Influence of side walls and undulated topography on viscous gravity-driven film flow
Full text
Influence of side walls and undulated topography on viscous gravity–driven film flow Von der Fakult¨ at f¨ ur Angewandte Naturwissenschaften der Universit¨ at Bayreuth zur Erlangung der W¨ urde eines Doktor-Ingenieurs (Dr.-Ing.) genehmigte Dissertation vorgelegt von Dipl.-Phys. Thilo Pollak aus Gummersbach Erstgutachter: Prof. Dr. rer. nat. N. Aksel Zweitgutachter: Prof. Dr. V. Bontozoglou Tag der m¨ undlichen Pr¨ ufung: 07. August 2012 Lehrstuhl f¨ ur Technische Mechanik und Str¨ omungsmechanik Universit¨ at Bayreuth 2012
1
List of Journal publications - Pollak T. and K¨ ohler W.: Critical assessment of diffusion coefficients in semidilute to concentrated solutions of polystyrene in toluene, The Journal of Chemical Physics, 130 (2009), 124905 - Wierschem A., Pollak T., Heining C. and Aksel N.: Suppression of eddies in films over topography, Physics of Fluids, 22 (2010), 113603 - Haas A., Pollak T. and Aksel N.: Side wall effects in thin gravity–driven film flow - steady and draining flow, Physics of Fluids, 23 (2011), 062107 - Pollak T., Haas A. and Aksel N.: Side wall effects on the instability of thin gravitydriven films - From long–wave to short–wave instability, Physics of Fluids, 23 (2011), 094110 - Heining C., Pollak T. and Aksel N.: Pattern formation and mixing in three–dimensional film flow, Physics of Fluids, 24 (2012), 042102 - Heining C., Sellier M. and Pollak T.,: Flow domain identification from free surface velocity in thin inertial films, Journal of Fluid Mechanics, submitted - Pollak T., Aksel A.: Experimental evidence of multiple instability branches of gravity– driven films over topography,in preparation 2
3
Abstract While a gravity–driven viscous film flow down an inclined flat plane of infinite extent can be described by an easy analytical solution, flow problems in nature, like glacier movements or the liquid film on the human eye are much more complex. Also to optimize a large number of technical applications, like coating applications or heat exchanger devices, one has to investigate and understand how different influencing factors, like topological features on the substrate or a finite width of the system, influence the flow and its stability isolated from each other. By introducing a wavy structure to the underlying topography, which could be for example a model for roughness, new effects emerge in the flow, which cannot be observed in flows over a flat incline. Eddies can separate from the main flow at the lee side of the undulation for kinematic reasons, or induced by inertial effects. In biological systems these eddies are dead water areas, which are cut off from nutrient supply, in heat exchanger applications their appearance has a strong impact on the convective heat transport within the liquid. Furthermore, the amplitude of free surface of the liquid can be amplified immensely when the liquid is in resonance with the undulation of the underlying topography. In this work we study experimentally as well as numerically the complex interaction of this resonance phenomenon with the appearing of eddy structures in the valleys of the undulation and show, that one can suppress flow separation selectively even at rather high Reynolds numbers when one exploits the resonance phenomenon specifically. Another part of this work deals with the question how the presence of side walls and the contact angle of the liquid there influences the free surface shape of the liquid, the velocity field and the globally transported volume flux. While an additional no–slip condition at the wall causes additional friction and leads thus to a lower volume flux, capillary elevation at the side walls can generate a velocity overshoot in the vicinity of the walls, depending on the film thickness and the wetting properties of the liquid, which counteracts the additional friction coming from the walls. An extensive theoretical parameter study, which is supplemented with experimental data, provides criteria for the first onset of the velocity overshoot and gives answer to the question when the counteracting influences on the global volume flux just cancel each other. An experimental study of the free surface shape of a draining flow shows that this configuration cannot be described by a series of quasi–steady states, even when a dynamic contact angle is taken into consideration, although the flow changes only very slowly in time. Additional time dependent numerical simulations of the draining flow reveal an indentation of the free surface in the vicinity of the side wall, which could promote film rupture in technical thin film applications. 4
Furthermore, side wall effects play an important role for the physical stability of the flow. Waves develop at the free surface of a gravity–driven flow and grow while they are traveling downstream, when a critical volume flux is exceeded. It is shown by experimental variation of the contact angle, the film thickness and the side wall distance, that the presence of side walls generates different effects which have competing influences on the stability of the flow. Capillary elevation leads to a pretensioning of the free surface, which tends to stabilize the free surface, just as the additional no–slip condition at the wall does. The emerging of a velocity overshoot in the capillary elevation on the other hand leads to a destabilization of the flow. In the system studied here the stabilizing influence of the side walls dominates over the destabilizing influence which is of comparatively short range, which means that this flow configuration is more stable than the corresponding flow of infinite extent. However, the results suggest that the destabilizing influences should dominate over the stabilizing influences in similar flow configurations when the film would become even thinner. While free surface film flows typically form long waves at first, we find for this flow configuration, that the type of instability changes from a long–wave type in the middle of the channel to a short–wave type instability, as it is well known for boundary layer flows, as the side wall distance is reduced. 5
Zusammenfassung W¨ ahrend sich eine schwerkraftsgetriebene viskose Str¨ omung, die eine unendlich ausgedehnte und glatte Ebene hinabfließt, durch eine einfache analytische L¨ osung beschreiben l¨ asst, sind die Str¨ omungsprobleme in der Natur, wie zum Beispiel eine Gletscherbewegung oder ein Fl¨ ussigkeitsfilm auf dem menschlichen Auge, weitaus komplizierter. Auch um zahlreiche technische Anwendungen, wie beispielsweise Beschichtungs- oder W¨ armetauschprozesse, optimieren zu k¨ onnen, m¨ ussen Einflussfaktoren, wie das Vorhandensein einer Struktur auf der Oberfl¨ ache des Bodens oder eine endliche Breite des Systems und deren Einfl¨ usse auf das Str¨ omungsfeld und die physikalische Stabilit¨ at der Str¨ omung isoliert untersucht und verstanden werden. Durch das Vorhandensein eines gewellten Untergrundes, der zum Beispiel ein Modell f¨ ur Rauheit sein k¨ onnte, entstehen neue Effekte in der Str¨ omung, die bei einem glatten Untergrund nicht beobachtet werden k¨ onnen. Sowohl aus rein kinematischen Gr¨ unden, aber auch durch tr¨ agheitsinduzierte Effekte kann die Str¨ omung auf der Windschattenseite der Bodenstruktur vom Boden abl¨ osen, so dass in den Bodenmulden Rezirkulationsgebiete entstehen. In biologischen Systemen stellen diese Regionen Totwassergebiete dar, die nicht mit N¨ ahrstoffen versorgt werden, in W¨ armetauscheranwendungen hat ihr Auftreten einen starken Einfluss auf den konvektiven W¨ armetransport. Neben dem Entstehen einer Str¨ omungsabl¨ osung kann durch Auftreten von Resonanz zwischen dem gewellten Boden und der Fl¨ ussigkeit die Amplitude der freien Fl¨ ussigkeitsoberfl¨ ache immens verst¨ arkt werden. In dieser Arbeit wird das komplizierte Zusammenspiel aus Resonanz und dem Entstehen von Rezirkulationsgebieten in den Bodenmulden sowohl numerisch als auch experimentell untersucht und es wird gezeigt, dass man durch geschickte Ausnutzung der Resonanz das Auftreten der Wirbelstrukturen auch bei relativ hohen Reynoldszahlen gezielt unterbinden kann. Ein weiterer Teil dieser Dissertation befasst sich mit der Frage, wie sich das Vorhandensein von Seitenw¨ anden und der Kontaktwinkel der Fl¨ ussigkeit dort auf die Form der freien Oberfl¨ ache, das Geschwindigkeitsfeld und den globalen Volumenstrom auswirkt. W¨ ahrend eine zus¨ atzliche Haftbedingung an der Wand zu zus¨ atzlicher Reibung und damit zu einem geringeren Volumenstrom f¨ uhrt, kann in Abh¨ angigkeit von Kontaktwinkel und Filmdicke durch kapillare Anhebung ein Geschwindigkeits¨ uberschuss in der N¨ ahe der Seitenwand entstehen, der dem Einfluss der Haftbedingung entgegenwirkt. Eine umfangreiche theoretische Parameterstudie, die durch experimentelle Ergebnisse erg¨ anzt wird, liefert Kriterien f¨ ur das erste Auftreten eines Geschwindigkeits¨ uberschusses und beantwortet die Frage, wann sich die entgegenwirkenden Einfl¨ usse auf den globalen Volumenstrom gerade gegenseitig aufheben. Die experimentelle Untersuchung der Form der freien Oberfl¨ ache einer Drainage- 6
str¨ omung zeigt, dass sich diese Str¨ omung auch durch Einf¨ uhrung eines dynamischen Kontaktwinkels nicht durch eine Folge quasistatischer Zust¨ ande beschreiben l¨ asst, obwohl sie sich zeitlich nur sehr langsam ver¨ andert. Zus¨ atzliche zeitabh¨ angige numerische Simulationen der Drainagestr¨ omung offenbaren eine Vertiefung der freien Oberfl¨ ache in der N¨ ahe der Seitenwand, die in technischen D¨ unnfilmanwendungen einen Abriss des Fl¨ ussigkeitsfilms hervorrufen k¨ onnte. Dar¨ uber hinaus spielen Seitenwandeffekte auch eine entscheidende Rolle f¨ ur die physikalische Stabilit¨ at der Str¨ omung. Auf der freien Oberfl¨ ache einer schwerkraftsgetriebenen Filmstr¨ omung bilden sich Wellen aus, die anwachsen, w¨ ahrend sie die Ebene hinabfließen, sobald ein kritischer Volumenstrom ¨ uberschritten wird. Es wird durch experimentelle Variation des Kontaktwinkels, der Filmdicke und des Seitenwandabstandes gezeigt, dass verschiedene Effekte, die durch das Vorhandensein von Seitenw¨ anden auftreten, miteinander konkurrierende Einfl¨ usse auf die Stabilit¨ at der Str¨ omung haben. So hat die kapillare Anhebung eine Vorkr¨ ummung der freien Oberfl¨ ache zur Folge, welche zusammen mit der zus¨ atzlichen Haftbedingung an der Wand zu einer Stabilisierung der Str¨ omung f¨ uhrt. Das Auftreten eines Geschwindigkeits¨ uberschusses in der kapillaren Anhebung f¨ uhrt hingegen zu einer Destabilisierung der Str¨ omung. Bei der hier untersuchten Str¨ omung ¨ uberwiegen die langreichweitigen stabilisierenden Einfl¨ usse den vergleichsweise kurzreichweitigen destabilisierenden Einfluss der Seitenwand, so dass dieses System insgesamt gegen¨ uber einer quer zur Hauptstr¨ omungsrichtung unendlich ausgedehnten Str¨ omung durch die Seitenw¨ ande stabilisiert wird. Die Ergebnisse legen jedoch nahe, dass f¨ ur ¨ ahnliche Str¨ omungskonfigurationen, die eine noch geringere Filmdicke aufweisen, der Nettoeinfluss der Seitenwand auf die Str¨ omung auch destabilisierend sein k¨ onnte. W¨ ahrend sich bei Filmstr¨ omungen typischerweise zuerst lange Wellen auf der freien Oberfl¨ ache ausbilden, finden wir f¨ ur diese Str¨ omung, dass sich durch eine Verringerung des Seitenwandabstandes ein ¨ Ubergang von einer Langwelleninstabilit¨ at zu einer Kurzwelleninstabilit¨ at, wie man sie typischerweise von Grenzschichtstr¨ omungen kennt, vollzieht. 7
Contents 1 Introduction 10 2 Experimental systems and setups 16 2.1 Liquids ..................................... 16 2.2 Flowfacilities.................................. 17 2.3 Tracerparticles................................. 18 2.4 Experimentalsetups .............................. 20 2.4.1 Flowrate ................................ 20 2.4.2 Detection of the free surface shape . . . . . . . . . . . . . . . . . . 21 2.4.3 Streamline detection . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.4.4 Velocity field measurements . . . . . . . . . . . . . . . . . . . . . . 24 2.4.5 Stability measurements . . . . . . . . . . . . . . . . . . . . . . . . 25 3 Two–dimensional film flow 32 3.1 Suppressionofeddies.............................. 32 3.1.1 Problem formulation . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.1.2 Experimental and numerical findings . . . . . . . . . . . . . . . . . 34 3.1.3 Physical interpretation and discussion . . . . . . . . . . . . . . . . 42 3.1.4 Conclusions............................... 44 4 Three–dimensional film flow 46 4.1 Basicflow.................................... 46 4.1.1 Governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.1.2 Flow type classification . . . . . . . . . . . . . . . . . . . . . . . . 50 4.1.3 Flowratestudy............................. 51 4.1.4 Velocityfield .............................. 54 4.1.5 Freesurfaceshape ........................... 57 4.1.6 Conclusions............................... 60 4.2 Stability near the side walls . . . . . . . . . . . . . . . . . . . . . . . . . . 62 4.2.1 Results ................................. 62 4.2.2 Conclusions............................... 65 5 Conclusions and outlook 68 8
Chapter 1. Introduction 15
Chapter 2 Experimental systems and setups 2.1 Liquids Three different silicone oils from Basildon and Elbesil with dynamic viscosities ηranging from approximately 10 mPas to 1000 mPas, which all showed Newtonian behavior within the considered shear rate and temperature range have been investigated. The main fluid properties at the mean measurement temperature Tare summarized in Table 2.1. The density, the dynamic viscosity, the kinematic viscosity and the surface tension are denoted by ρ,η,ν=η/ρ and σrespectively. Density measurements have been carried out with a Mohr Westphal balance from Gottl. Kern &Sohn GmbH with an absolute accuracy of ±0.3 kg/m3. The temperature of the liquid in the Mohr balance was controlled by a Lauda thermostat type ecoline RE204. Measurements of the surface tension σwere done with a Lauda ring–tensiometer type TE1CA-M whose fluid temperature was controlled by a Lauda thermostat type RC 6 CP. The resolution of the ring–tensiometer was 0.1 mN/m. The dynamic viscosity ηof the liquids has been measured with different Ubbelohde viscosimeter capillaries type 501 from Schott. The capillaries were plunged into a water bath whose temperature was controlled by a Schott thermostat within an accuracy of 0.05 ◦C. The precisions of the different viscosimeter capillaries were specified to be between 0.65% and 0.8%. All fluid property measurements have been carried out in a temperature interval from 20 −30 ◦C in 1 ◦C steps. The uncertainty of the liquid properties during an experimental run is essentially determined by the uncertainty of the liquid’s temperature and was thus calculated from the temperature dependence of the liquid properties. The temperature of the liquid flowing in the channel was measured downstream of the region of interest by Ahlborn Mess- und Regelungstechnik GmbH PT-100 and NTC Manufacturer Name T/ [◦C] ρ/ [kg/m3]η/ [mPas] ν/ [mm2/s] σ/ [mN/m] Basildon BC10cs 25 924.3 10.72 11.6 18.9 Basildon BC50 24 950.6 50.0 52.6 19.6 Elbesil B1000 24 969.0 1,076 1,110 20.4 Table 2.1: Liquid properties of the used silicone oils. 16
Chapter 2. Experimental systems and setups temperature sensors with an accuracy of 0.1◦C. The static contact angle θat the three–phase contact lines liquid/air/channel side wall has been measured with a contact angle goniometer from Dataphysics type OCA 20. All measured contact angles were found to be independent of the temperature within the measurement error of approximately 2◦and the temperature range investigated. 2.2 Flow facilities The experiments have been carried out in two different 170 ±1 mm broad channels with flat bottoms made of aluminum. The side walls of channel 1 were made of Plexiglas®and channel 2 was featured with side wall clamps which allowed to mount side walls made up of different materials to vary the contact angle θat the triple point air/liquid/side wall. In this work we have limited our contact angle study to the two extreme cases which are technically possible. Silicone oil, which was the only fluid used throughout all experimental runs in order to keep all material parameters (see section 2.1), in particular the surface tension σ, constant, shows nearly perfect wetting characteristics on the vast majority of substrates. Because of its excellent planarity we chose Plexiglass®as a side wall material to cover this case. The static contact angle θof silicone oil with plane Plexiglass®was measured with a sessile drop method using a static contact angle goniometer type OCA 20 from dataphysics to be 8◦±2◦. The second set of side walls was made up of glass which has been coated with pro.Glass®Clear 105 from nanogate to enlarge θto 52◦±2◦which was the largest contact angle we were able to achieve with the utilized fluid. The overall length of channel 1 was about two meters and of channel 2 was about half a meter. The inclination angle αof both channels could have been varied continuously between 0◦and 90◦and has been determined by a digital protractor with an accuracy of 0.1◦. The spanwise evenness of the channel has been checked by placing a water level with a display accuracy of 0.1 mm/m perpendicular to the side walls of the channel. Perpendicularity was assured with a 90 ◦–aluminum angle placed to the side wall. α g x z pump vibration isolating table liquid reservoir inflow tank ˙ V ˙ V ˙ V channel Figure 2.1: Sketch of the flow circuit including the channel which is mounted on a vibration isolating table and a pump which transports the liquid from a large liquid reservoir to a smaller inflow tank on top of the channel. Depending on the desired flow rate ˙qone of two different eccentric pumps from Johstadt provided a constant adjustable volume flux ˙ V= ˙qB, with Bbeing the channel width, 17
2.3. Tracer particles from a large liquid reservoir into a smaller inflow tank on top of the channel. Channel 1 was equipped with a pump type AFJ 15.1B which provided a volume flux in the range of 1 l/min to 10 l/min. Channel 2 was equipped with a smaller pump type AFJ 06B which provided a volume flux in the range of 0.025 l/min to 1 l/min. Smaller volume fluxes have been realized by an adjustable bypass in the tube system between the pump and the channel. From the smaller inflow tank on top of the channel the liquid flows gravity–driven, down the channel, back into the liquid reservoir to close the flow circuit as illustrated in in Figure 2.1. A sinusoidally undulated aluminum inlay, as illustrated in Figure 2.2, consisting of 50 periods with a wavelength λ= 10 mm and an amplitude a= 1 mm covering the whole width of the channel was inserted directly below the inflow of the channel 1. The gravitational acceleration is denoted by g, which can be written as g= (gsin α, −gcos α) in the (x, z)–coordinate system given in Figure 2.2. α g x z a λ Figure 2.2: Geometry of the two–dimensional undulated inlay. During all experimental runs the temperature of the liquid was controlled by a TC300 thermostat from Haake via a heat exchanger coil sitting in the large liquid reservoir. 2.3 Tracer particles We have used two different types of tracer particles. The mean diameter and the density of Red Fluorescent Polymer Microspheres from Duke Scientifics, which will be called fluorescent tracer particles in the following, have been specified by the manufacturer to 7µm and 1050 kg/m3, respectively. Additionally we have determined the volume weighted particle size distribution with a Mastersizer 2000 device from Malvern, which is plotted in Figure 2.3. The median particle size x50 and the grade of dispersity ξd, which is defined in[71] ξd=x84 −x16 2x50 ,(2.1) of the fluorescent tracer particles have been found to be x50 = 7.122 µm and ξd= 0.3058. The quantities x16,x50 and x84 denote the particle sizes, which are greater than or equal to 16%, 50% and 84% of all particles, respectively. The sedimentation speed used of small spheres falling in a viscous liquid can be calculated to be[72] used =2gr2 s 9η(ρs−ρ),(2.2) where rsand ρsare the radius and the density of the sphere. 18
Chapter 2. Experimental systems and setups 0.1 1 10 100 1000 Particle size / [µm] 0 5 10 15 20 Volume % Red Fluorescent Polymer Microspheres ZrO2/MgO powder Figure 2.3: Particle size distributions. The resulting sedimentation speeds of the fluorescent tracer particles used in the three different silicone oils BC10cs, BC50 and B1000 (see Table 2.1) were 3.2×10−4mm/s, 5.5×10−5mm/s and 2.1×10−6mm/s, respectively. All these sedimentation velocities are at each case orders of magnitudes smaller than the typical flow velocities measured. Therefore, the sedimentation distance during one experimental run did not exceed the particles diameter. Figure 2.4 shows the emission spectrum of the fluorescent tracer particles dissolved in silicone oil which has been measured with a Cary Eclipse fluorescence spectrophotometer from Agilent Technologies at an excitation wavelength of 532 nm. 500 550 600 650 700 750 800 λ / [nm] 0 100 200 300 400 Intensity / [a.u.] Figure 2.4: Emission spectrum of Red Fluorescent Polymer Microspheres from Duke Scientifics in silicone oil. Wavelength of the excitation light was 532 nm. The second type of tracer particles was made up of a powder of ZrO2/MgO from Goodfellow. The mean particle diameter is specified by the manufacturer to be 0.8µm. The density is specified to be 5700 kg/m3. The measured particle size distribution is plotted in Figure 2.3. The median particle size and the grade of dispersity have been measured to be x50 = 0.76 µm and ξd= 0.4682. These particles, which will be called scattering tracer particles in the following, were dissolved in the B1000 silicone oil from Elbesil, which is described in section 2.1, only. According to equation (2.2) the sedimentation speed used 19
2.4. Experimental setups of the ZrO2/MgO-particles in this oil was about 1.4×10−6mm/s, which was orders of magnitudes smaller, than all typical velocities measured. Therefore, the sedimentation distance during one experimental run did not exceed the particles diameter. 2.4 Experimental setups 2.4.1 Flow rate Determination of the flow rate ˙q, or the Reynolds number Re respectively, was done either by determining the film thickness dof the liquid flowing over a sufficiently long flat part in the middle of the channel or by a flow meter which measured the overall volume flux ˙ Vthrough the channel. Under perfectly stable flow conditions, at low Reynolds numbers, the film thickness has been measured by a micrometer screw with a needle tip. The micrometer screw was mounted to the channel in a way, that it pointed perpendicular to the free surface of the liquid or the bottom of the channel, respectively. By screwing the needle slowly towards the fluid, the position of the free surface can be detected, when the tip of the needle contacts the liquid and a capillary elevation forms instantaneously (See Figure 2.5). The position of the substrate has been determined in a similar fashion by screwing the needle further down until a small mechanical resistance was sensible. The accuracy of the film thickness determination is estimated to be better than 10 µm. (a) (b) Figure 2.5: Illustration of the tip of a needle which is less than 6.5µm above the surface of a flowing liquid film (a) and just touching it (b) what causes a capillary elevation to form instantaneously. The width of the needle illustrated is 400 µm. To determine the flow rate ˙qat intermediate Reynolds numbers or under weakly unsteady flow conditions, the overall average volume flux ˙ Vthrough the channel was measured by an analog flow meter which was installed between the outflow of the channel and the large liquid reservoir. For each volume flux measurement the averaging time was at least 600 s to reduce the statistical error of the volume flux measurement to less than 0.1 %. When side wall effects are neglected and the flow is steady, the velocity profile of a 20
Chapter 2. Experimental systems and setups liquid flowing down a flat incline is found to be parabolic[73] u(z) = ρg sin α 2η(2hn−z)z, (2.3) where zis the cartesian coordinate perpendicular to the bottom and hnis the film thickness of the well known Nusselt solution[9]. Integrating the velocity profile from the bottom (z= 0) to the free surface of the liquid (z=hn) yields the flow rate of the Nusselt film flow ˙q=Zhn 0 u(z)dz=ρg sin αh3 n 3η.(2.4) Integrating the flow rate ˙qover the channel width Byields the volume flux ˙ V. ˙ V= ˙qB =ρg sin αh3 nB 3η.(2.5) When the influence of side walls on the flow is neglected, the relation between the Reynolds number Re, the film thickness of a film flowing down a flat channel and the measured volume flux is in the following given by Re = 2ushn 3ν=¯uhn ν=˙q ν=˙ V νB ,(2.6) where usis the free surface velocity and ¯u= 2us/3 is the mean flow velocity. It has to be emphasized, that equations (2.5) and (2.6) are valid, only if the velocity field uis assumed to be independent of the spanwise y-coordinate, which is only the case when the impact of the presence of side walls on the flow filed is neglected. When the influence of side walls on the flow is investigated in chapter 4, this assumption has to be dropped. A detailed discussion of the relation between the film thickness d, the measured volume flux ˙ Vand the Reynolds number Re, when the influence of the presence of side walls is considered will follow in sections 4.1.2 and 4.1.3. 2.4.2 Detection of the free surface shape In both flow facilities the shape of the free surface has been measured by illuminating fluorescent tracer particles, which are described in section 2.3, in the bulk of the liquid with a laser sheet. The fluorescent light has been detected with a charged–coupled–device (CCD) camera. The laser sheet for channel 1 has been produced by a continuous–wave (cw) argon– Ion (Ar+) Laser from Spectra Physics emitting at a wavelength of 514.5 nm with an approximate output power of 100 mW (See Figure 2.6). The light sheet was aligned parallel to the side walls right in the middle of the channel to study the influence of the undulated bottom on the free surface shape of the liquid flowing above it. The fluorescent light from the fluorescent tracer particles in the liquid was detected with a JAI CV-M300 8-bit monochrome CCD camera with a resolution of 768x494 pixels, which was inclined by about 10◦with respect to the spanwise direction of the channel (See camera (2) in Figures 2.6 and 2.7). The much brighter scattered light from the underlying topography was blocked by an optical long pass filter with a 50% cut–off wavelength of 550 nm, which was 21
2.4. Experimental setups mounted on a 60 mm Nikkor micro lens. The cameras field of view covered approximately 2.5 periods of the bottom undulation. This camera setup resulted in a spatial resolution of about 30 µm/pixel. The image was calibrated spatially with a ceramic calibration scale with a point pattern of 4 pt/mm2. Liquid with Laser sheet 514.5 nm detection camera Surface Inclined channel puter Com− Argon Ion Laser Field camera fluorescent Filter detection tracers (2) (1) Figure 2.6: Experimental setup for the surface contour detection and the streamline detection in the troughs of the undulated inlay in channel 1. The eddy size is determined by detecting the path lines of the fluorescent tracer particles with the horizontal camera (1). The inclined camera (2) images the light sheet from the airside. The surface contour corresponds to the upper borderline of the bright sheet as seen by this camera. Reprinted with permission from [44]. ©2010, American Institute of Physics. Surface detection camera (2) H Hj Field detection camera (1) Figure 2.7: Photo of the experimental setup for the surface contour detection and the streamline detection in the troughs of the undulated inlay in channel 1. The experimental setup for the free surface detection in channel 2 is sketched in Figure 2.8. The laser sheet has been aligned perpendicular to the side walls of the channel and was expanded to illuminate the liquid in a region from the side wall to approximately 50 mm apart from it, to detect the shape of the capillary elevation of the liquid in the proximity of the side walls. The sheet was produced by a frequency doubled, pulsed 22
Chapter 2. Experimental systems and setups neodymium-doped yttrium aluminum garnet (Nd:Yag) laser from New wave research type Solo II 15Hz emitting at a wavelength of 532 nm. The pulse length and energy is specified to be 6 ns and 100 mJ. An optical device from Cosmicar/Pentax was attached directly to the laser head to create the light sheet which had a width of approximately 100 µm. The scattered primary light from the bottom was blocked in front of the camera’s lens using the same long pass filter as described above. The fluorescent light from the fluorescent tracer particles was detected by a monochrome CCD camera HiSense from Dantec with a resolution of 1280x1024 and a capturing rate of 8 Hz. The camera was inclined by about 15 ◦with respect to the channel inclination as illustrated in Figure 2.8. The spatial resolution of this camera setup was about 8 µm/pixel. Calibration of the image has been carried out a priori by placing a millimeter scale at the laser sheet position. Camera and laser have been synchronized by a triggering unit from Dantec. Nd:YAG Trigger PC camera CCD laser unit Light sheet Filter g Inlet Outlet Figure 2.8: Measurement setup for the free surface shape detection. Reprinted with permission from [66]. ©2011, American Institute of Physics. Capturing images from the liquid as described above resulted in grainy single images, because the fluorescent tracer particles sit at discrete points when one image is taken. Therefore, it was necessary to average or superimpose several images to get a uniformly bright sheet in the image whose upper border corresponds to the free surface contour of the liquid at the position of the laser sheet. Depending on the case the post–processing workflow differs slightly. Therefore, a more detailed description about the method of how the averaging or superimposing of the images was performed will be given where the corresponding results are presented. 2.4.3 Streamline detection A detection of the streamline pattern of the liquid in the throughs of the undulated inlay in channel 1 has been done in a similar way as described by Wierschem et al.[33, 35, 44]. The fluorescent tracer particles and light sheet are identical with the ones described in subsection 2.3 and 2.4.2. The scattered light from the illuminated particles was detected with a frame rate of up to 500 Hz depending on the mean flow velocity with a monochrome high–speed camera CamRecord 600 from Optronics. The camera was aligned perpendic- 23
2.4. Experimental setups ular to the channel side walls as illustrated in Figures 2.6 and 2.7 (camera (1)). The streamline patterns have been reconstructed by superposing 2048 single images taken in one run with the camera’s full resolution of 1280x1024 pixel by taking at each pixel position the brightest pixel of all images (see Figure 2.9(b)). The contour and position of the underlying topography could have been reconstructed in a similar fashion, but by taking not the brightest but the darkest pixel of all pictures at each pixel position (see Figure 2.9(a)). The bright lines below contour line of the underlying topography in Figure 2.9(b) came from reflections at the aluminum substrate. (a) Contour of the underlying topography reconstructed by taking the darkest pixel of a 2048 series images at each pixel position. (b) Streamline pattern reconstructed by taking the brightest pixel of a 2048 images series at each pixel position. Figure 2.9: Illustration of the evaluation method for the reconstruction of the shape of the underlying topography and the streamline pattern from experimental single image data. The resulting spatial resolution was about 12 µm/pixel. Spatial calibration of the images has been carried out with help of the ceramic calibration scale as described in subsection 2.4.2. 2.4.4 Velocity field measurements Velocity measurements have been done with a Laser-Doppler-Velocimeter (LDV) from Dantec/Invent. A detailed assessment of the general accuracy of the LDV-measurement technique when it is applied on film flows is provided by Aksel and Schmidtchen[74]. As tracer particles a powder of ZrO2/MgO with a mean particle diameter of 0.8µm, which is described more detailed in section 2.3, was used. The light source of the LDV–system was a Spectra Physics Argon–Ion (Ar+) Laser emitting light at three main wavelengths of 476.5 nm, 488 nm and 514.5 nm. A Dantec FiberFlow beam splitter divided the laser beam into two equally intense beams and coupled the three colors into 6 glass fibre optics. Additionally, a Bragg cell shifted one of the two laser beams by 40 MHz to higher frequencies before the beams are splitted into their different wavelengths for two reasons. One, to generate heterodyne detection signals with a sufficiently high frequency from slow scattering tracer particles and two, to obtain information about the direction of the flow. Because the LDV system has been used in the one–dimensional (1D) mode only, all wavelengths except for the most intense (514.5 nm) were blocked by mechanical shutters. An optical device (LDV-head) from Dantec with an extra focussing unit crossed the two remaining working frequency laser beams in an elliptical measurement volume which was specified to be 25 µm×24 µm×126 µm in size. 24
Chapter 2. Experimental systems and setups 31
Chapter 3 Two–dimensional film flow 3.1 Suppression of eddies 3.1.1 Problem formulation We consider a steady two–dimensional gravity–driven flow of a Newtonian liquid down an inclined topography which is sinusoidally undulated in the main flow direction. The profile of the substrate undulation which is described by b(x) = acos(2πx/λ),(3.1) is illustrated in Figure 3.1, where ais the amplitude and λthe wavelength of the periodic undulation. The gravitational acceleration gis given by g=g(sin α, −cos α) in the given x z h(x) b(x) α g Figure 3.1: Viscous film flow down a wavy incline. Reprinted with permission from [44]. ©2010, American Institute of Physics. (x, z)-coordinate system, with αbeeing the mean inclination angle of the channel. The position of the liquid’s free surface is denoted by h(x). The film thickness dcan easily be calculated by substracting the bottom contour b(x) from the free surface shape h(x): d(x) = h(x)−b(x).(3.2) The Navier–Stokes equations and the continuity equation for incompressible liquids ρ(u·∇)u=−∇p+η∆u+ρg,∇·u= 0 (3.3) 32
Chapter 3. Two–dimensional film flow are rewritten in a dimensionless form by introducing reference quantities. As a characteristic length we use the film thickness dnof the corresponding flow over a flat incline with the same flow rate ˙q. From the well known Nusselt solution[73] it can be calculated to dn=3 p3ν˙q/(gsin α).(3.4) Consequently, velocities are rescaled with the mean velocity of the corresponding Nusselt film flow which reads ¯un= ˙q/dn= (gd2 nsin α)/(3ν).(3.5) The pressure is rescaled with the dynamic pressure ρ¯u2 n. Inserting these scalings into (3.3) yields a dimensionless formulation of the Navier–Stokes and the continuity equations Re (˜ u·∇)˜ u=−Re∇˜p+ ∆˜ u+˜ g,∇· ˜ u= 0,(3.6) with the Reynolds number Re = ¯undn/ν = ˙q/ν (3.7) and the dimensionless gravity vector ˜ g= (3,−3 cot α), where a tilde˜ ·denotes a dimensionless quantity. The shape of the underlying topography is described by the dimensionless steepness parameter ξ=a/d and the dimensionless wave number k= 2πd/λ. At the bottom of the topography ˜y=˜ b(˜x) = ξcos(k˜x) the no–slip condition ˜ u=0 holds. Because the liquids free surface contour is a streamline the kinematic boundary condition d˜ h d˜x=˜v ˜u(3.8) has to be fulfilled at ˜y=˜ h, where ˜uand ˜vare the velocity components in ˜xand ˜y direction, respectively. The dynamic boundary condition, which takes care of the balance of stresses at the free surface, has to be fulfilled and reads n·T=3Bo−1 k2Re κn(3.9) when the viscosity of air is neglected. The outer normal unit vector of the free surface is denoted by n,Tis the stress tensor T=−(˜p−˜p0)I+ (1/Re)[∇˜ u+ (∇˜ u)T], Iis the identity matrix, p0is the ambient pressure and κis the curvature of the free surface which is given by κ=1 R= d2h dx2 h1 + dh dx2i3/2,(3.10) with Rbeing the radius of curvature of the free surface shape. The inverse Bond number Bo−1= 4π2σ/(ρgλ2sin α) is a measure for the ratio between surface tension stresses and gravitational stresses. We have now formulated the problem (3.3) with the given boundary conditions in a dimensionless form in a way that it is governed by five independent dimensionless parameters namely ξ,k, Bo−1, Re and cot α, only. Additionally, we claim the flow to be periodic as the underlying topography is. Therefore, we assume the free surface shape, the pressure and the velocity field to be periodic 33
3.1. Suppression of eddies in downstream- (˜x-) direction: ˜ h(˜x) = ˜ h(˜x+ 2π/k) (3.11) ˜p(˜x, ˜y) = ˜p(˜x+ 2π/k, ˜y) (3.12) ˜ u(˜x, ˜y) = ˜ u(˜x+ 2π/k, ˜y).(3.13) With the kinematic (3.8) and the dynamic (3.9) boundary conditions at the free surface and the no–slip boundary condition at the bottom, the periodic boundary conditions (3.11)-(3.13) we complete the set of dimensionless field equations (3.6). The fact that the position of the free surface, and therefore also the domain of solution, is not known a priori introduces a further degree of freedom, which has to be captured by a numerical procedure. This has been achieved by implementing an iterative procedure starting with an initial guess for the free surface shape. In each solution step the Navier– Stokes equations and the continuity equation (3.6) have been solved together with the no–slip condition at the substrate u=0and the dynamic boundary condition at the free surface (3.9). The kinematic boundary condition (3.8) cannot be fulfilled yet and is formally interpreted as a first order differential equation for the unknown free surface position ˜ h(˜x). With the new free surface position the iterative procedure is repeated until the difference between the solution of the previous and the current step is below a threshold value. Since the flow rate is still arbitrary we claimed a certain flow rate and therefore a certain Reynolds number to obtain a unique solution. In each iteration step, the velocities and the pressure are approximated using the Taylor–Hood element pair with piecewise quadratic velocity approximation and piecewise linear pressure. The resulting nonlinear equation for the nodal velocity and pressure is solved with Newton method, which typically converges in four to six iterations. The numerical procedure described above has been implemented and all numerical calculations presented in this chapter have been carried out by Christian Heining[44]. 3.1.2 Experimental and numerical findings The experiments have been carried out with Basildon silicone oil BC10cs, which is described in section 2.1, flowing over a sinusoidally undulated inlay with an amplitude a= 1 mm and a wavelength λ= 10 mm placed in channel 1 near its inflow as described in section 2.2. Measurements have been done at four different inclination angles ranging from 5 ◦to 14 ◦. The Reynolds number has been varied between 3 and 63. Streamline patterns have been recorded experimentally as described in section 2.4.3. Figure 3.2 shows a comparison of experimentally and numerically observed streamlines at different flow rates or Reynolds numbers, respectively. The bright sinusoidal line which is overlaid by a red one corresponds to the substrate geometry. The lines above are the numerically (green) and experimentally (black & white) determined streamlines. The uppermost red line corresponds to the numerically determined free surface contour. The bright lines below the bottom and the inversely bent lines above the free surface are reflections coming from the substrate or the free surface of the liquid. We find that the flow shows qualitatively different behavior depending on the flow rate or the Reynolds number, respectively. When the mean film thickness dis small compared to the wavelength λand the amplitude athe flow can locally be well described by the Nusselt solution with the local inclination angle[21]. Therefore, the liquid flows 34
Chapter 3. Two–dimensional film flow Figure 3.2: Comparison of experimental path lines to numerical streamlines. The images are rotated by the mean inclination angle of the channel. The volume flux is continuously increased form a) to d). a) Re = 9: no eddy at low Reynolds numbers. b) Re = 16: increasing inertia results in the generation of an eddy in the trough of the undulation. c) Re = 31: increasing inertia further, the eddy vanishes. d) Re = 48: flow separation reappears at even higher Reynolds numbers. Bottom contour: lower bright sinusoidal line; lines below and inversely bent lines in the upper part of the images are reflections of the path lines at the bottom and at the free surface. Channel inclination angle α= 8◦. Reprinted with permission from [44]. ©2010, American Institute of Physics. smoothly along the substrate contour when the Reynolds numbers are rather small (See Figure 3.2a)). When the flow rate is increased the flow begins to separate into a region where the flow recirculates within the trough of the undulation and into a main flow above (See Figure 3.2b)). In contrast to the vast majority of systems where an increase of the influence of inertia leads to a growth of recirculation areas[77], increasing the Reynolds number in this system leads to diminution of the eddies until they vanish completely as shown in Figure 3.2c). However, increasing the Reynolds number further, we find a critical Reynolds number at which the eddies reappear as depicted in Figure 3.2d). Above this critical Reynolds number the eddies grow monotonously in size with increasing Reynolds number. All streamline patterns shown in Figure 3.2 correspond to a mean channel inclination angle αof 8 ◦. Qualitatively similar results have been obtained for other inclination angles. We remark that closed path lines, the lack of jitter and the excellent agreement between experiment and numerics indicate that unsteady motion is negligible and that the periodicity (3.11)-(3.13) and two–dimensionality assumptions made in section 3.1.1 hold. Detection of the free surface shape of the liquid has been done as described in section 2.4.2. Because the fluorescent tracer particles described in section 2.3 sit at discrete 35
3.1. Suppression of eddies points when an image is taken, a single recording resulted in a grainy image of the liquid (see Figure 3.3(a)). Thus, the evaluation of the free surface shape has been done with an image, which has been averaged over 50 single images (See Figure 3.3(b)). We note, that the average film thickness in Figure 3.3 appears much thinner as it is because the image of the liquid below the free surface (and also of the underlying topography) is strongly distorted by the curved surface of the liquid. (a) Single image from surface detection camera (2). (b) Image averaged over 50 single recordings. Figure 3.3: Illustration of the averaging process for the detection of the free surface shape. The upper border of the bright sheet in each image corresponds to the contour of the free surface. As Figure 3.2 indicates not only the flow in the troughs of the undulation shows a strong Reynolds number dependence but also the free surface shape of the liquid changes considerably with the Reynolds number. Figure 3.4 provides a comparison of measured (symbols) and calculated (lines) free surface shapes at different Reynolds numbers for a mean channel inclination angle αof 8 ◦. The curves are staggered in z-direction to avoid overlapping. The Reynolds number ranges from 6.6 (lowermost curve) to 56.2 (uppermost curve). At low Reynolds numbers we find the free surface shape to be rather harmonic and of small amplitude. When Re is increased the free surface rapidly gains in amplitude, is shifted downstream and exhibits a sharp nonlinear indentation in the trough which becomes maximal at Re = 31.5 (See red ’x’-symbols in Figure 3.2). Increasing the Reynolds number just a little further, from Re = 31.5 to Re = 32.1 in Figure 3.2, causes this indentation to vanish very brusquely resulting in a smooth sinusoidal shape again (See green ’+’-symbols in Figure 3.2). Further increase of Re shifts the free surface contour further downstream while its amplitude decreases continuously. Qualitatively similar results have been obtained for other inclination angles. Recirculation areas Based on the experimentally and numerically obtained streamline patterns we have evaluated the size of the recirculation areas in the troughs of the undulation. Figure 3.5 shows the eddy area as a function of the Reynolds number at four different mean inclination angles α. Except for the steepest inclination angle of 14 ◦eddies appear with increasing Reynolds number at a first critical Reynolds number Re1≈11 which seems to be independent of the channel inclination angle α. Then the eddy size increases until it reaches a local maximum and shrinks again until it vanishes completely at a second critical Reynolds number Re2. Only beyond a third critical Reynolds number Re3the eddies reappear and grow monotonously in size with increasing Reynolds number in the investigated range. While the second critical Reynolds number for the disappearance of 36
Chapter 3. Two–dimensional film flow 0510 15 20 25 x / [mm] 0 1 2 3 4 free surface position / [mm] 56.2 48.8 42.6 36.4 32.1 31.5 24.1 18.9 11.6 6.6 Reynolds number Figure 3.4: Comparison of experimental and numerical free surface shapes at different Reynolds numbers. Experimental data are represented by symbols; numerical data are represented by lines. The free surface positions are shifted perpendicular to the mean flow direction to avoid overlapping. The vertical position augments with Reynolds number. Channel inclination angle α= 8 ◦. Reprinted with permission from [44]. ©2010, American Institute of Physics. the recirculation areas showed a strong channel inclination angle dependence, the critical Reynolds numbers Re1and Re3for the emerging of eddies seemed to be rather independent of αin the investigated inclination angle range. 10 20 30 40 50 Reynolds number 0 0.2 0.4 0.6 0.8 Eddy area / aλ α = 5° α = 8° α = 11° α = 14° Figure 3.5: Cross–sectional area of the eddy as a function of the Reynolds number at different inclination angles. Experimental and numerical data are represented by open and solid symbols, respectively. At least 40 measurements per inclination angle have been carried out from Re ≈6 to Re ≈62 in equidistant steps. Where no eddy was observed, most data points have been blanked out for clarity. Reprinted with permission from [44]. ©2010, American Institute of Physics. Thus, we find that eddies which appear at not too steep inclination angles at a critical Reynolds number Re1disappear again in an eddy–free window whose extent grows with increasing α. The window where eddies can be observed between Re1and Re2shrinks accordingly at the expense of the eddy–free window until it vanishes completely for an inclination angle α= 14 ◦. 37
3.1. Suppression of eddies Free surface shape Quantitative analysis of the free surface shape data shown in Figure 3.4 has been done by decomposing them into Fourier series by discrete Fourier transformation (DFT). Figures 3.6-3.8 show the Reynolds number dependence of the amplitudes of the zeroth, the first and the second Fourier modes for all inclination angles studied. The zeroth Fourier mode is illustrated in Figure 3.6 and corresponds to the film height1h(x) averaged over one period of the bottom contour. At Reynolds numbers below ≈25 and above ≈37 we find a monotonous increase of the average film thickness with increasing volume flux or Reynolds number, respectively for all investigated channel inclination angles, as it is common for gravity–driven film flows[33, 34, 35]. In the region at intermediate Reynolds numbers all data sets reveal a spontaneous drop in the average film thickness with increasing Reynolds number. The position where this drop takes place shifts with increasing channel inclination to smaller Reynolds numbers. 0 10 20 30 40 50 60 Reynolds number 1 1.5 2 2.5 3 3.5 4 Average film thickness / a α = 5° α = 8° α = 11° α = 14° Figure 3.6: Mean film thickness averaged over one bottom period as a function of the Reynolds number at different inclination angles. Experimental and numerical data are represented by symbols and lines, respectively. Reprinted with permission from [44]. ©2010, American Institute of Physics. The first harmonic of the Fourier transformed free surface shape corresponds to the wavelength of the bottom contour. Its amplitude shows a peak which grows and shifts its position from Re ≈35 to Re ≈27 to smaller Reynolds numbers when the channel inclination angle αbecomes steeper (See Figure 3.7). The presence of this peak reflects the fact that the free surface is strong undulated where the first harmonic peaks, but rather flat for low and for high Reynolds numbers as already illustrated in Figure 3.4. The amplitude of the second harmonic of the free surface shape characterizes its nonlinearity. We find that it grows with increasing Reynolds number and reaches a plateau before it drops discontinuously to a much smaller value and tends rapidly against zero for large Reynolds numbers, as visible from Figure 3.8. The growth and rapid drop of the second Fourier mode corresponds to the emerging of the sharp indentation in the troughs of the free surface shape, as can be seen most clearly from the red line in Figure 3.4, and the abrupt shape transition to a smooth sinusoidal one as represented exemplarily by the green line in Figure 3.4. The height of the plateau grows with steeper channel inclinations. 1or film thickness d(x) = h(x)−b(x). 38
Chapter 3. Two–dimensional film flow 0 10 20 30 40 50 60 Reynolds number 0 0.2 0.4 0.6 0.8 1 Amplitude of 1st harmonic / a α = 5° α = 8° α = 11° α = 14° Figure 3.7: Amplitude of the first harmonic of the free surface shape as a function of the Reynolds number at different inclination angles. Experimental and numerical data are represented by symbols and lines, respectively. Reprinted with permission from [44]. ©2010, American Institute of Physics. In conjunction to the growth, the position of the sharp drop is shifted to smaller Reynolds numbers with steepening the mean channel inclination. All higher harmonics, which are not shown here, showed qualitatively the same behavior as the second one but with much smaller amplitudes. 0 10 20 30 40 50 60 Reynolds number 0 0.05 0.1 0.15 0.2 0.25 Amplitude of 2nd harmonic / a α = 5° α = 8° α = 11° α = 14° Figure 3.8: Amplitude of the second harmonic of the free surface shape as a function of the Reynolds number at different inclination angles. Experimental and numerical data are represented by symbols and lines, respectively. Reprinted with permission from [44]. ©2010, American Institute of Physics. A comparison of Figures 3.6, 3.7 and 3.8 reveals that the drop of the average film thickness, the peak position of the first Fourier mode and the drop of the amplitude of the second Fourier mode seem to coincide at the same Reynolds number for each channel inclination angle. Figure 3.9 complies the first harmonics for the four investigated inclination angles. We find that for all investigated channel inclinations the first harmonic peaks, where the average film thickness drops and the shape of the free surface undergoes a sharp transition from an anharmonic shape to a smooth sinusoidal one. This position where a surface shape transition occurs is indicated by the dashed line in each diagram. The Reynolds number where this transition takes place shifts with steeper channel inclinations 39
3.1. Suppression of eddies to smaller Reynolds numbers. 10 20 30 40 50 60 Reynolds number 0 0.2 0.4 0.6 0.8 1 Amplitude of 1st and 2nd harmonic / a 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Average film thickness / a Average film thickness 1st Fourier mode 2nd Fourier mode (a) Channel inclination angle: 5 ◦. 10 20 30 40 50 60 Reynolds number 0 0.2 0.4 0.6 0.8 Amplitude of 1st and 2nd harmonic / a 0 0.5 1.0 1.5 2.0 2.5 3.0 Average film thickness / a Average film thickness 1st Fourier mode 2nd Fourier mode (b) Channel inclination angle: 8 ◦. Reprinted with permission from [44]. ©2010, American Institute of Physics. 10 20 30 40 50 60 Reynolds number 0 0.2 0.4 0.6 0.8 Amplitude of 1st and 2nd harmonic / a 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Average film thickness / a Average film thickness 1st Fourier mode 2nd Fourier mode (c) Channel inclination angle: 11 ◦. 10 20 30 40 50 60 Reynolds number 0 0.2 0.4 0.6 0.8 1 Amplitude of 1st and 2nd harmonic / a 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Average film thickness / a Average film thickness 1st Fourier mode 2nd Fourier mode (d) Channel inclination angle: 14 ◦. Figure 3.9: Average film thickness and amplitude of the first two Fourier components. The transition Reynolds number is indicated by the dashed line. The experimental and numerical data are represented by symbols and lines, respectively. Figure 3.10 shows a snapshot of a video2, which illustrates the transition of the liquid’s free surface shape while the Reynolds number has been increased continuously over the transition Reynolds number from Re ≈26 to Re ≈34 from two different perspectives at a fixed channel inclination angle of α= 8 ◦. The main frame depicts a slightly deformed picture of two and a half periods of the free surface shape. The main flow direction in the main frame is from right to left. Starting from low Reynolds numbers (Re ≈26) we find a free surface shape with sharp indentations which grow with Reynolds number. The position of the sharp indentation shifts slightly downstream while it is getting sharper until the transition Reynolds number (Re ≈32) is reached. Here the sharp indentation disappears and the free surface shape changes very abruptly into a smooth sinusoidal one. A further increase of the Reynolds number causes the amplitude of the free surface undulation to diminish continuously. The inset in the lower right corner shows the flow over the whole channel width from above. The green line in the middle of the inset corresponds to the laser sheet produced by the Ar+-laser described in section 2.4.2. The main flow direction there is from up 2Video available at http://www.tms.uni-bayreuth.de/videos/surfaces.mp4 40
Chapter 4. Three–dimensional film flow x z α u(z) g (a) Side view y z B θ ∆h H h(y) (b) Cross–sectional view Figure 4.1: Channel geometry illustrating side wall effects on the flow and the orientation and position of the (x, y, z)–coordinate system. Reprinted with permission from [66]. ©2011, American Institute of Physics. read u(y, z = 0) = 0,(4.5) u(y=±B/2, z)=0.(4.6) The no penetration condition at the rigid walls is trivially fulfilled because the flow has been postulated to be unidirectional. Additionally to the boundary conditions at the walls a kinematic boundary condition at the free surface n·uz=h(y)= 0 (4.7) basically demands the free surface contour to be a streamline, or in other words liquid particles must not leave the free surface. When the viscosity of air is neglected the dynamic boundary condition, which takes care of the balance of stresses at the free surface, reads h(p−p0)−σ Rin=T·nz=h(y),(4.8) with the stress tensor T=Tij =η∂ui ∂xj +∂uj ∂xi(4.9) and the outer unit normal vector n=∇h k∇hk=ˆez−∂h ∂y ˆey r1 + ∂h ∂y 2.(4.10) 47
4.1. Basic flow The pressure of the surrounding air is denoted by p0. The curvature of the free surface κ which is the inverse radius of curvature of the free surface Rcan be calculated to κ=1 R=∇·n=−d dy ∂h ∂y r1 + ∂h ∂y 2=− ∂2h ∂y2 1 + ∂h ∂y 23/2.(4.11) Inserting (4.10) and (4.11) into the dynamic boundary condition (4.8) leads to two components of the dynamic boundary condition normal and tangential to the free surface (p−p0)z=h(y)=−σd dy ∂h ∂y r1 + ∂h ∂y 2,(4.12) ∂u ∂z z=h(y)=∂u ∂y ∂h ∂y .(4.13) Dimensionless formulation To reformulate the problem in a dimensionless form we have to find some reference quantities for scaling. We take the free surface velocity ur=ρg sin αH2 2η(4.14) from the well known Nusselt solution [73] as a reference for all velocities. Hydrostatic pressure is taken as a reference for the pressure pr=ρg cos α. (4.15) We introduce a generalized capillary length L L=r2σ ρg cos α,(4.16) which takes care of a reduced gravitational acceleration perpendicular to the channel due to its inclination αand serves as a reference for all lengths to resolve effects within the capillary elevation. In the following all quantities which are labeled by a ˜ ·denote dimensionless variables which are scaled with the above reference quantities. Furthermore, we define a dimensionless capillary range l l=L B/2=2L B.(4.17) Applying the above scalings to the Navier–Stokes equations and the boundary conditions lead to a dimensionless formulation of the system in the following form. The Navier– Stokes equations read: 0 = ∂2˜u ∂˜y2+∂2˜u ∂˜z2+ 2,(4.18) 0 = −∂˜p ∂˜y,(4.19) 0 = −∂˜p ∂˜z−1.(4.20) 48
Chapter 4. Three–dimensional film flow The no–slip boundary conditions at the bottom and the side walls changes to ˜u(˜y, ˜z= 0) = 0,(4.21) ˜u(˜y=±1/l, ˜z) = 0 (4.22) and the tangential and normal components of the dynamic boundary condition now read ∂˜u ∂˜z˜z=˜ h(˜y)=∂˜u ∂˜y d˜ h d˜y,(4.23) (˜p−˜p0)˜z=˜ h(˜y)=−1 2 d d˜y ∂˜ h ∂˜y r1 + ∂˜ h ∂˜y2.(4.24) Free surface shape We introduce an additional boundary condition taking care of the contact angle θbetween the liquid and the side wall ∂˜ h ∂˜y˜y=±1/l =±cot θ(4.25) and a decomposition of the free surface shape ˜ h(˜y) into a constant part which is equal to the film height in the middle of the channel ˜ Hand a part ζdepending on ˜ywhich takes care of the capillary elevation in the vicinity of the side wall ˜ h(˜y) = ˜ H+ζ(˜y).(4.26) Evaluating ζat ˜y=±1/l by inserting the boundary condition (4.25) into (4.24) leads to the capillary elevation height depicted in Figure 4.1(b) [66] ∆˜ h=ζ(˜y=±1/l) = √1−sin θ. (4.27) The free surface shape can be obtained by integration of equation (4.24) as described in detail by Scholle and Aksel[63] or Haas et al.[66]. With the abbreviation G(x) := x−1 2√2ln √2 + x √2−x!,(4.28) the function for the film elevation ζ(˜y) can be written down in an implicit form ˜y(ζ) = ˜ h−1(ζ) = −1/l +hG√1 + sin θ−Gp2−ζ2i ˜y∈[−1/l, 0], 1/l −hG√1 + sin θ−Gp2−ζ2i ˜y∈[0,1/l].(4.29) Velocity field The velocity field can be described by the following ansatz which is a solution of equation (4.18) and already fulfills the no-slip boundary conditions at the side walls ˜u=4 l2X n∈N+Dnelcn˜z+Ene−lcn˜z−(−1)n c3 ncos(lcn˜y),(4.30) 49
4.1. Basic flow cn=n−1 2π. (4.31) The additional no-slip condition at the substrate (4.21) leads to nequations for the vector elements Dnand En Dn+En=(−1)n c3 n .(4.32) Inserting the ansatz for the velocity field (4.30) into the tangential part of the dynamic boundary condition (4.23) leads to an infinite system of algebraic equations for the coefficients DnX n∈N+ GnmDn=dm,(4.33) with the matrix elements Gnm =Z1/l −1/l 2 cosh(lcn˜ h(˜y)) sin(lcn˜y) sin(lcm˜y)d˜y(4.34) and the vector elements dm=X n∈N+Z1/l −1/l (−1)n c3 ne−lcn˜ h(˜y)+lcn˜ h(˜y)sin(lcn˜y) sin(lcm˜y)d˜y −l cmX n∈N+Z1/l −1/l (−1)n˜ h(˜y) cn [cos(lcn˜y) cos(lcm˜y)] d˜y. (4.35) The coefficients Dnand Enhave been calculated using MATLAB®[83] truncating the infinite system of algebraic equations to a certain order N∈N+. The accuracy of the power series expansion has been assured by the demand Dn, En<10−6.[66] The theoretical derivation and the implementation of the procedure of solution described above has been done and all theoretical results presented in this chapter have been calculated by Andr´e Haas[66]. 4.1.2 Flow type classification Depending on the magnitude of the capillary elevation compared to the film height H and the channel width Bit is useful to distinguish different flow types as illustrated in Figure 4.2. For a channel of infinite extent the flow configuration is equivalent the two–dimensional case and the solution of the velocity field equals the well known Nusselt solution. In this case we define the Reynolds number as Re(a)= Re2D =usH ν.(4.36) Because the free surface velocity of a Nusselt film flow can easily be calculated to us=gsin αH2 2ν,(4.37) the Reynolds number can also be expressed as a function of the film height H Re2D =gsin αH3 2ν2.(4.38) 50
Chapter 4. Three–dimensional film flow (a) Infinite broad channel or slip condition at the side walls (equals the 2D case). (b) No capillary elevation at the side walls (θ= 90◦ or ∆hH) (c) (θ < 90◦and ∆h∼H) (d) Capillary corner flow (θ < 90◦and ∆hH) Figure 4.2: Cross sectional velocity profiles of different channel flow types. The velocity is color coded: blue corresponds to slow and red corresponds to fast. Reprinted with permission from [66]. ©2011, American Institute of Physics. As soon as side walls are introduced the liquid in the vicinity of the side wall gets decelerated due to the additional no–slip condition at the boundary. Now, the free surface velocity is not independent of the crosswise coordinate yanymore and it is useful to define the Reynolds number in terms of the corresponding volume flux ˙ V Re(b)=3 2Bν Zf(y) 0ZB/2 −B/2 u(y, z)dydz=3˙ V 2Bν <Re2D.(4.39) Compared to the scenario depicted in Figure 4.2(a) less liquid is transported due to the additional drag at the side walls. Therefore, the Reynolds number of a film in this case is alway smaller than the Reynolds number of a two–dimensional film of the same film thickness H. When His decreased, the contact angle θbetween the liquid and the side wall becomes an important factor. The influence of capillarity leads to a capillary elevation of the liquid, when θis smaller than 90◦. Due to this locally thicker film a velocity overshoot close to the side walls may show up when the additional film thickness wins over the additional drag coming from the no–slip condition at the side walls (see Figure 4.2(c)). In the limit of vanishingly thin films (H→0) the flow degenerates to a capillary corner flow as depicted in Figure 4.2(d). Now most of the liquid is transported close to the side walls in the capillary elevation and the Reynolds number is obviously larger than in the two–dimensional case Re(d)=3˙ V 2Bν >Re2D →0, H →0.(4.40) When the contact angle between the liquid and the side wall is larger than 90◦the capillary elevation ∆hbecomes negative and no velocity overshoot can be observed. Therefore, we restrict our studies without loss of generality to liquids with wetting properties only. 4.1.3 Flow rate study Since the presence of side walls and the resulting capillary elevation has a significant impact on the flow rate, a study on the important parameters is to be carried out. The 51
4.1. Basic flow volume flux of a two–dimensional case ˙ V2D which is illustrated in Figure 4.2(a) is taken as a reference. Figure 4.3 illustrates the inequality (4.39) for two different capillary ranges, one matching the experimental setup (l= 0.028) and one referring to the capillary range of a narrower (and/or steeper) channel (l= 0.1). When no capillary elevation is present, as depicted in Figure 4.2(b) the additional drag at the side walls leads to an overall decrease of the transported liquid depending on the dimensionless film height ˜ Hand the capillary range l. The amount of missing volume flux increases with decreasing channel width B and increasing dimensionless film height ˜ H. Both an increasing dimensionless film height and a narrowing of the channel lead to an increase of the relative part of the side wall area which results in a stronger impact of the side wall presence on the volume flux. Therefore, especially for narrow channels and film thicknesses which are large compared to the capillary length Lthe influence of the side walls on the volume flux cannot be neglected. 0.9 1 0 0.2 0.4 0.6 0.8 1 ˜ H ˙ Vθ=90◦/˙ V2D l= 0.028 l= 0.1 Figure 4.3: Decrease of the volume flux due to the additional no–slip condition at the side walls (without capillarity), as it is depicted in Figure 4.2(b), compared to the two–dimensional case, which is depicted in Figure 4.2(a). Reprinted with permission from [66]. ©2011, American Institute of Physics. In the case depicted in Figure 4.2(c) it is not possible to make a similar general statement on the Reynolds number or the volume flux like in the equations (4.39) or (4.40), because both effets of one an additional drag coming from the side wall and two a velocity overshoot have a competing influence of the same order on the flow rate as illustrated in Figure 4.4. Figure 4.5 compares the amount of transported liquid when a capillary elevation due to the presence of side walls is considered with the two–dimensional case. The contact angles θare chosen to fit the experimental setup as described in section 2.2. For small film thickness ˜ Hthe influence of the capillary elevation and the resulting velocity overshoot becomes the most important transport mechanism. The ratio of ˙ V / ˙ V2D becomes larger than one and even diverges for ˜ H→0 because ˙ V2D then also tends to zero. For constant film height ˜ Hthe presence of a capillary elevation becomes obviously more important for narrower channels or larger capillary ranges l(see Figure 4.5(a)). Remember, that the two–dimensional case is equal to the case l→0. The influence of the capillary range on the volume flux increases with decreasing film height. One finds an explicit transition film height denoted by ˜ ht, which is independent of the capillary length, where the influences of the velocity overshoot due to capillary elevation and the influence of the no–slip condition 52
Chapter 4. Three–dimensional film flow u umax us,m Figure 4.4: Free surface velocity profile in flow direction showing velocity overshoot and defect compared to the plane flow with the same film height. Reprinted with permission from [70]. ©2011, American Institute of Physics.[70] at the wall on the normalized volume flux ˙ V / ˙ V2D just cancel each other. 1 1.5 2 0 0.2 0.4 0.6 0.8 1 ˜ H ˙ V / ˙ V2D l= 0.028 l= 0.1 ˜ ht (a) Variation of the capillary range lat fixed contact angle θ= 8◦. 1 1.5 2 0 0.2 0.4 0.6 1 ˙ V / ˙ V2D ˜ H θ= 8◦ θ= 52◦ ˜ ht˜ ht (b) Variation of the contact angle θat fixed capillary range l= 0.1. Figure 4.5: Influence of capillary effects at the side walls on the normalized volume flux. Reprinted with permission from [66]. ©2011, American Institute of Physics. The impact of the contact angle θon the volume flux is illustrated in Figure 4.5(b). Similar to the capillary range in Figure 4.5(a) also the influence of the contact angle on the volume flux is small for film heights ˜ Hof the order of one but gains in importance the thinner the liquid film gets. One finds that smaller contact angles lead to a larger velocity overshoot and therefore to a larger volume flux. The transition film height ˜ htis not independent of the contact angle. For large ˜ Hthe ratio of ˙ V / ˙ V2D becomes smaller than one because the influence of capillarity on the velocity field looses importance. The volume flux ˙ Vtends to ˙ Vθ=90◦for ˜ H→ ∞ which is always smaller than ˙ V2D as depicted in Figure 4.3. Figure 4.6 shows the dependence of the transition film height ˜ hton the contact angle θ. For a contact angle of θ= 90◦no capillary elevation is present and thus no velocity overshoot can be observed. Equality of the volume flux of the 2D case ˙ V2D and the volume flux ˙ Vcan only be reached, when also the drag influence of the side wall tends to zero which is only the case in the limit ˜ h→0. Decreasing the contact angle leads to a monotonous increase of the transition film thickness to finite values below one. Under perfect wetting conditions (θ= 0◦) the transition film height reaches a value of approximately ˜ ht≈0.92. To summarize: Treating a channel of finite width as two–dimensional always leads to 53
4.1. Basic flow 0 0.2 0.4 0.6 0.8 1 0 30 60 90 ˜ ht θ/[◦] Figure 4.6: Dependence of the transition film thickness ˜ hton the contact angle θ. Reprinted with permission from [66]. ©2011, American Institute of Physics. an overestimation of the volume flux (or the Reynolds number) when the film thickness H is larger than the generalized capillary length L(or above the red curve in Figure 4.6) due to the additional drag coming from the side walls. This discrepancy becomes large especially for narrow channels. When the film thickness is small or of the same order as the generalized capillary length the presence of a capillary elevation gains in importance one has to take the contact angle θinto account. The effect of a resulting velocity overshoot competes with the additional drag at the side wall. For thin films and small contact angles (below the red curve in Figure 4.6) treating the film as two–dimensional leads to an underestimation of the volume flux. 4.1.4 Velocity field Figure 4.7 shows a comparison of the theoretical and measured velocity profiles for three films of different heights Hand two different static contact angles θ. As liquid Elbesil silicone oil B1000 which is described in section 2.1 was used. For each flow configuration the side wall distance dsdependence of the velocity profiles was measured by a Laser Doppler Velocimeter described in section 2.4.4 at three different measurement heights Hm. The error bars of the measured data denote the root mean square error of the mean value of all detected velocity signals in each measurement volume. Because the number of evaluable counts per time decreases with the speed of the liquid in the measurement volume, the measurement time has been adopted to the flow velocity to get reasonable signal to noise ratios especially in near wall regions. Additionally, more points have been recorded in the vicinity of the wall to resolve the velocity overshoot. The overall agreement between the measured data and the calculated values for the flow velocities is excellent. The small deviations which never exceed the root mean square error bars are in the most cases of statistically nature. Systematic discrepancies such as in Figure 4.7(a) can be explained by errors in determining the distance between the channel bottom and the measurement volume Hmor in determining the film height in the middle of the channel H. For large side wall distances the measured and calculated velocity profile corresponds 54
Chapter 4. Three–dimensional film flow d u Hm= 0.7585 mm Hm= 0.5309 mm Hm= 0.3034 mm surface 0 5 0 10 20 [mm/ s] [mm] s (a) H= 0.965 mm, θ= 8◦. d u Hm= 0.7585 mm Hm= 0.5309 mm Hm= 0.3034 mm surface 0 5 0 10 20 [mm/ s] [mm] s (b) H= 0.960 mm, θ= 52◦. d u Hm= 1.0619 mm Hm= 0.7586 mm Hm= 0.3793 mm surface 0 5 10 0 10 20 [mm/ s] [mm] s (c) H= 1.550 mm, θ= 8◦. d u Hm= 1.0619 mm Hm= 0.7585 mm Hm= 0.3793 mm surface 0 5 10 0 10 20 [mm/ s] [mm] s (d) H= 1.550 mm, θ= 52◦. d u Hm= 1.8962 mm Hm= 1.5170 mm Hm= 1.1377 mm surface 0 5 10 15 0 10 20 [mm/ s] [mm] s (e) H= 2.208 mm, θ= 8◦. u d Hm= 1.8962 mm Hm= 1.5170 mm Hm= 1.1377 mm surface 0 5 10 15 0 10 20 [mm/ s] [mm] s (f) H= 2.204 mm, θ= 52◦. Figure 4.7: Comparison of measured (points) and calculated (lines) velocity profiles for different film heights H, measurement heights Hmand contact angles θ. Reprinted with permission from [66]. ©2011, American Institute of Physics. to the Nusselt solution. In the near wall region a velocity overshoot is observed whose magnitude, quantified by the ratio of the highest velocity umax and the surface velocity in the middle of the channel us,m, depends strongly on the film thickness Hand the contact angle θ. Figure 4.8(a) shows the dependence of the magnitude of the velocity overshoot on the film thickness ˜ Hfor both contact angles θmeasured. For thick films the ratio of umax/us,mtends to one, meaning that no velocity overshoot can be observed. When the film thickness ˜ His approximately 0.5 or less the magnitude of the velocity overshoot becomes considerable and depends on the contact angle θas also shown in Figure 4.7. As the film thickness is decreased further the velocity overshoot diverges, because the free 55
4.1. Basic flow surface velocity in the middle of the channel tends to zero. The flow then degenerates to a capillary corner flow as illustrated in Figure 4.2(d). 0 2 4 6 0 0.5 1 1.5 umax/us,m ˜ H θ= 8◦ θ= 52◦ (a) Normalized maximal free surface velocity. 0 0.5 1 1.5 ˜ymax 0 1/2l 1/l θ= 8◦ θ= 52◦ ˜ hc˜ hc ˜ H (b) Position of the maximum free surface velocity. Figure 4.8: Influence of the film thickness on the velocity overshoot. Reprinted with permission from [66]. ©2011, American Institute of Physics. Not only the magnitude of the velocity overshoot but also its location depends on the film height ˜ has well as on the contact angle θ. Figure 4.8(b) shows film height dependence of the ˜y-position ˜ymax where the largest free surface velocity umax is located. An increase of the film height leads to a shift of the maximum free surface velocity towards the middle of the channel until the velocity overshoot disappears and the maximal velocity is located in the middle of the channel at ˜y= 0. The critical film thickness at which the velocity overshoot disappears, labeled in Figure 4.8(b) by ˜ hc, shifts with increasing contact angle θto smaller values. Both the position and the magnitude of the velocity overshoot depend on the contact angle θand the film height H. Therefore, we introduce a combined dimensionless parameter rwhich can be attributed to a certain shape of the free surface velocity profile and depends on the dimensionless capillary elevation height ∆˜ h(θ) and the dimensionless film height ˜ Hin the form of[66] r=∆˜ h(θ)c1 ˜ Hc2.(4.41) The free parameters c1and c2were obtained by fitting the results of additional simulations: c1= 0.0435 ±0.002 and c2= 0.9814 ±0.0176.[66] With these parameters inserted in equation (4.41) all experimental setups with the same parameter rshow the same behavior for the velocity overshoot. Thus, we can now find a critical ratio rcfor the onset of a velocity overshoot to rc= 0.733 ±0.002. Inserting rcinto the equation (4.27) for the capillary elevation height leads to an empirical threshold for the onset of a velocity overshoot in terms of a critical film thickness ˜ hc=r−1/c2 c(1 −sin θ)c1/(2c2)(4.42) which is illustrated in Figure 4.9. The critical film thickness ˜ hcis larger than transition film thickness ˜ ht(compare Figures 4.6 and 4.9), because ˜ hcdescribes the film thickness where a velocity overshoot just emerges. The magnitude of the velocity will not become sufficiently strong to balance the no–slip condition at the until the film height is decreased further to the transition film height ˜ htwhich, therefore, has always to be smaller than ˜ hc. 56
Chapter 4. Three–dimensional film flow studied system to be more stable in the middle of the channel than the two–dimensional case, especially at high excitation frequencies. The neutral curve in the middle of the channel shows no contact angle dependence. Vlachogiannis et al. [68] and Georgantaki et al. [69] studied the influence of a finite channel width on the stability of the flow at different inclination angles and fluids properties at very low excitation frequency (fe= 0.167 Hz). They also found the flow to be more stable under the presence of side walls, but only if the Kapitza number Ka is sufficiently high (see Figure 4.15). The Kapitza number is defined as Ka = σ/(ρg1/3ν4/3) and represents the ratio of capillary stresses to viscous stresses. It is a dimensionless material property only and does not depend on flow properties. They found the ratio R∗= Rec/Rec,2D to depend only on the channel width and on the Kapitza number, but not on the channel inclination or material properties like for example the surface tension σ. When the Kapitza number is of the order of one or smaller they found no influence of the channel width on the stability of the flow when the channel was at least 100 mm broad. Although all our measurements were done at a Kapitza number of approximately five we found a strong stabilizing influence of the side walls on the flow which is not in line with the data shown in Figure 4.15. Yet, we suppose that these findings are not in conflict with each other, because an extrapolation of our data shown in Figure 4.14 yields that it might coincide with the two–dimensional case at the limit of very low frequency fe→0 as proposed by the work of Georgantaki et al. [69]. (a) Ratio R∗as a function of Ka for several inclination angles, fluids and channel widths. The dimension of the given surface tension is mN/m. The upper curve corresponds to a channel width of B= 100 mm. The lower curve corresponds to a channel width of B= 250 mm. (b) Ratio R∗as a function of Ka for several channel widths B,α= 3◦. Figure 4.15: Ratio R∗as a function of Ka. Reprinted with permission from [69]. ©2011, American Physical Society. URL: http://pre.aps.org/abstract/PRE/v84/i2/e026325 In the vicinity of the side walls all our measurements showed a further stabilization of the flow compared to the data from the center–line measurements. Furthermore, the contact angle θ, which did not play a role in the middle of the channel, gains in importance when the side wall distance dsis reduced. Measurements done in the channel with the coated glass side walls (θ= 52 ◦) show neutral curves which are significantly shifted 63
4.2. Stability near the side walls to lower Reynolds numbers than the measurements done in the channel with untreated Plexiglas®side walls (θ= 8 ◦). We account two effects for this phenomenon. One, the smaller contact angle causes a larger capillary elevation height ∆h(see e.g. Figure 4.10 or equation (4.27)) and thus to a larger contact area between the liquid and the retarding side wall. Two, when a finite surface tension of the liquid is considered, the smaller contact angle leads to a stronger curvature and therefore to a stronger pretensioning of the free surface in the vicinity of the side wall which hinders free surface waves to develop and thus tends to stabilize the flow. We find a remarkable range of the contact angle influence on the stability of the flow. At a side wall distance of 10 mm, which is about four times the capillary length Lor the film height H, the difference between the neutral curves at θ= 8 ◦and θ= 52 ◦is up to 25%. Even up to a side wall distance of 40 mm, which is about 17 times the capillary length Lor the film height H, the difference between the neutral curves at θ= 8 ◦and θ= 52 ◦is still more than 7%. Additionally, the shape of the neutral curves changes when the side wall distance is decreased. In the middle of the channel we observe the longest waves to become initially unstable as predicted by Benjamin[46] and Yih[47] for the two–dimensional case. For side wall distances of 10 mm in the case of uncoated side walls and 20 mm in the case of coated side walls the type of the instability changes from a long–wave type to a short–wave type instability in the investigated frequency range. This type of instability is well known for boundary layer flows as observed experimentally by Schubauer and Skramstad[86] for a plate which is aligned parallel to a plane flow. This configuration was later described in detail by Schlichting and Gertsen[87]. The mechanisms for the instability are quite different because in the present work an instability of a free surface near a side wall and not the instability of a bulk is investigated. Typical critical Reynolds numbers found for this bulk instability in a boundary layer are about two orders of magnitude larger than for the free surface flow investigated here[87]. However the similarity of the shape of the neutral curves close to the side wall suggests to treat the near wall region as a capillary boundary layer with a range of four to eight times the capillary length L. Other gravity–driven free surface flows showing a short–wave instability are described in a two–dimensional theoretical framework by D’Alessio et al.[59] for Newtonian liquids at very high inverse Bond numbers, which means that capillary forces dominate over gravity, or by Heining and Aksel[60] for power–law liquids flowing down a sinusoidally undulated incline. At intermediate side wall distances the neutral curves neither show the character of a typical long–wave instability nor the typical short–wave instability. In this transition region we observe, that the neutral curves have an inflection point in the investigated frequency range. The size of the transition region seems to be larger for smaller contact angles θ. Figure 4.16 shows the side wall distance dependence of the neutral points for both investigated contact angles and two different excitation frequencies. Because the measurements were done at slightly different excitation frequencies the data shown in Figure 4.14 have been interpolated linearly to provide comparability. In the middle of the channel we do not observe a contact angle dependence as a comparison of the Figures 4.14(a) and 4.14(b) already revealed. Reducing dsleads at first to a monotonous increase of the Reynolds number at which free surface waves are neither damped, nor amplified while 64
Chapter 4. Three–dimensional film flow travelling downstream, which means, that the flow is getting more and more stable due to the retarding influence of the side wall and the pretensioning of the free surface coming into play. Especially at low excitation frequency fewe observe the large amplitude and range of the contact angle influence. 0 10 20 30 40channel center ds/L 1 1.5 2 2.5 3 Re fe = 2Hz; θ = 8° fe = 2Hz; θ = 52° fe = 6Hz; θ = 8° fe = 6Hz; θ = 52° Figure 4.16: Side wall distance dependence of the neutral points at two different contact angles θ and excitation frequencies fe. Reprinted with permission from [70]. ©2011, American Institute of Physics. However, this monotonous behavior is broken very close to the side wall, although the stabilizing effects of the side wall should be strongest here. This can also be seen in Figure 4.14: Especially in the case of uncoated side walls the neutral curve for ds= 5 mm is left of the neutral curve for ds= 10 mm over the whole investigated excitation frequency range. Obviously another (competing) effect, namely the presence of a velocity overshoot, gains in importance there. When the film thickness is smaller than the critical film thickness hc, which is about ∼1.3Lfor small contact angles θ, not only the film thickness in the vicinity of the side wall is larger than the film thickness in the middle of the channel H, but also a velocity overshoot is observed due to the capillary elevation (see section 4.1). Both, the higher film thickness as well as the higher velocity at the free surface cause the local Reynolds number Reloc(ds) = h(ds)us(ds)/ν to exceed the (global) Reynolds number Re = 3 ˙ V /(2νB) at some ds(See Figure 4.17). In those regions the onset of waves at the free surface is promoted and the flow tends to be more unstable. Due to the more pronounced velocity overshoot and capillary elevation for smaller contact angles the magnitude of the local Reynolds number overshoot is larger for θ= 8 ◦. That explains why the peaks in the side wall distance dependence of the neutral points in Figure 4.16 are more pronounced for θ= 8 ◦than for θ= 52 ◦. Compared to the stabilizing effects, the destabilizing influence of the local Reynolds number overshoot seems to be of a much shorter range. 4.2.2 Conclusions We have shown that the neutral curve for the onset on a primary instability in gravity– driven free surface flows depends strongly on the distance to the side wall of the channel. In the studied system the flow in the vicinity of the side wall was always more stable 65
4.2. Stability near the side walls Reloc ∼us·h us h ds Figure 4.17: Sketch of the side wall distance dependence of the local Reynolds number Reloc when a velocity overshoot is present. than the flow in the middle of the channel. A direct consequence is, that a flow may be unstable at some regions, but shows other regions where the free surface waves coming from the unstable regions are damped at the same time. Of course a flow has to be treated as unstable as soon as first surface waves appear somewhere for most applications. Nevertheless, we found that a coexistence of stable and unstable regions is possible, which is important to understand the influence of the side walls on the instability of the flow and the underlying mechanisms responsible.[70] In the middle of the channel we observe a long–wave type instability as found by Benjamin[46] and Yih[47] for a purely two–dimensional flow. When the distance to the side wall is reduced the instability undergoes a transition from the long–wave type to a short–wave type as it is typical for boundary layer flows[86, 87]. Georgantaki et al.[69] restricted their studies on the influence of the channel width on the instability of film flow to the limit of very long waves. They find the ratio R∗, which is the critical Reynolds number normalized with the critical Reynolds number of a film flowing down a plane of infinite extent, to be a function of the channel width and the Kapitza number. Since we have shown, that the most unstable wave may also have a finite wavelength in near wall regions, we propose that a excitation frequency variation has to be carried out to determine the ratio R∗properly for all channel widths and Kapitza numbers. One has to consider different competing effects of the side walls on the instability of the free surface, some tend to stabilize and some tend to destabilize the flow. The additional no slip condition at the wall and the pretensioning of the free surface due to capillary elevation tend to stabilize the flow. These effects are more pronounced when the contact angle between the liquid and the side wall θis small. For thin films the capillary elevation leads to a velocity overshoot at the free surface and thus to an overshoot of the local Reynolds number which promotes the onset of free surface waves. Compared to the effects which tend to stabilize the flow, which are still significant up to side wall distances of 17L, the influence of the local Reynolds number overshoot on the stability of the free surface seems to be of rather short range. Also the magnitude of the impact of the Reynolds number overshoot in the vicinity 66
Chapter 4. Three–dimensional film flow of the side wall is clearly of minor importance compared to the stabilizing effects for the investigated flow configuration. But we remark that a further reduction of the film height will amplify the magnitude of the Reynolds number overshoot until its influence might overcome the stabilizing effects. Therefore, we speculate that there should exist a critical film height h0 cat which a flow initially becomes unstable close to the side walls before the Reynolds number in the middle of the channel reaches the classical result of Rec= (5/4) cot α. This would imply that a film of thickness h<h0 cwhich is confined by side walls initially becomes unstable at a smaller Reynolds number than a channel flow of infinite extent (B→ ∞) with the same film thickness h. To prove this assumption that the presence of side walls have an overall destabilizing influence on very thin gravity–driven film flows further experiments have to be carried out. 67
Chapter 5 Conclusions and outlook In the present work we study viscous gravity–driven film flow down an inclined channel numerically, analytically and experimentally. Particular focus lies on the influence of a periodic two–dimensional sinusoidally undulated topography on the flow and the appearance and disappearance of eddies in the valleys of the undulation. The corresponding results are presented in chapter 3. Chapter 4 deals with the question how the presence of side walls and different contact angles between the liquid and the side wall influences the flow structure, the overall volume flux and the physical stability of the flow. The main findings are summarized in the following. First, we consider numerically and experimentally a steady flow down an inclined periodic two–dimensional sinusoidally undulated topography. We find, that the free surface undergoes a sharp transition from a strongly anharmonic shape with a strong indentation just above the valleys of the undulation to a smooth harmonic shape as the Reynolds number is increased above a transition Reynolds number. At this transition Reynolds number a hydraulic jump which has been established continuously with increasing Reynolds number disappears abruptly because the flow changes its type from a sub- to a supercritical flow in terms of the Froude number. Furthermore, we find that eddies, which are formed with increasing Reynolds number in the valleys of the underlying undulation, disappear again just in the vicinity of the transition Reynolds number. However, the eddies are suppressed when the hydraulic jump is present below the transition Reynolds number as well as above it. Hence, we conclude, that not the transition itself is responsible for the suppression of the eddy structures, but an amplification of the free surface amplitude, which comes along with the transition and is well known as a resonance phenomenon in literature. Such systematic suppression of eddies is of particular relevance to open up new optimized process windows for many industrial applications where their formation is desired or undesired, depending on the purpose, since they have significant influence on macroscopic system properties. In heat exchanger applications for example the formation of such eddy structures has a major impact on the convective heat transport in the liquid while they lead to drag reduction, which might be useful for bearings or any kind of material transport. In environmental systems particles captured in the recirculating flow are cut off from subsequent delivery of potentially necessary nutrient substances and so is the wall which is in contact with the eddy. The possibility to generate or destroy eddy formations in the valleys of an undulation just by damping or exciting resonance seems 68
Chapter 5. Conclusions and outlook to be a good framework in general, different from applying external forces to the system. In the second part of the present work the influence of side walls and the contact angle between the liquid there on the velocity profile and the flow rates has been studied experimentally and theoretically. In the case of a wetting liquid a capillary elevation is formed at the side walls, which may lead to a velocity overshoot in the vicinity of the wall. While this velocity overshoot leads to an increase of the transported liquid in the channel an additional no–slip condition at the side walls has a contrary influence on the flow rate. Here, a criterion for the first onset of a velocity overshoot is presented, as well as a criterion where the influences of the velocity overshoot and of the additional no–slip condition on the overall volume flux just cancel each other. As long as they do not cancel each other, neglecting the three–dimensionality of the flow would lead to an over– or underestimation of the flow rate, what becomes especially important in the case of a very thin film, where most liquid is transported in the capillary corner of the channel. In addition to the steady flow case findings for the case of a slowly draining flow are presented. Experiments show, that the free surface shape of the liquid can not be modeled by a series of quasi–steady states, even if a dynamic contact angle is taken into consideration, although the settling speed of the liquid is orders of magnitude smaller than the mean flow velocity. Time dependent numerical simulations of the slowly draining flow case revealed an indentation of the free surface near the side walls, which might promote film rupture in industrial thin film applications. Finally, we investigate experimentally the influence of the presence of side walls on the primary instability of the free surface of the flow and find, that the neutral stability curve shows a strong dependence on the distance to the side wall and on the wetting properties of the liquid. In the middle of the channel, far away from the side walls, we observe a long–wave type instability of the free surface, which is independent of the contact angle between the liquid and the side wall. However, when the side wall distance is reduced the type of the instability changes to a short–wave type instability, as it is well known from boundary layer flows at much higher Reynolds numbers, and shows a strong contact angle dependence. We find, that one has to consider different competing effects of the side walls on the stability of the flow. The pretensioning of the free surface due to the capillary elevation and the additional no–slip boundary condition at the side wall tend to stabilize the flow and are sensible up to large side wall distances of about 17 times the generalized capillary length. The formation of a velocity overshoot in the capillary elevation, on the contrary, tends to destabilize the flow, because the local Reynolds number in the near wall region can exceed the (global) Reynolds number of the channel flow. However, compared to the stabilizing effects coming from the side wall the destabilizing effect of the side walls seems to be of rather short range and plays a minor part only for the investigated flow configuration. Thus, the presence of side walls leads to an overall stabilization of the system studied here. Nevertheless, we speculate that for even thinner films the increasing magnitude of the velocity overshoot might cause the destabilizing effect to win over the stabilizing effects, which would implicate that very thin films bounded by side walls would be more unstable than the corresponding films of infinite extent. In industrial applications which deal with very thin films like curtain coating processes for example, one would like to optimize the process in terms of a fast processing and an effective exploitation of the coating material. Since side walls lead to a nonuniform coating at the edges one usually cuts the edges of the coated material away to get a 69
uniformly coated result. Here, we show that especially in thin films most of the material is transported just where edges are located when the contact angle of the liquid at the side wall is below 90◦. The speed of processing in industrial applications is often limited by the stability of the flow, which is also strongly affected by the side walls and the wetting properties of the liquid. Therefore, we suggest that the contact angle should be considered for the design of such devices since it holds some relevant potential for optimization in coating and other thin film applications. To summarize, the present work deals with the influence of a corrugated topography and of side walls on film flows. Nevertheless, several aspects concerning the stability of gravity–driven film flows remain open questions. Since our experiments on the influence of the side wall on the stability of the flow were restricted to side wall distances of 5 mm, we suggest a full numerical study of the Navier–Stokes equations involving extensive parameter studies especially for the case of very thin films to get a closer look into the liquid near the side wall and thus a better understanding of the underlying physics. However, we remark, that the computational effort for such simulations will quickly become a key issue, since for the investigation of convective instability it is necessary to simulate rather large domains which are fully three–dimensional and time dependent on a rather precise mesh to resolve capillary effects properly. In addition to the question how side walls influence the flows’ stability, the flow over substrates with finite corrugations has been studied by several authors over the last years theoretically as well as experimentally. However, the question how a formation of eddies in steep substrate undulations might influence the stability of the flow is very interesting but still open and is worth to be addressed in future works. 70
Chapter 5. Conclusions and outlook 71
List of symbols (x, y, z) spatial coordinates (˜x, ˜y, ˜z) dimensionless spatial coordinates (ˆex,ˆey,ˆez) unit vectors in x-, y- and z-direction (X, Y, Z) spatial coordinates in the reference frame of the traverse (x0, y0, z0) spatial coordinates in the reference frame of the screen (u, v, w) velocity components in x-, y- and z-direction (˜u, ˜v, ˜w) dimensionless velocity components in x-, y- and z-direction uvelocity vector ttime ρ, ν, η, σ fluid properties: density, kinematic viscosity, dynamic viscosity and surface tension nouter normal unit vector Tstress tensor Iidentity matrix g, ˜gdimensional and dimensionless acceleration of gravity g,˜ gdimensional and dimensionless gravity vector p, ˜pdimensional and dimensionless pressure p0,˜p0dimensional and dimensionless ambient pressure a, λ amplitude and wavelength of the channel undulation αchannel inclination angle Bchannel width b,˜ bdimensional and dimensionless contour of the channel topography l1, l2laser beam travel distances h, ˜ hdimensional and dimensionless position of the free surface d, ˜ ddimensional and dimensionless film thickness hn, dnNusselt film thickness H, ˜ Hdimensional and dimensionless film height in the middle of the channel Hmmeasurement height usfree surface velocity us,mfree surface velocity in the middle of the channel umax maximal free surface velocity ¯umean flow velocity ¯unmean velocity of a Nusselt film flow uloc local free surface velocity ymax y-location of the maximal free surface velocity 72
Bibliography [29] Heining, C. Velocity field reconstruction in gravity-driven flow over unknown topography. Physics of Fluids 23(3), 032101 (2011). [30] Heining, C., Sellier, M., and Aksel, N. The inverse problem in creeping film flows. Acta Mechanica 223, 1–7 (2012). [31] Pozrikidis, C. The flow of a liquid film along a periodic wall. Journal of Fluid Mechanics Digital Archive 188, 275–300 (1988). [32] Scholle, M., Rund, A., and Aksel, N. Drag reduction and improvement of material transport in creeping films. Arch. Appl. Mech 75, 93–112 (2006). [33] Wierschem, A., Scholle, M., and Aksel, N. Vortices in film flow over strongly undulated bottom profiles at low Reynolds numbers. Physics of Fluids 15(2), 426–435 (2003). [34] Scholle, M., Wierschem, A., and Aksel, N. Creeping films with vortices over strongly undulated bottoms. Acta Mechanica 168(3), 167–193 (2004). [35] Wierschem, A. and Aksel, N. Influence of inertia on eddies created in films creeping over strongly undulated substrates. Physics of Fluids 16(12), 4566–4574 (2004). [36] Scholle, M., Haas, A., Aksel, N., Wilson, M. C. T., Thompson, H. M., and Gaskell, P. H. Competing geometric and inertial effects on local flow structure in thick gravitydriven fluid films. Physics of Fluids 20(12), 123101 (2008). [37] Trifonov, Y. Y. Viscous liquid film flows over a periodic surface. International Journal of Multiphase Flow 24(7), 1139–1161 (1999). [38] Zhao, L. and Cerro, R. Experimental characterization of viscous film flows over complex surfaces. International Journal of Multiphase Flow 18(4), 495–516 (1992). [39] Bontozoglou, V. and Papapolymerou, G. Laminar film flow down a wavy incline. International Journal of Multiphase Flow 23(1), 69–79 (1997). [40] Bontozoglou, V. Laminar film flow along a periodic wall. Computer Modelling in Engineering & Sciences 1, 133–142 (2000). [41] Wierschem, A. and Aksel, N. Hydraulic jumps and standing waves in gravity-driven flows of viscous liquids in wavy open channels. Physics of Fluids 16(11), 3868–3877 (2004). [42] Wierschem, A., Bontozoglou, V., Heining, C., Uecker, H., and Aksel, N. Linear resonance in viscous films on inclined wavy planes. International Journal of Multiphase Flow 34(6), 580–589 (2008). [43] Heining, C., Bontozoglou, V., Aksel, N., and Wierschem, A. Nonlinear resonance in viscous films on inclined wavy planes. International Journal of Multiphase Flow 35(1xperim), 78–90 (2009). [44] Wierschem, A., Pollak, T., Heining, C., and Aksel, N. Suppression of eddies in films over topography. Physics of Fluids 22, 113603 (2010). 79
Bibliography [45] Nguyen, P. K. and Bontozoglou, V. Steady solutions of inertial film flow along strongly undulated substrates. Physics of Fluids 23, 052103 (2011). [46] Benjamin, T. B. Wave formation in laminar flow down an inclined plane. Journal of Fluid Mechanics 2(06), 554–573 (1957). [47] Yih, C.-S. Stability of liquid flow down an inclined plane. Physics of Fluids 6(3), 321–334 (1963). [48] Liu, J., Paul, J. D., and Gollub, J. P. Measurements of the primary instabilities of film flows. Journal of Fluid Mechanics 250, 69–101 (1993). [49] Liu, J. and Gollub, J. P. Solitary wave dynamics of film flows. Physics of Fluids 6(5), 1702–1712 (1994). [50] Vlachogiannis, M. and Bontozoglou, V. Observations of solitary wave dynamics of film flows. Journal of Fluid Mechanics 435, 191–215 (2001). [51] Chang, H. C. and Demekhin, E. A. Complex wave dynamics on thin films. Elsevier, Amsterdam, (2002). [52] Craster, R. V. and Matar, O. K. Dynamics and stability of thin liquid films. Rev. Mod. Phys. 81(3), 1131–1198 (2009). [53] Wierschem, A. and Aksel, N. Instability of a liquid film flowing down an inclined wavy plane. Physica D: Nonlinear Phenomena 186(3-4), 221–237 (2003). [54] D´avalos-Orozco, L. A. Nonlinear instability of a thin film flowing down a smoothly deformed surface. Physics of Fluids 19(7), 074103 (2007). [55] D´avalos-Orozco, L. A. Instabilities of thin films flowing down flat and smoothly deformed walls. Microgravity Science and Technology 20(3), 225–229 (2008). [56] Benney, D. J. Long waves on liquid films. Journal of Mathematical Physics 45(2), 150–155 (1966). [57] Trifonov, Y. Y. Stability and nonlinear wavy regimes in downward film flows on a corrugated surface. Journal of Applied Mechanics and Technical Physics 48(1), 91–100 (2007). [58] Trifonov, Y. Y. Stability of a viscous liquid film flowing down a periodic surface. International Journal of Multiphase Flow 33, 1186–1204 (2007). [59] D’Alessio, S. J. D., Pascal, J. P., and Jasmine, H. A. Instability in gravity-driven flow over uneven surfaces. Physics of Fluids 21(6), 062105 (2009). [60] Heining, C. and Aksel, N. Effects of inertia and surface tension on a power-law fluid flowing down a wavy incline. International Journal of Multiphase Flow 36(11-12), 847–857 (2010). [61] Vlachogiannis, M. and Bontozoglou, V. Experiments on laminar film flow along a periodic wall. Journal of Fluid Mechanics 457, 133–156 (2002). 80
Bibliography [62] Argyriadi, K., Vlachogiannis, M., and Bontozoglou, V. Experimental study of inclined film flow along periodic corrugations: The effect of wall steepness. Physics of Fluids 18(1), 012102 (2006). [63] Scholle, M. and Aksel, N. An exact solution of visco-capillary flow in an inclined channel. Zeitschrift f¨ ur Angewandte Mathematik und Physik 52(5), 749–769 (2001). [64] Hopf, L. Turbulenz bei einem Flusse. Annalen der Physik 337(9), 777–808 (1910). [65] Scholle, M. and Aksel, N. Thin film limit and film rupture of the visco–capillary gravity–driven channel flow. Zeitschrift f¨ ur Angewandte Mathematik und Physik 54(3), 517–531 (2003). [66] Haas, A., Pollak, T., and Aksel, N. Side wall effects in thin gravity–driven film flow - steady and draining flow. Physics of Fluids 23, 062107 (2011). [67] Leontidis, V., Vatteville, J., Vlachogiannis, M., Andritsos, N., and Bontozoglou, V. Nominally two-dimensional waves in inclined film flow in channels of finite width. Physics of Fluids 22(11), 112106 (2010). [68] Vlachogiannis, M., Samandas, A., Leontidis, V., and Bontozoglou, V. Effect of channel width on the primary instability of inclined film flow. Physics of Fluids 22(1), 012106 (2010). [69] Georgantaki, A., Vatteville, J., Vlachogiannis, M., and Bontozoglou, V. Measurements of liquid film flow as a function of fluid properties and channel width: Evidence for surface-tension-induced long-range transverse coherence. Phys. Rev. E 84(2), 026325 (2011). [70] Pollak, T., Haas, A., and Aksel, N. Side wall effects on the instability of thin gravitydriven films – From long-wave to short-wave instability. Physics of Fluids 23, 094110 (2011). [71] Tropea, C., Yarin, A., and Foss, J. F., editors. Springer Handbook of Experimental Fluid Mechanics, p. 681. Springer, Berlin, (2007). [72] Happel, J. and Brenner, H. Low Reynolds number hydrodynamics. Martinus Nijhoff Publishers, Dordrecht, (1983). [73] Spurk, J. H. and Aksel, N. Fluid Mechanics. Springer, Berlin, 2nd edition, (2008). [74] Aksel, N. and Schmidtchen, M. Analysis of the overall accuracy in ldv measuremet of film flow in an inclined channel. Meas. Sci. Technol. 7, 1140–1147 (1996). [75] Schmid, P. J. and Henningson, D. S. Stability and Transition in Shear Flows, volume 142. Springer, New York, (2001). [76] Drazin, P. G. Introduction to Hydrodynamic Stability. Cambridge University Press, Cambridge, (2002). 81
[77] Taneda, S. Experimental investigation of the wakes behind cylinders and plates at low Reynolds numbers. Journal of the physical society of Japan 11(3), 302–307 (1956). [78] Bohr, T., Ellegaard, C., Hansen, A. E., and Haaning, A. Hydraulic jumps, flow separation and wave breaking: An experimental study. Physica B: Condensed Matter 228(1-2), 1–10 (1996). [79] Bohr, T., Putkaradze, V., and Watanabe, S. Averaging theory for the structure of hydraulic jumps and separation in laminar free-surface flows. Physical Review Letters 79(6), 1038–1041 (1997). [80] Bush, J. W. M. and Aristoff, J. M. The influence of surface tension on the circular hydraulic jump. Journal of Fluid Mechanics 489, 229–238 (2003). [81] Lugt, H. J. Vortex flow in nature and technology. G. Braun, New York, (1983). [82] Dyke, M. V. An Album of Fluid Motion. The Parabolic Press, Stanford, (1997). [83] Matlab®2010b The Mathworks, Inc. [84] OpenFOAM 1.6. The open source CFD Toolbox, OpenCDFD Ltd. Reading, Berkhire, UK, (2009). See http://www.openfoam.com for more information. [85] Nyquist, H. Certain topics in telegraph transmission theory. Trans. IEEE 47, 617– 644 (1928). [86] Schubauer, G. B. and Skramstad, H. K. Laminar-boundary-layer oscillations and transition on a flat plate. J. of the Aeronautical Sciences 14, 69–78 (1947). [87] Schlichting, H. and Gersten, K. Boundary-layer theory. Springer, Berlin, 8th edition, (2003). 82
Lebenslauf Thilo Pollak Geboren am 26. Dezember 1982 in Gummersbach (NRW) Familienstand: Ledig Staatsb¨ urgerschaft: Deutschland Sprachen: Deutsch, Englisch Parsifalstrasse 27a 95445 Bayreuth Tel.: 0172 9569022 E-mail: [email protected] Schulbildung 08/89 - 07/93 Grundschule Steinenbr¨ uck bei Gummersbach 08/93 - 06/99 St¨ adt. Gymnasium Moltkestrasse in Gummersbach 07/99 Umzug nach Hirschaid (Bayern) 09/99 - 06/02 Dientzenhofer–Gymnasium Bamberg (Bayern) Abschluss: Abitur (Leistungskurse: Mathematik und Physik) Studium 10/03 - 12/08 Physikstudium an der Universit¨ at Bayreuth (Bayern) Abschluss Diplom Physik (Thema der Abschlussarbeit: ”L¨ angenskalenabh¨ angigkeit der Diffusion im bin¨ aren Glasbildner Polystyrol/Toluol“) 05/09 - heute Promotionstudium am Lehrstuhl f¨ ur Technische Mechanik und Str¨ omungsmechanik an der Universit¨ at Bayreuth Sonstige T¨ atigkeiten 07/99 - 08/99 Ferienarbeit bei INA Motorenwerke in Hirschaid 08/01 - 09/01 Ferienarbeit bei BOSCH in Bamberg 07/02 - 03/03 Grundwehrdienst in Roth (Bayern), Lenggries (Bayern) und Leipheim (Bayern) 10/05 - 09/07 Gew¨ ahlte Studentenvertretung Fachbereich Mathe/Physik/Informatik der Universit¨ at Bayreuth 03/09 - 04/11 Gr¨ undungsmitglied und 2. Vorsitzender des Absolventen- und F¨ ordervereins Mathe/Physik/Informatik Uni Bayreuth e.V. 19. April 2012, Thilo Pollak 83
Danksagung Mein Dank gilt allen Mitarbeitern des Lehrstuhls f¨ ur Technische Mechanik und St¨ omungsmechanik der Universit¨ at Bayreuth die mich bei der Erstellung dieser Arbeit tatkr¨ aftig unterst¨ uzt haben. Hier m¨ ochte ich mich insbesondere beim Lehrstuhlinhaber Herrn Prof. Dr. Nuri Aksel bedanken, der meine Arbeit ¨ uber den gesamten Zeitraum sehr geradlinig betreut hat, mir stets hilfsbereit zur Seite stand und sich immer Zeit f¨ ur Diskussionen ¨ uber offene Fragestellungen nahm. Regelm¨ assig r¨ aumte er mir auch die M¨ oglichkeit ein, an diversen internationalen Fachtagungen teilzunehmen bei denen ich sehr interessante Kontakte zu Forschungskollegen aus der ganzen Welt aufbauen und vertiefen konnte. Außerdem hervorzuheben sind meine Kollegen Dr. Andr´e Haas, Dr. Christian Heining und Dr. Riccardo Puraglesi die mir durch zahlreiche Fachdiskussionen h¨ aufig helfen konnten neue Ideen und L¨ osungsans¨ atze zu erarbeiten. Großer Dank gilt insbesondere auch Carola Lepski und Marion M¨ arkl, da durch ihre unterst¨ utzende Laborarbeit ein z¨ ugiges und effektives wissenschaftliches Arbeiten ¨ uberhaupt erst erm¨ oglicht wurde. Weiterhin m¨ ochte ich Dr. Lutz Heymann und Katja Helmrich f¨ ur ihre bereitwillige Hilfe nicht nur bei organisatorischen Fragestellungen einen großen Dank aussprechen. Selbstverst¨ andlich haben dar¨ uber hinaus auch alle anderen Mitarbeiter des Lehrstuhls zu einer Arbeitsatmosph¨ are beigetragen, die ich immer als sehr angenehm und kollegial empfand. Nicht zuletzt m¨ ochte ich mich bei meinen Eltern bedanken, die mich w¨ ahrend meiner gesamten Ausbildung uneingeschr¨ ankt unterst¨ utzt haben und mir so jeden nur erdenklichen Freiraum geschaffen haben um mein Studium und das Verfassen dieser Arbeit erfolgreich abzuschließen. 84
Selbstst¨ andigkeitserkl¨ arung Hiermit erkl¨ are ich, dass die vorliegende Dissertation ohne fremde Hilfe eigenst¨ andig und ausschließlich unter Verwendung der angegebenen Hilfsmittel und Quellen verfasst wurde. Ich versichere, dass ich die vorliegenden Arbeit erstmalig einreiche und keine fr¨ uheren Versuche einer Promotion unternommen habe. (Dipl.-Phys. Thilo Pollak) Bayreuth, 19. April 2012 85