Liquid crystalline oligofluorenes and their derivatives

Homo-oligofluorenes (OFn), polyfluorenes (PF2/6) and oligofluorenes with one fluorenenone group in the center (OFnK) were synthesized. They were used as model compounds to understand of the structure-property relationships of polyfluorenes and the origin of the green emission in the photoluminescence (after photooxidation of the PFs) and the electroluminescence (EL) spectra. The electronic, electrochemical properties, thermal behavior, supramolecular self-assembly, and photophysical properties of OFn, PF2/6 and OFnK were investigated. Oligofluorenes with 2-ethylhexyl side chain (OF2-OF7) from the dimer up to the heptamer were prepared by a series of stepwise transition metal mediated Suzuki and Yamamoto coupling reactions. Polyfluorene was synthesized by Yamamoto coupling of 2,7-dibromo-9,9-bis(2-ethylhexyl)fluorene. Oligofluorenes with one fluorenone group in the center (OF3K, OF5K, OF7K) were prepared by Suzuki coupling between the monoboronic fluorenyl monomer, dimer, trimer and 2, 7-dibromofluorenone. The electrochemical and electronic properties of homo-oligofluorenes (OFn) were systematically studied by several combined techniques such as cyclic voltammetry, differential pulse voltammetry, UV-vis absorption spectroscopy, steady and time-resolved fluorescence spectroscopy. It was found that the oligofluorenes behave like classical conjugated oligomers, i.e., with the increase of the chain-length, the corresponding oxidation potential, the absorption and emission maximum, ionization potential, electron affinity, band gap and the photoluminescence lifetime displayed a very good linear relation with the reciprocal number of the fluorene units (1/n). The extrapolation of these linear relations to infinite chain length predicted the electrochemical and electronic properties of the corresponding polyfluorenes. The thermal behavior, single-crystal structure and supramolecular packing, alignment properties, and molecular dynamics of the homo-oligofluorenes (OFn) up to the polymer were studied using techniques such as TGA, DSC, WAXS, POM and DS. The OFn from tetramer to heptamer show a smectic liquid crystalline phase with clearly defined isotropization temperature. The oligomers do show a glass transition which exhibits n-1 dependence and allows extrapolation to a hypothetical glass transition of the polymer at around 64 °C. A smectic packing and helix-like conformation for the oligofluorenes from tetramer to heptamer was supported by WAXS experiments, simulation, and single-crystal structure of some oligofluorene derivatives. Oligofluorenes were aligned more easily than the corresponding polymer, and the alignability increased with the molecular length from tetramer to heptamer. The molecular dynamics in a series of oligofluorenes up to the polymer was studied using dielectric spectroscopy. The photophysical properties of OFn and PF2/6 were investigated by the steady-state spectra (UV-vis absorption and fluorescence spectra) and time-resolved fluorescence spectra both in solution and thin film. The time-resolved fluorescence spectra of the oligofluorenes were measured by streak camera and gate detection technique. The lifetime of the oligofluorenes decreased with the extension of the chain-length. No green emission was observed in CW, prompt and delayed fluorescence for oligofluorenes in m-THF and film at RT and 77K. Phosphorescence was observed for oligofluorenes in frozen dilute m-THF solution at 77K and its lifetime increased with length of oligofluorenes. A linear relation was obtained for triplet energy and singlet energy as a function of the reciprocal degree of polymerization, and the singlet-triplet energy gap (S1-T1) was found to decrease with the increase of degree of polymerization. Oligofluorenes with one fluorenone unit at the center were used as model compounds to understand the origin of the low-energy (“green”) emission band in the photoluminescence and electroluminescence spectra of polyfluorenes. Their electrochemical properties were investigated by CV, and the ionization potential (Ip) and electron affinity (Ea) were calculated from the onset of oxidation and reduction of OFnK. The photophysical properties of OFnK were studied in dilute solution and thin film by steady-state spectra and time-resolved fluorescence spectra. A strong green emission accompanied with a weak blue emission were obtained in solution and only green emission was observed on film. The strong green emission of OFnK suggested that rapid energy transfer takes place from higher energy sites (fluorene segments) to lower energy sites (fluorenone unit) prior to the radiative decay of the excited species. The fluorescence spectra of OFnK also showed solvatochromism. Monoexponential decay behaviour was observed by time-resolved fluorescence measurements. In addition, the site-selective excitation and concentration dependence of the fluorescence spectra were investigated. The ratio of green and blue emission band intensities increases with the increase of the concentration. The observed strong concentration dependence of the green emission band in solution suggests that increased interchain interactions among the fluorenone-containing oligofluorene chain enhanced the emission from the fluorenone defects at higher concentration. On the other hand, the mono-exponential decay behaviour and power dependence were not influenced significantly by the concentration. We have ruled out the possibility that the green emission band originates from aggregates or excimer formation. Energy transfer was further investigated using a model system of a polyfluorene doped by OFnK. Förster-type energy transfer took place from PF2/6 to OFnK, and the energy transfer efficiency increased with increasing of the concentration of OFnK. Efficient funneling of excitation energy from the high-energy fluorene segments to the low-energy fluorenone defects results from energy migration by hopping of excitations along a single polymer chain until they are trapped on the fluorenone defects on that chain or transferred onto neighbouring chains by Förster-type interchain energy transfer process. These results imply that the red-shifted emission in polyfluorenes can originate from (usually undesirable) keto groups at the bridging carbon atoms-especially if the samples have been subject to photo- or electro-oxidation or if fluorenone units are present due to an improper purification of the monomers prior to polymerization.

Echtzeitsuche nach Neutrinoausbrüchen von Supernovae mit dem AMANDA-II-Detektor

Der AMANDA-II Detektor ist primär für den richtungsaufgelösten Nachweis hochenergetischer Neutrinos konzipiert. Trotzdem können auch niederenergetische Neutrinoausbrüche, wie sie von Supernovae erwartet werden, mit hoher Signifikanz nachgewiesen werden, sofern sie innerhalb der Milchstraße stattfinden. Die experimentelle Signatur im Detektor ist ein kollektiver Anstieg der Rauschraten aller optischen Module. Zur Abschätzung der Stärke des erwarteten Signals wurden theoretische Modelle und Simulationen zu Supernovae und experimentelle Daten der Supernova SN1987A studiert. Außerdem wurden die Sensitivitäten der optischen Module neu bestimmt. Dazu mussten für den Fall des südpolaren Eises die Energieverluste geladener Teilchen untersucht und eine Simulation der Propagation von Photonen entwickelt werden. Schließlich konnte das im Kamiokande-II Detektor gemessene Signal auf die Verhältnisse des AMANDA-II Detektors skaliert werden. Im Rahmen dieser Arbeit wurde ein Algorithmus zur Echtzeit-Suche nach Signalen von Supernovae als Teilmodul der Datennahme implementiert. Dieser beinhaltet diverse Verbesserungen gegenüber der zuvor von der AMANDA-Kollaboration verwendeten Version. Aufgrund einer Optimierung auf Rechengeschwindigkeit können nun mehrere Echtzeit-Suchen mit verschiedenen Analyse-Zeitbasen im Rahmen der Datennahme simultan laufen. Die Disqualifikation optischer Module mit ungeeignetem Verhalten geschieht in Echtzeit. Allerdings muss das Verhalten der Module zu diesem Zweck anhand von gepufferten Daten beurteilt werden. Dadurch kann die Analyse der Daten der qualifizierten Module nicht ohne eine Verzögerung von etwa 5 Minuten geschehen. Im Falle einer erkannten Supernova werden die Daten für die Zeitdauer mehrerer Minuten zur späteren Auswertung in 10 Millisekunden-Intervallen archiviert. Da die Daten des Rauschverhaltens der optischen Module ansonsten in Intervallen von 500 ms zur Verfgung stehen, ist die Zeitbasis der Analyse in Einheiten von 500 ms frei wählbar. Im Rahmen dieser Arbeit wurden drei Analysen dieser Art am Südpol aktiviert: Eine mit der Zeitbasis der Datennahme von 500 ms, eine mit der Zeitbasis 4 s und eine mit der Zeitbasis 10 s. Dadurch wird die Sensitivität für Signale maximiert, die eine charakteristische exponentielle Zerfallszeit von 3 s aufweisen und gleichzeitig eine gute Sensitivität über einen weiten Bereich exponentieller Zerfallszeiten gewahrt. Anhand von Daten der Jahre 2000 bis 2003 wurden diese Analysen ausführlich untersucht. Während die Ergebnisse der Analyse mit t = 500 ms nicht vollständig nachvollziehbare Ergebnisse produzierte, konnten die Resultate der beiden Analysen mit den längeren Zeitbasen durch Simulationen reproduziert und entsprechend gut verstanden werden. Auf der Grundlage der gemessenen Daten wurden die erwarteten Signale von Supernovae simuliert. Aus einem Vergleich zwischen dieser Simulation den gemessenen Daten der Jahre 2000 bis 2003 und der Simulation des erwarteten statistischen Untergrunds kann mit einem Konfidenz-Niveau von mindestens 90 % gefolgert werden, dass in der Milchstraße nicht mehr als 3.2 Supernovae pro Jahr stattfinden. Zur Identifikation einer Supernova wird ein Ratenanstieg mit einer Signifikanz von mindestens 7.4 Standardabweichungen verlangt. Die Anzahl erwarteter Ereignisse aus dem statistischen Untergrund beträgt auf diesem Niveau weniger als ein Millionstel. Dennoch wurde ein solches Ereignis gemessen. Mit der gewählten Signifikanzschwelle werden 74 % aller möglichen Vorläufer-Sterne von Supernovae in der Galaxis überwacht. In Kombination mit dem letzten von der AMANDA-Kollaboration veröffentlicheten Ergebnis ergibt sich sogar eine obere Grenze von nur 2.6 Supernovae pro Jahr. Im Rahmen der Echtzeit-Analyse wird für die kollektive Ratenüberhöhung eine Signifikanz von mindestens 5.5 Standardabweichungen verlangt, bevor eine Meldung über die Detektion eines Supernova-Kandidaten verschickt wird. Damit liegt der überwachte Anteil Sterne der Galaxis bei 81 %, aber auch die Frequenz falscher Alarme steigt auf bei etwa 2 Ereignissen pro Woche. Die Alarm-Meldungen werden über ein Iridium-Modem in die nördliche Hemisphäre übertragen, und sollen schon bald zu SNEWS beitragen, dem weltweiten Netzwerk zur Früherkennung von Supernovae.

Fluorescence close to dielectric interfaces

The aim of the work was to study the correlation between the orientation and excited-state lifetimes of organic dyes close to dielectric interfaces. For this purpose, an experimental setup was designed and built, guiding the light through a prism in total internal reflection geometry. Fluorescence intensities and lifetimes for an ensemble of dye molecules were analyzed as a function of the excitation and detection polarizations. Working close to the total internal reflection angle, the differences between polarization combinations were enhanced. A classical electromagnetic model that assumes a chromophore as a couple of point-like electrical dipoles was developed. A numerical method to calculate the excitation and emission of dye molecules embedded in a multilayer system was implemented, by which full simulation of the time resolved fluorescence experiments was achieved. Free organic dyes and organic dyes covalently bound to polyelectrolyte chains were used. The polymer functionalization process avoided aggregation and provided control over the dyes position, within a few nanometers to the interface. Moreover, by varying the pH, the polymer chains could be deposited on different substrates with different conformations and the resulting fluorescence characteristics analyzed. Initially the fluorescence of organic dyes embedded in a polymer matrix was studied as a function of the distance between the fluorophores and the polymer-air interface. The non-radiative decay rate, vacuum decay rate and the relative angle between the excitation and emission dipoles of the chromophores could be determined. Different free organic dyes were deposited onto different dielectric spacers, as close as possible to the air-dielectric interface. Surprisingly, the fluorescence characteristics of dyes deposited onto polyelectrolyte layer were in good agreement with theoretical predictions of dyes in a polymer matrix, even when the layer was only 2 nm thick. When functionalized chains were deposited at low pH, on top of a polyelectrolyte spacer, the fluorescence had the characteristics of emitters embedded in a polymer matrix as well. Surface deposition at high pH showed an intermediate behaviour between emitters embedded in polymer and on top of the surface, in air. In general, for low pH values, the chains are deposited on a substrate in a train-like conformation. For high pH values, the chains are deposited in a loop-like conformation. As a consequence at low pH the functionalized polymer strongly interdigitates with the polyelectrolyte chains of the spacer, bringing most of the dyes inside the polymer. Thus, the fluorophores may experience the polymer as surrounding environment. On the other hand, for high pH values the dye-loaded chains adsorbed have a conformational arrangement of dense loops that extend away from the surface. Therefore many fluorophores experience the air as surrounding environment. Changing the spacer from polyelectrolyte to negatively charged silane produced contradictory results for lifetimes and intensities. The fluorescence intensities indicated the behaviour of emitters embedded in a polymer matrix, regardless of the pH value. On the other hand, for low pH values, the excited-state lifetimes showed that the emitters behaved as in air. For higher pH values, an intermediate behaviour between fluorophores located within and above of a dielectric film was observed. The poor agreement between theoretical and experimental data may be due to the simplified model utilized, by which the dipoles are assumed either in one side or in the other with respect to a geometrical air-dielectric interface. In the case when the dielectric film is constituted by the functionalized polymer chains themselves, reality is more complex and a different model may apply. Nevertheless, possible applications of the technique arise from a qualitative analysis.

Monte Carlo simulations of Potts glasses

A complete understanding of the glass transition isstill a challenging problem. Some researchers attributeit to the (hypothetical) occurrence of a static phasetransition, others emphasize the dynamical transitionof mode coupling-theory from an ergodic to a non ergodicstate. A class of disordered spin models has been foundwhich unifies both scenarios. One of these models isthe p-state infinite range Potts glass with p>4, whichexhibits in the thermodynamic limit both a dynamicalphase transition at a temperature T_D, and a static oneat T_0 < T_D. In this model every spins interacts withall the others, irrespective of distance. Interactionsare taken from a Gaussian distribution.In order to understand better its behavior forfinite number N of spins and the approach to thethermodynamic limit, we have performed extensive MonteCarlo simulations of the p=10 Potts glass up to N=2560.The time-dependent spin-autocorrelation function C(t)shows strong finite size effects and it does not showa plateau even for temperatures around the dynamicalcritical temperature T_D. We show that the N-andT-dependence of the relaxation time for T > T_D can beunderstood by means of a dynamical finite size scalingAnsatz.The behavior in the spin glass phase down to atemperature T=0.7 (about 60% of the transitiontemperature) is studied. Well equilibratedconfigurations are obtained with the paralleltempering method, which is also useful for properlyestablishing static properties, such as the orderparameter distribution function P(q). Evidence is givenfor the compatibility with a one step replica symmetrybreaking scenario. The study of the cumulants of theorder parameter does not permit a reliable estimation ofthe static transition temperature. The autocorrelationfunction at low T exhibits a two-step decay, and ascaling behavior typical of supercooled liquids, thetime-temperature superposition principle, is observed. Inthis region the dynamics is governed by Arrheniusrelaxations, with barriers growing like N^{1/2}.We analyzed the single spin dynamics down to temperaturesmuch lower than the dynamical transition temperature. We found strong dynamical heterogeneities, which explainthe non-exponential character of the spin autocorrelationfunction. The spins seem to relax according to dynamicalclusters. The model in three dimensions tends to acquireferromagnetic order for equal concentration of ferro-and antiferromagnetic bonds. The ordering has differentcharacteristics from the pure ferromagnet. The spinglass susceptibility behaves like chi_{SG} proportionalto 1/T in the region where a spin glass is predicted toexist in mean-field. Also the analysis of the cumulantsis consistent with the absence of spin glass orderingat finite temperature. The dynamics shows multi-scalerelaxations if a bimodal distribution of bonds isused. We propose to understand it with a model based onthe local spin configuration. This is consistent with theabsence of plateaus if Gaussian interactions are used.

Strukturfaktoren, Modenkopplungsgleichungen und Glasübergang molekularer Kristalle

In dieser Arbeit wird der Orientierungsglasübergang ungeordneter, molekularer Kristalle untersucht. Die theoretische Behandlung ist durch die Anisotropie der Einteilchen-Verteilungsfunktion und der Paarfunktionen erschwert. Nimmt man ein starres Gitter, wird der reziproke Raum im Gegenzug auf die 1. Brillouin-Zone eingeschränkt. Der Orientierungsglasübergang wird im Rahmen der Modenkopplungsgleichungen studiert, die dazu hergeleitet werden. Als Modell dienen harte Rotationsellipsoide auf einem starren sc Gitter. Zur Berechnung der statischen tensoriellen Strukturfaktoren wird die Ornstein-Zernike(OZ)-Gleichung molekularer Kristalle abgeleitet und selbstkonsistent zusammen mit der von molekularen Flüssigkeiten übernommenen Percus-Yevick(PY)-Näherung gelöst. Parallel dazu werden die Strukturfaktoren durch MC-Simulationen ermittelt. Die OZ-Gleichung molekularer Kristalle ähnelt der von Flüssigkeiten, direkte und totale Korrelationsfunktion kommen jedoch wegen des starren Gitters nur ohne Konstantanteile in den Winkelvariablen vor, im Gegensatz zur PY-Näherung. Die Anisotropie bringt außerdem einen nichttrivialen Zusatzfaktor. OZ/PY-Strukturfaktoren und MC-Ergebnisse stimmen gut überein. Bei den Matrixelementen der Dichte-Dichte-Korrelationsfunktion gibt es drei Hauptverläufe: oszillatorisch, monoton und unregelmäßig abfallend. Oszillationen gehören zu alternierenden Dichtefluktuationen, führen zu Maxima der Strukturfaktoren am Zonenrand und kommen bei oblaten und genügend breiten prolaten, schwächer auch bei dünnen, nicht zu langen prolaten Ellipsoiden vor. Der exponentielle monotone Abfall kommt bei allen Ellipsoiden vor und führt zu Maxima der Strukturfaktoren in der Zonenmitte, was die Tendenz zu nematischer Ordnung zeigt. Die OZ/PY-Theorie ist durch divergierende Maxima der Strukturfaktoren begrenzt. Bei den Modenkopplungsgleichungen molekularer Kristalle zeigt sich eine große Ähnlichkeit mit denen molekularer Flüssigkeiten, jedoch spielen auf einem starrem Gitter nur die Matrixelemente mit l,l' > 0 eine Rolle und es finden Umklapps von reziproken Vektoren statt. Die Anisotropie bringt auch hier nichtkonstante Zusatzfaktoren ins Spiel. Bis auf flache oblate Ellipsoide wird die Modenkopplungs-Glaslinie von der Divergenz der Strukturfaktoren bestimmt. Für sehr lange Ellipsoide müssen die Strukturfaktoren zur Divergenz hin extrapoliert werden. Daher treibt nicht der Orientierungskäfigeffekt den Glasübergang, sondern Fluktuationen an einer Phasengrenze. Nahe der Kugelform ist keine zuverlässige Glasline festlegbar. Die eingefrorenen kritischen Dichte-Dichte-Korrelatoren haben nur in wenigen Fällen die Oszillationen der statischen Korrelatoren. Der monotone Abfall bleibt dagegen für lange Zeiten meist erhalten. Folglich haben die kritischen Modenkopplungs-Nichtergodizitätsparameter abgeschwächte Maxima in der Zonenmitte, während die Maxima am Zonenrand meist verschwunden sind. Die normierten Nichtergodizitätsparameter zeigen eine Fülle von Verläufen, besonders tiefer im Glas.

Factorization and non-local 1/m b corrections in the decay B X s gamma

In this thesis, a systematic analysis of the bar B to X_sgamma photon spectrum in the endpoint region is presented. The endpoint region refers to a kinematic configuration of the final state, in which the photon has a large energy m_b-2E_gamma = O(Lambda_QCD), while the jet has a large energy but small invariant mass. Using methods of soft-collinear effective theory and heavy-quark effective theory, it is shown that the spectrum can be factorized into hard, jet, and soft functions, each encoding the dynamics at a certain scale. The relevant scales in the endpoint region are the heavy-quark mass m_b, the hadronic energy scale Lambda_QCD and an intermediate scale sqrt{Lambda_QCD m_b} associated with the invariant mass of the jet. It is found that the factorization formula contains two different types of contributions, distinguishable by the space-time structure of the underlying diagrams. On the one hand, there are the direct photon contributions which correspond to diagrams with the photon emitted directly from the weak vertex. The resolved photon contributions on the other hand arise at O(1/m_b) whenever the photon couples to light partons. In this work, these contributions will be explicitly defined in terms of convolutions of jet functions with subleading shape functions. While the direct photon contributions can be expressed in terms of a local operator product expansion, when the photon spectrum is integrated over a range larger than the endpoint region, the resolved photon contributions always remain non-local. Thus, they are responsible for a non-perturbative uncertainty on the partonic predictions. In this thesis, the effect of these uncertainties is estimated in two different phenomenological contexts. First, the hadronic uncertainties in the bar B to X_sgamma branching fraction, defined with a cut E_gamma > 1.6 GeV are discussed. It is found, that the resolved photon contributions give rise to an irreducible theory uncertainty of approximately 5 %. As a second application of the formalism, the influence of the long-distance effects on the direct CP asymmetry will be considered. It will be shown that these effects are dominant in the Standard Model and that a range of -0.6 < A_CP^SM < 2.8 % is possible for the asymmetry, if resolved photon contributions are taken into account.

Condensational growth of atmospheric aerosol particles in an expanding water saturated air flow

Aerosol particles and water vapour are two important constituents of the atmosphere. Their interaction, i.e. thecondensation of water vapour on particles, brings about the formation of cloud, fog, and raindrops, causing the water cycle on the earth, and being responsible for climate changes. Understanding the roles of water vapour and aerosol particles in this interaction has become an essential part of understanding the atmosphere. In this work, the heterogeneous nucleation on pre-existing aerosol particles by the condensation of water vapour in theflow of a capillary nozzle was investigated. Theoretical and numerical modelling as well as experiments on thiscondensation process were included. Based on reasonable results from the theoretical and numerical modelling, an idea of designing a new nozzle condensation nucleus counter (Nozzle-CNC), that is to utilise the capillary nozzle to create an expanding water saturated air flow, was then put forward and various experiments were carried out with this Nozzle-CNC under different experimental conditions. Firstly, the air stream in the long capillary nozzle with inner diameter of 1.0~mm was modelled as a steady, compressible and heat-conducting turbulence flow by CFX-FLOW3D computational program. An adiabatic and isentropic cooling in the nozzle was found. A supersaturation in the nozzle can be created if the inlet flow is water saturated, and its value depends principally on flow velocity or flow rate through the nozzle. Secondly, a particle condensational growth model in air stream was developed. An extended Mason's diffusion growthequation with size correction for particles beyond the continuum regime and with the correction for a certain particle Reynolds number in an accelerating state was given. The modelling results show the rapid condensational growth of aerosol particles, especially for fine size particles, in the nozzle stream, which, on the one hand, may induce evident `over-sizing' and `over-numbering' effects in aerosol measurements as nozzle designs are widely employed for producing accelerating and focused aerosol beams in aerosol instruments like optical particle counter (OPC) and aerodynamical particle sizer (APS). It can, on the other hand, be applied in constructing the Nozzle-CNC. Thirdly, based on the optimisation of theoretical and numerical results, the new Nozzle-CNC was built. Under various experimental conditions such as flow rate, ambient temperature, and the fraction of aerosol in the total flow, experiments with this instrument were carried out. An interesting exponential relation between the saturation in the nozzle and the number concentration of atmospheric nuclei, including hygroscopic nuclei (HN), cloud condensation nuclei (CCN), and traditionally measured atmospheric condensation nuclei (CN), was found. This relation differs from the relation for the number concentration of CCN obtained by other researchers. The minimum detectable size of this Nozzle-CNC is 0.04?m. Although further improvements are still needed, this Nozzle-CNC, in comparison with other CNCs, has severaladvantages such as no condensation delay as particles larger than the critical size grow simultaneously, low diffusion losses of particles, little water condensation at the inner wall of the instrument, and adjustable saturation --- therefore the wide counting region, as well as no calibration compared to non-water condensation substances.

Simulation studies of correlation functions and relaxation in polymeric systems

We investigate the statics and dynamics of a glassy,non-entangled, short bead-spring polymer melt with moleculardynamics simulations. Temperature ranges from slightlyabove the mode-coupling critical temperature to the liquidregime where features of a glassy liquid are absent. Ouraim is to work out the polymer specific effects on therelaxation and particle correlation. We find the intra-chain static structure unaffected bytemperature, it depends only on the distance of monomersalong the backbone. In contrast, the distinct inter-chainstructure shows pronounced site-dependence effects at thelength-scales of the chain and the nearest neighbordistance. There, we also find the strongest temperaturedependence which drives the glass transition. Both the siteaveraged coupling of the monomer and center of mass (CM) andthe CM-CM coupling are weak and presumably not responsiblefor a peak in the coherent relaxation time at the chain'slength scale. Chains rather emerge as soft, easilyinterpenetrating objects. Three particle correlations arewell reproduced by the convolution approximation with theexception of model dependent deviations. In the spatially heterogeneous dynamics of our system weidentify highly mobile monomers which tend to follow eachother in one-dimensional paths forming ``strings''. Thesestrings have an exponential length distribution and aregenerally short compared to the chain length. Thus, arelaxation mechanism in which neighboring mobile monomersmove along the backbone of the chain seems unlikely.However, the correlation of bonded neighbors is enhanced. When liquids are confined between two surfaces in relativesliding motion kinetic friction is observed. We study ageneric model setup by molecular dynamics simulations for awide range of sliding speeds, temperatures, loads, andlubricant coverings for simple and molecular fluids. Instabilities in the particle trajectories are identified asthe origin of kinetic friction. They lead to high particlevelocities of fluid atoms which are gradually dissipatedresulting in a friction force. In commensurate systemsfluid atoms follow continuous trajectories for sub-monolayercoverings and consequently, friction vanishes at low slidingspeeds. For incommensurate systems the velocity probabilitydistribution exhibits approximately exponential tails. Weconnect this velocity distribution to the kinetic frictionforce which reaches a constant value at low sliding speeds. This approach agrees well with the friction obtaineddirectly from simulations and explains Amontons' law on themicroscopic level. Molecular bonds in commensurate systemslead to incommensurate behavior, but do not change thequalitative behavior of incommensurate systems. However,crossed chains form stable load bearing asperities whichstrongly increase friction.

Decay of imprinted surface waves in polymers

CONCLUSIONS The focus of this work was the investigation ofanomalies in Tg and dynamics at polymer surfaces. Thethermally induced decay of hot-embossed polymer gratings isstudied using laser-diffraction and atomic force microscopy(AFM). Monodisperse PMMA and PS are selected in the Mwranges of 4.2 to 65.0 kg/mol and 3.47 to 65.0 kg/mol,respectively. Two different modes of measurement were used:the one mode uses temperature ramps to obtain an estimate ofthe near-surface glass temperature, Tdec,0; the other modeinvestigates the dynamics at a constant temperature aboveTg. The temperature-ramp experiments reveal Tdec,0 valuesvery close to the Tg,bulk values, as determined bydifferential scanning calorimetry (DSC). The PMMA of65.0 kg/mol shows a decreased value of Tg, while the PS samples of 3.47 and 10.3 kg/mol (MwDecay experiments were performed at a variety ofconstant temperatures. Master plots are produced by shifting the decay curves, on the logarithmic time scale,with respect to a reference curve at Tref. From thisprocedure shift factors were extracted. An Arrhenius analysis of the shift factors reveals adecreasing non-Arrhenius (fragile) behavior with molecular weight for PMMA. PS is fragile for all Mw asexpected for linear polymers. Non-Arrhenius behavior allowsone to fit the shift factors to the William-Landel-Ferry(WLF) equation. The WLF parameters for the varying molecular weights ofPMMA and PS were extracted and compared to the values frombulk rheology measurements. Assuming cg1=16+/-2 at Tg, assuggested by Angell, the glass temperature was determinedfrom the dynamic decay experiments. Within the experimentalerrors, the values for Tg,surf(c1=16) and T_g,bulk(c1=16)tend to be smaller than Tdec,0 and Tg,bulk fromtemperature-ramp and DSC measurements, but confirm thecourse of the values with increasing Mw. The comparison of the fragilities (temperaturedependence of the polymer properties at Tg) near the surfaceand in the bulk shows a higher fragility for PS near thesurface, a lower one for PMMA with molecular weights of 4.2and 65.0 kg/mol. The different surface behavior of PS istraced back to a lower degree of cooperation and a largerfree volume fraction.

Reinforcement Lernen mit Regularisierungsnetzwerken

Die Arbeit behandelt das Problem der Skalierbarkeit von Reinforcement Lernen auf hochdimensionale und komplexe Aufgabenstellungen. Unter Reinforcement Lernen versteht man dabei eine auf approximativem Dynamischen Programmieren basierende Klasse von Lernverfahren, die speziell Anwendung in der Künstlichen Intelligenz findet und zur autonomen Steuerung simulierter Agenten oder realer Hardwareroboter in dynamischen und unwägbaren Umwelten genutzt werden kann. Dazu wird mittels Regression aus Stichproben eine Funktion bestimmt, die die Lösung einer "Optimalitätsgleichung" (Bellman) ist und aus der sich näherungsweise optimale Entscheidungen ableiten lassen. Eine große Hürde stellt dabei die Dimensionalität des Zustandsraums dar, die häufig hoch und daher traditionellen gitterbasierten Approximationsverfahren wenig zugänglich ist. Das Ziel dieser Arbeit ist es, Reinforcement Lernen durch nichtparametrisierte Funktionsapproximation (genauer, Regularisierungsnetze) auf -- im Prinzip beliebig -- hochdimensionale Probleme anwendbar zu machen. Regularisierungsnetze sind eine Verallgemeinerung von gewöhnlichen Basisfunktionsnetzen, die die gesuchte Lösung durch die Daten parametrisieren, wodurch die explizite Wahl von Knoten/Basisfunktionen entfällt und so bei hochdimensionalen Eingaben der "Fluch der Dimension" umgangen werden kann. Gleichzeitig sind Regularisierungsnetze aber auch lineare Approximatoren, die technisch einfach handhabbar sind und für die die bestehenden Konvergenzaussagen von Reinforcement Lernen Gültigkeit behalten (anders als etwa bei Feed-Forward Neuronalen Netzen). Allen diesen theoretischen Vorteilen gegenüber steht allerdings ein sehr praktisches Problem: der Rechenaufwand bei der Verwendung von Regularisierungsnetzen skaliert von Natur aus wie O(n**3), wobei n die Anzahl der Daten ist. Das ist besonders deswegen problematisch, weil bei Reinforcement Lernen der Lernprozeß online erfolgt -- die Stichproben werden von einem Agenten/Roboter erzeugt, während er mit der Umwelt interagiert. Anpassungen an der Lösung müssen daher sofort und mit wenig Rechenaufwand vorgenommen werden. Der Beitrag dieser Arbeit gliedert sich daher in zwei Teile: Im ersten Teil der Arbeit formulieren wir für Regularisierungsnetze einen effizienten Lernalgorithmus zum Lösen allgemeiner Regressionsaufgaben, der speziell auf die Anforderungen von Online-Lernen zugeschnitten ist. Unser Ansatz basiert auf der Vorgehensweise von Recursive Least-Squares, kann aber mit konstantem Zeitaufwand nicht nur neue Daten sondern auch neue Basisfunktionen in das bestehende Modell einfügen. Ermöglicht wird das durch die "Subset of Regressors" Approximation, wodurch der Kern durch eine stark reduzierte Auswahl von Trainingsdaten approximiert wird, und einer gierigen Auswahlwahlprozedur, die diese Basiselemente direkt aus dem Datenstrom zur Laufzeit selektiert. Im zweiten Teil übertragen wir diesen Algorithmus auf approximative Politik-Evaluation mittels Least-Squares basiertem Temporal-Difference Lernen, und integrieren diesen Baustein in ein Gesamtsystem zum autonomen Lernen von optimalem Verhalten. Insgesamt entwickeln wir ein in hohem Maße dateneffizientes Verfahren, das insbesondere für Lernprobleme aus der Robotik mit kontinuierlichen und hochdimensionalen Zustandsräumen sowie stochastischen Zustandsübergängen geeignet ist. Dabei sind wir nicht auf ein Modell der Umwelt angewiesen, arbeiten weitestgehend unabhängig von der Dimension des Zustandsraums, erzielen Konvergenz bereits mit relativ wenigen Agent-Umwelt Interaktionen, und können dank des effizienten Online-Algorithmus auch im Kontext zeitkritischer Echtzeitanwendungen operieren. Wir demonstrieren die Leistungsfähigkeit unseres Ansatzes anhand von zwei realistischen und komplexen Anwendungsbeispielen: dem Problem RoboCup-Keepaway, sowie der Steuerung eines (simulierten) Oktopus-Tentakels.