Publications
Permanent URI for this collection
Members of TU Berlin can publish different kinds of (text) publications on DepositOnce:
- publications with academic content such as monographs and collected editions, journal articles and book chapters, conference proceedings and conference papers, research reports,
- publications that must be made public in connection with examination regulations (doctoral theses and post-doctoral theses),
- theses of TU Berlin students if the thesis has received the grade "very good" and the publication is endorsed by the supervisor or referee.
The following publication types can be selected: Doctoral thesis, Habilitation, Master thesis, Bachelor thesis, Book, Conference Proceedings, Periodical Part, Preprint, Report, Research Paper, Article, Book Part, Conference Object
For further information on requirements and workflows see Checklists and Hints for Publishing.
Contact:- Dissertation Office (theses): dissertationen@ub.tu-berlin.de
- University Press: publikationen@ub.tu-berlin.de
- Other text publications (reports, open access self-archiving for journal articles, etc.): openaccess@ub.tu-berlin.de
Browse by
Recent Submissions
- 2024-03-04Article
Ground‐dwelling mammal and bird diversity in the southern Annamites: Exploring complex habitat associations and the ghost of past hunting pressure
The Langbian Plateau, a biodiversity hotspot in the southern Annamites of Viet Nam, has undergone extensive hunting pressure. However, the limited information on the effects of overexploitation on the current status and community composition of wildlife hinders effective conservation efforts, including the implementation of targeted patrols to reduce snaring. In this study, we conducted a camera-trapping survey across the Langbian Plateau, consisting of a broadleaf evergreen and coniferous habitat mosaic. We recorded 46 ground-dwelling mammals and birds, including several threatened Annamite endemics. Using multi-species Royle-Nichols model and landscape covariates, we found higher richness in broadleaf evergreen forest located in more remote and less rugged areas. We then used species responses to covariates to predict species distribution and identify high-priority areas for conservation. Furthermore, we constructed diversity profiles that indicated higher biodiversity in broadleaf evergreen forest compared to the coniferous forest. Finally, we used a dissimilarity index to assess the level of defaunation, revealing 16% of the community had been lost, with higher levels of defaunation for threatened and larger-sized species. Our findings provide insights into the status, distribution, and occurrence of the ground-dwelling mammal and bird communities in the Langbian Plateau, and can help stakeholders design more effective conservation strategies to protect existing populations. - 2024-01-17Article
Computed tomographic study analysing functional biomechanics in the thoracolumbar spine of horses with and without spinal pathology
To better understand physiological and pathological movement patterns in the equine thoracolumbar spine, investigation of the biomechanics on a segmental level requires a constant moment. A constant moment along the spinal column means that the same torque acts on each vertebral segment, allowing the range of motion of different segments to be compared. The aims of this study were to investigate the range of motion of the equine thoracolumbar spine in horses with and without spinal pathology and to examine whether the pressure between the spinous processes depends on the direction of the applied moment. Thoracolumbar spine specimens (T8-L4) of 23 horses were mounted in a custom-made mechanical test rig to investigate spinal biomechanics during lateral bending, axial rotation, flexion and extension using computed tomographic imaging. Results were compared between horses with spondylosis, overriding spinous processes and specimens free of gross pathology. The interspinous space pressure was additionally determined using a foil sensor. The median lateral bending between T9 and L3 was 3.7°–4.1° (IQR 5.4°–8.0°). Maximum rotational movement with inconsistent coupled motion was observed at T9–T16 (p < 0.05). The dorsoventral range of motion was greatest in segments T9–T11 (p < 0.05). Spondylosis and overriding spinous processes restricted spinal mobility, depending on the severity of the condition. There was no significant difference in interspinous pressure during motion (p = 0.54). The biomechanical study confirmed that the range of motion of intervertebral joints depends on the anatomical position of the joint and the direction of the moment applied. Restricted mobility was evident in the presence of different grades of overriding spinous processes or spondylosis. A better understanding of equine spinal biomechanics in horses with spinal pathology facilitates individual rehabilitation. - 2023-07-14Article
iKNOWgynetics – A web‐based learning concept to empower primary care gynecologists to participate in the care of patients with a family history of breast and ovarian cancer
Familial cancer burden and genetics play an increasingly important role in the early detection and prevention of gynecological cancers. However, people with hereditary cancer risks are often identified late when they already have cancer. We aimed at developing and evaluating a training concept for primary care gynecologists—iKNOWgynetics—to improve their knowledge and awareness of genetic cancer syndromes and their ability to identify patients with increased familial cancer risks based on up-to-date evidence and current guidelines (in Germany, primary care includes all doctors treating patients on an outpatient basis without a clear separation of the expertise of the doctor or of their specialty). Starting off with a needs assessment among primary care gynecologists, we developed and evaluated an online training concept—using a web-based learning platform in combination with a live virtual seminar—to convey practice-relevant knowledge about familial cancer. After registration, participants get access to the web-based learning platform (www.iknowgynetics.de) to prepare for the virtual seminars and to use it as online reference to re-access the contents after the training. Evaluation included multiple-choice (MC) questions on knowledge and participants' self-efficacy to implement the acquired knowledge, which were administered in a pre-post design. Of 109 participants, 103 (94.5%) filled out pre- and post-questionnaires. Eighty-five participants were gynecologists in primary care from Berlin (81.2%) and Brandenburg (18.8%) and had an average of 24.1 years (SD = 8.5 years) of professional experience. After the training, participants answered significantly more knowledge questions correctly (M = 15.2 of 17, SD = 1.3) than before (M = 13.8 of 17, SD = 1.7) (p < 0.01) and felt more confident to be able to apply referral criteria for specialized counseling in practice (p < 0.001). The online-based training iKNOWgynetics considers the busy schedule of primary care gynecologists and supports them in acquiring practice-relevant information on familial cancer risks and on how to identify healthy persons at risk, which may ultimately help to improve the prevention of gynecological cancers. In future studies, the reported concept could be transferred to other entities. - 2023-12-06Article
Minimizing Interfacial Recombination in 1.8 eV Triple‐Halide Perovskites for 27.5% Efficient All‐Perovskite Tandems
Yang, Fengjiu; Tockhorn, Philipp; Musiienko, Artem; Lang, Felix; Menzel, Dorothee; Macqueen, Rowan; Köhnen, Eike; Xu, Ke; Mariotti, Silvia; Mantione, Daniele; Merten, Lena; Hinderhofer, Alexander; Li, Bor; Wargulski, Dan R.; Harvey, Steven P.; Zhang, Jiahuan; Scheler, Florian; Berwig, Sebastian; Roß, Marcel; Thiesbrummel, Jarla; Al‐Ashouri, Amran; Brinkmann, Kai O.; Riedl, Thomas; Schreiber, Frank; Abou‐Ras, Daniel; Snaith, Henry; Neher, Dieter; Korte, Lars; Stolterfoht, Martin; Albrecht, SteveAll-perovskite tandem solar cells show great potential to enable the highest performance at reasonable costs for a viable market entry in the near future. In particular, wide-bandgap (WBG) perovskites with higher open-circuit voltage (VOC) are essential to further improve the tandem solar cells’ performance. Here, a new 1.8 eV bandgap triple-halide perovskite composition in conjunction with a piperazinium iodide (PI) surface treatment is developed. With structural analysis, it is found that the PI modifies the surface through a reduction of excess lead iodide in the perovskite and additionally penetrates the bulk. Constant light-induced magneto-transport measurements are applied to separately resolve charge carrier properties of electrons and holes. These measurements reveal a reduced deep trap state density, and improved steady-state carrier lifetime (factor 2.6) and diffusion lengths (factor 1.6). As a result, WBG PSCs achieve 1.36 V VOC, reaching 90% of the radiative limit. Combined with a 1.26 eV narrow bandgap (NBG) perovskite with a rubidium iodide additive, this enables a tandem cell with a certified scan efficiency of 27.5%. - 2024-02-08Article
Isomaltulose Enhances GLP‐1 and PYY Secretion to a Mixed Meal in People With or Without Type 2 Diabetes as Compared to Saccharose
Scope: Secretion of the gut hormones glucagon-like peptide (GLP-1) and peptide YY (PYY) are induced by nutrients reaching the lower small intestine which regulate insulin and glucagon release, inhibit appetite, and may improve β-cell regeneration. The aim is to test the effect of a slowly digested isomaltulose (ISO) compared to the rapidly digested saccharose (SAC) as a snack given 1 h before a standardized mixed meal test (MMT) on GLP-1, PYY, glucose-dependent insulinotropic peptide (GIP), and metabolic responses in participants with or without type 2 diabetes (T2DM). Methods and results: Fifteen healthy volunteers and 15 patients with T2DM consumed either 50 g ISO or SAC 1 h preload of MMT on nonconsecutive days. Clinical parameters and incretin hormones are measured throughout the whole course of MMT. Administration of 50 g ISO as compared to SAC induced a significant increase in GLP-1, GIP, and PYY responses over 2 h after intake of a typical lunch in healthy controls. Patients with T2DM showed reduced overall responses of GLP-1 and delayed insulin release compared to controls while ISO significantly enhanced the GIP and almost tripled the PYY response compared to SAC. Conclusion: A snack containing ISO markedly enhances the release of the metabolically advantageous gut hormones PYY and GLP-1 and enhances GIP release in response to a subsequent complex meal. - 2024-02-20Article
Resource asynchrony and landscape homogenization as drivers of virulence evolution: The case of a directly transmitted disease in a social host
Throughout the last decades, the emergence of zoonotic diseases and the frequency of disease outbreaks have increased substantially, fuelled by habitat encroachment and vectors overlapping with more hosts due to global change. The virulence of pathogens is one key trait for successful invasion. In order to understand how global change drivers such as habitat homogenization and climate change drive pathogen virulence evolution, we adapted an established individual-based model of host–pathogen dynamics. Our model simulates a population of social hosts affected by a directly transmitted evolving pathogen in a dynamic landscape. Pathogen virulence evolution results in multiple strains in the model that differ in their transmission capability and lethality. We represent the effects of global change by simulating environmental changes both in time (resource asynchrony) and space (homogenization). We found an increase in pathogenic virulence and a shift in strain dominance with increasing landscape homogenization. Our model further indicated that lower virulence is dominant in fragmented landscapes, although pulses of highly virulent strains emerged under resource asynchrony. While all landscape scenarios favoured co-occurrence of low- and high-virulent strains, the high-virulence strains capitalized on the possibility for transmission when host density increased and were likely to become dominant. With asynchrony likely to occur more often due to global change, our model showed that a subsequent evolution towards lower virulence could lead to some diseases becoming endemic in their host populations. - 2024-02-24Article
Working memory signals in early visual cortex are present in weak and strong imagers
It has been suggested that visual images are memorized across brief periods of time by vividly imagining them as if they were still there. In line with this, the contents of both working memory and visual imagery are known to be encoded already in early visual cortex. If these signals in early visual areas were indeed to reflect a combined imagery and memory code, one would predict them to be weaker for individuals with reduced visual imagery vividness. Here, we systematically investigated this question in two groups of participants. Strong and weak imagers were asked to remember images across brief delay periods. We were able to reliably reconstruct the memorized stimuli from early visual cortex during the delay. Importantly, in contrast to the prediction, the quality of reconstruction was equally accurate for both strong and weak imagers. The decodable information also closely reflected behavioral precision in both groups, suggesting it could contribute to behavioral performance, even in the extreme case of completely aphantasic individuals. Our data thus suggest that working memory signals in early visual cortex can be present even in the (near) absence of phenomenal imagery. - 2023-08-17Article
Global disparity of camera trap research allocation and defaunation risk of terrestrial mammals
Quantifying and monitoring the risk of defaunation and extinction require assessing and monitoring biodiversity in impacted regions. Camera traps that photograph animals as they pass sensors have revolutionized wildlife assessment and monitoring globally. We conducted a global review of camera trap research on terrestrial mammals over the last two decades. We assessed if the spatial distribution of 3395 camera trap research locations from 2324 studies overlapped areas with high defaunation risk. We used a geospatial distribution modeling approach to predict the spatial allocation of camera trap research on terrestrial mammals and to identify its key correlates. We show that camera trap research over the past two decades has not targeted areas where defaunation risk is highest and that 76.8% of the global research allocation can be attributed to country income, biome, terrestrial mammal richness, and accessibility. The lowest probabilities of camera trap research allocation occurred in low-income countries. The Amazon and Congo Forest basins – two highly biodiverse ecosystems facing unprecedented anthropogenic alteration – received inadequate camera trap research attention. Even within the best covered regions, most of the research (64.2%) was located outside the top 20% areas where defaunation risk was greatest. To monitor terrestrial mammal populations and assess the risk of extinction, more research should be extended to regions with high defaunation risk but have received low camera trap research allocation. - 2024-02-19Article
Causally‐Informed Deep Learning to Improve Climate Models and Projections
Climate models are essential to understand and project climate change, yet long-standing biases and uncertainties in their projections remain. This is largely associated with the representation of subgrid-scale processes, particularly clouds and convection. Deep learning can learn these subgrid-scale processes from computationally expensive storm-resolving models while retaining many features at a fraction of computational cost. Yet, climate simulations with embedded neural network parameterizations are still challenging and highly depend on the deep learning solution. This is likely associated with spurious non-physical correlations learned by the neural networks due to the complexity of the physical dynamical system. Here, we show that the combination of causality with deep learning helps removing spurious correlations and optimizing the neural network algorithm. To resolve this, we apply a causal discovery method to unveil causal drivers in the set of input predictors of atmospheric subgrid-scale processes of a superparameterized climate model in which deep convection is explicitly resolved. The resulting causally-informed neural networks are coupled to the climate model, hence, replacing the superparameterization and radiation scheme. We show that the climate simulations with causally-informed neural network parameterizations retain many convection-related properties and accurately generate the climate of the original high-resolution climate model, while retaining similar generalization capabilities to unseen climates compared to the non-causal approach. The combination of causal discovery and deep learning is a new and promising approach that leads to stable and more trustworthy climate simulations and paves the way toward more physically-based causal deep learning approaches also in other scientific disciplines. - 2024Doctoral Thesis
Fluctuations and exciton dynamics in molecular semiconductors
This work describes experimental studies of the fluctuation-dominated electronic structure of molecular semiconductors and of the exciton dynamics in the same materials. The method employed in this thesis is time- and angle-resolved photoemission spectroscopy. In the first part, the electronic structure of molecular semiconductors at equilibrium is considered. We show that the experimental band structure of several representative molecular crystals can be described with a common tight-binding model and analyze how both the crystal structure and the π-topology impact the electronic structure. With this consistent description of the electronic states, it is discussed how fluctuations act on different materials and how they impact charge transport properties. These deliberations lead to non-trivial design rules for future materials with high charge mobilities that have been independently verified in charge transport experiments. The subsequent chapter concerns the spatial structure of the electronic states captured in momentum space. By applying the previously established tight-binding model, a strong connection between the anatomy of momentum maps, i.e., constant-energy cuts through the photoemission intensity, the electronic structure, and the real space description of electrons as Bloch states is built up. It is also discussed whether properties arising from fluctuations are visible in these maps. Most importantly, this part provides the basis for interpreting momentum maps of ground and excited states. In the second part, we analyzed the exciton dynamics in some of the same compounds that staged in the previous part. For pentacene single crystals, we obtained the first momentum maps of singlet and triplet excitons. These demonstrate the predicted, but never observed, similar orbital character and different localization properties of the two states and the images allow to study singlet exciton fission in unprecedented detail. Previous experimental results pointed to an instantaneous, coherent excitation of an intermediate bitriplet state. Using the exciton momentum maps, the dynamics could be projected to the orbital character of the underlying states and which proves that bitriplet states are not instantaneously excited, but rather created by a charge-transfer mediated transition from the photoexcited singlet exciton. That establishes the charge-transfer mediated mechanism of the primary step of singlet exciton fission. The last chapter describes the observed exciton dynamics in rubrene and tetracene single crystals, where singlet exciton fission is endothermic. We find that the momentum maps in tetracene are highly similar to those of the structural cousin pentacene. However, the dynamics differ substantially. We demonstrate that one key to understanding singlet exciton fission lies in the relative energy and the coupling strength of the charge-transfer states. In tetracene, the charge-transfer states stabilize the bitriplet state; that facilitates the ultrafast <200 fs creation of the bitriplet exciton but slows down its separation into two independent triplet excitons which occurs in 6 ps. Lastly, it is shown that singlet exciton fission in rubrene takes a different path due to the highly symmetric crystal structure which prohibits a coupling between bitriplet and charge-transfer states. Hence, the separated bitriplet exciton is directly created from the singlet without the involvement of charge-transfer states. We conclude by discussing the possible mechanisms of an ultrafast, but energetically steeply uphill, process, which highlights the need for theories that are able to treat both fluctuations and exciton dynamics. - 2024-02-29Article
Infrared spectroscopy reveals metal-independent carbonic anhydrase activity in crotonyl-CoA carboxylase/reductase
The conversion of CO2 by enzymes such as carbonic anhydrase or carboxylases plays a crucial role in many biological processes. However, in situ methods following the microscopic details of CO2 conversion at the active site are limited. Here, we used infrared spectroscopy to study the interaction of CO2, water, bicarbonate, and other reactants with β-carbonic anhydrase from Escherichia coli (EcCA) and crotonyl-CoA carboxylase/reductase from Kitasatospora setae (KsCcr), two of the fastest CO2-converting enzymes in nature. Our data reveal that KsCcr possesses a so far unknown metal-independent CA-like activity. Site-directed mutagenesis of conserved active site residues combined with molecular dynamics simulations tracing CO2 distributions in the active site of KsCCr identify an ‘activated’ water molecule forming the hydroxyl anion that attacks CO2 and yields bicarbonate (HCO3−). Computer simulations also explain why substrate binding inhibits the anhydrase activity. Altogether, we demonstrate how in situ infrared spectroscopy combined with molecular dynamics simulations provides a simple yet powerful new approach to investigate the atomistic reaction mechanisms of different enzymes with CO2. - 2024-02-21Article
Thermal segment microwell plate control for automated liquid handling setups
Automated high-throughput liquid handling operations in biolabs necessitate miniaturised and automatised equipment for effective space utilisation and system integration. This paper presents a thermal segment microwell plate control unit designed for enhanced microwell-based experimentation in liquid handling setups. The development of this device stems from the need to move towards geometry standardization and system integration of automated lab equipment. It incorporates features based on Smart Sensor and Sensor 4.0 concepts. An enzymatic activity assay is implemented with the developed device on a liquid handling station, allowing fast characterisation via a high-throughput approach. The device outperforms other comparable devices in certain metrics based on automated liquid handling requirements and addresses the needs of future biolabs in automation, especially in high-throughput screening. - 2024-02-06Article
Infrared spectra of SinH4n−1+ ions (n = 2–8): inorganic H–(Si–H)n−1 hydride wires of penta-coordinated Si in 3c–2e and charge-inverted hydrogen bonds
SinHm+ cations are important constituents in silane plasmas and astrochemical environments. Protonated disilane (Si2H7+) was shown to have a symmetric three-centre two-electron (3c–2e) Si–H–Si bond that can also be considered as a strong ionic charge-inverted hydrogen bond with polarity Siδ+–Hδ−–Siδ+. Herein, we extend our previous work to larger SinH4n−1+ cations, formally resulting from adding SiH4 molecules to a SiH3+ core. Infrared spectra of size-selected SinH4n−1+ ions (n = 2–8) produced in a cold SiH4/H2/He plasma expansion are analysed in the SiH stretch range by complementary dispersion-corrected density functional theory calculations (B3LYP-D3/aug-cc-pVTZ) to reveal their bonding characteristics and cluster growth. The ions with n = 2–4 form a linear inorganic H–(Si–H)n hydride wire with adjacent Si–H–Si 3c–2e bridges, whose strength decreases with n, as evident from their characteristic and strongly IR active SiH stretch fundamentals in the range 1850–2100 cm−1. These 3c–2e bonds result from the lowest-energy valence orbitals, and their high stability arises from their delocalization along the whole hydride wire. For SinH4n−1+ with n ≥ 5, the added SiH4 ligands form weak van der Waals bonds to the Si4H19+ chain. Significantly, because the SinH4n−1+ hydride wires are based on penta-coordinated Si atoms leading to supersaturated hydrosilane ions, analogous wires cannot be formed by isovalent carbon. - 2023-11-23Book
Mobilität und Verkehr
Die traditionelle Verkehrsplanung konzentrierte sich bisher darauf, wachsende Verkehrsmengen immer schneller über immer größere Distanzen zu organisieren - mit bis heute schwerwiegenden Folgen für Natur, Mensch und Klima. Demgegenüber stellt die moderne Verkehrsplanung die Mobilität der Menschen ins Zentrum, mit dem Ziel, neben dem Klima- und Umweltschutz auch ihre gesellschaftliche Teilhabe zu gewährleisten. Der vorliegende Band diskutiert diesen Perspektivwechsel anhand zentraler Begrifflichkeiten, wie Mobilität und Verkehr. Damit unterstützt er praktizierende und angehende Planer:innen, ebenso wie Wissenschaftler:innen und Bürger:innen dabei, die Zusammenhänge zwischen Mobilität und Verkehr besser nachzuvollziehen. Der Band ist das Ergebnis eines kollektiven Selbstverständigungsprozesses am Fachgebiet Integrierte Verkehrsplanung der Technischen Universität Berlin. - 2024-01-04Article
Controlling active turbulence by activity patterns
By patterning activity in space, one can control active turbulence. To show this, we use Doi's hydrodynamic equations of a semidilute solution of active rods. A linear stability analysis reveals the resting isotropic fluid to be unstable above an absolute pusher activity. The emergent activity-induced paranematic state displays active turbulence, which we characterize by different quantities including the energy spectrum, which shows the typical power-law decay with exponent −4. Then, we control the active turbulence by a square lattice of circular spots where activity is switched off. In the parameter space lattice constant versus surface-to-surface distance of the spots, we identify different flow states. Most interestingly, for lattice constants below the vorticity correlation length and for spot distances smaller than the nematic coherence length, we observe a multi-lane flow state, where flow lanes with alternating flow directions are separated by a street of vortices. The flow pattern displays pronounced multistability and also appears transiently at the transition to the isotropic active-turbulence state. At larger lattice constants a trapped vortex state is identified with a non-Gaussian vorticity distribution due to the low flow vorticity at the spots. It transitions to conventional active turbulence for increasing spot distance. - 2024-01-09Article
LA-ICP-MS/MS-based Rb–Sr isotope mapping for geochronology
This study introduces a new approach for in situ Rb–Sr dating that utilizes rapid line scans instead of static spot ablation, enabling the creation of two-dimensional 87Rb/86Sr and 87Sr/86Sr isotope ratio and Rb–Sr age maps. The data acquisition is conducted utilizing an ICP-MS/MS instrument with N2O as the reaction gas, coupled to a 193 nm excimer laser via a low-aerosol-dispersion interface. This configuration allows for high repetition rates (>100 Hz) and sensitivities, enabling data acquisition at a high scanning speed and small laser beam size (3–4 μm). Notably, this approach requires just about 1/30 of the sample volume typically utilized in conventional spot ablation mode, while achieving similar levels of precision and accuracy. Line scan ablation is tested and compared to spot ablation on age-homogeneous crystalline muscovite and biotite, for which reference Rb–Sr age data is acquired through ID-TIMS. Results show that a key requirement for accurate Rb–Sr ages based on line scan analyses is matrix correction using chemically matched crystalline mica. By presenting Rb–Sr age maps of three naturally deformed mica samples, we highlight the potential of Rb–Sr mapping for extracting age data from rocks that exhibit complex metamorphic-metasomatic histories and microscale dynamic recrystallization. Additionally, we show that quantitative elemental information (Al, Fe, Si, Li) can be collected alongside Rb–Sr isotope data. This advancement offers a distinctly more insightful assessment of isotope mobility in natural systems, the timing of element enrichment processes and enables, in high-Rb/Sr rock systems, precise and accurate isotopic dating of intricate geological processes at small scales. - 2024Doctoral Thesis
Plasmonic nanostructures: a microscopic perspective
The fields of plasmonics and two-dimensional semiconductors represent some of the most rapidly advancing areas in current solid-state research, driven by the immense potential of these materials for numerous technological applications. This work is situated at the intersection of these two dynamic fields. It focuses on the development of the microscopic framework to describe the electron dynamics in metal nanostructures and the theoretical understanding of exciton dynamics in hybrid metal-semiconductor nanostructures. The methodological umbrella is provided by the Heisenberg equations of motion. It is employed to derive a spatio-temporal, momentum-resolved phase space description of the electron dynamics in metal nanostructures as well as a momentum-resolved description of excitons in hybrid systems consisting of transition metal dichalcogenide (TMDC) monolayers and metal nanostructures. The first part of this thesis focuses on studying a system comprising a metal nanoparticle and a TMDC monolayer using self-consistent Maxwell-Bloch theory. The combined system yields an effective eigenvalue equation governing the center-of-mass motion of dressed excitons in a plasmon-induced potential. Bound states with negative eigenenergies are found in the dynamical equation of the exciton-plasmon hybrid, indicating exciton localization in the plasmon-induced potential. The coupling regime is quantified by computing the scattered light in the near-field, revealing strong exciton-plasmon coupling with an avoided crossing behavior and an effective Rabi splitting of tens of meV. The key achievement of this work lies in the successful development of a comprehensive microscopic approach for studying spatio-temporal, momentum-resolved electron and phonon dynamics in metals in a Wigner phase space representation. This approach accurately reproduces macroscopic equations in both local and non-local formulations and incorporates geometrical effects and multi-band processes, enabling a description of interband transitions based on microscopic parameters. To investigate the potential of actively tuning the optical response of metal nanoparticles using strong THz fields, we develop a fully numerical method combining the three-dimensional momentum-resolved microscopic Boltzmann scattering equations for the electronic Wigner function with a three-dimensional finite-difference time-domain solver. This approach allows for a spatio-temporal treatment of microscopic dynamics, including non-equilibrium, non-perturbative, and nonlocal phenomena, as well as interband transitions. The same scenario is described using an analytical model, revealing that the additional THz pulse effectively renormalizes the electronic ground state of the system through pressure renormalization of the electron gas within the nanoparticle. This leads to a blue shift of the plasmon resonance, which is also observed in experiments. Finally, the framework is applied to study the onset of radial oscillations in spherical nanoparticles. This microscopic approach incorporates direct electron-coherent phonon interaction, which reveals an additional driving source for nanoparticle oscillations. In particular, we identify spatial electron density gradients as the dominant driving source for the onset of radial oscillations on short time scales beyond thermal contributions, providing a compelling explanation for recent experimental findings. - 2024Doctoral Thesis
Energy efficient hardware architectures for memory prohibitive deep neural networks
Deep Neural Networks (DNN) form the backbone of modern Artificial Intelligence (AI) systems. However, due to the high computational complexity and divergent shapes and sizes. Dedicated hardware accelerators are required to achieve very high performance and energy efficiency across various DNNs to enable AI in real-world applications. To address this problem and improve the DNN processor’s energy efficiency, we introduce the multi-bit accelerator. The multi-bit accelerator achieves the energy-efficient goals for a low-power DNN processor by truncating the preceding layer’s partial sums (PSums) before feeding it as an input to the next layer. The Row Stationary (RS) dataflow method is used to implement the design. We start inferencing with the high order bit-width, like 32 bits for the first convolution layers, and sequentially truncate the bits on the MSB/LSB of the integer and the LSB of the fraction part. Even with the sequential truncation, the processor could achieve a top-1 accuracy of up to 14 bits and a top-5 accuracy of up to 10 bits. The proposed truncation scheme helped in reducing the resource utilization by 73.25% for LUTs (Lookup tables), 68.76% for FFs (Flip Flops), and 74.60% for BRAMs (Block RAMs), and 79.425% for DSPs (Digital Signal Processors). The multi-bit accelerator could achieve an overall throughput of 223.39 GOPS on a Virtex Ultra Scale FPGA. The corresponding ASIC version implemented on the GF 22nm FDSOI could achieve an overall throughput of 2.03 TOPS/W with a total power consumption of 791mW and an overall area of 1.2mm x 1.2mm. In order to further improve energy efficiency and area efficiency, we introduce a hardware-software co-designed FantastIC4 accelerator to handle the compact representations of the fully connected layers and reduce the total number of required multipliers to four. In order to make the DNN models amenable for efficient execution, the DNN models are trained to handle the 4-bit quantization. The FantastIC4 accelerator achieves a very high throughput of 2.45 TOPS due to the extreme compression of the models with an overall power consumption of 3.6W on a Virtex Ultrascale FPGA. The equivalent ASIC version implemented on a GF 22nm FDSOI achieves a very high energy efficiency of 20.17 TOPS/W. When compared to other accelerators designed for the Google Speech Command (GSC) dataset, FantastIC4 is better by 51x in terms of throughput and 145x in terms of area efficiency (GOPS/mm2). - 2024Doctoral Thesis
Ladungsträgertransport und Ladungsträgerrekombination in AlGaN-basierten DUV-LEDs
The present work analyses the recombination and the transport of charge carriers in AlGaN-based light-emitting diodes (LEDs) with emission wavelengths (λ) in the deep UV spectral range. In this context, the vertical electrical conductivity (σ_V) of Mg-doped p-AlGaN layers and short-period AlGaN:Mg superlattices (SPSL) with high Al mole fraction was investigated. It was found that σ_V exhibits a strong temperature (T) and electric field strength (F) dependence in both AlGaN:Mg-layers and AlGaN:Mg-SPSLs. This behavior could be attributed to the field-assisted thermal ionization of Mg-acceptors described by the 3D Poole-Frenkel effect (3D-PFE). By fitting the measured σ_V-F-T-curves using a 3D-PFE model, several important material parameters of the investigated structures could be determined, i.e. the ionization energy and the acceptor density. Using drift-diffusion simulations and calculations based on the transfer matrix method the ionization and the transport of charge carriers in AlGaN:Mg-SPSLs were investigated on the example of a superlattice with an average Al mole fraction of 68%. The calculations show that σ_V in this SPSL can only exceed the electrical conductivity of an Al0,68Ga0,32N:Mg-layer for extremely high Al mole fraction differences between wells and barriers (≥ 20%) and extremely long period lengths (≥ 12 nm). The measured σ_V of an Al0,71Ga0,29N:Mg/Al0,65Ga0,35N:Mg-SPSL with a 2 nm period is lower than the vertical conductivity in a Al0,68Ga0,32N:Mg-layer, which is in quantitative agreement with the simulation results. Finally, it was shown that taking 3D-PFE into account is essential for proper selection of the Mg-doping concentration and maximization of the electrical conductivity in AlGaN:Mg-layers during LED operation. Next, the efficiency of 265 nm LEDs was investigated. Here, LEDs on four different AlN/sapphire substrates, i.e. planar, high-temperature-annealed (HTA) AlN, epitaxial laterally overgrown (ELO) AlN and HTA-ELO AlN with different threading-dislocation densities (TDD), were analyzed. The measured external quantum efficiency of each characterized LED was separated into the carrier injection efficiency (η_CIE), the radiative recombination efficiency (η_RRE), and the light extraction efficiency (η_LEE). This is achieved by combining an ABC-model based method and calibrated Monte Carlo ray tracing simulations. LEDs grown on ELO and on HTA-ELO substrates have a maximum η_RRE (η_RRE^max) of (57±5)% and (55±6)%, respectively. The η_RRE^max for devices with growth on HTA ((45±5)%) and the planar ((12±6)%) substrates are lower. These results revealed that for the investigated devices the TDD is not the only parameter responsible for the quantity of the η_RRE, as LEDs on ELO substrate have a comparable η_RRE^max with a much higher TDD (1.6×109 cm-2) compared to LEDs on HTA-ELO substrate with lower TDD (0.95×109 cm-2). The discovered discrepancy is attributed to the formation of dislocation half-loops (DHL). The DHLs are generated due to compressive strain in AlN layers and are induced during the high-temperature process of the HTA and HTA-ELO substrates. In addition, the Shockley-Read-Hall recombination constants (A = (2±1)×107 s-1 on ELO and HTA-ELO-AlN/sapphire and A = (6±2)×107 s-1 on HTA-AlN/sapphire) are determined, as well as the Auger coefficient C = (4±2)×1030 cm6s-1. As it was theoretically demonstrated by simulations, at current densities which are relevant for applications the investigated LEDs cannot achieve a η_RRE^max of 100% due to the high C value even when grown on AlN substrates with extremely low TDD. Moreover, for the investigated heterostructure such a low TDD would only lead to a significant increase of η_RRE^max for relatively low current densities (≤ 25 Acm-2). Additionally, it was shown that η_CIE of the investigated LEDs ranges from 40% to 55% regardless of the substrate. Finally, η_CIE = (26±8)%, η_RRE^max = (43±7)% and η_LEE = (8.8±0.8)% of LEDs with λ = 233 nm were determined. These results indicate that the light extraction efficiency and the carrier injection efficiency offer a great potential for optimization of UV LEDs with particularly short emission wavelengths. - 2024Doctoral Thesis
Three essays on the impact of coding training on students' learning behavior and outcomes
Motivated by the substantial changes instigated by the Fourth Industrial Revolution and its significant implications for education, especially in developing countries like Vietnam, this dissertation meticulously explores the influence of coding education within Vietnam's educational reforms. This comprehensive study consists of three distinct essays, each addressing crucial aspects: (1) evaluating the impact of coding courses on secondary school students' engagement in self-directed coding activities, (2) assessing the effect of teaching coding through visual programming languages on students' computational thinking abilities, and (3) establishing and validating the Coding Motivation Scale for Middle School Students (CMS-M), a reliable instrument designed to measure students' motivation, attitudes, and interests in coding within Vietnam's educational context. The study conducted multiple rounds of surveys in Vietnam to collect data for its analysis. These data sets were then analyzed using both quantitative and qualitative approaches. In detail, the study employed panel data estimation and multivariate statistical techniques, as well as conducted focus group discussions and in-depth interviews. The core findings reveal that coding education exerts a substantial impact. Mastery of coding not only fosters self-directed learning behaviors but also enhances students' computational thinking skills. These outcomes are primarily attributed to the solidification of foundational knowledge and the students' increased interest cultivated through hands-on coding projects.