis reveal an antiferromagnetic coupling between the two ferromagnetic layers. Due to the superposition of the measured signal of both layers, the Co hysteresis exceeds any possible residual EuO contribution to the magneto-optical signal above its bulk T_C. Time-resolved MOKE measurements show a transient enhancement of the EuO magnetization.
It is still present when the Co layer is selectively photoexcited by tuning the photon energy of the pump beam below the EuO band gap energy. This behavior is attributed to the generation of a superdiffusive spin current of majority electrons upon demagnetizing the Co layer. It propagates towards the EuO layer to populate its 5d states, inducing a similar magnetization enhancement compared to direct photoexcitation of the rare earth oxide. The two layers of the investigated sample system exhibit a contribution to the transient spin dynamics with opposite signs. Therefore, the EuO/Co bilayer provides a system in which the transient magneto-optical Kerr rotation can be tuned by varying external parameters such as the sample temperature, the applied magnetic field and the pump beam fluence. By strongly exciting the Co layer and thus quenching its magnetization, the EuO hysteresis becomes accessible in transient hysteresis measurements. It persists up to a temperature of 300 K, pointing to an experimental evidence for room temperature magnetic order in EuO induced by proximity to Co.2023-01-01T00:00:00ZSmall beams, fast predictions: a comparison of machine learning dose prediction models for proton minibeam therapy
http://hdl.handle.net/2003/42239
Title: Small beams, fast predictions: a comparison of machine learning dose prediction models for proton minibeam therapy
Authors: Mentzel, Florian; Kröninger, Kevin; Lerch, M.; Nackenhorst, Olaf; Rosenfeld, A.; Tsoi, A. C.; Weingarten, Jens; Hagenbuchner, M.; Guatelli, S.
Abstract: Background:
Dose calculations for novel radiotherapy cancer treatments such as proton minibeam radiation therapy is often done using full Monte Carlo (MC) simulations. As MC simulations can be very time consuming for this kind of application, deep learning models have been considered to accelerate dose estimation in cancer patients.
Purpose:
This work systematically evaluates the dose prediction accuracy, speed and generalization performance of three selected state-of-the-art deep learning models for dose prediction applied to the proton minibeam therapy. The strengths and weaknesses of those models are thoroughly investigated, helping other researchers to decide on a viable algorithm for their own application.
Methods:
The following recently published models are compared: first, a 3D U-Net model trained as a regression network, second, a 3D U-Net trained as a generator of a generative adversarial network (GAN) and third, a dose transformer model which interprets the dose prediction as a sequence translation task. These models are trained to emulate the result of MC simulations. The dose depositions of a proton minibeam with a diameter of 800μm and an energy of 20–100 MeV inside a simple head phantom calculated by full Geant4 MC simulations are used as a case study for this comparison. The spatial resolution is 0.5 mm. Special attention is put on the evaluation of the generalization performance of the investigated models.
Results:
Dose predictions with all models are produced in the order of a second on a GPU, the 3D U-Net models being fastest with an average of 130 ms. An investigated 3D U-Net regression model is found to show the strongest performance with overall 61.0%±0.5% of all voxels exhibiting a deviation in energy deposition prediction of less than 3% compared to full MC simulations with no spatial deviation allowed. The 3D U-Net models are observed to show better generalization performance for target geometry variations, while the transformer-based model shows better generalization with regard to the proton energy.
Conclusions:
This paper reveals that (1) all studied deep learning models are significantly faster than non-machine learning approaches predicting the dose in the order of seconds compared to hours for MC, (2) all models provide reasonable accuracy, and (3) the regression-trained 3D U-Net provides the most accurate predictions.2022-10-30T00:00:00ZApplication of machine learning in glow curve deconvolution
http://hdl.handle.net/2003/42094
Title: Application of machine learning in glow curve deconvolution
Authors: Lienau, Evelin
Abstract: Routine dosimetry aims to estimate the radiation dose of occupationally exposed persons for a monitoring interval of one month. The Material Prüfungsamt NRW (MPA NRW) provides a thermoluminescence (TL) dosimeter based on LiF:Mg,Ti (TL-DOS). Thermal fading causes a time-dependent signal loss when using a TL dosimeter. This signal change is used to gain information about the irradiation event beyond the dose estimate, which can help to improve the radiation protection concept of occupationally exposed persons.
In this work, multivariate analysis techniques for glow curve analysis using deep learning approaches are developed to estimate the irradiation day within a monitoring interval of 40 days with single-dose irradiation using a Cs-137 source with a prediction uncertainty of two days.
To create a data basis for training the application of deep learning, over 10 000 measurements were performed in cooperation with the MPA NRW and the TL-DOS project. Furthermore, a technique to generate realistic glow curves based on generative adversarial networks (GANs) is presented, which makes it possible to expand the measured data set artificially and thus create a larger database for the deep learning approaches.2023-01-01T00:00:00ZMicrobeams - quick and dirty
http://hdl.handle.net/2003/42003
Title: Microbeams - quick and dirty
Authors: Mentzel, Florian
Abstract: Microbeam radiation therapy (MRT) is a promising yet preclinical radiotherapy treatment for several tumour diagnosis such as gliosarcoma and radioresistant melanoma for which even modern clinical treatments such as intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT) yield poor outcome perspectives. The dose prediction during MRT treatment planning, as for most other novel radiotherapies, is mostly performed with very time-consuming Monte Carlo (MC) simulations. This slows down preclinical research processes and renders treatment plan optimization infeasible.
In this thesis, several milestones for the introduction of a fast machine learning (ML) dose calculation method for MRT are presented. First, a 3D U-Net-based ML dose engine is developed using MC training data obtained with Geant4 simulations of a synchrotron broadbeam incident on different bone slab models and a simplified human head phantom as a proof of concept. The developed model is shown to produce dose predictions within less than 100ms which is substantially faster than the used MC simulations with up to 20hours and also the currently fastest approximative MRT dose prediction approach, called HybridDC, with approximately 30minutes. The model is also shown to be superior to a dose prediction approach using generative adversarial networks (GANs) and also a novel transformer-based ML model called Dose Transformer (DoTA), with which it is compared for application in proton minibeam radiation therapy (pMBRT) in a subsequent study. Secondly, the developed ML model and the MC simulations for data generation are extended to account for the spatially fractionated nature of MRT. For this, a novel MC scoring method is developed which is able produce separate dose estimations for the high-dose peak regions where the microbeams traverse the phantoms and the low-dose valley regions in-between those beams. Finally, the developed ML model and the MC scoring method are deployed in a first application of an ML dose prediction method in a preclinical MRT study in collaboration with the University of Wollongong, Australia, conducted at the Imaging and Medical Beamline (IMBL) at the Australian Synchrotron which aimed at treating rats after implanting gliosarcoma cells. It is shown that the ML model can be trained to provide unbiased dose estimations in complex target phantoms even when trained on high-noise MC data, in important finding for the acceleration of future developments of ML models as such datasets can be produced significantly faster. The ML predictions in the rat phantoms deviate at most 10% from the MC simulations, rendering the proposed model a suitable candidate for fast dose predictions during treatment plan optimization in the future.; C (MRT) ist eine vielversprechende vorklinische Strahlentherapie für einige Tumordiagnosen, wie beispielsweise Gliosarcome und radioresistente Melanome, für die auch moderne Therapiemethoden wie intensity-modulated radiation therapy (IMRT) und volumetric modulated arc therapy (VMAT) schlechte Therapieaussichten haben. Die Dosisvorhersage während der Behandlungsplanung für MRT, ebenso wie für viele andere neue Strahlentherapien, wird meistens mit sehr zeitaufwändigen Monte Carlo (MC) Simulationen durchgeführt. Dies zieht die Forschungsschritte in vorklinischen Studien in die Länge und verhindert vor allem die Optimierung von Behandlungsplänen. In dieser Arbeit werden mehrere Meilensteine für die Einführung einer schnellen MRT-Dosisberechnungsmethode auf der Basis von ML präsentiert. Zuerst wird ein machine learning (ML)-Dosisberechnungsmodell auf der Grundlage eines 3D U-Nets entwickelt. Dazu werden zunächst MC Trainingsdaten mithilfe von Geant4 Simulationen erzeugt, die die Dosisverteilung in verschiedenen Knochenscheibenphantomen und einem vereinfachten Kopfphantom nach Bestrahlung mit einem sogenannten Synchrotron broadbeam vorhersagen. Das entwickelte Modell erzeugt Dosisvorhersagen innerhalb von weniger als 100ms, was signifikant schneller als die Laufzeit der verwendeten MC Simulationen (bis zu 20Stunden) und ebenfalls die zur Zeit schnellsten MRT Dosisberechnungsmethode mithilfe von Approximationen, der sogenannten HybridDC Methode (ca. 30Minuten). Darüber hinaus wird gezeigt, dass das vorgestellte Modell sowohl bessere Vorhersageergebnisse als ein alternativer ML-Ansatz auf Basis von generative adversarial networks (GANs), als auch ein neues Transformer-basiertes ML-Modell namens Dose Transformer (DoTA) erreicht. Der Vergleich mit dem DoTA-Modell erfolgt in einer Studie zur Dosisvorhersage einer anderen neuen Strahlentherapiemethode, der proton minibeam radiation therapy (pMBRT). Anschließend wird das entwickelte ML-Modell und die MC Simulationen weiterentwickelt, um der räumlich fraktionierten Natur von MRT gerechnet zu werden. Dazu wird eine neue MC Scoringmethode entwickelt, welche separate Dosisverteilungen für den Peakbereich, in dem die Microbeams die Phantome durchqueren und eine hohe Dosis deponieren, und für den Valleybereich mit deutlich geringeren Dosisdepositionen dazwischen erstellt. Abschließend werden das entwickelte ML-Modell und die neue MC Scoringmethode in einer ersten Anwendung von ML-Dosisvorhersagemethoden in einer vorklinischen MRT-Studie einer Forschungsgruppe der University of Wollongong angewendet, in der mit Gliosarcomen implantierte Ratten an der Imaging and Medical Beamline (IMBL) am Australian Synchrotron bestrahlt wurden. Es wird gezeigt, dass das ML-Modell nach dem Training Dosisvorhersagen ohne Bias erzeugen kann, obwohl es mithilfe von MC Simulationen mit einer hohen statistischen Unsicherheit trainiert wird. Dies ist eine wichtige Erkenntnis für die beschleunigte Entwicklung zukünftiger ML-Modelle, da solche Daten deutlich schneller erzeugt werden können. Die produzierten Dosisvorhersagen weichen zumeist höchstens 10% von den MC Simulationen ab, daher wird das entwickelte Modell als geeigneter Kandidat für zukünftige schnelle Dosisvorhersagen für die Planungsoptimierung von MRT-Bestrahlungen eingeordnet2023-01-01T00:00:00ZPhysics-based generative model of curvature sensing peptides; distinguishing sensors from binders
http://hdl.handle.net/2003/41365
Title: Physics-based generative model of curvature sensing peptides; distinguishing sensors from binders
Authors: Hilten, Niek van; Methorst, Jeroen; Verwei, Nino; Risselada, Herre Jelger
Abstract: Proteins can specifically bind to curved membranes through curvature-induced hydrophobic lipid packing defects. The chemical diversity among such curvature “sensors” challenges our understanding of how they differ from general membrane “binders” that bind without curvature selectivity. Here, we combine an evolutionary algorithm with coarse-grained molecular dynamics simulations (Evo-MD) to resolve the peptide sequences that optimally recognize the curvature of lipid membranes. We subsequently demonstrate how a synergy between Evo-MD and a neural network (NN) can enhance the identification and discovery of curvature sensing peptides and proteins. To this aim, we benchmark a physics-trained NN model against experimental data and show that we can correctly identify known sensors and binders. We illustrate that sensing and binding are phenomena that lie on the same thermodynamic continuum, with only subtle but explainable differences in membrane binding free energy, consistent with the serendipitous discovery of sensors.2023-03-17T00:00:00ZQuantum spin dynamics of quasi-one-dimensional Heisenberg-Ising magnets in a transverse field: confined spinons, E8 spectrum, and quantum phase transitions
http://hdl.handle.net/2003/41252
Title: Quantum spin dynamics of quasi-one-dimensional Heisenberg-Ising magnets in a transverse field: confined spinons, E8 spectrum, and quantum phase transitions
Authors: Amelin, Kirill; Viirok, Johan; Nagel, Urmas; Rõõm, Toomas; Engelmayer, Johannes; Dey, Tusharkanti; Agung Nugroho, Agustinus; Lorenz, Thomas; Wang, Zhe
Abstract: We report on high-resolution terahertz spectroscopic studies of quantum spin dynamics in the quasi-one-dimensional Ising-like ferromagnet CoNb2O6 and antiferromagnet BaCo2V2O8 as a function of an applied transverse magnetic field. In the ordered phases stabilized by inter-chain couplings, we reveal characteristics for confined spinon excitations, E8 dynamical spectrum, and field-induced quantum phase transitions. The connections between these characteristic dynamical features are found in the field-dependent evolution of the excitation spectra.2022-12-07T00:00:00ZComparing biological effectiveness guided plan optimization strategies for cranial proton therapy: potential and challenges
http://hdl.handle.net/2003/41246
Title: Comparing biological effectiveness guided plan optimization strategies for cranial proton therapy: potential and challenges
Authors: Hahn, Christian; Heuchel, Lena; Ödén, Jakob; Traneus, Erik; Wulff, Jörg; Plaude, Sandija; Timmermann, Beate; Bäumer, Christian; Lühr, Armin
Abstract: Background:
To introduce and compare multiple biological effectiveness guided (BG) proton plan optimization strategies minimizing variable relative biological effectiveness (RBE) induced dose burden in organs at risk (OAR) while maintaining plan quality with a constant RBE.
Methods:
Dose-optimized (DOSEopt) proton pencil beam scanning reference treatment plans were generated for ten cranial patients with prescription doses ≥ 54 Gy(RBE) and ≥ 1 OAR close to the clinical target volume (CTV). For each patient, four additional BG plans were created. BG objectives minimized either proton track-ends, dose-averaged linear energy transfer (LETd), energy depositions from high-LET protons or variable RBE-weighted dose (DRBE) in adjacent serially structured OARs. Plan quality (RBE = 1.1) was assessed by CTV dose coverage and robustness (2 mm setup, 3.5% density), dose homogeneity and conformity in the planning target volumes and adherence to OAR tolerance doses. LETd, DRBE (Wedenberg model, α/βCTV = 10 Gy, α/βOAR = 2 Gy) and resulting normal tissue complication probabilities (NTCPs) for blindness and brainstem necrosis were derived. Differences between DOSEopt and BG optimized plans were assessed and statistically tested (Wilcoxon signed rank, α = 0.05).
Results:
All plans were clinically acceptable. DOSEopt and BG optimized plans were comparable in target volume coverage, homogeneity and conformity. For recalculated DRBE in all patients, all BG plans significantly reduced near-maximum DRBE to critical OARs with differences up to 8.2 Gy(RBE) (p < 0.05). Direct DRBE optimization primarily reduced absorbed dose in OARs (average ΔDmean = 2.0 Gy; average ΔLETd,mean = 0.1 keV/µm), while the other strategies reduced LETd (average ΔDmean < 0.3 Gy; average ΔLETd,mean = 0.5 keV/µm). LET-optimizing strategies were more robust against range and setup uncertaintes for high-dose CTVs than DRBE optimization. All BG strategies reduced NTCP for brainstem necrosis and blindness on average by 47% with average and maximum reductions of 5.4 and 18.4 percentage points, respectively.
Conclusions:
All BG strategies reduced variable RBE-induced NTCPs to OARs. Reducing LETd in high-dose voxels may be favourable due to its adherence to current dose reporting and maintenance of clinical plan quality and the availability of reported LETd and dose levels from clinical toxicity reports after cranial proton therapy. These optimization strategies beyond dose may be a first step towards safely translating variable RBE optimization in the clinics.2022-10-22T00:00:00ZExplanation, Reduction, Progress
http://hdl.handle.net/2003/40887
Title: Explanation, Reduction, Progress
Authors: Scheibe, Erhard1987-01-01T00:00:00Z