Refine
Year of publication
- 2020 (246) (remove)
Document Type
- Article (177)
- Postprint (35)
- Doctoral Thesis (31)
- Monograph/Edited Volume (1)
- Other (1)
- Review (1)
Is part of the Bibliography
- yes (246)
Keywords
- diffusion (16)
- anomalous diffusion (7)
- model (6)
- perovskite solar cells (6)
- random diffusivity (5)
- dynamics (4)
- impact (4)
- methods: numerical (4)
- photoluminescence (4)
- subdwarfs (4)
Institute
- Institut für Physik und Astronomie (246) (remove)
Lattice dynamics
(2020)
In this thesis I summarize my contribution to the research field of ultrafast structural dynamics in condensend matter. It consists of 17 publications that cover the complex interplay between electron, magnon, and phonon subsystems in solid materials and the resulting lattice dynamics after ultrafast photoexcitation. The investigation of such dynamics is necessary for the physical understanding of the processes in materials that might become important in the future as functional materials for technological applications, for example in data storage applications, information processing, sensors, or energy harvesting.
In this work I present ultrafast x-ray diffraction (UXRD) experiments based on the optical pump – x-ray probe technique revealing the time-resolved lattice strain. To study these dynamics the samples (mainly thin film heterostructures) are excited by femtosecond near-infrared or visible light pulses. The induced strain dynamics caused by stresses of the excited subsystems are measured in a pump-probe scheme with x-ray diffraction (XRD) as a probe. The UXRD setups used during my thesis are a laser-driven table-top x-ray source and large-scale synchrotron facilities with dedicated time-resolved diffraction setups. The UXRD experiments provide quantitative access to heat reservoirs in nanometric layers and monitor the transient responses of these layers with coupled electron, magnon, and phonon subsystems. In contrast to optical probes, UXRD allows accessing the material-specific information, which is unavailable for optical light due to the detection of multiple indistinguishable layers in the range of the penetration depth.
In addition, UXRD facilitates a layer-specific probe for layers buried opaque heterostructures to study the energy flow. I extended this UXRD technique to obtain the driving stress profile by measuring the strain dynamics in the unexcited buried layer after excitation of the adjacent absorbing layers with femtosecond laser pulses. This enables the study of negative thermal expansion (NTE) in magnetic materials, which occurs due to the loss of the magnetic order. Part of this work is the investigation of stress profiles which are the source of coherent acoustic phonon wave packets (hypersound waves). The spatiotemporal shape of these stress profiles depends on the energy distribution profile and the ability of the involved subsystems to produce stress. The evaluation of the UXRD data of rare-earth metals yields a stress profile that closely matches the optical penetration profile: In the paramagnetic (PM) phase the photoexcitation results in a quasi-instantaneous expansive stress of the metallic layer whereas in the antiferromagnetic (AFM) phase a quasi-instantaneous contractive stress and a second contractive stress contribution rising on a 10 ps time scale adds to the PM contribution. These two time scales are characteristic for the magnetic contribution and are in agreement with related studies of the magnetization dynamics of rare-earth materials.
Several publications in this thesis demonstrate the scientific progress in the field of active strain control to drive a second excitation or engineer an ultrafast switch. These applications of ultrafast dynamics are necessary to enable control of functional material properties via strain on ultrafast time scales.
For this thesis I implemented upgrades of the existing laser-driven table-top UXRD setup in order to achieve an enhancement of x-ray flux to resolve single digit nanometer thick layers. Furthermore, I developed and built a new in-situ time-resolved magneto-optic Kerr effect (MOKE) and optical reflectivity setup at the laser-driven table-top UXRD setup to measure the dynamics of lattice, electrons and magnons under the same excitation conditions.
Galaxies are gravitationally bound systems of stars, gas, dust and - probably - dark matter. They are the building blocks of the Universe. The morphology of galaxies is diverse: some galaxies have structures such as spirals, bulges, bars, rings, lenses or inner disks, among others. The main processes that characterise galaxy evolution can be separated into fast violent events that dominated evolution at earlier times and slower processes, which constitute a phase called secular evolution, that became dominant at later times. Internal processes of secular evolution include the gradual rearrangement of matter and angular momentum, the build-up and dissolution of substructures or the feeding of supermassive black holes and their feedback. Galaxy bulges – bright central components in disc galaxies –, on one hand, are relics of galaxy formation and evolution. For instance, the presence of a classical bulge suggests a relatively violent history. In contrast, the presence of a disc-like bulge instead indicates the occurrence of secular evolution processes in the main disc. Galaxy bars – elongated central stellar structures –, on the other hand, are the engines of secular evolution. Studying internal properties of both bars and bulges is key to comprehending some of the processes through which secular evolution takes place. The main objectives of this thesis are (1) to improve the classification of bulges by combining photometric and spectroscopic approaches for a large sample of galaxies, (2) to quantify star formation in bars and verify dependencies on galaxy properties and (3) to analyse stellar populations in bars to aid in understanding the formation and evolution of bars. Integral field spectroscopy is fundamental to the work presented in this thesis, which consists of three different projects as part of three different galaxy surveys: the CALIFA survey, the CARS survey and the TIMER project.
The first part of this thesis constitutes an investigation of the nature of bulges in disc galaxies. We analyse 45 galaxies from the integral-field spectroscopic survey CALIFA by performing 2D image decompositions, growth curve measurements and spectral template fitting to derive stellar kinematics from CALIFA data cubes. From the obtained results, we present a recipe to classify bulges that combines four different parameters from photometry and kinematics: The bulge Sersic index nb, the concentration index C20;50, the Kormendy relation and the inner slope of the radial velocity dispersion profile ∇σ. The results of the different approaches are in good agreement and allow a safe classification for approximately 95% of the galaxies. We also find that our new ‘inner’ concentration index performs considerably better than the traditionally used C50;90 and, in combination with the Kormendy relation, provides a very robust indication of the physical nature of the bulge. In the second part, we study star formation within bars using VLT/MUSE observations for 16 nearby (0.01 < z < 0.06) barred active-galactic-nuclei (AGN)-host galaxies from the CARS survey. We derive spatially-resolved star formation rates (SFR) from Hα emission line fluxes and perform a detailed multi-component photometric decomposition on images derived from the data cubes. We find a clear separation into eight star-forming (SF) and eight non-SF bars, which we interpret as indication of a fast quenching process. We further report a correlation between the SFR in the bar and the shape of the bar surface brightness profile: only the flattest bars (nbar < 0.4) are SF. Both parameters are found to be uncorrelated with Hubble type. Additionally, owing to the high spatial resolution of the MUSE data cubes, for the first time, we are able to dissect the SFR within the bar and analyse trends parallel and perpendicular to the bar major axis. Star formation is 1.75 times stronger on the leading edge of a rotating bar than on the trailing edge and is radially decreasing. Moreover, from testing an AGN feeding scenario, we report that the SFR of the bar is uncorrelated with AGN luminosity. Lastly, we present a detailed analysis of star formation histories and chemical enrichment of stellar populations (SP) in galaxy bars. We use MUSE observations of nine very nearby barred galaxies from the TIMER project to derive spatially resolved maps of stellar ages and metallicities, [α/Fe] abundances, star formation histories, as well as Hα as tracer of star formation. Using these maps, we explore in detail variations of SP perpendicular to the bar major axes. We find observational evidence for a separation of SP, supposedly caused by an evolving bar. Specifically, intermediate-age stars (∼ 2-6 Gyr) get trapped on more elongated orbits forming a thinner bar, while old stars (> 8 Gyr) form a rounder and thicker bar. This evidence is further strengthened by very similar results obtained from barred galaxies in the cosmological zoom-in simulations from the Auriga project. In addition, we find imprints of typical star formation patterns in barred galaxies on the youngest populations (< 2 Gyr), which continuously become more dominant from the major axis towards the sides of the bar. The effect is slightly stronger on the leading side. Furthermore, we find that bars are on average more metal-rich and less α-enhanced than the inner parts of the discs that surrounds them. We interpret this result as an indication of a more prolonged or continuous formation of stars that shape the bar as compared to shorter formation episodes in the disc within the bar region.
This thesis is focused on a better understanding of the formation mechanism of bulk birefringence gratings (BBG) and a surface relief gratings (SRG) in photo-sensitive polymer films. A new set-up is developed enabling the in situ investigation how the polymer film is being structured during irradiation with modulated light. The new aspect of the equipment is that it combines several techniques such as a diffraction efficiency (DE) set-up, an atomic force microscope (AFM) and an optical set-up for controlled illumination of the sample. This enables the simultaneous acquiring and differentiation of both gratings (BBG and SRG), while changing the irradiation conditions in desired way.
The dissertation is based on five publications. The first publication (I) is focused on the description of the set-up and interpretation of the measured data. A fine structure within the 1st-order diffraction spot is observed, which is a result of the inhomogeneity of the inscribed gratings.
In the second publication (II) the interplay of BBG and SRG in the DE is discussed. It has been found, that, dependent on the polarization of a weak probe beam, the diffraction components of the SRG and BBG either interfere constructively or destructively in the DE, altering the appearance of the intensity distribution within the diffracted spot.
The third (III) and fourth (IV) publications describe the light-induced reconfiguration of surface structures. Special attention is payed to conditions influencing the erasure of topography and bulk gratings. This can be achieved via thermal treatment or illumination of the polymer film. Using the translation of the interference pattern (IP) in a controlled way, the optical erase speed is significantly increased. Additionally, a dynamic reconfigurable surface is generated, which could move surface attached objects by the continuous translation of the interference pattern during irradiation of the polymer films.
The fifth publication (V) deals with the understanding of polymer deformation under irradiation with SP-IP, which is the only IP generating a half-period topography grating (compared to the period of the IP) on the photo-sensitive polymer film. This mechanism is used, e.g. to generate a SRG below the diffraction limit of light. It also represents an easy way of changing the period of the surface grating just by a small change in polarization angle of the interfering beams without adjusting the optical pass of the two beams. Additionally, complex surface gratings formed in mixed polarization- and intensity interference patterns are shown.
I J. Jelken, C. Henkel and S. Santer, Applied Physics B, 125 (2019), 218
II J. Jelken, C. Henkel and S. Santer, Appl. Phys. Lett., 116 (2020), 051601
III J. Jelken and S. Santer, RSC Advances, 9 (2019), 20295
IV J. Jelken, M. Brinkjans, C. Henkel and S. Santer, SPIE Proceedings, 11367 (2020), 1136710
V J. Jelken, C. Henkel and S. Santer, Formation of Half-Period Surface Relief Gratings in Azobenzene Containing Polymer Films (submitted to Applied Physics B)
Perovskite solar cells have become one of the most studied systems in the quest for new, cheap and efficient solar cell materials. Within a decade device efficiencies have risen to >25% in single-junction and >29% in tandem devices on top of silicon. This rapid improvement was in many ways fortunate, as e. g. the energy levels of commonly used halide perovskites are compatible with already existing materials from other photovoltaic technologies such as dye-sensitized or organic solar cells. Despite this rapid success, fundamental working principles must be understood to allow concerted further improvements. This thesis focuses on a comprehensive understanding of recombination processes in functioning devices.
First the impact the energy level alignment between the perovskite and the electron transport layer based on fullerenes is investigated. This controversial topic is comprehensively addressed and recombination is mitigated through reducing the energy difference between the perovskite conduction band minimum and the LUMO of the fullerene. Additionally, an insulating blocking layer is introduced, which is even more effective in reducing this recombination, without compromising carrier collection and thus efficiency. With the rapid efficiency development (certified efficiencies have broken through the 20% ceiling) and thousands of researchers working on perovskite-based optoelectronic devices, reliable protocols on how to reach these efficiencies are lacking. Having established robust methods for >20% devices, while keeping track of possible pitfalls, a detailed description of the fabrication of perovskite solar cells at the highest efficiency level (>20%) is provided. The fabrication of low-temperature p-i-n structured devices is described, commenting on important factors such as practical experience, processing atmosphere & temperature, material purity and solution age. Analogous to reliable fabrication methods, a method to identify recombination losses is needed to further improve efficiencies. Thus, absolute photoluminescence is identified as a direct way to quantify the Quasi-Fermi level splitting of the perovskite absorber (1.21eV) and interfacial recombination losses the transport layers impose, reducing the latter to ~1.1eV. Implementing very thin interlayers at both the p- and n-interface (PFN-P2 and LiF, respectively), these losses are suppressed, enabling a VOC of up to 1.17eV. Optimizing the device dimensions and the bandgap, 20% devices with 1cm2 active area are demonstrated. Another important consideration is the solar cells’ stability if subjected to field-relevant stressors during operation. In particular these are heat, light, bias or a combination thereof. Perovskite layers – especially those incorporating organic cations – have been shown to degrade if subjected to these stressors. Keeping in mind that several interlayers have been successfully used to mitigate recombination losses, a family of perfluorinated self-assembled monolayers (X-PFCn, where X denotes I/Br and n = 7-12) are introduced as interlayers at the n-interface. Indeed, they reduce interfacial recombination losses enabling device efficiencies up to 21.3%. Even more importantly they improve the stability of the devices. The solar cells with IPFC10 are stable over 3000h stored in the ambient and withstand a harsh 250h of MPP at 85◦C without appreciable efficiency losses. To advance further and improve device efficiencies, a sound understanding of the photophysics of a device is imperative. Many experimental observations in recent years have however drawn an inconclusive picture, often suffering from technical of physical impediments, disguising e. g. capacitive discharge as recombination dynamics. To circumvent these obstacles, fully operational, highly efficient perovskites solar cells are investigated by a combination of multiple optical and optoelectronic probes, allowing to draw a conclusive picture of the recombination dynamics in operation. Supported by drift-diffusion simulations, the device recombination dynamics can be fully described by a combination of first-, second- and third-order recombination and JV curves as well as luminescence efficiencies over multiple illumination intensities are well described within the model. On this basis steady state carrier densities, effective recombination constants, densities-of-states and effective masses are calculated, putting the devices at the brink of the radiative regime. Moreover, a comprehensive review of recombination in state-of-the-art devices is given, highlighting the importance of interfaces in nonradiative recombination. Different strategies to assess these are discussed, before emphasizing successful strategies to reduce interfacial recombination and pointing towards the necessary steps to further improve device efficiency and stability. Overall, the main findings represent an advancement in understanding loss mechanisms in highly efficient solar cells. Different reliable optoelectronic techniques are used and interfacial losses are found to be of grave importance for both efficiency and stability. Addressing the interfaces, several interlayers are introduced, which mitigate recombination losses and degradation.
In this dissertation we introduce a concept of light driven active and passive manipulation of colloids trapped at solid/liquid interface. The motion is induced due to generation of light driven diffusioosmotic flow (LDDO) upon irradiation with light of appropriate wavelength. The origin of the flow is due to osmotic pressure gradient resulting from a concentration gradient at the solid/liquid interface of the photosensitive surfactant present in colloidal dispersion. The photosensitive surfactant consists of a cationic head group and a hydrophobic tail in which azobenzene group is integrated in. The azobenzene is known to undergo reversible photo-isomerization from a stable trans to a meta stable cis state under irradiation with UV light. Exposure to light of larger wavelength results in back photo-isomerization from cis to trans state. The two isomers have different molecular properties, for instance, trans isomer has a rod like structure and low polarity (0 dipole moment), whereas cis one is bent and has a dipole moment of ~3 Debye. Being integrated in the hydrophobic tail of the surfactant molecule, the azobenzene state determines the hydrophobicity of the whole molecule: in the trans state the surfactant is more hydrophobic than in the cis-state. In this way many properties of the surfactant such as the CMC, solubility and the interaction potential with a solid surface can be altered by light. When the solution containing such a surfactant is irradiated with focused light, a concentration gradient of different isomers is formed near the boundary of the irradiated area near the solid surface resulting in osmotic pressure gradient. The generated diffusioosmotic (DO) flow carries the particles passively along.
The local-LDDO flow can be generated around and by each particle when mesoporous silica colloids are dispersed in the surfactant solution. This is because porous particles act as a sink/source which absorbs azobenzene molecule in trans state and expels it when it is in the cis state. The DO flows generated at each particle interact resulting in aggregation or separation depending upon the initial state of surfactant molecules. The kinetic of aggregation and separation can be controlled and manipulated by altering the parameters such as the wavelength and intensity of the applied light, as well as surfactant and particle concentration. Using two wavelengths simultaneously allows for dynamic gathering and separation creating fascinating patterns such as 2D disk of well separated particles or establishing collective complex behaviour of particle ensemble as described in this thesis.
The mechanism of l-LDDO is also used to generate self-propelled motion. This is possible when half of the porous particle is covered by metal layer, basically blocking the pores on one side. The LDDO flow generated on uncapped side pushes the particle forward resulting in a super diffusive motion. The system of porous particle and azobenzene containing surfactant molecule can be utilized for various application such as drug delivery, cargo transportation, self-assembling, micro motors/ machines or micro patterning.
Giant unilamellar vesicles are an important tool in todays experimental efforts to understand the structure and behaviour of biological cells. Their simple structure allows the isolation of the physical elastic properties of the lipid membrane. A central physical
property is the bending energy of the membrane, since the many different shapes of giant vesicles can be obtained by finding the minimum of the bending energy. In the spontaneous curvature model the bending energy is a function of the bending rigidity as well as the mean curvature and an additional parameter called the spontaneous curvature, which describes an internal preference of the lipid-bilayer to bend towards one side or the other. The spontaneous and mean curvature are local properties of the membrane.
Additional constraints arise from the conservation of the membrane surface area and the enclosed volume, which are global properties.
In this thesis the spontaneous curvature model is used to explain the experimental observation of a periodic shape oscillation of a giant unilamellar vesicle that was filled with a protein complex that periodically binds to and unbinds from the membrane.
By assuming that the binding of the proteins to the membrane induces a change in the spontaneous curvature the experimentally observed shapes could successfully be explained. This involves the numerical solution of the differential equations as obtained from the minimization of the bending energy respecting the area and volume constraints, the so called shape equations. Vice versa this approach can be used to estimate the spontaneous curvature from experimentally measurable quantities.
The second topic of this thesis is the analysis of concentration gradients in rigid conic membrane compartments. Gradients of an ideal gas due to gravity and gradients generated by the directed stochastic movement of molecular motors along a microtubulus were considered. It was possible to calculate the free energy and the bending energy analytically for the ideal gas. In the case of the non-equilibrium system with molecular motors, the characteristic length of the density profile, the jam-length, and its dependency on the opening angle of the conic compartment have been calculated in the mean-field limit.
The mean field results agree qualitatively with stochastic particle simulations.
In der vorliegenden Arbeit wird untersucht, in wie weit physikalische Experimente ein flow-Erleben bei Lernenden hervorrufen. Flow-Erleben wird als Motivationsursache gesehen und soll den Weg zu Freude und Glück darstellen. Insbesondere wegen dem oft zitierten Fachkräftemangel in naturwissenschaftlichen und technischen Berufen ist eine Motivationssteigerung in naturwissenschaftlichen Unterrichtsfächern wichtig. Denn trotz Leistungssteigerungen in internationalen Vergleichstests möchten in Deutschland deutlich weniger Schüler*innen einen solchen Beruf ergreifen als in anderen Industriestaaten. Daher gilt es, möglichst früh Schüler*innen für naturwissenschaftlich-technische Fächer zu begeistern und insbesondere im regelrecht verhassten Physikunterricht flow-Erleben zu erzeugen.
Im Rahmen dieser Arbeit wird das flow-Erleben von Studierenden in klassischen Laborexperimenten und FELS (Forschend-Entdeckendes Lernen mit dem Smartphone) als Lernumgebung untersucht. FELS ist eine an die Lebenswelt der Schüler*innen angepasste Lernumgebung, in der sie mit Smartphones ihre eigene Lebenswelt experimentell untersuchen.
Es zeigt sich, dass sowohl klassische Laborexperimente als auch in der Lebenswelt durchgeführte, smartphonebasierte Experimente flow-Erleben erzeugen. Allerdings verursachen die smartphonebasierten Experimente kaum Stressgefühle.
Die in dieser Arbeit herausgefundenen Ergebnisse liefern einen ersten Ansatz, der durch Folgestudien erweitert werden sollte.
We consider the first-passage problem for N identical independent particles that are initially released uniformly in a finite domain Ω and then diffuse toward a reactive area Γ, which can be part of the outer boundary of Ω or a reaction centre in the interior of Ω. For both cases of perfect and partial reactions, we obtain the explicit formulas for the first two moments of the fastest first-passage time (fFPT), i.e., the time when the first out of the N particles reacts with Γ. Moreover, we investigate the full probability density of the fFPT. We discuss a significant role of the initial condition in the scaling of the average fFPT with the particle number N, namely, a much stronger dependence (1/N and 1/N² for partially and perfectly reactive targets, respectively), in contrast to the well known inverse-logarithmic behaviour found when all particles are released from the same fixed point. We combine analytic solutions with scaling arguments and stochastic simulations to rationalise our results, which open new perspectives for studying the relevance of multiple searchers in various situations of molecular reactions, in particular, in living cells.
We consider the first-passage problem for N identical independent particles that are initially released uniformly in a finite domain Ω and then diffuse toward a reactive area Γ, which can be part of the outer boundary of Ω or a reaction centre in the interior of Ω. For both cases of perfect and partial reactions, we obtain the explicit formulas for the first two moments of the fastest first-passage time (fFPT), i.e., the time when the first out of the N particles reacts with Γ. Moreover, we investigate the full probability density of the fFPT. We discuss a significant role of the initial condition in the scaling of the average fFPT with the particle number N, namely, a much stronger dependence (1/N and 1/N² for partially and perfectly reactive targets, respectively), in contrast to the well known inverse-logarithmic behaviour found when all particles are released from the same fixed point. We combine analytic solutions with scaling arguments and stochastic simulations to rationalise our results, which open new perspectives for studying the relevance of multiple searchers in various situations of molecular reactions, in particular, in living cells.
The plasmasphere is a dynamic region of cold, dense plasma surrounding the Earth. Its shape and size are highly susceptible to variations in solar and geomagnetic conditions. Having an accurate model of plasma density in the plasmasphere is important for GNSS navigation and for predicting hazardous effects of radiation in space on spacecraft. The distribution of cold plasma and its dynamic dependence on solar wind and geomagnetic conditions remain, however, poorly quantified. Existing empirical models of plasma density tend to be oversimplified as they are based on statistical averages over static parameters. Understanding the global dynamics of the plasmasphere using observations from space remains a challenge, as existing density measurements are sparse and limited to locations where satellites can provide in-situ observations. In this dissertation, we demonstrate how such sparse electron density measurements can be used to reconstruct the global electron density distribution in the plasmasphere and capture its dynamic dependence on solar wind and geomagnetic conditions.
First, we develop an automated algorithm to determine the electron density from in-situ measurements of the electric field on the Van Allen Probes spacecraft. In particular, we design a neural network to infer the upper hybrid resonance frequency from the dynamic spectrograms obtained with the Electric and Magnetic Field Instrument Suite and Integrated Science (EMFISIS) instrumentation suite, which is then used to calculate the electron number density. The developed Neural-network-based Upper hybrid Resonance Determination (NURD) algorithm is applied to more than four years of EMFISIS measurements to produce the publicly available electron density data set.
We utilize the obtained electron density data set to develop a new global model of plasma density by employing a neural network-based modeling approach. In addition to the location, the model takes the time history of geomagnetic indices and location as inputs, and produces electron density in the equatorial plane as an output. It is extensively validated using in-situ density measurements from the Van Allen Probes mission, and also by comparing the predicted global evolution of the plasmasphere with the global IMAGE EUV images of He+ distribution. The model successfully reproduces erosion of the plasmasphere on the night side as well as plume formation and evolution, and agrees well with data.
The performance of neural networks strongly depends on the availability of training data, which is limited during intervals of high geomagnetic activity. In order to provide reliable density predictions during such intervals, we can employ physics-based modeling. We develop a new approach for optimally combining the neural network- and physics-based models of the plasmasphere by means of data assimilation. The developed approach utilizes advantages of both neural network- and physics-based modeling and produces reliable global plasma density reconstructions for quiet, disturbed, and extreme geomagnetic conditions.
Finally, we extend the developed machine learning-based tools and apply them to another important problem in the field of space weather, the prediction of the geomagnetic index Kp. The Kp index is one of the most widely used indicators for space weather alerts and serves as input to various models, such as for the thermosphere, the radiation belts and the plasmasphere. It is therefore crucial to predict the Kp index accurately. Previous work in this area has mostly employed artificial neural networks to nowcast and make short-term predictions of Kp, basing their inferences on the recent history of Kp and solar wind measurements at L1. We analyze how the performance of neural networks compares to other machine learning algorithms for nowcasting and forecasting Kp for up to 12 hours ahead. Additionally, we investigate several machine learning and information theory methods for selecting the optimal inputs to a predictive model of Kp. The developed tools for feature selection can also be applied to other problems in space physics in order to reduce the input dimensionality and identify the most important drivers.
Research outlined in this dissertation clearly demonstrates that machine learning tools can be used to develop empirical models from sparse data and also can be used to understand the underlying physical processes. Combining machine learning, physics-based modeling and data assimilation allows us to develop novel methods benefiting from these different approaches.