Introduction to Vortex Theory by Hans J. Lugt Vortex Flow Press Potomac, MarylandDr. Hans J. Lugt was a Senior Research Physicist and Consultant at the Naval Surface ‘Warfare Center, Carderock Division, (formerly David Taylor Research Center) in ‘West Bethesda, Maryland 20817-5700, until his retirement in 1995. ' Copyright © 1996 by Hans J. Lugt Published by Vortex Flow Press, Inc. P.O. Box 61414, Potomac, Maryland 20859-1414, Fax No. (301) 983-3843. Printed in The United States of America by Hagerstown Bookbinding & Printing Co., Inc. P.O. Box 190, 952 Frederick Street, Hagerstown, Maryland 21741. Fax No. (301) 733-6586. Library of Congress Catalog Card Number: 97-90428 ISBN 0-9657689-0-2Ef Preface Introduction to Vortex Theory is intended as a textbook for graduate students of fluid dynamics. It evolved from a course of the same title given at the George Washington University in the spring semester of 1988. Vortex Theory is part of the larger discipline of fluid dynamics. What justifies a course in such a specialized field? It is the tremendous role that vortices play in fiuid dynamics: the development, maintenance, and interaction ‘of ordered flow structures which efficiently transport quantities like mass, momentum, and energy. These vortical structures are particularly important in the understanding of turbulent flows, which contain large-scale coherent patterns in an environment of irregular motion. Moreover, vortices serve as a paradigm for the development of dynamic pattems in general since their theory is so well advanced. This rather lengthy description of the role of vortices reflects the complexity of the subject area. Dietrich Kiichemann’s well-known metaphor that ‘‘vortices are the sinews and muscles of fluid motion’? captures the essence of vortices and may justifiably even include the ‘‘skeleton’” (another metaphor which is sometimes used for three-dimensional vortex filaments) and the “‘voice of the flow’ that reflects the role of vortices in aeroacoustics. The importance of vortices and their role in fluid dynamics will be thorougly covered in this book. The increasing interest in vortex theory during the last two decades is closely linked to the availability of large computers and to a lesser extent to experiments based on the non-invasive technique of Laser Doppler Velocimetry (LDV). Before that time, the study and teaching of fluid dynamics covered essentially the theories of potential flow, inviscid fluid flow, boundary-layer flow, and slow motion, areas which permitted the truncation of the equations of motions, the so-called Navier-Stokes equations, to soluble partial differential equations. With the advent of high-speed computers it became possible to solve the nonlinear Navier-Stokes equations numerically for moderate Reynolds numbers and thus to investigate the nature of vortices: LDV made it possible to study vortices without destroying or even distorting the flow field to be measured and to complement and verify the numerical results. The study of vortices has also stimulated and embraced mathematical disciplines which are indispensible for the understanding of the nature of vortices: topology and nonlinear theories like chaos and bifurcation. In this context and in the treatment of vortices in general, the concept of vorticity plays such an essential role that an acquaintance with it is mandatory. The significance of vortices goes far beyond fluid dynamics, for vortices play an important role not only in other areas of science and engineering, but also in mythology, history, art, and philosophy. An overview of these aspects of vortices was - iii -presented in Vortex Flow in Nature and Technology, Wiley-Interscience 1983 and reprinted by Krieger Publishing Co. 1995. This textbook is addressed to students with a background in fluid dynamics and mathematics. It is organized in a format that conveys the function of vorticity in the generation, behavior, and decay of vortices, the role of vortical patterns in fluid flow, and the importance of vortices in engineering, biology, meteorology, and oceano- graphy, with preliminary remarks on the semantics of the words vortex and vorticity. The greatest difficulty in writing this book was the selection of material from the immense literature. owe thanks to many friends and colleagues. Acknowledgement is especially due to Profs. K. Biihler, University of Karlsruhe; P. Freymuth, University of Colorado; H. Mueller, American University; P. Orlandi and R. Verzicco, Universita di Roma; W.R.. Phillips, Clarkson University; A. Powell, University of Houston; T. Sarpkaya, Naval Postgraduate School; Drs. Ch. Bricker, Technische Hochschule Aachen; G.S. Copeland, J. Gorski, and H.J. Haussling, Naval Surface Warfare Center, Carderock; U. Dallmann, Deutsche Forschungsanstalt fiir Luft- und Raumfahrt, Gottingen; L. Schmid, National Institute of Standards and Technology; and A. Weigand, California Institute of Technology. I am pleased to specially mention Prof. F. Hussain, Univer- sity of Houston, who went over most parts of the book. Particular gratitude goes to Mrs. Alice Phillips, who has helped to improve the readability of many publications of mine for more than two decades and made valuable improvements in this text. The book is dedicated to my wife Anneliese, whose encouragement and support have been indispensable. Hans J. Lugt Potomac, Maryland, spring 1995 -iv-Contents Frequently Used Symbols Preliminary Remarks 1. Vortices as Flow Patterns Ll. =. 13. 14. 15. 1.6. Flow Patterns in Lagrangian and Bulerian Frames Definition of Vortices and their Role in Fluid Motion Streamline, Pathline, Timeline, and Streakline ‘Topological Aspects of Flow Patterns ‘The Invariance of Vorticity Fields Patterns which Exist only on the Average 2. Basic Concepts and Laws » 24. 22. aa 2.4. 25. 2.6. 2.7. 3.1. 3.2. ae 3.4. 3.5. 3.6. The Modeling of Fluid Motion Three Theorems of Vector Fields Conservation Laws Constitutive Equations The Basic Equations of Motion Initial and Boundary Conditions Conservation Law of Angular Momentum |. The Navier-Stokes Equations Various Forms and Properties Basic Equations in Different Coordinate Systems ‘An Example of Local Boundary Conditions The Concept of Similarity Basic Equations in Dimensionless Form ‘Truncated Versions of the Navier-Stokes Equations 4. Symmetric Solutions of the Navier-Stokes Equations for Vortices 41. 4.2. 43. 44. Universal Solutions Similarity Solutions for Decaying Vortices Vortices Perpendicular to a Slip Surface Assessment xiv Babuae 21 es | 22 26 30 32 BE) 36, 39 39 43 49 51 53 2 59 65 76 855. a 2 2 2 Vorticity 5.1. The Vorticity-Induction Equation 5.2. Vorticity-Transport Equation 5.3. Vorticity Theorems of Helmholtz 5.4, Vorticity Generation 5.5. Maximum Principles 5.6. Forces Acting on a Body 5.7. Circulation 5.8. Summary Potential Flow I 6.1. General Properties 6.2. Bernoulli’s Theorems 6.3. Complex-Function Theory Applied to Two-Dimensional Flows 6.4. Two-Dimensional Wing Theory 6.5. Quantized Vortices in Superfluid Helium Potential Flow II 7.1, Two Point Vortices 7.2. Symmetric Systems of Point Vortices 7.3. Excursus into Classical Mechanics 7.4, The Hamiltonian System of Point Vortices 7.5. Chaotic Motion of Point Vortices 7.6. Point Vortices near a Solid Boundary Potential Flow III 8.1. Discontinuity Lines 8.2. Vortex-Wake Models for Wings 8.3. The Roll-Up of a Discontinuity Line 8.4. Conical Vortex Sheets 8.5. Vortex Lines in Three Dimensions 8.6. Assessment Inviscid Vortices I 9.1. Integral Invariants in Two-Dimensional Flow 9.2. Elliptical Vortices 9.3. Nonuniform Vortices of Constant Shape 9.4. Vortices of Deformable Shape -vi- 89 92 96 101 107 il a 115 119 119 122 123 131 134 137 137 142 147 151 156 158 165 166 169 174 184 188 192 ee 196. 197 202 20310. Inviscid Vortices II 10.1. Integral Invariants in Three-Dimensional Flow 10.2. Vortex Rings 10.3. The Local Induction Equation 10.4. Vortex Filaments Moving with Constant and Deformable Shapes 10.5. Vortex Filaments with Axial Velocity 10.6. Assessment 11. Instability 11.1. The Concept of Instability 11.2. Linear Instability of Inviscid Flows 11.3. Nonlinear Instability of Inviscid Flows 11.4. Unstable Viscous Shear Layers 11.5. Taylor-Couette Flow 11.6. Self-Rotation 11.7. Concluding Remarks 12. Flow Separation and Vortex Generation 12.1. Two-Dimensional Flow Separation 12.2. Preliminary Remarks on Three-Dimensional Flow Separation 12.3. Three-Dimensional Flow Near Singular Points on a Nonslip Surface 12.4, Three-Dimensional Flow Near Singular Points on a Slip Surface 12.5. Flow Topology of a Finite Region 12.6, Towards a Vortex Definition in a Viscous Fluid 13. Properties of Viscous Vortices I 13.1. Stokes Vortices 13.2. Vortices Attached to a Surface 13.3, Vortex Shedding 14, Properties of Viscous Vortices II 14.1. Viscous Vortex Pairs and Rings 14.2. Vortex Merging, Splitting, and Reconnection 14.3. Two Examples of Three-Dimensional Vortices on a Body Surface 14.4, The Complexity of Vortex Patterns 15. Vortices near a Boundary 15.1. Two-Dimensional Vortices near a Nonslip Surface -vii- 214 rah 228 231 238 241 245 246 249 256 260 265 275 277 279 280 286 290 — 300 304 309 310 319 325 339 ee 344 350 356 361 36115.2. Two-Dimensional Vortices near a Free Surface 15.3, Vortex Rings Normal to a Surface 15.4, Vortex Rings approaching a Surface Obliquely 15.5. Rotating Fluids Normal to a Nonslip Surface 16. Swirling Motion 16.1. Inviscid Swirling Motion with Straight Axial Flow 16.2. Viscous Swirling Motion with Straight Axial Flow 16.3. Waves in Swirling Motion 16.4. Vortex Breakdown 16.5. Rotating Fluids in a Container 16.6. Intake Vortices 17. Vortex Shedding from Oscillating and Rotating Bodies, Vortex Sound 17.1. Preliminary Remarks on Pendulum Motion 17.2. The “‘Black-Box’* Approach 17.3. Autorotation of Plates 17.4. Vortex-Induced Oscillation 17.5. Vortex Sound 18. Remarks on Turbulent Vortices 18.1. Some Basic Concepts 18.2. The Hypothesis of Eddy Viscosity 18.3, Turbulent Vortex Rings 18.4, Coherent Structures 18.5. Closing Remark 19. Vortices in a Rotating Frame . Equations of Motion in a Rotating Frame Hyperbolicity and Taylor-Proudman Theorem .3. Inviscid Vortices in a Rotating Frame . Large-Scale Circulation 20. Miscellaneous Topics 20.1. Symmetry and Conservation of Vorticity 20.2. Generalized Vorticity Equations for Viscous Fluid Flows 20.3. Stratification and Buoyancy 20.4. Cellular Convection 363 375 377 — 403 442 447 451 459 467 468 472 477 480 487 489 489 490 499 509 509 ao 520 532Epilogue 543 Appendix A: Useful Relations among Differential Operators, Integral Formulas 547 Appendix B: Classification of Singular Points 549 References a Subject Index 599. ~ix-a,b,c A,B,C Sporn s coe 8 So we lL 00 ty mp EAR Qo Q Frequently Used Symbols arbitrary constant, distance initial position vector components of @ (Lagrangian coordinates), arbitrary constants area, matrix arbitrary constants, Lagrangian multipliers vector potential unit binormal vector flexion potential magnetic induction specific heat, concentration velocity of sound lift coefficient torque coefficient diameter, drag, dissipation deformation tensor =-Dy internal energy, enstrophy exponential integral electric field Eckert number surface function, arbitrary function, Coriolis parameter mean value of f fluctuating part of f frequency arbitrary scalar function external force Froude number gravitational constant, arbitrary function scalar function with dG/dt = 0 Grashof number height, helicity density, Planck’s constant, arbitrary functionx cya a TR hp ds ot SOR as* ® BY aes Bele PS “oO _ ores Fy Hamiltonian, helicity, total head, height Hamiltonian in a rotating system magnetic field Cartesian unit vectors action integral fluid impulse unit tensor modified Bessel function electric current density Jacobian Bessel function, action variable thermal conductivity thermal diffusivity = k/c modified Bessel function mixing length element of vortex line direction cosines natural logarithm length, lift, Lagrangian function Lamb vector = @xV mass, source strength, integer torque integer unit normal vector integer, Brunt-Vaisil frequency moment of impulse pressure, roll parameter generalized momentum integration constants canonically conjugate variables Prandtl number heat flux, acoustic source term generalized coordinate generalized velocity volume flux oxi-ar 10,2 ROA R Ra Re Pawan oaaae ® 4,0, u,0,0 uV,Ww radius position vectors cylindrical polar coordinates spherical polar coordinates radius of curvature Rayleigh number Reynolds number Rossby number arc length =7-?" entropy =pgv:8 Strouhal number time : surface force, unit tangent vector temperature, kinetic energy characteristic temperature stress tensor Taylor number velocity components in Cartesian coordinates average velocity components constant velocity components velocity component normal to a surface velocity volume fixed in space, potential energy volume moving with the fluid complex potential = 6 + iy Weber number Cartesian coordinates position vectors force components Bessel function saxty =x-iy angle of attack, expansion coefficient = dffdy (f Coriolis parameter), = - 3T/z xii -1 4 a gums They have invariance properties and are intimately related to vortices. We shall dwell on this thought in the text that follows. Let us retum to the vortex concept. We want to use it in a scientific context. After all, the title of this book refers to a ‘‘vortex theory”’. The linguistic definition cited above implicitly contains features that are obvious to a non-scientific observer, but these features gain special importance and need mentioning in a scientific frame: (1) A vortex occupies a finite space and, while we speak of a “‘large-scale’’ flow structure or a flow pattern, the vorticity vector is defined at a point. (2) In the linguistic definition the notion of a material particle is not specified. In order to relate the vorticity concept (or any other mathematically defined quantity) to that of the vortex, we must make the vortical flow field compatible with the vorticity, that is, the vortex must be modeled as a motion in a continuum. Then we can show rigorously (Chapter 6) that a vortex requires vorticity for its existence. We may also say metaphorically that vorticity is the building block for vortices. The reverse statement, however, that a vorticity field represents a vortex, is not true. A simple counter-example is the straight shear layer which has vorticity but is not a vortex. Although we have given the term vortex a scientific look by describing it with exact physical terms such as continuum and vorticity, its definition is still not precise because so far it has defied a mathematical formulation. Nevertheless, a quotation from Lord Rayleigh® may justify such a deficiency: ‘I hold that if the introduction of a quantity promotes clearness of thought, then even if at the moment we have no means of determining it with precision, its introduction is not only legitimate but desirable.” 5. Lighthill, MJ, in ‘Laminar Boundary Layers,” ed. L. Rosenhead. Oxford University Press 1963. 6, Lord Rayleigh, “The Life of Sir J.J. Thomson.” Cambridge University Press 1942. - xvii -A relation between vortex and vorticity makes it easy to study mathematically the generation, behavior, and decay of vortices. Since these processes are nonlinear and, hence, capable of abrupt, drastic changes within a field, and since they reveal the detailed generation of patterns from certain unstructured initial conditions (‘‘self- organization’’), their study is important for other branches of physics, in which order prevails over chaos:” Vortex theory has ‘become a paradigm for pattern development in general. ne 7. Prigogine, I. ‘“From being to becoming.”” Freeman, San Francisco, 1980. i1. Vortices as Flow Patterns 1.1. Flow Patterns in Lagrangian and Eulerian Frames We observe patterns of fluid motions in the form of waves and vortices.* In this book we are dealing with vortical patterns which include the generation, behavior, and decay of a single vortex as well as a system of vortices. We also shall see that vortices can sustain waves so that a combination of vortex and wave patterns can occur. Patterns are here defined as geometric representations of certain charac- teristics of a flow field that contain regular features. Important examples are patterns of streamlines, which are an assembly of streamlines, and patterns of velocity, which are direction fields of the velocity vector. Before we continue to dwell on this idea of patterns as indicators of certain flow Properties, let us spend a few thoughts on the concept of patterns which are more general than we have just defined. We can observe patterns immediately through our senses, for instance, visual patierns in the form of certain geometric arrangements or acoustic patterns in the form of tonal successions. Patterns can also result from an intellectual process, for instance, the plotting of curves in phase space. Patterns confront us as factual observations without deeper understanding. However, patterns as manifestations of regularities suggest an underlying order or design. Since the patterns in which we are interested are of a general physical nature, this order reflects physical laws. Physical laws themselves can be inferred from symmetry properties which are in the general sense invariance properties, Here are some examples from geometry and physics that have invariance properties: The geometrical rotation of a figure is an invariant transformation, the class of ellipses is topologically invariant, and the conservation laws of classical mechanics are invariant in time and space. Patterns produced by the movement of fluids are different from the static patterns of ice crystals and gem stones. Particles or molecules in a solid have a permanent place, but in fluids they move. The paths of the particles form patterns of flux or flow patterns. In a burning candle the visible part of the flame, although it may flicker, maintains its shape, while the journeys of the individual molecules and their chemical transformation in the flame are fleeting episodes. The last sentence contains the idea that fluid motions can be described essentially in two ways, depending on the kind of information we want: (1) by following the * Other patterns like those due to density changes also occur, but they are, for the purpose of this. book, of less concern than waves and vortices.paths of the individual particles and (2) by describing the velocity field in a certain reference frame, for instance, one that is fixed in space. The latter method is appealing since it provides pattems with a certain permanency as indicated in the shape of the flame in the example cited above. This possibility implies also the profound philosophical idea that even in flux certain features can be unchanging. On the other hand, the second method of describing only the velocity field is not as general as the first one which keeps a record of each individual particle.* Before we express this distinction mathematically, we must decide which model we want to use for the fluid, In fluid dynamics it is not practical in most cases to describe the fluid as an assembly of particles; rather the fluid is idealized as a continuum and the particle itself is replaced by a fluid element. This means that the medium is considered infinitely divisible (in contrast to quantized unities like atomic or molecular particles). A fluid element is then that part of the continuum which occupies the infinitesimal volume dV. By assigning a value of a physical quantity to each fiuid element in the continuum we obtain a “‘field’” of this quantity. The method of describing moving individual fluid elements is called Lagrangian and corresponds to the description of a Newtonian mass point except that the mass point is here replaced by a fluid element. This replacement is crucial in a continuum theory, since the new model has a higher level of sophistication than the model of a mass point because the volume element can deform and rotate (see Section 2.2). The full implication of this replacement will be recognized when we discuss the basic physical laws. The concept of a particle as a mass point, however, will not be aban- doned. Point singularities will be permitted which can play the role of “‘particles’” under certain situations (This concept of a particle, of course, has nothing to do with that on the atomic level). The other way, that of describing the velocity field in a frame fixed in space or fixed to a suitable reference frame, is called “Eulerian’’. A volume element in such a frame is designated by dV (in contradistinction to a moving Lagrangian fluid element dV). Mathematically, the difference between Lagrangian and Eulerian frames can be expressed in the following way (Prandtl and Tietjens 1934): In a Lagrangian frame the individual fluid elements must be identified by names. Each must be labeled by means of three parameters a, b, c which can be the position vector @ with its three components a, b, c at the initial time t= 0. The position 7 of the fluid element is then a function of @ and t: 7= 7°(@, t). The velocity is given by * The mathematical foundation for the reduced description of a flow field is outlined by Salmon (1988). This reduction is connected with the fact that individual particles are indistinguishable by the description of the velocity field only or, that the flow field has “‘relabeling symmetry” (Chapter 20). iP= IP@ t/t. In the Bulerian frame the velocity is defined at the position? fixed in space: T= VY, t) = d?/dt. Now 7 (with x, y,z its Cartesian coordinates) is an independent variable. For an arbitrary function F(x, y, z, #) the derivatives with regard to time are related in both frames by: FOF OF. OF t* Ox Dar ors a (1.1) OE pt where the subscripts L and E denote Lagrangian and Eulerian, respectively. The velocity components u, 0, z are in the x, y, z- directions. The left side of Eq. (1.1) is called the material derivative of F. It is a total differential: aE dt Le. of +@-V)F. (1.2) The first term on the right side of Eq. (1.2) is the local time derivative, and the second term, written in vector form, is a nonlinear term in the Eulerian frame. Subscripts are omitted. Eq. (1.2) is valid also for any vector quantity which replaces the scalar function F. In a steady-state flow (or steady flow), defined as fluid motion unchanging with time, the local time derivative is zero. The concept of “steady-state” is, hence, only meaningful in an Eulerian frame. This statement is very important because it makes the Eulerian approach so attractive: It is possible to describe many fluid motions in patterns which do not change with time. Even in time-dependent flows the Eulerian method is helpful as it provides a means to study the change of flow patterns. This and the fact that fluid motions in a Lagrangian frame are difficult to compute are reasons that the Lagrangian description is not often used. However, under certain conditions, the following property of a Lagrangian scheme makes the Lagrangian description appealing: If the quantity F adheres to a fluid element, the Lagrangian time derivative or, according to Eq. (1.2), the total differential is zero: aE ace . (1.3) ‘This important relation enables us to follow the quantity F through the movement of the fluid element as a carrier of that quantity. Obviously, the label of a fluid element? must remain attached: . d@ “7 (4) With Eq. (1.3), which is a conservation law for F, the methods of Hamiltonian or classical mechanics can be exploited. This will be demonstrated in Chapter 7.| 4 1.2. Definition of Vortices and Their Role in Fluid Motion We now define a vortex as the rotation of fluid elements around a common center (Lugt 1983). These fluid elements cover a finite space so that a vortex is a macro- scopic or "large-scale" flow structure in contrast to the movement of a fluid element with the infinitesimal volume dV. .The center, around which the fluid elements rotate, does not need to be a line or a point but can be a finite region. Before we discuss this vortex definition further, let us address the following question: Why do vortices develop? The answer to this question is closely linked to the role vortices play in fluid motions. We may give three reasons for vortex generation that are not independent of each other: (1) Any motion in a closed region, whether the boundary of this region is rigid or fluid, must be of a rotatory nature. This follows from the conservation of mass and is immediately clear for an incompressible fluid, because fluid elements pushed away must be replaced by others. A vacuum is excluded. This is also true for a steady, compressible fluid flow. But in an unsteady, compressible fluid flow, temporal non- rotational (radial) motion can occur. Motion in closed region does not imply, however, that a well-defined center of rotation exists. We will show later in Chapters 7, 17, and 20 that fluid motions can be “‘chaotic,”” a state in which the paths of the fluid elements will be different after each orbit. Then, the meaning of center must be extended. (2) An accumulation of fiuid elements with vorticity, which is defined by B= curld= Vxv, (1.5) can lead to a vortex analogous to the accretion of mass particles in celestial bodies. Those fluid elements with vorticity are the building blocks of vortices. We will discuss the concept of vorticity at length in Chapter 5. (3) Vortices ‘can develop out of flow instability. This phenomenon enables the transition from one flow configuration to another (Chapter 11). If the efficient transport of mass, energy, and momentum can no longer be guaranteed by diffusion and simple convection alone, vortices appear which carry those flow quantities in concentratéd form in closed fluid pockets. The: motion in such pockets must be vortical according to item (1). These listed mechanisms of vortex generation, which together constitute the foundation for the development of vortical patterns (‘‘self-organization’’), will be analyzed in detail in the course of this book.‘The ubiquity of vortices in nature and technology may be indicated by a few examples which stand for the whole spectrum of vortices from tiny eddies on the microscopic scale to rotating systems of galaxies (see photograph on page 598): Quantized vortices in liquid helium Smallest turbulent eddies Vortices generated by insects Vortex rings of squids Bathtub vortex Dust whirl on streets Vortices in machinery Whirls behind weirs Wing-tip vortices Whirlpools in turbine intakes and in tidal currents Dust devils ‘Tornadoes and waterspouts Vortices shed from the Gulf Stream Hurricanes High- and low-pressure systems Ocean circulations General circulation of the atmosphere Convection cells inside the earth Planetary atmospheres Great Red Spot of Jupiter ‘Sun spots Rotation inside of stars Galaxies 1.3. Streamline, Pathline, Timeline, and Streakline Vortices, or flow patierns in general, can be displayed and categorized in many ways. Two methods of understanding flow properties are used in this book: The study of the solutions themselves, obtained by integration of the equations of motion, and the search for topological properties. In this section we will deal with the first method by looking at the various modes of displaying flow fields. This will lead us to the discussion of the limitation and deficiencies of the vortex definition given in Section 12. ‘We can compute or measure the velocity field and plot the direction field of the velocity vectors at selected points (Fig. 1.1). This is an Eulerian viewpoint and is expressed mathematically by UxdP= 0, (1.6) or in Cartesian coordinates by(7) Integration yields a streamline, which is a line tangent to the velocity vectors at a given time. Patterns of streamlines and patterns of velocity vectors thus describe the instantaneous state of flow. _——_— ee Fig. 1.1; Direction field of velocity vectors and streamlines. The position of a fluid element is given by r= 8@ Dp. (1.8) Al the positions between the initial time t= fo, at which the fluid element was at the location @, and the time f= t form a pathline for that time interval ty —to (Lagrangian viewpoint). Often the velocity V of a fluid element is given. Then, ‘ Pit)= [VE dt , tS tS ty. (1.9) ty For a fixed t= ty, on the other hand, Eq. (1.8) gives the change of the initial positions % of i fluid elements to the positions 7} at t= ty. In particular, if the initial positions form a continuous line or a line consisting of discrete elements, the new line at £= ty is’called a timeline (sometimes called a material line). For a given velocity field, Eq. (1.9) changes to h R@)= |P@, Hat. (1.10) to A simple example in a two-dimensional flow will illustrate this idea: At f= 0 a straight line across the vortex vg = rd /dt = I/r (which is called a potential vortex of unit strength) is considered. At t = f; the line has the shape: gr? = ty (Fig. 1.2). i i t qFig. 1.2: Timeline for a potential vortex (sketch). If the velocity field changes in time, we obtain pathlines that obviously no longer coincide with the streamlines. Only for steady flows do the fluid elements move along streamlines, and it follows that pathlines and streamlines coincide only in a steady, that is, time-independent flow. Timelines, however, are intrinsically time-dependent. Going back to the definition of a vortex we conclude that in steady flows closed or spiraling streamlines indicate vortices. The center of the vortex is the center of nested closed streamlines in two-dimensional flows and the center of spiraling (or, in special cases, of nested closed) streamlines in three-dimensional flows. Sometimes a time-dependent flow can be made time-independent by changing the position of the observer with respect to the motion of the body (Truesdell and Toupin 1960). This transformation is important because the time-invariance in the new reference frame makes it possible to reveal the presence of vortices. We will demonstrate this property in the text that follows. Fig. 1.3a shows the constant movement of a sphere through a fluid. The position of the observer is fixed in space and, as the sphere moves to the left, it passes the observer. The sphere pushes the fluid elements in front of it ahead as it moves, whereas it drags the fluid elements in the rear along with it. The instantaneous streamline patterns consist of closed curves, since far away from the sphere the fluid is at rest. A moment later the sphere is moved to the left. At this instant the streamlines look as before, but they are at a different location. Therefore, the streamlines change in space and time, and the flow is time-dependent (the flow in this example is considered very viscous). The pathlines are loops near the sphere (Fig. 1.3b). However, if the observer moves with the sphere, ot which is the same, if a parallel fiow is superposed, the flow becomes steady (Fig. 1.3c). The closed streamlines have vanished. Instead, they are parallel far away from the sphere, Of course, the conservation of matter is valid. The space in this case is no longer closed. Fluid enters from the left and leaves at the right. Since the flow is steady, pathlines and streamlines coincide.@ ) & © Fig. 1.3: (a) A sphere moves in a fluid at rest from right to left. The reference frame is fixed in the fluid at rest, that is, the sphere passes by the observer. (b) For case (a) the pathlines near the sphere form loops. They do not coincide with the streamlines. (c) The flow in case (@) becomes steady if the reference frame is fixed with the sphere or, which is the same, if a parallel flow from left to right is superposed in such a way that the sphere does not move relative to the observer. Fig. 1.3 reveals something strange: Although the physical process of the moving sphere does not change when the position of observation is altered, the streamlines (and pathlines) are completely different in the two reference frames. Since the frames move relative to each other with constant translational velocity, they are inertial frames. Thus the following theorem is true: Streamlines and pathlines aré not invariant when changing the inertial system. a q jSomething else in Fig. 1.3 is astonishing. Since the fluid far away from the sphere in Fig. 1.3a is at rest, the streamlines must be closed according to the conservation law of matter. However, they do not exhibit a vortex in the sense of the definition, because the fluid elements do not rotate about a common center (Fig. 1.3b). The steady-state case (Fig. 1.3) demonstrates more clearly that no vortex exists in the flow. On the other hand, a vortex is not always characterized by closed streamlines. Fig. 1.4a shows a vortex whose axis of rotation is at rest relative to an observer. If a parallel flow is superposed or if the observer moves relative to the center of the vortex, the streamlines open up to a wave (Figs. 1.4b and 1.4c). Zw — @) ) © Fig. 1.4: Ona vortex (a) a parallel flow is superposed. In (b) the velocity of the parallel flow is small with respect to the tangential velocity of the vortex; in (c) the velocity of the parallel flow is large. Fig. 1.5 shows an unsteady flow situation which cannot be changed to a steady state. Closed and wavy streamlines occur simultaneously. Behind a plate obliquely placed in a stream, vortices form with alternating direction of rotation (vortex street, see Chapter 13). The observer moves with the plate so that the streamlines far away from the plate are parallel, and the surface of the plate itself is a streamline. The closed separation region behind the front edge of the plate is a vortex since the body and separation region do not move relative to the reference frame (the separation region will deform, however, with time). The flow inside the separation region, however, is so weak that no streamline other than the one which marks the boundary is computed. To the right is a strong vortex with four closed streamlines. We recognize from the form of the neighboring streamlines that this vortex rotates counterclockwise and has a small translational motion relative to the plate (Fig. 15b). This vortex originated at the trailing edge of the plate. The third vortex rotates clockwise and has already moved away from the front edge of the plate. The difficulty of defining a vortex is evident. We note that the definition of a vortex as the rotation of fiuid elements around a common center has the great, disadvantage that it is valid only in a reference frame that does not move relative to the center of the vortex. In addition, Fig. 1.3 tells us that closed streamlines do not necessarily represent a vortex, and Fig. 1.4, that wavy streamlines may manifest a -9-vortex (but not always, as flows near a water surface show). The instantaneous streamline patterns are, therefore, not suited in general to define and identify a vortex ‘inction to steady-state flows). SNe Sa ) LG O- Fig. 1.5: (a) Streamlines around an oblique plate behind which a vortex street develops (computer-generated picture by Lugt and Havssling 1974). In this time-dependent flow three | vortices can be recognized whose movemebt relative to the plate is indicated in (b). Pathlines, like streamlines, are not invariant in general when the reference frame is changed. However, a reference frame can be sought in which closed or spiral pathlines are revealed (if they do in fact exist). Then, the very nature of pathlines, as obtained by a time-integration process (in contrast to instantaneous streamlines), ensures that the flow is observed in the reference frame fixed to the vortex over the period under consideration, Such closed or spiraling pathlines reveal the existence of vortices according to the definition. ‘Alternatively, a time sequence of instantaneous streamline patterns can be used in the following way: All possible inertial frames are checked to locate regions with closed streamlines. Then at another instant, the closed streamline regions in their respective reference frames are examined to see if their centers have moved. When the centers of such regions do not change (or barely change) with time, these regions represent vortices. Tn summary: The identification of a vortex in an, unsteady flow requires knowledge of the history of the motion. ‘We may point out that the definition of a vortex based on streamlines or pathlines is very general and is valid for any viscous fluid, as long as we are aware of the above -10-mentioned dependence on the reference frame. This shortcoming is actually not surprising as it is already implicit in the definition itself. The identification of a vortex requires that the center of rotation be fixed, that is, in an observer-dependent frame. Sometimes the concept of vorticity is used in the vortex definition since the vorticity is observer-invariant (Section 1.5), but this approach must also be treated with great caution (Chapter 13). Mathematically precise definitions exist for irrotational fluid flows (those without vorticity except at singularities, Chapter 6) and for rotational, inviscid fluid flows (frictionless flows with vorticity, Section 1.5 and Chapter 9). Attempts are under way to find also a vortex definition for viscous fluid flows (Section 12.6) ‘When we observe vortices in an experiment or in nature, we usually see neither streamlines nor pathlines. Streamlines cannot easily be made visible, and it is not always possible to follow the path of an individual fluid clement. We often see marked timelines like the spiral cloud bands of a low-pressure system. But what do we see when paint is injected into water over an interval of time at precisely the same spot and paint streaks are formed? These streaks are neither pathlines nor timelines but streaklines. We understand the difference among streamlines, pathlines, and streaklines most easily if we study the development of the streakline at various times. In Fig. 1.6 a paint particle, identified by a cross, is brought into the flow from a nozzle at time fp. At time #; this particle has traveled a certain distance, and a new paint particle, marked by a small circle, leaves the injection nozzle. At time f this particle also has left the nozzle and has covered a certain distance. Its path is different from that of the first particle at time ty because the flow has changed in the meantime. The pathlines of the individual paint particles are, therefore, different. If we take an instantaneous picture, for instance at time fz, we do not see the pathlines (dashed) but only the paint particles at time fs, which together form a streakline (solid line). In time-independent flows the pathlines do not change, and thus the statement can be made: In time-independent flows pathlines, streamlines, and streaklines coincide. = \ \ a yy = —~ et —— — >-- o 1 a 3 Fig. 1.6: A sketch for explaining the definition of a streakline (see text), This situation can be expressed mathematically by a Lagrangian method: At each instant a particle @ passes through the fixed point ’= 7, = rel #) or, when solved for@ @= Faw. In Fig. 1.6, the point? is the location of the nozzle. Over a time span, say from 0 to #, these particles @ form a streakline with the position (Prandtl and Tietjens 1934): -ll-P= CHL), 1. (uy So far we have studied means to display flow patterns, in particular vortical patterns, which we have obtained from computer solutions in the form of streamlines, direction fields of velocity vectors, pathlines, timelines, and streaklines (Freymuth 1993). If, on the other hand, flow patterns are given in the form of photographs or measured data, we must first determine the kind of flow lines and then analyze them. Satellite photos for weather forecast are an example. With some experience, we can also predict patterns to a certain extent for a given flow problem. In many situations an aerodynamicist knows what kind of flow configuration to expect behind an airfoil. The examples which we have discussed so far are those for essentially two- dimensional motions. Three-dimensional flow patterns can be very complex, and this subject will be taken up throughout the book with some preliminary ideas in the section that follows. 1.4. Topological Aspects of Flow Patterns ‘A more recent endeavor in vortex theory in the context of three-dimensional flows is to extract topological information from fluid motion, theoretically as well as experimentally (Moffatt and Tsinober 1990). We shall find topological statements throughout this book with various degrees of difficulty. In this section we are content with presenting some basic ideas which will require little knowledge of the material to be given later. In simple words we call topological those properties of a geometrical figure which are not affected by any deformation without tearing the figure apart or joining it with others. Or to be more precise, these properties are said to be topologically invariant under continuous transformation. For instance, a sphere which is deformed to an oblate spheroid is topologically invariant, whereas an oblate spheroid which becomes a ring is not. Examples of topological properties are connectivity, boundaries, and singular points. In contrast, geometrical properties are the shape and the size of a figure. The study of a flow field with its multitude of streamlines or pathlines can be cumbersome, particularly if the flow field is three-dimensional. Such a study can be simplified by finding so-called critical or singular points, where the velocity, the shear stress (defined in Section 2.3), or the vorticity vanishes (Appendix B). Thesé singular points enable us to obtain the essentials of a flow field by connecting singular points with lines which must not cross (Helman and Hesselink 1990). For this method exact ‘and detailed knowledge of the flow field is not necessary. The topological properties suffice. The analysis of singular points is not restricted to flow fields but can also be applied to other displays such as those in phase space. -12-‘There are various types of singular points on a two-dimensional surface which are classified in the following way. Locally we assume a flat surface at z= 0 in a Cartesian coordinate system x, y, z. On this flat surface, with the coordinates x and y, the velocity components are u and v, respectively. The location of the singular point is chosen, without loss of generality at x= y= 0 and the linear terms of the Taylor expansions are considered: u=axtby , v= cx+dy, (1.12) with a= (Qu/dx)o, b= (Ou/dy), c= (v/@x)p, and d= (@v/Ay)p. The differential equation for the streamlines is, according to Eq. (1.7), v_ dy cxtdy u dx ax+by c We introduce the Jacobian J= ad—be and the discriminant D = (a—d)? + 4be and distinguish the following types of singular points (Kaplan 1958):* Saddle point: J <0, D> 0, example: y au oc ee, x Nodal point: J>0,D>0 or D= 0, example: y= > y= Cx, Spiral point: [>0,D<0,a+d# 0, au . ata example: y= Y—* + VF +y? = Ce is y+x Center point: J>0,D<0,a+d= 0, example: y’=- ; > y= ‘The four types of singular points are illustrated in Fig. 1.7. Infinitely many lines go through a nodal point. The same holds for a spiral point except that it does not have a common tangent line. Only two lines pass through a saddle point, and none through a center point. If we interprete the lines as stream- + In most texts two types of singular points are distinguished, those characterized by positive (nodal) and negative (saddle) Jacobians (Appendix B). The other two are a subclass of the nodal type. Because of the importance of spiral and center points for vortices, the distinction of the singular points as four different types is here preferred. iselines, then spiral and center points are evidence that a vortex is present. This follows from the definition of a vortex in Section 1.2 and will be a basis in Section 12.6 for an attempt to find a mathematical vortex definition for a viscous fluid flow. KAS OY Fig. 1.7: The four types of singular points: (a) saddle, (b) nodal, (c) spiral, and (d) center points. The distinction between a spiral and a center point is important since it enables us to recognize local three-dimensional and two-dimensional flows, respectively. The condition a+d= 0 means that the vector field 7 with its components w and v is divergence-free: u/dx + du/éy = 0. The fluid flow cannot have a source or a sink and is, thus, two-dimensional, whereas a fluid with a source or a sink in the plane (u, 0) must have a three-dimensional structure. An example is furnished in Fig. 1.8. (a) (b) Fig. 1.8: (@) Streamlines in a two-dimensional flow inside a corner with a center point in the vortex and (b) symmetry plane with a spiral point of a flow around a cylinder on a flat surface. A= saddle separation point and B = saddle attachment point. On solid walls, where the fluid adheres and the velocity vanishes, the velocity ‘components and v in Eqs. (1.12) and (1.13) must be replaced by the gradients u/0z and dv/2z, that is, by the vorticity or the shear stress. However, itis still permissible to -14- : ' | i |speak of streamlines on the body surface in the sense of limiting streamlines. We can approach the surface to within an arbitrarily small distance, and there is still a velocity field from which streamlines can be constructed. Fig. 1.5 gives us an example of a two-dimensional flow field, in which singular points occur both inside the flow and at the body surface. We notice two center points, one clearly visible, the other in the separation pocket behind the leading edge that is not visible because the flow is so weak. There is also one unmarked saddle point above the second vortex. Four saddle points are at the body surface at which streamlines end or begin.* We distinguish between points which the flow approaches and points at which the flow leaves the body. They are called attachment and separation points, respectively. We observe a rather trivial topological property: the arrangement of attachment and separation points in a two-dimensional flow must alternate. In three-dimensional flows attachment and separation lines, so called bifurcation lines, can occur in addition to singular points. An example of singular points and bifurcation lines on the surface of a body in a steady three-dimensional flow is shown in Fig. 1.9 for the nose region of a cigar-shaped body. Such singular points and bifurcation lines are not restricted to the body surface but can occur in the symmetry planes of a fluid or, in fact, on any surface of interest. An example is furnished in Fig. 1.8a for a two-dimensional flow inside a comer and in Fig. 1.8b for the symmetry plane of a flow around an obstacle on a flat surface. In Fig. 1.8a the pivot of the vortex is a center point and in Fig. 1.8b a spiral point. Notice that the flow vanishes at the spiral point as it moves perpendicularly away from the plane in both directions. This behavior is a typical three-dimensional flow property in contrast to the two-dimensional flow in Fig. 1.8a. The distinction between a two-dimensional and a three-dimensional flow is important for the description of vortex patterns. More on this subject will be presented in Chapter 14. Fig. 1.9: Streamlines with singular points on a body surface. A = nodal attachment point, B = saddle point, and C = spiral separation point (Legendre 1965, reprinted by permission of La Recherche Aérospatiale), * These saddle points are actually ‘“half-saddle points" at the surface, because the other half inside the body has no physical meaning. ise1.5. The Invariance of Vorticity Fields We have briefly alluded in Section 1.3 to the fact that a vorticity field is invariant with respect to changes of the reference frame. We want to show this mathematically and demonstrate with an example the advantage of an invariant vorticity field over a velocity field. ‘The vorticity field is invariant with respect to the change of the inertial frame, or the vorticity field @ does not change if we superpose a constant translational velocity, say V. The vorticity in the new reference frame is @ Gr= Vx@+V)= VxB=3. (1.14) ‘We superpose a constant rotation, say a constant angular velocity G on the vorticity field @. This superposition leaves the new vorticity field @z unchanged except for a constant: B= Vx +x = B+28 . (1.15) In two-dimensional flow, we can distinguish four different reference frames if the fluid moves with constant — LI and constant + Q around the body and if Q* and U* are the angular and translational motions of the reference frame: (1) U*= 0, Q*=-Q. The frame is fixed to the body with regard to translation, but the body rotates relative to the frame. (2) U*= U, Q*=-Q. The body translates and rotates relative to the frame which is fixed to the fluid at rest at infinity. (3) U*= 0, Q*= 0. The frame is fixed to the body. (4) U*= U, Q*= 0. The frame does not rotate relative to the body, but the body has a translational motion relative to the frame. The behavior of the streamlines for the four different cases is illustrated in Fig. 1.10 for a viscous fluid with U 0 and Q# 0. As in the situation of Fig. 1.5, the flow patterns cannot be reduced to a steady-state case because of vortex shedding, a phenomenon that will be explained in Chapter 13. The use of frame 1 is appealing because of its analogy to instantaneous streamline photos of a rotating body in a wind tunnel. Frame 3 is advantageous since the body contour is a streamline, and the flow behavior near the body can be examined, as is done in Fig. 1.5. Frames 2 and 4 are of less interest. -16-Fig. 1.10: Computer-generated streamline patterns for the flow past a rotating elliptic cylinder (wing) in four different reference frames (Lugt and Ohring 1977). ‘We want to briefly interpret Fig. 1.10 on the basis of our experience in the study of Fig. 1.5. The patterns in frame 3 reveal that a small attached vortex develops behind the advancing edge of the wing (visible as a little bubble below the left edge). Wavy streamlines farther away to the left indicate a vortex moving relative to the wing. This interpretation is confirmed by the closed-streamline patterns of frames 1 and 2. The closed nested streamlines in frames 3 and 4 are due to the rotating frame (which itself is a vortical motion) and do not represent a vortex shed from the wing. This center of the rotating frame is not located in the middle of the wing because the center is displaced by the parallel flow. By contrast, the vorticity field in Fig. 1.11 is invariant and differs under rotational transformations only by a constant. The preeminence of the vorticity field over the streamline pattems for explaining flow behavior is clearly. demonstrated. Unfortunately, this advantage can only be partially exploited in the search for vortices. We shall see in Chapter 13 that a unique identification of a vortex in a homogeneous real flow based on vorticity is possible only for vortices detached from a body. They are characterized by closed contour lines. Attached vortices, however, have tongue-shaped contour lines. The lines in Fig. 1.11 are equivorticity lines. Since @ degenerates in two- dimensional flows to one component perpendicular to the flow plane, that is, the vorticity vector is always perpendicular to that plane, the vorticity field is displayed -17-| | | | by connecting points of equal vorticity to form an equivorticity line. This technique can be extended to three-dimensional flow fields, those with certain symmetries like axisymmetry. In general three-dimensional flow fields, vorticity is represented by vorticity vectors or vorticity lines which are parallel to the vector field (analogous to the velocity vectors and streamlines in Fig. 1.1). However, it is difficult to interpret such a field. During the course of this book we will lear how to read graphs of equivorticity lines and vorticity lines. Fig. 1.11: Equivorticity lines for the flow situation displayed in Fig. 1.10 (Lugt and Obring 1977). ‘We may mention that a vortex can be defined exactly and in an invariant way for an inviscid fluid: According to Safiman and Baker (1979) a vortex is a finite, simply connected,* but deformable region of vorticity surrounded by an irrotational fluid flow. This is a topological definition since the actual shape of the region is irrelevant (see Chapter 9 for its limitations). 1.6. Patterns which Exist only on the Average We are familiar now with various forms of flow representations and their corresponding pattems, at least for two-dimensional motion. These patterns can be of the kind that we immediately observe in reality, such as marked streaklines and timelines, or they can be quite abstract, such as equivorticity lines. There is yet another type of flow pattern that exists only when time-dependent patterns are averaged. Turbulent motion comes to mind. * A closed line or a closed surface which can shrink to a point. -18- i -At this point we must introduce two basic concepts of fluid dynamics. So far we have assumed that fluid elements travel on their prescribed paths all the time. However, it is possible (and this is the most common case in nature and technology) that fiuid elements fluctuate around an average path in an irregular way and move along the exact pathlines only on the average. Such a flow is called turbulent as distinguished from the /aminar motion considered so far. In the statistical sense the considerations for streamlines, pathlines, etc, are valid for turbulent motions also. Turbulent flow properties, however, differ from those of laminar flow as explained in more detail in Chapter 18. For instance, a turbulent vortex in most cases will decay faster than a laminar one. A time-dependent flow can have various distinct time-scales, and averaging always involves the smaller of two time-scales. The continuum model, which is being used in this book, has already averaged out fluctuations of the time-scale of irregular motion on the microscopic level. Averaging then the turbulent fluctuations leaves us with time-dependent turbulent flows. They in turn may be averaged, as is being done in atmospheric and oceanic circulations. The flow patterns which evolve from different scale averaging can be quite surprising since they may be “‘hidden’’ in the flow fluctuations. The discipline which evaluates measured data and extracts underlying flow patterns from them is called pattern recognition (Wallace et al. 1977). Fig. 1.12: (a) Periodic vortex shedding behind a cylinder at a certain instant for Re = 100. (b) ‘When averaged over time, the pattern of a steady vortex pair “‘attached’’ to the cylinder appears (computer-generated pictures from Karniadakis and Triantafyllou 1989). -19-Even periodic laminar fiows, with -frequencies small compared to those of _ turbulence, can be averaged to steady flows with familiar patterns (Kamiadakis and Triantafyllou 1989). A periodic vortex street behind a cylinder reduces to a steady flow with a typical wake consisting of a vortex pair attached to the cylinder (Fig. 1.12). We will read more on vortex shedding in Chapter 13 and on averaging proce- dures in Chapter 18. Exercises 1. In polar coordinates 7,6 with x= rcos,y= rsing, the corresponding velocity components v,, v4 have the form v, = cr* and vg = dr” with c, d and 0) Az Ag A3 At Ay Fig. 2.1: Flows in a converging tube (a) and in a branching tube (b). -26-‘We see immediately that streamtubes, which consist of a bundle of streamlines (those passing through a finite area), cannot end inside a fluid but can end only at its boun- dary.* In order to express the conservation of linear momentum in a continuum we write an equation, analogous to Newton’s second law for a volume either fixed in space or moving with the fluid (not restricted to constant p), Io@av= JoFav+[faa, (2.18) where dV and dA refer here to volume and area elements at a point fixed in space. On such a volume element, body forces F’ as well as surface forces f can act. The inclusion of surface forces makes Eq. (2.18) more general than Newton's law for mass points. The first term represents the change of momentum and is also called the inertial force, although it is only a fictitious force, to emphasize the balance among inertial, body, and surface forces. Examples of body forces are gravity and electro- magnetic forces. The surface force Fcan be written according to Fig. 2.2 in the form Pah +R +E, (2.19) with > > > fy = Fax + fay + Rte » >> > By = Hye + ity + Rye » (2.20) > > 2 ty = Mey + flay + Kee, 237 . reoti : where i, j, F are the unit vectors in the x, y, z - directions. The components of this scheme form the matrix of the stress tensor TT, with 1; the components of the normal stress and ty, i#j, those of the shear stress: ‘tex Tay Taz ye Tyy Tye] - (2.21) ‘tox Tay Tez * This boundary (which consists of the openings in Fig. 2.1) cannot be a solid wall, because the fluid flow must be able to enter or leave the flow domain. ‘The streamtube thus ‘‘continues” into the space outside the flow domain proper. 27In symbolic form we may write for Eq. (2.20) f= %-T. Applying Gauss’ divergence theorem to fPaa = f2-Taa = Jaiv Tav, we obtain for a volume element the differential form of Eq. (2.18), valid also fora compressible fluid, according to Cauchy (1822) fe pF+divT. (2.22) The last term in Eq. (2.22) can be derived also by finding the net stress components on the surfaces of a cubic volume element dV (Fig. 2.2). For the two surface elements perpendicular to the x-direction the difference between the surface forces is: dt iy pie 2% ax ax) ae 7" Ox’ 7 B+ 2) R= at gy and for the other directions correspondingly. Their summation results in div T. ‘The constitutive equation to be discussed in Section 2.4 requires that the tensor T be symmetric, that is, ty = Tj. : 6 > at te tae ox Fig. 2.2: Components of the stress tensor of a fluid element. -28-The first law of thermodynamics may be written in the form rate of change of total energy = rate of work done + heat flux The total energy is the sum of kinetic energy and internal energy. We obtain, for constant density and for a volume fixed in space, joe where E and7@ are the specific internal energy and the heat flux, respectively. Heat flux comprises the microscopic energy transfer, for instance, through heat conduction and radiation. The dimension of each term in Eq. (2.23) is energy per unit time ( = power). The work done on the surface (per unit time) is [dA = J div (T-D dV, and from Fig. 2.2 (in which tj should be replaced by ut,, etc.), div @-T) yields in Cartesian coordinates a Bay 4 fof PaV= fP-vaa+JoF-vav-[R?dA, (2.23) a a 2 Fe het May # Wie) + By ay + yy # Why) + Bag + ye +H) _ te, , Nay | Mee , tw , Me Oty tye OTe Gores oyecrtoz ya ty tate tay tae” a ov, ou ow , a Ou, dw + aye St) +e Be) tel ar + ay)! (2.24) or in symbolic form div (T D=8-divT+T:D. (2.25) The deformation tensor D is defined by Eq. (2.6), and the scalar product of T : Dis T:D tae + TeyDry + texDz + TyDyx + Dy + eDyz + tegDax + GyDey + taDz The work done on the surface of a fluid element is the sum of mechanical work and friction. Eq. (2.23) can be simplified by considering Eq. (2.22) which, when multiplied by, yields. -29-pF-0+0 divT. (2.26) ‘The change of kinetic energy is due to work done by body forces and surface forces. For a volume element dV we obtain from Eq. (2.23) oe = -divq+T:D, (2.27) or with dE = cdT,c being the specific heat, oar -div@+T:D. (2.28) Eq, (2.28) is called simply the energy equation. Heat convection is balanced by heat fiux and dissipation. In general, the dissipation term T-D can be neglected. 2.4. Constitutive Equations ‘The three conservation laws Eqs. (2.16), (2.22), and (2.28) form a system of partial differential equations which contain more dependent variables than equations. Stress tensor and heat flux must be expressed by the velocity and temperature fields. Cauchy's equation of motion (2.22) is valid for any continuous medium, no matter what kind of form the stress tensor T may take (Serrin 1959). We must specify T according to the type of medium we want to study. Such a specification is called a constitutive equation. A simple example is an inviscid fluid with T = — pl, where Tis the unit tensor. The description of a viscous fiuid in its simplest form goes back to Newton. Although he did not express this mathematically, the one-dimensional shear stress is, in Cartesian notation, ey = oe Newton (2.29) where j1 is the dynamic viscosity which is assumed to be a constant. However, the significance of Eq. (2.29) as a relation that cannot be derived from classical mechanics and that is a postulate for an irreversible, molecular process, was recognized first by Fourier in 1811 when he introduced the analogous form for the heat flux G= -KVT. Fourier (2.30) -30-This heat flux is called heat conduction, and the coefficient k thermal conductivity, which also is assumed to be a constant. Heat radiation is not considered. The tensor relation between stress and deformation was given by Cauchy and Poisson (Poisson 1831) for constant p T=-pl+2uD. Cauchy-Poisson (2.31) Since D is a symmetric tensor, the linear relation (2.31) requires that the stress tensor be symmetric too. A fluid described by Eq. (2.31) is called a Newtonian fluid. The most important fluids like water, air, and most oils behave as Newtonian fiuids. We may emphasize that Eqs. (2.29) through (2.31) are postulates or hypotheses, supported by experiments. With Eqs. (2.30) and (2.31) the terms div@ and T : D in Eq. (2.28) can be replaced by -divq= kVT, (2.32) T:D= —pdiv0+2uD:D. (2.33) Because div@ = 0 for an incompressible fluid and because D:D= 30? ~div@x@) + 5V77, (2.34) we obtain T:D=0=p [B?-2div xd) +v27? | : (2.35) The term ® is called the dissipation function. This is the mechanical energy lost by frictional heat. The total dissipation over a certain space is 2 Joav = ala avn) [ 2000, - 2°] dA. (2.36) For the derivation of the last two terms Gauss’ divergence theorem has been used. We see immediately from Eq. (2.35) that dissipation does not vanish in an irrotational fiuid flow with @ = 0, and that according to Eq. (2.36) the total dissipation of an irrotational fluid flow is determined by the velocity on the boundary of the flow domain, that is, by the last term of Eq. (2.36). This term represents the work of the shear stress on the boundary. If the velocity on the boundary vanishes, @ depends only on the vorticity. For instance, in a closed container, a fluid motion comes to rest only through dissipation of its vorticity. -31-2.5. The Basic Equations of Motion ‘The three conservation laws can now be written in the following form: divt= 0, (2.37) B +@-WP= a¥ +wG+F, (2.38) at @-WT= KVIT+ Kia? -2 dio @xB) +V?F71. (2.39) Here, v= wp is the kinematic viscosity and K = Kipc is the thermal diffusivity. Eqs, (2.37) through (2.39) are five partial differential equations for @, p, and T with given, Eq. (2.38) is called the Navier-Stokes equation or, considering its three Components, the Navier-Stokes equations. Often Eq. (2.38) and the continuity equation (2.37) are designated as Navier-Stokes equations. Mathematically, the Navier-Stokes equations for an incompressible medium can exhibit elliptic, parabolic, and hyperbolic characteristics. ‘The terms with the highest derivatives include’ the Laplacian which is an elliptic operator. The first-order time derivatives signal parabolic behavior, and hidden in those equations are hyperbolic features (wave motions) that become apparent in a rotating system (Chapter 19). ‘Additional information on the properties of the Navier-Stokes equations will be given in Chapter 3. For the difficult ‘questions of existence and uniqueness of the Navier- Stokes solutions we refer to the books by Ladyzhenskaya (1969) and Témam (1977). {At this point, we may mention only two main physical properties. The nonlinear terms on the left of Eqs. (2.38) and (2.39) are called convection terms and the Laplace terms V2 and V2T on the right the conduction terms. We may remember the meaning of these two terms by thinking ‘of a water heater. The heat transported by the water is convection, whereas the heat flowing in a piece of metal through molecular Uifusion is conduction, In what follows, the energy equation will be considered only occasionally. Laminar fluid motions are well described by the Navier-Stokes equations, as many experiments have attested their validity, at feast within the range of accuracy of today’s experiments and numerical calculations. Examples will be given in this book.* ® ‘Truesdell (1974) may be quoted here: ““I think our confidence in the Navier-Stokes equations Jeives ultimately, not from any particular class of experiments, but from the rules of similitude those equations imply” (reproduced, with permission, from the Annual Review of Fluid Mechanics, Vol. 6, 1974, by Annual Reviews Inc.). See Chapter 3. a2e2.6. Initial and Boundary Conditions An initial-boundary value problem is completely defined if we prescribe initial and boundary conditions. These conditions actually determine a special problem from the multitude of possibilities. Initial conditions must be prescribed at the beginning t= fp for the dependent variables 7 and T with the constraint that the mass conservation div@= 0 be preserved. If we think of all the varieties possible in nature and technology, we will not be surprised to encounter great difficulties with respect to the collection of permissible boundary conditions. These difficulties have surfaced in the context of the numerical techniques used to solve the Navier-Stokes equations on computers. The issue has not yet been settled (Gresho 1991). Let us distinguish for the time being three types of boundaries: (1) the boundary of the total flow field that may include the infinite periphery, (2) the surface of a solid body, and (3) the interface between two different immiscible fluids. To set up boundary conditions for the first two types seems to be no problem. We prescribe the velocity vector and the temperature (or the gradient of velocity and temperature on part of the boundary), while no boundary conditions are required for the pressure. The mass conservation, of course, must be fulfilled along the boundary, that is, J#-VdA=0 for #20. (2.40) 3 Here, we are already confronted with a problem. What kind of boundary conditions shall we prescribe at the downstream side of a domain, where the flow is still disturbed (for instance, by a long vortex street)? In this case, actually, knowledge of the solution itself is required. This problem will not be addressed here (see Gresho 1991). Rather let us discuss the conditions for types 2 and 3. First, we should return to the statement that the pressure variable does not require boundary conditions. The reason is that the velocity field can be used to determine the initial and boundary conditions for the pressure. To make this clear, we replace the continuity equation (2.37) with an equivalent equation which we obtain by applying the div-operator on Eq. (2.38) and by considering Eq. (2.37): Vp = -pdiv@-V)7. (2.41) This Poisson equation for p can be solved with the boundary conditions for dp/én given by the velocity field. The pressure is determined up to an additive constant except for an interface where the pressure level is set. 33We define a boundary: between a fluid and a solid or between two immiscible fluids by the condition that no fluid shall cross it. Mathematically, such a boundary is desotibed by f(x,y,2,t) = 0. The velocity of the boundary normal to itself* a VER +R must be equal to the fluid velocity normal to the boundary Uf, + Ofy + Whe VE +h th (af ore Ofer Ol eee P= TH 5e +05, te (2.42) ‘Thus, Eq, (2.42) is the kinematic boundary condition. Often we find the equation for the boundary written in the form y = 11 (x, z, #) orf= y-1 (x, 2, #)= 0. Then, a. fy-W= 0, dt di One oe Sen Oy ae (2.43) Inareference frame fixed to the boundary, the kinematic condition reduces to R-U= 0. (2.44) In dynamic boundary conditions surface forces are involved. Experiments have shown that,a fluid adheres to a solid surface. It obeys the nonslip condition. We cannot obtain this information from the basic laws of a continuum. Rather it must follow from a microscopic theory (Koplik et al. 1989). The tangential velocity on a nonslip surface must vanish: RxB= 0. (2.45) * Inthe text that follows the abbreviations af/0x = fe, 0°//0%" = fax, ete. are used. 234.On the interface of two immiscible fluids A and B the nonslip condition requires the tangential velocity components to be continuous. In addition, the surface forces must be in equilibrium, that is, the stress must balance the surface tension (capillary forces). This balance can be expressed in the following way. If we abbreviate (only in this context) the difference of a function g by 4 ~ gp = [g] at the interface and if we restrict ourselves to two-dimensional flows because of the complexity of the problem, we postulate for the balance of the surface forces TATA ae ae +Vo, (2.46) where 77 is the outward normal and G is the coefficient of surface tension. Eq. (2.46) yields in the x- and y- directions (Wehausen and Laitone 1959, Levich and Krylov 1969) [eggl + tym] = a +0, (2.47) [eyl+ Gym] = fm + oy. (2.48) with 1= —n,/V1 +12 and m= 1/\1 +12. These two equations yield, for the stress components in the normal and tangential directions, [pl (ES toy 2g — Maloy + uy) = -o- - LEEEE | 249) 1+n2 y Y 7 ee [y. 2nel(0y — tee) + (Uy + O12) = (Ox +Ne)V1402 , (2.50) where the curvature 1/R is Nex Fr: 2.51 (1+ np? eb Surface tension on the interface becomes effective when the interface is curved and when the surface tension varies from point to point on the interface. If one fluid, say fluid B, is negligible in the sense that Hp « ta, but still wp # 0, we call the boundary of fluid A a free surface, and we can omit the subscript A. The pressure above the free surface is usually assumed to be a constant: p= pp. Eqs. (2.46) and (2.47) yield -35-Te 2, 2 et P-Po- fo, + yn} —Ne(Uy + VJ} = — je , (2.52) a es (+n)? ten? W2ne(Dy ~ Me) + (Uy + g)(1 — TB) = (e+ 1x6y) VI+ng (2.53) Experiments show that for many fluids 6 can be expressed as a linear function of temperature o= 6 -Y(T-To) , Y29, (2.54) where ‘is the temperature coefficient of 6 and To is a reference temperature. In order to see clearly the effect of a gradient of surface tension on the fluid motion, we assume a flat free surface that reduces Eq. (2.53) to (2.55) Puy = Ox = ‘The temperature change along the free surface creates a shear stress which, in turn, danse # fluid flow from the warmer to the cooler part of the surface. This flow is called Marangoni convection of thermocapillary convection. The exactly analogous form of Eq, (2.55) exists for fluids which contain other ingredients such as impurities or surface active agents. Then, T is replaced by the concentration ¢, or the formula for cis added to that for T (Levich and Krylov 1969, Schwabe 1988). 2.7. Conservation Law of Angular Momentum ‘The conservation law of angular momentum is not an independent axiom but can be derived from the conservation law of linear momentum, Eq. (2.18), if the stress tensor is symmetric, that is, if tj = ty Sertin 1959, p. 136). The cross product of position? ‘with each term of the momentum equation in integral form yields* J pc? #) a= £ Jom av= JorxPavelPxtda. 056) The change of angular momentum is equal to the torque due to body forces and surface forces, We can use Eq. (2.56) to derive an equation of practical importance, the so-called turbine equation of Euler. tn Fig, 2.3 two blades of a turbine form a Bq. (2.56) is also valid for a compressible fluid. Po6echannel which rotates with constant angular velocity Q around the shaft. The relative velocities at the entrance 1 and the exit 2 are @; and V%, respectively (the absolute velocity is the sum of the tip velocity of the blade and the entrance or exit velocity). The radial and tangential components of @are v, and Up. We neglect the torque due to frictional forces and write for the rate of change of angular momentum between entrance 1 and exit 2 of the turbine blades of width I, with p = const, ol fre av= of Davey fe VW) RxD dV (2.57) which, for a steady-state flow, results in p[@-v) PxBav= kp fo + vgr ldo = B pl Q (11%, — 12) (2.58) The volume fiux per unit width Q is Q= Abi, = 12A020;), (2.59) with Ad the angle of the opening between two blades, and 3, and U the radial and tangential velocity components averaged over the opening Ad. Vo, vee des Abe Fig. 2.3: Sketch for the derivation of Euler’s turbine equation. This change of angular momentum (per unit width) must be balanced by the torque M acting on the blades (Buler’s turbine equation): M= PQ (11091 — 72042) - (2.60) aieThe power of the turbine due to the pair of blades is P= QM (Vavra 1960, Streeter 1961). Exercises Verify that a pseudo-vector in four-dimensional space has six components. Find the free-surface vorticity with the aid of Eq. (2.53) for Nx < 1 and compare the result with Eq. (2.55). Show with Eq. (2.12) that Bit = 0 yields dffdt = 0 and dg/dt = 0. Under which condition does a turbine provide optimal power?3. The Navier-Stokes Equations 3.1. Various Forms and Properties The Navier-Stokes equations for an incompressible fluid are the foundation for our investigation of vortices. Some properties of the Navier-Stokes equations with various boundary conditions have been mentioned in the previous chapter. The derivation of criteria for existence and uniqueness goes beyond the scope of this book, and we must refer this part of the Navier-Stokes theory to other books like those of Ladyzhenskaya (1969) and ‘Témam (1977). However, one important property of the Navier-Stokes equations that has profound consequences must be cited without proof. This is the non-uniqueness of those equations unless the fluid is very viscous. Non- uniqueness is the key to the variety of solutions which reflect the complexity of nature. Mathematically, non-uniqueness is linked to the concepts of instability, bifurcation, and symmetry-breaking that will be discussed in Chapter 11. The success of the Navier-Stokes equations as a realistic model of nature may lie in the combination of two different physical ideas: the axioms of classical mechanics characterized by the property of conservation or reversibility, that is, the solutions of the equations of motion do not change (except for a possible change of sign) when the time is reversed (t is replaced by — £); and the irreversibility of the dissipative process of diffusion. With this distinction in mind let us look at the Navier-Stokes equations (2.37) and (2.38) which we write here again for convenience div@= 0. Gl) we 1 2 # .g.yyee —1y, toe prt , (3.2) External forces are not considered here. Nonlinearity is associated with the inertial terms whereas the viscous terms are linear. This means that the nonlinear behavior with all its consequences* is contained in the reversible part of the equation, and the irreversible part is linear. However, the combination of the two is more than its parts as the complexity of the solutions reveals. We will discuss the various aspects during the course of this book. “Chaotic motion in classical mechanics (many-body problem) is a nonlinear effect although the ‘motion is reversible (the system is conservative or non-dissipative). 308)We can easily determine whether Eq. (3.2) remains unchanged by the time reversal {> ~# (Kochin et al. 1964). According to their definitions, velocity @ and pressure p at t change at — f to; and p,, whereby BE N=-7R-1, 3.3) p@)=p, 7-2). G4) In an inviscid fluid flow, each fluid element would return along the same path to its initial position. This is not the case for a viscous fluid.*’@, and p,, then, would not be solutions of the equations of motion. In order to prove this, we examine the individual terms of the Navier-Stokes equation (3.2) Bw Or = ior VW= Vp, VR=-VH. (3.5) OVB= GV, We observe that’, and p, do not satisfy the Navier-Stokes equation. In an analogous way we can prove that the energy equation is time-irreversible due to heat conduction and dissipation. We will obtain a better insight into the properties of the equations of motion for the text that follows if we consider various forms of the Navier-Stokes equations and their individual terms. The rules of differential operators in Appendix A are helpful here. ‘The viscous term vV? Van also be written in the form (because div@’= 0) w?V= -veurlcurl@= —vcurlB, (3.6) and the nonlinear term (@- V)V in the form zw @-V)8= Bxt+ ws (3.7) Exceptions include the potential flow; see end of this section.In both alternative expressions the vorticity vector appears. The cross product @x7 plays a special role in hydro-acoustics and in turbulence (see Chapters 17 and 18) so that a new symbol, the Lamb vector, is justified: = Oxt. (3.8) We have already become acquainted with one alternative form of the Navier- Stokes equations in Section 2.6, when we replaced the continuity equation (3.1) by the Poisson equation for the pressure, Eq. (2.41). The original set of equations (3.1) and (3.2) is, by the way, known as the primitive-variable formulation, whereas the combination of Eq. (3.2) and (2.41) is the pressure-Poisson formulation. We will derive now a third version of great importance for vortex motion, the so-called vorticity-transport equation. We obtained the pressure-Poisson equation (2.41) from Eqs. (2.37) and (2.38) by applying the div-operator to Eq. (2.38), but now we use the curl-operator and arrive at Sewtt +BY) curl P- (cut: W8= vVPeul?, G9) or because curl@= @ Bie. V)G-G@:V@=vVS. (3.10) Eq, (3.10) is the vorticity-transport equation. The pressure has been eliminated. As with the pressure-Poisson equation, for which no boundary conditions for the pressure are needed, there are no boundary conditions for the vorticity either, when ‘we want to solve Eq. (3.10). The vorticity at the boundaries is determined by the velocity field and is part of the solution. ‘We may illustrate the role of velocity in the boundary conditions for the special case of two-dimensional flows. We convert Eq. (3.10) with the aid of @= curl = VFX Vg into a scalar equation of fourth order for a single variable, the so-called stream function w B= curl (yR= VwxVz , y= -o,, (3.11) and in Cartesian coordinates aaa. \This stream function automatically fulfills the continuity equation, and the vorticity- transport equation (3.10) assumes the form ay aw Vw) _ ve ae amy YY oa We now have before us a single differential equation with the elliptic biharmonic operator V4 which determines the number and form of the boundary conditions. For a well-posed problem, y and dy/én must be prescribed with the aid of the velocity components. The vorticity will then be computed from Vy = - @, of Eq, (3.11). ‘We will discuss the vorticity-transport equation in Chapter 5. Without actually solving the Navier-Stokes equations, we can identify’ an impor- tant class of solutions for which the viscous term vanishes automatically, that is, the solutions are independent of the kinematic viscosity v. These solutions are called universal solutions (Gortler and Wieghardt 1942, Marris 1981). The condition for their existence is, according to Eqs. (3.6) and (3.10), 2B cunt ed = 0, curlcurl@= 0. (3.14) We see immediately that the potential flow with @= 0 is a universal solution. The potential flow of a viscous fiuid with v > 0 is dissipative because the shear stress is not zero. We can see this in Eq. (2.36) for the total dissipation. However, the restriction of @= 0 does not permit us to prescribe the boundary condition for the velocity component tangential to the boundaries, for instance, the nonslip condition.* We have to accept whatever the solution will give us. This does not mean, however, that this class of Navier-Stokes solutions cannot conform to a physical flow. The question is whether the boundary data imposed on us have some physical meaning. In the case of the Rankine vortex (explained in Chapter 4), the boundary condition happens to coincide with the nonslip condition; a rare situation indeed, Let us choose the usual situation of a potential flow, for instance, the flow past a cylinder. This flow causes no drag, a result that is known as the d'Alembert paradox. Is it really a paradox? If we let pieces of surface areas glide with locally varying velocities, such that the potential flow is modeled, we inject into the fluid energy that dissipates. The sums of normal and shear stresses at the surface are zero. The flow would be irrotational (Hamel 1941). “An important exception is the Rankine vortex, Chapter 4. -42-However, the significance of potential flow in fluid dynamics rests on another definition which contains an additional restriction to @ = 0: the restriction to inviscid fluids, that is, v= 0. Now, d’Alembert’s paradox is a real paradox, because a body without drag cannot be explained in this fluid model. Throughout this book we will make a distinction between the two definitions of potential flow: (1) potential flow of a viscous fluid (for details see Hamel 1941, Ackeret 1952, and Zierep 1983), and (2) potential flow of an inviscid fluid. The latter model is usually found in the literature as an approximation to fluid flows with very small friction. Solutions with curl curl @= 0, B+ 0, that is, with curl@= VB, are universal solutions of the Navier-Stokes equations too (provided that the left-side equation of the double Eq. (3.14) is also fulfilled). The function B is called the flexion potential. Examples of universal solutions with @ # 0 will be described in Chapter 4. For steady flows, the left side of equation (3.14) is a condition for ‘gnishing nonlinear inertial terms in the vorticity-transport equation. A flow with L= Beltrami flow, and a flow with curlL = 0 a generalized Beltrami now. Thus, steady universal flows are generalized Beltrami flows, but unsteady generalized Beltrami flows,are not universal flows. In steady two-dimensional motions, the condition curlL = 0 is reduced to (a, WAAlx, y) = 0. We notice that @, = @= fly) is a solution of this equation (Wang 1991). 3.2. Basic Equations in Different Coordinate Systems For the description of vortices it is not convenient in many cases to use Cartesian coordinates. Circular motions can best be described by cylindrical polar coordinates 7, 9,2 or by spherical polar coordinates R, , 2. There are three reasons for this change in coordinates: (a) The number of dependent and independent variables can be reduced. This is the idea of introducing “generalized coordinates and momenta” in classical mechanics: to have the number of coordinates limited to the degrees of freedom of the system. As an example we may mention the plane pendulum, Instead of the two Cartesian coordinates x, y we need only the polar coordinate ¢ as the generalized coordinate to describe the motion of the pendulum. (b) The change of coordinates can simplify the equations of motion so that they are much easier to solve (for instance, by separating dependent variables), and so that they reveal essential physical features. Examples will be furnished in Section 4.1. (©) The border line of the flow can be a coordinate line that makes it easier to impose boundary conditions. 2435For completeness and reference, the continuity equation (2.37), the Navier-Stokes equation (2.38), and the energy equation (2.39) are given first in Cartesian coordinates. The components of the external force F are omitted. Cartesian coordinates ou ow ax * Oy 70 (3.15) mu 2H dt, We wy Pu, Pu Fe Be oe Tay th on 7 par ae * ay? * a” (3.16) 20 20,982, p22 1a, ao, Bo, Fo at ae tay oe 9 ay "Ya? * ag? ae” G17) dw, dw , dw, 30 le, ew, Pw, Pw Seto oy oral Biocee tetas (3.18) OT, yay) _ (SL, ST, YT Pelayo thay tM 9g) RG + et ae) t ee (3.19), with 9 py , ,82y2 , BW , 1,dw , a» 1/du wy 1a, mu Dam LSE? +S) HGP + Oy 8328 2 Ge te 2 Ge * By (3.20) The components of the vorticity vector are = 6 _ oD = By 92" @y = au ee (3.21) fn us ous = ax ay” The components of the deformation tensor are Py.Boe ou ae Dae 5p 7 Dar 55 + Gy)? = -1/9u , Dyy= Se, Dya= Ge + $2), 22) = = Lau , dw. Da= ap Pe 90 ae tae For two-dimensional flow in the (x, y) - plane, the stream function is defined by -& 5. mu ua ay PRB (3.23) Cylindrical polar coordinates x= reosd, y= rsing, (3.24) Zaz. The equations of motion are 80, , Br, 1 Oty | Be Ope are Ari osh ae (3.25) Bor |, or | Mer tyr | By - +5 = rd or * Oz (3.28)ar, % aT YT tar 1 eT, #T + + ar aT) PL | pelt ase tag tt ae) EG ty an te ag ete (3.29) with By (1% — Pate spe Gt 5 1 ep a =—=—S—sés—sesCr aire a don 3.30) The components of the vorticity vector are i q pio oat i 7 de Or | te Me ae i | = oar’ ey | | = 12 (m,)-1% rT na 0) OH” | The components of the deformation tensor are | | 2, 1 dM), LOM) Pen Daw Ses EL. | 18% .% pp _ 12% 10% Die= 5 36 1 De= Sr tT 39)” 332) a "1m Bee, ( | Dam a1 Dam 9G tar ih | | For two-dimensional flow in the (r, 6) - plane, the stream function is defined by | - 1% ow: | m= Pay ae! (3.33) i and in the (r, z) - plane | 1a 19 |Spherical polar coordinates x= Rsin@cosi, Rsin@ sind, (3.35) z= Reos0. r= Rsin6, , (3.36) z= Reos@. The equations of motion are (Re, Dy Dey sin@) + —1_ = 3.37) a —_ rm Rsiné 30° Rsin@ 0h : Box BRM Rm DR Veoh ay i TORTS: RMON Rs 0s Ree aap oR a Ca 2 a ay Le (ye 8 1B ping PR), —1 _2R _ APR _ 2 So ROR” OR’ * Rising 00°” 80°” Rsin’@ BA? -R? —-R? 08 2ogcot® 2a, oe R ” Resind al a av, ave, 6 du md , Udy whoo! ap it NRT RST Rao HR Ra gap Roo ey lnee cele ea Bey, 1_ 0 , 2 den R? OR OR R’sind 30 R’sin’@ an?" R? 00 2% 2cos@_ 9%, - > (3.39) R’sin’@ —R’sin?@ 2 a an an HAH HAH BYR, HCHO 1 1 ~ ap ar’? R * R88 * Rsind OW” R * Rp Rsind ar 2, pee 19), 0m 13m 2 dR 7 i aOR? * sing 00" 29” * Rlsinto on? Resin’o * Resind Oh 2c0s0_ 2%) + Rsin’o ah 3.40) aie4g SE. HE, : ROR *R a6” Rsind dA OT 2a, 108T, oto aT, 1 FT Hope * RAR RO ag?” Re 00 * Resinvo arte, ee with ® _ WR tp 19K» 19% | MR», 1,sin@ dO) Ar a a” PRR R R00! * Ra TR? TDR BO Gnd 1 2 du, +78 —_ dvR a Ryo * Rind ey Raa ee ta Rae eR (3.42) The components of the vorticity vector are . 2 _ 306 ne Tana 5 529) al? _ o= Rend an RaR®'” ee a 1 d0R . Rae ee fe 30° ‘The components of the deformation tensor are dop Foy Dar = SE, Dro= FIRE + RSE. Dog = 22% 7% py = Lysine 2M), 0 ca) 9° RR’ 2° R 2 sind nae 5 7 av 10 a Dy = oe eR, OO py = LR RAY, Rsind 2k RR” R 2°)Rsind o& ORR For two-dimensional flow in the R, 8) - plane, the stream function is defined by —1i_ [oie Ovi Rsino 00’ Rsin@ aR” 2 ae uyIn order to define a specific flow problem, we must prescribe initial and boundary conditions. They have been presented in Chapter 2. In the next section we would like to repeat the boundary conditions for a local piece of the boundary to demonstrate clearly their physical significance. .3. An Example of Local Boundary Conditions If we are interested in studying only local flow behavior, it might be sufficient to use Taylor expansions in Cartesian coordinates (as we have done in Chapter 2 in the derivation of the fundamental theorem of kinematics) or, if curvature is involved, either intrinsic coordinates* or Taylor expansions in cylindrical or spherical polar coordinates. An example of Taylor expansions in cylindrical coordinates is given in this section. Let us express mathematically the boundary of a two-dimensional flow by a circular arc r= r, that may be considered a local approximation of an arbitrary curve with radius of curvature 7, (Fig. 3.1). The boundary is fixed with respect to the reference frame and does not change with time, but the fluid beneath can be unsteady. y LZ \ ™4 ut Ae, Fig. 3.1: Polar coordinate system (r, 6) for describing the series expansions of the Navier- Stokes equations about the surface point P. The curvature of the boundary is 1/r.. The kinematic boundary condition on a circular arc is then = 0 on r= Te. (3.46) * Intrinsic coordinates in two dimensions are coordinates that coincide with the streamlines and their ‘orthogonal trajectories (Serrin 1959, p. 155 and p. 250). Both streamlines and their normals can be curved, whereas in polar coordinates the normal (which is the radial component) is straight. If we accept this restriction, the use of polar coordinates is instructive and much easier than intrinsic coordinates. -49-|, The dynamic poundary conditions specify the tangential velocity at the boundary. We distinguish between solid and fluid boundaries. On a solid surface the nonslip | condition holds: p= 0 on r= fe. G47) The surface vorticity is then @, = 909/01 and is part of the solution. The interface between two immiscible fluids A and Bisa fluid boundary. On this fluid boundary the normal velocity components of A and B must be the same because they define the interface, and the tangential velocity components of A and B must be the same due to the adherence of the fluids. The shear stresses of the two fluids differ at the interface by the gradient of the surface tension, and the normal stresses differ by the surface tension and its gradient (see Eqs. (2.52) and (2.53)). Then, the ‘boundary conditions are (o)a-O)p = 0, (3.48) | (vg) —(p)p = 0, 49) i Wy % avy a Bre My SM _ Me). OF Bale 7,24 BBE Fe B= Tag’ (3.50) Ber dy) 9 30 | (P- 2n)a ~~ 3 B= aon (3.51) For a free surface we omit the subscript A and obtain, ina reference frame fixed to the surface, (v,)= 0, (3.52) a0 %, a0 2% MH) _ WP T= Fe86 ’ (3.53) | ae, s_a6 a Cs with po the pressure outside the fluid. Condition (3.49) is now superfluous. Without surface tension, Eq, (3:53) represents the perfect-slip condition: . | a | tgs WEE = 0. 55) -50-Considering this perfect-slip condition, we obtain for the vorticity @, = v9/0r + Ug/re at the free surface = 20. (3.56) The free-surface vorticity is equal to twice the curvature times the velocity tangential to the surface, With surface tension the vorticity at the free surface is 2 1_do ro Wr.) * 3.57) ‘The second term to the right does not depend on the curvature I/r, since the length element of the surface is d(r,9). 3.4. The Concept of Similarity Because of the complexity and generality of the Navier-Stokes equations it is advantageous to simplify either the Navier-Stokes equations by restricting the number of dependent and independent variables, or to truncate the equations in a meaningful way. The first task will be based on the concepts of symmetry and similarity. * Trunca- tion then will be facilitated by using the concept of similarity. We have used symmetry as a general concept of order that expresses invariance properties (Chapter 1). Here, we shall use symmetry in the simple form of parameter and variable reduction. Solutions of the Navier-Stokes equations with such as- sumptions are called symmetric solutions. An example of most importance in the study of vortices is the assumption of axisymmetry, in terms of cylindrical polar coordinates: 9/2 = 0. The solutions are independent of 6 or invariant with respect tod. Another group of symmetric solutions is based on the concept of similarity. ** We are familiar with geometric similarity. Two bodies are geometrically similar when the shape of the bodies is the same but their magnitude is different. For instance, the similarity between an original airplane and a model is expressed by the ratio of their size (Fig. 3.2a). In the same way the ratio of forces acting on a body, or in general the * Sometimes, especially in older literature, the synonym “‘similitude’” is used. *® A rigorous treatment of similarity can be found in Birkhoff (1960) and Panton (1984). hile| ratio of forces within a fluid, can be defined. This ratio permits us to make judgments about the influence of those forces in the basic equations of motion. If the similarity concept is applied to the same “‘object,’” we use the word self- similarity. For instance, a spiral vortex behind a sharp edge can have the same form but different size at two different times (Fig. 3.2b). There is no natural reference scale. Fig. 3.2: (@) Geometric similarity of two airfoils. (b) Self-similarity of the spiraling edge vortex at two different times. The flow fields at these instants are self-similar. In contrast, the above mentioned differently-sized airplanes are not self-similar but similar. The concept of similarity may also be applied to independent and dependent variables. For instance, solutions of the heat equation with T = T(r, t) may exist for one independent variable ¢ which is composed of the two variables r and t in such a way that = r?/4Kt: OE py SL, Lar. Bg a oi arene (3.58) Separation of variables leads to the solution T = [eyJoQr) + c2¥o(Aryle**, (3.59) where Jo and Yo are the zeroth-order Bessel function of the first and second kind, respectively, cy and é2, are integration constants, and 2 is the eigenvalue of the boundary-value problem. For an infinite domain with c2 = 0 (since Yq is infinite at r= 0), integration over the eigenvalues 2 yields T* = cy |Ao(ar) e** da = oa eK (3.60) 0 esaPAAKt. The 2 before Jp is a ‘weighting’ factor. The introduction of similarity parameters and variables has a number of advantages: The new similarity variable is 1. Similarity parameters like the Reynolds number (see next section) form the basis for the use of models in wind tunnels and water tanks. 2. Similarity parameters and variables reduce the number of flow constants and variables. They can simplify the number of experiments or the complexity of the Navier-Stokes equations. 3. The latter simplification of the Navier-Stokes equations can become a means to solve these equations. 4. Similarity parameters can give information on a flow problem without actually solving the partial differential equations. For instance, the Reynolds number as the ratio of inertial over viscous forces can give us information on their relative influence, and based on this, the Navier-Stokes equations may be truncated to a form amenable to solution. Unfortunately, we have to pay a price for the simplicity of symmetric solutions. We are restricted in the selection of the initial and boundary conditions, and sometimes we have no choice of the boundary data at all. Moreover, statements which hold for three-dimensional flows are not necessarily valid for two-dimensional flows. For instance, three-dimensional disturbances with vanishing velocity far away from them are less a disturbance than two-dimensional ones with non-vanishing velocity in the third dimension. Worse, symmetric solutions may enforce uniqueness and prevent the transition to a new flow configuration through symmetry breaking such as the transition from a steady symmetric wake flow to a periodic vortex street behind cylindrical bodies. In the section that follows, similarity parameters are derived by making the Navier-Stokes equations dimensionless. Truncations based on such dimensionless equations will be discussed in Section 3.6. General symmetric solutions of relevance to vortical motions will be presented in Chapter 4 and Section 13.1. 3.5. Basic Equations in Dimensionless Form ‘We introduce the characteristic length L, the time L/U, the velocity U, and the temperature T, and make the dimensional dependent and independent variables in the equations of motion (2.37) through (2.39) dimensionless. If we designate, for the time being, these dimensional variables with a prime, that is, the dimensional length x’, etc., then the dimensionless variables are introduced by -53-ee xs Ly y= Ly, 22 lz, f= 4, UG, p’'=pl’p,T=TT. (G61) ‘The scaling for p is guided by V(p/p +7°/2) in the Navier-Stokes equation (3.2) with the alternate term (3.7). With Eq. (3.61) we can write the equations of motion (2.37) you (2.39) in dimensionless form by including as external force F only the gravity = gk div@= 0, (3.62) 1 tyy,+ Re” C+ (3.63) Roos. 2 3? VW)t= -Vp+ fe’ 2 @-wr= avr B?-2 din @xB+ V7? | , (3.64) with the new flow parameters: Reynolds number Re, Froude number Fr, Prandtl number Pr, and Eckert number Ec: Re= —= (3.65) (3.66) (3.67) (3.68) Dynamic similarity with regard to the Reynolds number means that this parameter must be the same for the original and the model. For the two airfoils in Fig. 3.2a, say 1 and 2, itis nae bu, Ro (3.69) V1 Vo If, as an example, the geometric similarity is given by L;/Lz and the fluids for models 1 and 2 are the same, that is, vy = V2, the ratio of the velocities U;/Up must be L2/L;. ‘We notice that the Prandtl number Pr is a material constant. ‘As indicated earlier, the flow parameters may be introduced directly as ratios of forces or ratios of other physical quantities. For instance, the Reynolds number is the ratio of inertial force over viscous force, and the Froude number the ratio of inertial force over gravitational force. -54-3.6. Truncated Versions of the Navier-Stokes Equations Another advantage of working with dimensionless quantities is that they permit us to estimate which terms in the Navier-Stokes equations may be neglected. Such a truncation of the set of partial differential equations must be made with great caution since it can change the character of the partial differential equations. For instance, the elliptic Navier-Stokes equations for steady flows change to parabolic partial differen- tial equations if the boundary-layer assumptions are made. In the following text the ‘most important truncations are introduced. Stokes’ Slow-Motion Equations For small Reynolds numbers, that is, for Re <1, the Navier-Stokes equations can be drastically simplified because the nonlinear terms are eliminated. The new equa- tions are called either slow-motion, creeping-flow, ot Stokes equations. The Reynolds number tells us that this simplification applies either to very slow motion, to very viscous fluids, that is, v very large, or to the movement of very small objects. Applications of the Stokes equations include lubrication, movement of micro- organisms, and creeping flow in geophysics. Slow motion occurs also in the final phase of decay of a fluid flow. If we put Re = 0 in Eqs. (3.63) and (3.64), we obtain universal solutions because V°@= 0. However, we know from experience that such flows can be time dependent and pressure dependent. This leads us to the conclusion that the scaling made in Section 3.5 must be modified. Indeed, if we make the time dimensionless by t’ = Pt and the pressure by p’= LAWUP, we arrive at the slow-motion equations Bap. , (3.70) at lea or at PY T, (3.71) with the continuity equation (3.62) which remains unchanged. The dissipation function is neglected in Eq. (3.71). ‘This linearization has the advantage that superposition is possible, but it intro- duces new problems such as the Stokes paradox (Birkhoff 1960). For vortices the most interesting property concerns the pressure and the vorticity. The Poisson equation for the pressure (Eq. 2.41) becomes the Laplace equation V2p = 0 and the vorticity for steady flow is also governed by the Laplace equation V?@= 0, obtained from Eq. (3.10). The significance of these two Laplace equations will be discussed in Chapter 13. ~55-Boundary-Layer Equations The inability of potential flows to satisfy the nonslip-boundary condition on a solid surface led Prandtl in 1904 to introduce the concept of boundary layer: A thin vorticity layer along the solid surface in which the fluid flow decreases to nonslip (Fig. 3.3). For this thin boundary layer the Navier-Stokes equations can be truncated according to the following assumptions: (1) A scale analysis eliminates certain viscous terms. (2) The pressure across the boundary layer is constant or is balanced by the centrifugal force. Distance from the wall Boundary layer Velocity Vorticity Fig. 3.3: The concept of boundary layer. ‘The procedure is illustrated in cylindrical polar coordinates for two different flow situations: 1. The boundary layer exists along the solid surface z = 0. 2. The thin core of the vortex forms the boundary layer. In the first situation, we estimate the order of magnitude for each term: a. 2 ody, Re~ Ok £- OW), $5 ~O(5), Re~ OF), 0, ~ O(L), U ~ O(L), 0, ~ O08), (3.72) where 8 <1 is the thickness of the boundary layer. For a steady axisymmetric flow the dimensionless Navier-Stokes equations (without external forces) are Ca oor a =0, (3.73) 56-a, om me 1 ar tae ar Re ax’ G74) dey ay my 1 HD paid E hy te A - x +0, ae _ r 7 Re a’ (3.75) ox B, (3.76) az If the diameter of the thin core of a vortex is of order 6, then for steady, axisymmetric flows Oy oe ~or O05), 52 7 OW, Re~ OG), ar # ~ O(5), v% ~ O(1), vz ~ O(1), (3.77) and a 4.78 ort a @78) a eer oz) Pas Fa? (3.79) 20% By Dy am So, me oy oe: St tase rat ae ta (3.80) av, oom ane 1 oe Dz eee oa er tae ae Re ar ror The Euler Equations for Inviscid Fluid Flows If we neglect the viscous-force term in the equations of motion, that is, if we put v= 0, we arrive at the Euler equation (without external forces) B g.yp- tb a +@-Vie= — oP. (3.82) Solutions of this equation are more general than those of the potential-flow equation because they permit non-vanishing vorticity fields. We shall see in Chapter 5, however, that vorticity is attached to the fluid elements and cannot diffuse. Moreover, solutions of the Euler equation are not solutions of the Navier-Stokes -57-equation as are those of the potential-flow equation. ‘The Euler equation shares with the potential-low equation the disadvantage that the tangential velocity in the boundary conditions cannot be prescribed. Nevertheless, the Boler equation plays an important role in vortex dynamics as we shall see in Chapters 9 and 10. Exercises ‘Show that the energy equation (2.39) is time-irreversible. Use the solutions (3.59) and (3.60) to verify that they do not satisfy Eq. (3.58) when time is reversed. Demonstrate that the solution of Eq. (3.13) has for steady circular streamlines the form ye ata log r+ ay? +037 logr. (3.83) Verify that the flexion potential for the Poiseuille flow in a pipe is B= (p1 —p)/a with py the pressure at the entrance of the pipe. Write the free-surface conditions, Eqs. (2.52) and (2.53), in dimensionless form and find the relevant flow parameters. 6. Use Bq, (3.27) to discuss the significance of the velocity components which contribute to the nonlinear terms in an axisymmetric flow with 0/06 = 0. 7. A ship model is studied in a water tunnel. The kinematic viscosity v of the fluids for model and original is the same, namely that for water, and so is the gravitational constant g. Verify that the constancy of the similarity parameters Re and Fr cannot be preserved. ane4. Symmetric Solutions of the Navier-Stokes Equations for Vortices In the last chapter we truncated the Navier-Stokes equations to make it easier to find solutions. It would be of great value to have analytic solutions (in closed-form or expressed in series expansions) of the complete equations of motion that would not require computer runs to construct solutions numerically. Because of the complexity of the Navier-Stokes equations only a few analytic solutions for laminar vortex motions of relative simplicity are known. Since these solutions are so rare, we will discuss and evaluate them in detail to learn as much as possible about the behavior of vortices. The restriction of our discussion to laminar flows has both advantages and disadvantages. The main advantage is that we have confidence that solutions to the Navier-Stokes equations for laminar flows describe reality with great accuracy. The main disadvantage is the fact that most fluid motions in nature and technology are turbulent, a state for which only less accurate, restricted, or’ semi-empirical methods are available. These methods will be used and applied in later chapters. Nevertheless, whenever laminar flows occur in praxis, their analytic and numerical descriptions are most reliable.* Analytic solutions owe their simplicity to certain symmetry properties. Symmetry is a general concept in physics, to which we alluded in previous chapters. We must caution again that symmetric solutions can become unstable beyond a certain parameter value, An example is the steady laminar pipe flow of Hagen-Poiseuille which is stable up to a Reynolds number of 3200 (based on the diameter of the pipe) and which becomes turbulent beyond that threshold. We will study symmetric solutions of two kinds: universal solutions and self- similar solutions. Their mathematical simplicity matches certain fluid dynamical features. Universal solutions simulate flow patterns that are independent of the viscosity coefficient v or, in dimensionless form, independent of the Reynolds number. * We may call a solution method numerical if the field equations are still in the original form of partial differential equations and are solved directly by numerical methods such as finite-difference, finite-element, finite-volume, or spectral techniques. Analytic solutions are either those in closed form or those expressed in series expansions (as mentioned above); however, more recently solutions of the field equations have also been called analytic, after they have been reduced to ordinary differential equations. -59-Self-similar solutions represent flow patterns without a natural length scale (Fig. 3.2b). 4.1. Universal Solutions Universal solutions are those solutions of the Navier-Stokes equations which fulfill the two conditions in Eq. (3.14). These conditions are expressed by the vorticity vector and do not contain the pressure term. If we want to write the conditions using the velocity-pressure formulation, we are confronted with two possibilities: Big Des A¥p= vV¥= 0 and (4.1) « +V DB= a +vV2 (4.1a) In Eq. (4.1a) the kinematic viscosity can be absorbed in the pressure term. However, the two conditions of Eq. (4.1) would give a multi-valued pressure and are thus ignored (see Exercise 1). For the study of vortices we assume that the fluid motion takes place in the 7, 6 - plane, that is, the fluid flow does not change in the z - direction (‘‘two-dimensional flow"”). Moreover, we assume that the fluid flow is steady, axisymmetric, and that the velocity components v, and v, are zero. Then Eqs. (3.27) and (3.26) reduce to the differential equations uy 1% 2% aaa oe ee 42 Pt op eo (4.2) 2 2% _ Lap Pia pion (4.3) These equations satisfy Eq. (4.1), and their solutions thus represent both universal flows and generalized Beltrami flows. The general solution of Eq. (4.2) is the circular motion % = 7 +hr. (44) ‘The integration constants a and b are determined by the boundary conditions. We note that the nonlinear inertial terms vanish in this coordinate system. This linear property of Eq. (4.2) is physically significant, as it means that only a diffusion process is involved. Eq. (4.3), however, contains a quadratic term which is important for the behavior of the pressure at the vortex center. With v given by Eq. (4.4), we compute the pressure with the aid of Eq. (4.3) -60-“2 4) +20 tos a+ Eee -AD; (4.5) with r, a reference radius, say the radius of a circular boundary line. We could have chosen Cartesian coordinates or other coordinates and obtained the same solution (Exercise 2). However, the simplicity of the set of equations (4.2) and (4.3) would have been lost because we would have been confronted with the two conditions of Eq. (4.1) which in this case are four differential equations for the velocity components and the pressure gradient. The advantage of the special form of Eqs. (4.2) and (4.3) will become even more apparent for the self-similar solutions in Sections 4.2 and 4.3 and for slow motion in Section 13.1. Flow between two rotating cylinders (circular Couette flow) For the flow between two concentric cylinders of radii rz and rp (72 > rz), rotating with constant angular velocities Q1 and Q2, the constants a and b are nH a= (Qy - 20) 3-13’ a b= (Qgr} -Qyr4) (4.6) B-7 Of practical interest is knowledge of the torque which the rotating cylinders impose on the fluid. Let us consider, for simplicity, the inner cylinder at rest, that is = 0. Then, the torque per unit length of the outer cylinder is (see Section 5.6) Mz = (tyg)22nry72 a 2 =(Se- ye 2nr} = dre oid | (472 4 The torque of the fluid exerted on the fixed inner cylinder is the same as M3. Potential vortex For b= 0, Eq. (4.4) represents the potential vortex = (48) +i with a = « being the strength of the vortex. The vorticity is zero except at r= 0, and thus Eq. (4.8) is a potential-flow solution, hence the name potential vortex. Since olevorticity is concentrated at the singularity r= 0, the potential vortex is also called a point vortex or a line vortex. We shall use all three words and will distinguish among them later. The pressure distribution in a potential vortex can be obtained from Eq. (4.5): -pX 4b P-Pe= Poy te -@. (4.9) The potential vortex dissipates in a viscous fluid with > 0. From Eq. (3.30) we obtain the dissipation function = a. =| (40) x Solid-body rotation If we seta = 0 in Eq. (4.4), we obtain the solid-body rotation dp = br 11) with b= © the constant angular velocity. The vorticity is @, = @= 20, and the corresponding pressure distribution is (4.12) Q? Ml The dissipation function ® is zero, and the energy equation is independent of the velocity field if the temperature is T= T(r). Rotating bucket In a rotating bucket, half-filled with water (Fig. 4.1), the water rotates like a solid body. If z is the axis of rotation perpendicular to the earth’s surface, the gravitational force acts with F= —g. Then, Egs. (3.26) and (3.28) reduce to fA Oly ee se | Ia = Ordr—gdz or (4.13) BLigp_ oh er gz+tconst. 02/7/2 can be interpreted as a centrifugal force if the reference frame is fixed to the rotating bucket. At the free surface the pressure is constant according to Eq. (3.51). Hence, the surface is a paraboloid with the origin at the lowest point of the surface: _ ar oir (4.15) <-1--3 Centrifugal force Resultant Fig. 4.1: Water in an open, rotating container. Rankine vortex The two solutions, Eqs. (4.8) and (4.11), can be combined to describe the Rankine vortex with k= Or? (Fig. 4.2) % = Or, @= 20, rSre, Qn? m= SH, WF 0, > Ke. (4.15) If we imagine that the solid-body rotation in the Rankine vortex is produced by a solid rod, we have a means of generating a potential vortex by rotating that rod. The frictional heat produced by the potential vortex must be balanced by the power of the shear stress provided by the rod. According to Eqs. (2.36) and (4.10) the total dissipation which equals the rotation power is, Joav= am. (4.16) Te -63-The total Kinetic energy of the Rankine vortex is only finite, if we limit its extension to 72 < «with 7 > T-. Then, n 2 | ofanrdr= pax’ (log 2,4). (4.17) a ~ 4 @ Velocity CO) qe Distance from | Distance from the axis I the axis Fig. 4.2: Rankine vortex. (a) Velocity distribution and (b) vorticity distribution. We can also imagine the solid-body rotation as a column of fiuid and consider the Rankine vortex as an approximation of a vortex with a core of vorticity. To avoid the question of how the steady vortex is maintained, we assume an inviscid fluid. The Rankine vortex is often applied as a simple model for a real vortex because it has a velocity distribution with 0 = 0 at bothr = 0 andr Helical flow in a straight tube We can relax assumption v,= 0 to v= ¥,(r) without affecting Eq. (4.4). Supetposing v, = const on Yp(r), we obtain screw-type streamlines. For 0,(7), Eas. (3.28) and (3.26) yield peleen a= ay Ser +Alogr+B, (4.18) = 2& 2 dr p= Px ploy toe (4.19) where A, B, C, and ap/az are arbitrary constants, Equations (4.4), (4.18), and (4.19) desoribe helical or spiraling flows in rotating concentric tubes. Rotating flows, which contain an axial velocity component, are called swirling flows or swirling motions. They will be discussed in Chapter 16. | |Elliptical vortex Next to circular fluid motions, elliptical vortices are of great importance as we shall see throughout this book. We consider a linear velocity field in the (x, y)-plane and write with the aid of Eq. (28) by putting Djy=Dy,=—e= const and ©, = 2y= const u=-(y+ey , v= (y-e)x. (4.20) The constants Dy; =~ Dyy have been eliminated through a suitable rotation, We see immediately that Eq. (4.20) is a universal solution because V? 7= 0. The pressure is p oe -e)(x? +42) + const . (421) To study the shape of the streamlines, we compute them from Eq. (3.12), which gives the stream function —2y = (y—e)x? + (y+ ey? = cons! x. or =sS- (4.22) The streamlines are either ellipses or hyperbolas, depending on whether y>e or ‘y 1. If we interprete the two terms on the right of Eq, (4.37) as two vortex fields of different vorticity distribution, we can determine the total vorticity fields and arrive at the result that the two fields cancel each other: cH er Bi s 2 yw . Tres ze riba ~ So aa rdodr =0, (4.38) . 2c _ Lg. t : with 2C/t= T. This means that the Taylor vortex can be interpreted as the super- position of two vortices with equal but opposite circulation. The total energy and the total dissipation are finite. Decay of a vortex with an arbitrary initial vorticity We want to generalize the problem of a decaying potential vortex and ask for a formulation that describes the decay of any circular motion (that is, any motion with * In fact, the n-th derivatives with respect to £ are solutions of Eq. (4.23). This fact is another ‘example of how new classes of solutions can be found by “‘superposition.”” -70- 1circular streamlines) with an arbitrary initial condition. For this reason let us consider a fluid element at the location , x, (Fig. 4.5) with the strength @p(p)pdpdx/2n, where p and x are polar coordinates.* After the time ¢ the vorticity in the area element pdpdy is (Goldstein 1932, Kochin et al. 1964) ps1 —2orcosy, (9) ~ aoe =o iv ay. ‘avi ¢ appay, y p f x Fig. 4.5: The location of a fiuid element described by p, 1. Integration yields 1.3% ptt? ~2prcosy 10 Gane - ay. ot,0= 7 J Joop) pdpdy With the formula 2 cosy . ‘vt - pT Pr ! ot Nay = Info RE) = Anlg( O), ‘we obtain _ 2 _ Tae py Bt => dp. or, )= 578 Joootor lo Fi eap The corresponding velocity field is * The notation p, 7 for a second set of polar coordinates is used only in this section. ooo (4.39) (4.40) (4.41) (4.42)2 2 ryt = Fe 7 how (ple phe #8 1g Pda « (4.43) In(x) = J(ix) is the modified Bessel function of order zero. Decay of a Rankine vortex ‘As an example let us consider the decay of a Rankine vortex with the initial vorticity «99 which is equal to @g = const for p re. The vorticity field at fis oy rr, 20 ae fa pF alr, )= Fe fe lol Fr IpAo . (4.44) with the solution, which is given here without derivation (see Goldstein 1932, Kochin etal. 1964), : Pei a rer alr, )= ae Pi Wy for r2 re, Pan alr, f) = 7 ee 5 o o) for rS1,- (4.45) Both solutions of Eq. (4.45) yield for r= 7, ® ca (te, ) = oe eee (4.46) The vorticity is plotted in Fig. 4.6. The corresponding velocity field is fed ope oy EB SK )| sor rz, 0 Ol oer opr ! ive oe | cota. i 2% = @ toate we SEG ie for rSr;,. (4.47) 720 |s 0 O05 10 15 20 25 30 Fig. 4.6: Decay of the vorticity of a Rankine vortex (Kochin et al. 1964). Decay of a circular discontinuity line ‘A second example* of the application of Bg. (4.42) is the decay of a circular discontinuity line. The vorticity field is zero except at the line r= 7, where the vorticity is infinite but x=" Joppdp = finite. The vorticity at time t is according to 0 Eq, (4.42) fer at er I . (4.48) ol) 4.48) — alr, D= 3 This function is plotted in Fig. 4.7. The velocity field is, according to Eq. (4.43), 2 nd, = KW fp Bt pe mir = Tee fre Tog ar (4.49) * Other examples can be found in Goldstein (1932), such as the decay of a potential vortex inside a circular boundary, the decay of a circular discontinuity line inside a coaxial circular boundary, the decay of a circular discontinuity line outside a coaxial circular boundary, and the decay of a vortex with an arbitrary initial vorticity inside a circular boundary. -TB-Fig. 4.7: Decay of a circular discontinuity line (Kochin et al. 1964). This integral agrees with that in Eq. (4.44), except for the exchange of r and 7 and and x/r, and the solution is thus corresponding to Eq, (4.45) | coe Ter | air t)= Se 2G WG fer rSte, } Pan Joos el, AR a tey tet ar, = *|1-e EEG] for rere. k=0 Decay of a vortex grid (4.50) | > Anexample of symmetric solutions, which are periodic in space, is given for the = +Acosxsinye™, | decay of a vortex grid. G.I. Taylor (1923a) gives v= —Asinxcosye™", lines are not circular and are shown in Fig. 4.8. -74- (4.51) with A an arbitrary constant, and with 7 the size of the quadratic grid. The stream-Fig. 4.8: Decay of a vortex grid (from Whitham 1963, by permission of Oxford University Press). ‘The stream function and the vorticity are w= —Acos xcos ye , (4.52) @, = —2A cos xcos ye™ . (4.53) Thus, streamlines and equivorticity lines always coincide. ‘The pressure is P-Pe= fate (sin?x + sin?y) . (4.54) Cartesian coordinates are a natural choice for constructing a solution of the decaying vortex grid. Since the nonlinear inertial terms are not separated from the diffusion terms as in the case of Eqs. (4.23) and (4.3) for polar coordinates, we can’ study here decaying processes with different rates in the same differential equation. The nonlinear terms again balance the pressure gradient. They decay as fast as exp (-4vé). The linear diffusion terms, on the other hand, decay at the slower rate exp (-2v#).* These different decay rates within one equation reveal clearly the role of the Reynolds number in the dimensionless formulation of the equations of motion. The statement in Section 3.6 is confirmed that slow motion with Re —> 0 is described by neglecting the nonlinear inertial terms. The role of the pressure gradient, however, must be examined anew (Section 13.1). * When the solutions are being derived, nonlinear terms can be avoided by using the stream-function formulation, Eq. (3.13) (Taylor 1923a). The nonlinear terms drop out because w= f(y). a54.3. Vortices Perpendicular to a Slip Surface If we look for further axisymmetric solutions of the simple form % = v9(, 1), we must keep in mind that the restriction v9(r, #) forces also v, and 0, to be of the form 2,(r, t) and 0, = 2f(r, 1). This follows from Eqs. (3.25) and (3.27). Let us look at the Navier-Stokes equations (3.25) through (3.28) in cylindrical coordinates with the assumption of axisymmetry: a, 2% Ov, tte or” a =0, (4.55) dv, av, 097 av, Lap Pu 1a, % Fy or + Y oF “2 + ow) ) (4.56) ay, Bu Orr ao, Ho, au my Fy Be | Me rte, AM _ TM 19% _ MH SM a ar te tae Ge te ee et ae OD Fu, 19% , Py at yr ae ar at ). (458) Although , = zr, ) indicates that the plane z= 0 may represent a surface, it certainly cannot be a nonslip surface because of the nonvanishing velocity com- ponents v, and % at z= 0. However, if we accept a slip surface (whatever that slip may be), we can start the investigation with the simplest solutions, that is, the universal solutions: 7% = K/r and Up = Qr. In addition to V*ug = 0, these universal solutions must satisfy the equation au du DyD Be en weds Ye (4.59) at or FT 0. (4.59) The stagnation-point flow v, =~ ar, 0, = 2az together with point vortex v9 = K/r fulfills Eg. (4.59). Thus, Ug = 0, Sar, 0, = 2az. (4.60) xn The arbitrary constant a can be either positive or negative, meaning radial iriflow or outflow, respectively. Solution (4.60), however, is not realistic for a viscous fluid because of the singularity at r = 0. The other universal solution, that is, the solid-body rotation 0 = Q(¢)r superposed on the same stagnation-point flow, also fulfills Eq. (4.59), from which follows 168on = 2Q > Q= Qe, (4.61) with Qo the initial value at t= 0. Here, the radial inflow with a> 0 increases the angular velocity, whereas radial outflow (a < 0) eliminates the fluid rotation. We can use this solution only locally because it is unbounded for r— - and t > 0». The plane = @, is called the meridional plane. It is a streamsurface for the stagnation-point flow with the stream function y= azr?, This part of the fluid motion is called the meridional flow (Fig. 4.9). Fig. 4.9: Definition of meridi- onal plane and meridional flow. In order to find other solutions, more realistic than those of Eqs. (4.60) and (4.61), we investigate self-similar solutions with the aid of a recipe by Bellamy-Knights (1970) on the’ basis of equations (4.55) through (4.58). The restrictions o,(r, #) and 2; = 2f(r, t) suggest a stream function g(r, t) of the form =k pu 2 We Ue Riana (4.62) which satisfies Eq. (4.55) automatically. Zero radial velocity at r= 0 requires the boundary conditions ¢ = dg/dr= 0. Introduction of g into Eqs. (4.56) and (4.58) yields Lm _ dvdr ie), % 13% pe arrae 22)? + ra’ 463) 2, ix _, - Yop 2 hey, LHe, 821M) (Ap (4.64) paz yor or y or r oret or rar The right-hand side of Eq. as 63) is a function of r and t only, so that by differentiation with respect to z we obtain 0*p/drdz = 0. Then, Bq. (4.63) can be written Tielope paz 7 2F- (4.65) If we compare Eq. (4.65) with Eq. (4.64) we obtain vd, 9 Lig, 1g, gd/1dg) Laep_ yo aes ary Brot * torr a rar FO OS Eq. (4.57) yields, when expressed by the strength K = r 2128 OK, £9K_y, arr ar at + oe ). (4.67) ‘We look now for a group of self-similar solutions with the similarity parameter € (r, #) that obeys the ordinary differential equation for «(Q) 2. oe +0(0) 7 =0, (4.68) to which Eg. (4.67) can be reduced if the condition (1%, % got _ aby Wary ar at rar ONG? ) is fulfilled. We obtain this condition when we replace the derivatives 3«/0t and OK/Or by dx/@l and consider 0: as defined in Eq. (4.68). The procedure for finding this group of self-similar solutions involves the following steps: Eq. (4.66) provides the stream function g and, with g, we obtain v, and v, from Eq, (4.62). We search now for solutions of Eq. (4.69) for ¢ that also determines o(¢). Finally, Eq. (4.68) yields x, and Eqs. (4.63) and (4.65) yield the pressure. As a simple example, let us assume a= 1. A similarity variable that satisfies Eq. (4.69) is ¢= r”/4vt, and this leads to Lamb's solution (4.26). The one-cell solution of Burgers ‘A solution of Eqs. (4.55) through (4.58), which remedies the shortcoming of Eq. (4.60) and avoids the singularity at r= 0, is Burgers’ solution (1948) %= —(1-e""), a> 0 (4.70) r Dy = —ar, V2 = 2az. 4.71) ‘We can obtain this solution from Bellamy-Knights’ scheme by assuming the existence Seof the meridional flow, Eq. (4.71), and hence g= —ar?. The special form of g suggests a similarity variable C= cr’ which we insert into Eq. (4.69) and obtain c= a2vanda= 1. ‘The tangential velocity, Eq. (4.70), now has a finite value in the vortex core, which reduces for the new similarity variable = ar? /2v = to the potential vortex. For a finite value of C, vorticity of the amount @, = aes + O= @ (4.72) diffuses and is balanced by the radial influx and axial outflow of the meridional flow, Eq. (4.71). The constant a can have only positive values and is a measure of the radial influx. The meridional flow, depicted in Fig. 4.9 with its image counterpart, is independent of vg, that is, the strength x of the rotation does not affect the infiux and outflow. However, the stronger the inflow, the faster is the diffusion of angular momentum over the distance from the axis. Itis noteworthy that the meridional flow, Eq. (4.71), is a potential flow that - in the context of Burgers’ solution - is meaningful only for v > 0. In other words, it is an example of a laminar potential flow (see Section 3.1). If v= 0, however, Eqs. (4.70) and (4.71) represent a potential flow for an inviscid fluid, that is, Eq. (4.60). The meridional flow, Eq. (4.71), may be considered the lower part of a circulatory motion, called a “‘cell.”” At the center of the vortex r= 0, @, has an extremum of ak/, which is twice the angular velocity of the solid-body rotation of the core a = er or Q= oa for we ae (4.73) The pressure field is given by 1- x ax} j A P(r, 2) = pO, +p J ( Pde a a’(r? + 427). (4.74) ‘The temperature field with dissipation was discussed by Rott (1959). He also studied the transient state (Rott 1958) Pd , [s-e0|-3- char 1, 0,5 ar, 0, = 2az, (4.75) with b an arbitrary constant. Eq. (4.75) can be found with Bellamy-Knights’ assumption C= c(t)r? which transforms Eg. (4.69) into an ordinary differential equation of first order. Solving this equation leads to Eq. (4.75). -79-For a meridional influx toward the z-axis a> 0, it follows that exp —2at < 1 and 1+bz 0. Burgers’ steady-state solution is obtained when >, Thus, a small initial value of @ at f= 0 results in the development of a concentrated vortex. If the meridional flow turns away from the z-axis, that is, if @< 0, then exp—2at2 1 and 1+4b< 0. Diffusion of angular momentum is in the same direction as the convection of angular momentum. The rotation ceases for t > »», since Up > 0. The two-cell solution of Sullivan ‘The meridional flow of a vortex that has two stagnation points between 0S 1 < 2 may be considered the lower part of two circulations or cells (Fig. 4.10). Such a “two-cell vortex”” is described by Sullivan's solution (1959)* Z Fig. 4.10: The meridional flow of Sullivan's solution. venars Lae), (477) 0, = 2az (1-36), (4.78) with = “Pos -x43 j (1-e*)slds | dx, (4.79) * A three-cell solution was discussed by Donaldson and Sullivan (1960). -80-H,, = 37.905. As in Burgers’ solution, the similarity variable is ¢= ar”/2v. The function H(0) is plotted in Fig. 4.11. 40 HO 20 0 10 ¢ 20 Fig. 4.11: The function H(C) according to Leslie and Snow (Copyright © 1980 AIAA - Reprinted with permission). We can rewrite Eqs. (4.77) and (4.78) in the following way 2, =ar a-Za- “y), 4.80) 0, = 2az (1-3€%). (4.81) ‘The cylindrical area, where v, = O for r> 0, is from Eq.(4.80): 1= 2(1~e), that is, at C= 2.821. Eq. (4.81) determines the location of v, = 0 to be at C= log3= 1.099. "The advantage of using ( is evident. As in Burgers’ solution, z= 0 is a slip surface, and vp is decoupled from the me- ridional flow, that is, the meridional flow is independent of vg. The vorticity field is 2 : ba —*o_ oH. = 0, og=— Ane ™ , Oe = (4.82) The vorticity lines are of spiral type as depicted in Fig. 4.12. -81-Fig. 4.12: Typical vorticity line of Sullivan's flow (Perry and Chong 1987, reproduced, with permission, from the Annual Review of Fluid Mechanics, Vol. 19, 1987, by Annual Reviews Inc,). The pressure is A ptr, 2)= pO, 0)— © | data? +027? + fq -ewmy +of % ar. (4.83) 0 It is worthwhile to compare Sullivan’s two-cell solution with Burgers’ one-cell solution (Fig. 4.13). @ ©) o5 1 4 1 1 tL 7 Fig. 4.13: (a) The meridional flows of the one-cell (dashed lines) and the two-cell (solid lines) vortices of Burgers and Sullivan and (b) the corresponding tangential velocity distributions (from Donaldson and Sullivan 1960). ocFor largé ¢ Sullivan's solution approaches Burgers’ solution. For small ¢ the tangential velocity distributions differ considerably. The rotation within the inner cell is very weak. The solid-body rotation near the axis is a 0% = a for Ce 1. (4.84) This velocity is H..-times weaker than the local rotation of the one-cell vortex, Eq. (4.73). We notice that the two-cell solution has features which are known from observations of tropical storms such as hurricanes. The “eye” of a hurricane has a two-cell structure and is cloudless because of tHe downward direction of the meridional flow. The calm center can also be explained by the very small value of vy near r= 0. We must keep in mind, however, that the real mechanism of a hurricane, which is a heat engine embedded in the general atmospheric circulation, is much more complicated (Section 20.3). The unsteady two-cell flow of Bellamy-Knights The unsteady case of Sullivan’s solution was presented by Bellamy-Knights (1970), Journal of Fluid Mechanics, Cambridge University Press. This solution is of particular interest because all previous solutions are included and can be interpreted as asymptotic cases. The discussion reveals some basic flow behavior with possible applications to columnar atmospheric vortices. We try the following form for ¢ g= alt? + b()l1 ee" (4.85) and introduce it into Eq. (4.66). This yields a yg Ley a 2a? ZF = 0, (4.86) fb _ bay We = ap 700 2Pe= 0, 487) a +2ac +4vc? = 0. (4.88) As in the previous cases, we assume ¢ = c(t)r. If we substitute and Eq, (4.85) into Eg. (4.69), we arrive at 1 bye) e age tm -e) = Ave, 6| | | from which we get, with Eq. (4.88) and 6, o (4.89) Since we have assumed that o.= oC), b must be a constant. Eq. (4.87) yields a=- Fonte) . : (4.90) Eqs. (4.88) and (4.90) determine 1 de oe (4.91 ar (491) where the constant 1 = 2/3(6v —b). The solution of Eq. (4.91) is ae c= Mop (4.92) with 1 as an integration constant. The similarity variable is then C= 7/(AE+W). With Eqs. (4.85), (4.90), and (4.92), the solution for the meridional flow is 2 -»|-22 ater =| Sire tte (4.93) 1 Doonan =p|-t—t-44q_ eee 2, | gen tr e i), (4.94) » (1 350 2 AL, Men oon [2 e |. (4.95) The tangential velocity is obtained from Eq. (4.67) with ot given by Eq. (4.89) 0 Hypv(O) = eae 4.96) 88 T Hypv(o) 496) with % yes Hypy(Q) = Jexp wnt bh fas | ax. (497) i 2a 8 oa j { | |We compute the pressure from Eqs. (4.63) and (4.64) as ee = fhe o? pO, 0, t)] = [e4 (At + yw)? Foye ep 2 : -5 e Mtn +) Ear. (4.98) 0 With 2= 0, that is, b= 6v, we retrieve the steady-state solution of Sullivan. In this case, it is w= a/2v. If b= 0, Eq, (4.89) yields = 1, and we have the one-cell solutions of Burgers and Rott. If, in addition, t= 0, the coefficient for the time- dependent term, 2 assumes the value 4v, and we recover Lamb's solution (4.26). For 1.# 0 and t= 0, Sullivan's solution may be considered the initial state of a decaying process that ends, for € > e, at the potential flow v, = b/rand vg = Ko/t. As in the steady-state case, the time-dependent solution yields, for the location of O and 0, = 0, the conditions 0, = 0: C= 2821, m= 0: C= log3, (4.99) where now (= rear With the passing of time, the cylindrical boundary of the inner cell expands outward at the speed dr/dt = 2.821 Wr. 4.4, Assessment ‘Symmetric solutions of the Navier-Stokes equations, presented in this chapter, can be divided into those which describe a pure diffusion process and are thus generalized Beltrami flows (Sections 4.1 and 4.2), and those which display a balance between nonlinear convection and linear diffusion (Section 4.3). Circular motions proper,* which depend only on 7 and #, are determined by pure diffusion governed by the linear partial differential equation * Noncircular motions like the vortex grid can have vanishing nonlinear terms (Weinbaum and O'Brien 1967). -85-‘The pressure force is balanced by the quadratic centripetal acceleration term a _ 2% oor’ ‘These equations immediately reveal the following properties for laminar circular vortices: In steady motions, the vorticity flux vdw/dr vanishes, and only solid-body rotation is possible, since a singular solution at r= 0 is excluded, Circular boundaries are necessary either to limit the flow to a finite region r= r, $ oe, or to provide the inner boundary for the potential flow (Rankine vortex). In unsteady motions, there is a vorticity flux toward or away from the center r= 0 that brings or carries away vorticity. Hence, a time-dependent extremum of vorticity exists, and the flow rotates like a solid body near r= 0. A particular feature of curved flows is the occurrence of the centripetal accel- eration term v4/r in the r - component of the Navier-Stokes equations. This is the only (quadratic) convection term in pure circular motions since the other convection terms are zero because they include the vanishing velocity components v, and 0, The linearity of the diffusion equation does not mean that the flow is restricted to slow motion, that is, to Re < 1. For steady flows the flow field is independent of Re (universal solutions) and virtually independent of Re for unsteady flows, except that the rate of decay depends on Re. However, the rate of decay for a laminar vortex with a large Reynolds number is extremely slow (see Exercise 7). The most important solution is that for the decaying potential vortex, Eq. (4.26), since decaying straight laminar vortices, observed in experiments and numerical computations, follow that law. We cannot infer from the linearity of the vorticity-transport equation that the original Navier-Stokes equations are linear. They do not have to be, but the differ- ential equation for the pressure gradient is separated from that for the velocity. Although the difference between these two formulations seems to be insignificant, it becomes important for the limit of vanishing Reynolds number,.that is, for slow- motion vortices as discussed in Section 13.1. If no rotating walls are available to provide the energy for maintaining the fluid flow, convection is necessary to counterbalance diffusion. Examples are Burgers’ and Sullivan's solutions. The spatial diffusion is r~ Yv/a. These similarity solutions do not have a natural length scale due to their similarity behavior. The characteristic length for the Burgers and Sullivan solutions is Ww/a . -86- |The existence of two-cell flow structures is of practical interest since they occur in atmospheric vortices and in vortex chambers. The usefulness of these solutions, however, is limited by the independence of the meridional flow from vg, by the free slip on the horizontal plane z= 0, and by the unboundedness of the meridional flow (see Chapter 15). We have pointed out on several occasions that we must pay a price for constructing symmetric solutions. In their simplicity due to the imposed symmetry restrictions, there is built in either the | inability to satisfy all boundary conditions and to predict flow instabilities, or the | occurrence of flow singularities. Despite these shortcomings, symmetric solutions are instructive if we recognize their limitations. Exercises 1. Explain the discrepancy between Eq. (4.4) and Eq. (3.83). 2. Write Eqs. (4.2) and (4.3) in Cartesian coordinates and verify the solution 09 = br, p= 0.5 pb? (r? — 72). 3. Compute the volume flux Q through a straight circular pipe which rotates with the constant angular velocity Q. The average velocity is w. 4. Find the temperature field of a Rankine vortex. 5. Determine the constants A and B in Eq. (4.18) by assuming nonslip at the walls 7 and 72. 6. Determine the streamlines of the swirling flow vg(r) and v, = const. Discuss the difference between vg = Gr and vg = K/r. 7. Compute the half-life of a decaying potential vortex for r) = 1.cm and 10 m in air with v = 15:10 m/s at T = 20°C. 8. Show that the radius of a decaying vortex, in which half of the strength is located, is r= 1.665WE. 9. Subtract Ko/r from vg in Eq. (4.26) and discuss the new transient state. 10. Derive a formula for the time span in which the Rankine vortex decays from «9 to @/2 at the center r= 0 with the aid of Eq. (4.45) and compare this formula with Eq. (4.29) for the half-life of a decaying potential vortex. -87-Hermann von Helmholtz (1821-1894) during his stay in Washington, D.C., September 1893 (photograph by M. Brady, courtesy Inter Nationes e.V., Bonn, Germany). Pes. Si ee5. Vorticity In several places we have mentioned the important role of vorticity in fluid dynamics and particularly its significance for vortex motion. In this chapter we will investigate this importance in a more systematic manner. First, let us summarize the knowledge of vorticity we obtained in previous chapters: i: 2, Vorticity is defined as the curl of the velocity vector: @= curl@. Vorticity appeairs in the decomposition of a velocity field as the rotation term in Eq, (2.8) and indirectly in the vector potential in Eq. (2.1). Vorticity may be interpreted as a measure of the rate of rotation of a fluid element. The Navier-Stokes equations can be written in an alternate form as the vorticity-transport equation (3.10). In contrast to the velocity field, the vorticity field is invariant with respect to a Galilean transformation. It is also invariant with respect to a constant rotational transformation except for a constant (Section 1.5). From the definition @ = curI@ it follows that the vorticity field is solenoidal: dv@= 0. (6.1) Analogous to the continuity equation div@= 0, Eq. (5.1) reveals that vorticity tubes (like stream tubes) cannot end inside the fluid, that vorticity tubes, therefore, obey the law stated as Eq. (2.17), and that they can branch (Fig. 2.1). We will have more discussions on this subject in Section 5.3. Potential-fiow solutions, defined by @= 0, are solutions of the Navier-Stokes equations. However, they .cannot satisfy (with a few exceptions) boundary conditions for the tangential velocity such as those on nonslip and free surfaces. The inability to prescribe these boundary conditions indicates that vorticity is necessary to overcome this deficiency. Hence, a study of potential flows in the absence of vorticity can give, indirectly, insight into the meaning of vorticity. But, more importantly, certain singularities in a potential-flow field can be considered as places of concentrated vorticity. Such cases make it possible to study aspects of vortical flows that do not require detailed knowledge of the vortex structure, but to explore, as a unit, an assembly of mahy vortices with their interactions (Chapter 7). A vortex in a viscous fluid rotates like a solid body at the axis and thus requires vorticity: There is no vortex without vorticity. -89-In the following text we will extend our knowledge of vorticity by examining more closely its field properties, in particular vorticity generation and spreading. Before we' proceed, let us agree on the following designations: A vorticity line is a line tangent to the vorticity vector (analogous to the streamline in a velocity vector field). In two-dimensional flows the vorticity line is perpendicular to the plane of motion. If we connect in this plane points of equal vorticity, we obtain an equivorticity line, In three-dimensional flows a bundle of vorticity lines is enclosed in a vorticity tube. If the whole field is irrotational except for an isolated vorticity line, this line is then called a vortex line. An isolated vorticity tube is a vortex tube, which encloses a vortex filament. The Rankine vortex is an infinitely long straight vortex filament. 5.1. The Vorticity-Induction Equation If the velocity field is given, the definition of vorticity @ = curl provides us with the vorticity field through differentiation. What does the reverse function look like, if the vorticity field is given, and the velocity field is sought? The answer requires an integration process, and the solution is known in electrodynamics as the Biot-Savart law, in which the electric current density and the magnetic field play the roles of vorticity and velocity, respectively. We have seen in Section 2.2, Eq. (2.1) that the velocity field can be written in the terms of a scalar potential and a vector potential. Applying the curl-operator to Eq. (2.1) results in Eq. (2.3), repeated here VA= -3. (5.2) For three-dimensional flows the solution to this equation is (Courant and Hilbert 1962, p. 245) AM= %! BEI gy, (53) with?’ the position vector of a field point,’ the position vector of a volume element of vorticity, and ¥= 7-7” the distance between field point and volume element of vorticity withs= IP1 (Fig. 5.1). The integration is carried out over the region of vorticity. For two-dimensional flows, with A, = y, the solution to Eq. (5.2) is aL L W= 3p! a: log 4A (5.4) -90-Fig. 5.1: Sketch for the description of the vorticity-induction equation. (it may be recalled that dA is the surface element and is not to be confused with the vector potential A ). The velocity induced by the vorticity field is for three-dimensional flows because C= curl 1 an curl J Bw, (5.5) and for two-dimensional flows correspondingly. According to Eq. (2.1), an irrotational and solenoidal velocity vector V@ must be added to Eq. (5.5) to satisfy boundary conditions. We can write Eq. (5.5), for vanishing vorticity at infinity, in the form po ay 2 B= -3 Jaxwav = 5 Jax aw. (5.6) This is the vorticity-induction equation. We may point out that the relation between velocity and vorticity vectors is based on the definition of vorticity and that it has nothing to do with the equations of motion or the constitutive equation. In the text that follows we will make ample use of the vorticity-induction equation. ‘We may also mention that the vorticity field may occupy only a part or the whole of the space over which the velocity field is nonzero. For instance, in a potential vortex, U4 = W/r, the vorticity field is confined to the singular point r= 0 with @, = «. Another example is the boundary layer in which the vorticity is concentrated in an otherwise potential-flow field. In pipe flow, on the other hand, the vorticity and velocity fields coincide. A confined vorticity field, which exponentially diminishes away from its boundary, is called a compact field. -91-ee 5.2. Vorticity-Transport Equation If the flow field is unknown, we must solve the Navier-Stokes equations (2.37) and (2.38). They can be written in another form by eliminating the pressure through the jatroduction of the vorticity vector. This means that the flow-field description which is based on the vorticity field does not require explicit discussion of the pressure field. This replacement has been shown in Eq. (3.10) and is repeated here. ‘Applying the curl-operator to Eq. (2.38) means that the terms with the grad- operator vanish, that is, Vx V (p/p +3” 2) = 0, and we obtain: BB cui x= (5.7) a +@-V)G-G-V)V= (6.8) B_@.vr- vv? G+culF. (9) ‘This equation is called the vorticity-transport equation (or ‘‘vorticity equation"). The complete set of incompressible fluid-fiow equations which are based on the vorticity includes div@= 0, B= curl, and the boundary conditions. The velocity vector can be replaced by @ with Eq. (5.6). This formulation is called vortex dynamics. Thus, for describing incompressible fluid motions either the velocity-pressure formulation (Navier-Stokes equation) or the vorticity formulation (vorticity-transport equation) may be used. They are equivalent if we accept the conservation property of momentum and vorticity as equivalent.* ‘The vorticity equation reveals immediately the following properties: 1. Forcurl P= 0, vorticity cannot be generated within the fluid. 2. We can explore the meaning of the nonlinear term (@ V)@ when we consider that @-VW8=B-D. (5.10) + The question of which system of fluid dynamical axioms i more fundamental is an idle one as long as they cover the same range of validity. For an appreciation of the merits of vortex dynamics see the Epilogue. -92-This relation can be verified immediately in Cartesian coordinates for the x- component: au a fon ge oo une ow ae +2 ae * ay) 2 ae * 2) @-V)u= ong + oy 5H + 0,58 (5.11) The y- and z-components are expressed correspondingly. Eq. (5.11) can be interpreted physically as the stretching and twisting of vorticity lines. The first term is in the direction of @, and indicates stretching (which is a change of @, itself). The other two terms represent twisting to generate @y ‘The appearance of the nonlinear term (”V)7 in the vorticity-transport equation (5.9) is unique in the sense that it does not occur in the other forms of the Navier-Stokes equations. The term (@- V)@ is responsible for important features in vorticity dynamics that will be outlined in the course of this book. ‘We may add that this term vanishes for two-dimensional flows since @- V =0 so that the flow features attributed to it get lost. Therefore, two-dimensional flows can present only an incomplete picture of the vorticity concept. For general orthogonal coordinates, the components of (@- V)@ are not as easy to obtain as those for Cartesian coordinates. Because of the importance of this term, the components of (@-V)7 are given for both cylindrical polar coordinates and spherical polar coordinates. In cylindrical polar coordinates (r, 0, Z) they are: ayy Lote _ BH) yg a OT ay ta: a a (@B- Wy = oS He 4 oye 05%, (6.12) a, 1%, &, (B- WB, = 0, S* + oy Pr eae and in spherical polar coordinates (R, 8, 4): 93 -avp 12R % 1 dR (@B-VIAg = OR SR + OCR Gg RT Kein’ an R’ 8 | og ¢ A 220. 4 PR 1_ 2% [(@-V) Clq = Op OR 1K 59 Rt Ring ae 7 Roe OI) FR gh SR 1% | PR, Me F [@- Wth = ORF + OR GG) + Ring et RT RM Now it is not difficult to write the vorticity equation in those coordinates by using the corresponding parts of the Navier-Stokes equations, given in Section 3.2. According to Eqs. (5.7) and (5.8), the two nonlinear terms in the vorticity, transport equation (5.8) are combined in the Lamb vector, by curlL = curl(@x). Nonlinearity thus can be suppressed by reducing T, that is, by trying to make the two vectors @ and @ parallel. This procedure for reducing nonlinearity plays an important role in the understanding of pattern develop- ment in turbulent flows (Chapter 18). In Section 3.1 flows with vanishing Lamb vector have been called Beltrami flows and those with curl L = 0 generalized Beltrami flows. From the identity 3 OC VG= BxB+ vw (5.14) it follows that 4 feav= J [Pan ace dA. (5.15) We obtain this result by applying the curl-operator to dit = dat + (@-V)7, that is, curl a = - + curl (BxD) , (5.16) and by using Gauss’ divergence theorem. For vanishing velocity at-the boun- daries of a finite space or, for vanishing normal components of velocity and vorticity at infinity, Eq. (5.15) yields 4faw= 0 or [@av= const. (6.17) -94-The total vorticity in a two-dimensional region is constant, and this constant is zero for a three-dimensional region. An example is the total vorticity field of a decaying potential vortex, Eq. (4.25). We note that the word decay is used in the sense of temporal spreading of the core and not in the sense of vanishing vorticity which occurs when positive and negative vorticity annihilate each other. Spatial spreading, as exemplified by Eq. (4.72), is called simply spreading. Condition (5.17) can be tightened for two-dimensional flows initially without total vorticity. Then, Jeav=o. (9.18) ‘An example is the occurrence of two decaying potential vortices of opposite rotation. Over the course of time, Eq. (5.18) remains valid. The Taylor vortex, Eq. (4.35), whose total vorticity is zero, can be interpreted as the superposition of two vortices of opposite sign but with different decay rates, Eq. (4.38). The transfer of vorticity from one fluid element to another is only possible through viscous diffusion. In two-dimensional flow the vorticity-transport equation is reduced to a single scalar equation 2g. Vio=vV2o, (5.19) which, together with the Poisson equation for the stream function, Vey = ©, (5.20) constitutes the stream function-vorticity formulation of the Navier-Stokes equations. . These two scalar partial differential equations played an important role in computational fluid dynamics when almost all calculations were performed for two-dimensional flows because of the limited size of computers at the time. We notice that the change of the flow field is controlled by dw/at, that is, by the finite change of vorticity compared to the infinite change of the pressure (which is the infinite sound velocity). Because of its importance, we will write Eq. (5.20) in cylindrical and spherical polar coordinates. In the , z-plane of a cylindrical polar coordinate system we obtain # 1a 1% er ete ag (5.21)in the r, z-plane of a cylindrical polar coordinate system al 12 PbO) EEE aay: (622) in the R, @-plane of a spherical polar coordinate system 1a [pa],12 [rw Pein & [rz] +33 sind a | _ (5.23) 5.3. Vorticity Theorems of Helmholtz For an inviscid fluid, the vorticity-transport equation (5.9) is truncated to Helmholtz’s equation (1858) Bes 0, (5.24) if we consider only conservative forces. A fluid element with zero vorticity at f= 0 cannot acquire vorticity later. The solenoidal relation divB= 0 (5.25) is not affected by the inviscid-fluid restriction. If we write the integral version of Eq. (6.25) fora tube, analogous to Eq. (2.17) for the velocity, we arrive at the statement that the integral over the product of the vorticity component normal to the entrance area of the tube times this area must equal the corresponding integral at the tube’s exit (Fig. 2.1). This is the First Helmholtz Theorem (1858). For thin tubes with constant vorticity in the entrance area A; and exit area Az the theorem yields @,A1 = @2A2 (5.26) It has often been stated that vorticity tubes can ‘“‘end’” only at a boundary. This statement requires closer scrutiny. When we consult Fig. 2.1 (which will now represent a vorticity tube) and apply the ‘“‘continuation’’ idea to the vorticity tbe according to Helmholtz (1858), we can imagine the boundary only as an interface or as a free surface. A waterspout over a sea surface is an example. The vorticity tube of the atmospheric vortex continues into the water beneath. Another example is the vorticity field of a Totating stick of finite length. The stick is a vorticity tube with solid-body rotation that continues into the surrounding medium by spreading out the vorticity lines which then close. How do we interprete a vorticity tube which -96-approaches a nonslip wall? We commonly speak of a tornado ending on the ground (See Section 15.5). In this situation, only a singular line ends at the nonslip boundary while the rest of the tube spreads out to a thin sheet of vorticity (or vice versa, the sheet concentrates to a tube). In this sense we may speak of a tube ‘‘ending”” at a boundary, but this interpretation is not generally accepted.* Eq. (5.24) simplifies for two-dimensional flows to do _ 7 o (6.27) Eqs. (5.24) and (5.27) say that vorticity can neither be created nor destroyed in an inviscid fluid. Moreover, vorticity is attached to a fluid element. However, Eq. (5.24) does not reveal this conservation property immediately (except for zero vorticity) as does the two-dimensional case, Eq. (5.27). This conservation can be better presented in a Lagrangian form. In fact, with this approach we will arrive at a profound explanation for the conservation of vorticity. For this reason we multiply Eq. (5.24) by -VG, where G is an arbitrary quantity which is conserved, that is, dG/dt = 0. Because 4G _ da, “VO vie. fyc+Vv0-ve=0, (5.28) we obtain £@-¥G)=0 (6.29) 4 . : Notice that @- VG is a scalar in which the component of @ parallel to VG is involved. We extend Eq, (5.29) in that we replace the scalar G by a vector G with dG/dt = 0. ‘We may also recall from Section 1.1 that the total derivative d/dt is identical to the Lagrangian (2/0t),. Then, according to Eckart (1960), 4¢8.v@)=(2)@-V@= ae v@= Sue VG)=0. (5.30) Eq. (5.30) is the most general form of the conservation law of vorticity for an * A real dilemma occurs if we assume a solid slip boundary. Then the continuation concept cannot be applied. A possible explanation might be sought on the molecular level that the dynamic viscosity 1 changes. -97-incompressible inviscid fluid. G can be interpreted as a {"label’’ or a “‘tracer’” when we keep in mind that dG/dt = 0 means attachment of G@ to a fluid element (Salmon 1988). The quantity @- VG may be called generalized vorticity. As any conservation law in classical mechanics, the conservation law of vorticity can be related to an invariance (symmetry) property with the aid of Noether’s theorem, in this case the invariance with respect to relabeling of fluid elements. However, we anticipate this important relation and will pursue its derivation in Section 20.1. Relabeling invariance means that the velocity is not affected by change in labeling of fluid elements. This property makes the Eulerian flow description much simpler than the Lagrangian description. Simplicity here refers not only to the advantage of Eulerian patterns, but also to the shortened flow description in an Eulerian frame. ‘An important special case* of Eq. (5:30) is G = 7 that is, the identification of the components of G with Lagrangian coordinates. Then, Eq. (5.30) reads 41@-v= | Va=0, (5.31) which is the sought equivalent to Eq, (5.24), Eq, (5.31) can be integrated immediately and yields B-Ve= GVO, (5.32) or written in Cartesian coordinates 22, 9 ML Or 5y + Ov ay Or ag 7 Ome” Oo cee oe = or Sy ty 5y te g2 = OW” (633) a 0,8 + 08 + 0.55 = On. The three initial components @g., @gp, and ge are integration constants. Exchanging the dependent and independent coordinates of Eq. (5.33) yields B= Vir. (5.34) > The designation “special” is actually superfluous since all G's are “Lagrangian coordinates.” -98-This is Cauchy's formula from 1815 (1827), although it was not written with the symbol @, and its importance for vorticity theory was not yet recognized. Eq. (5.34) is in Cartesian coordinates ®, Es cog LE + 09, 2% = age Bap + OM Ae’ = oo oy ou Oy = Ope Sh + Ow Sy + Ove SE» (5.35) oe ops % Or = Ooe'ag * Ora + POE AC * At E= 0, the initial data are yp = @pa, Op = py, and Wy, = Wp,. If we compare the components in Eq. (5.35) with those of the position vector of a fluid element Ox Ot, x(a, b, 6 #)-200,0,0,1)= Eas Mos He, va, b, c, t)-y(0, 0,0, #)= May My, OY, oe, (6.36) = ag B54 Ke, 2a, b,c) 20, 0,0,)= Fas Bos He we observe that the components of the vorticity vector could be used as Lagrangian coordinates of a fluid element. This means that the vector @ is attached to the fluid elements and that the length of the vorticity vector changes with the distance of neighboring fluid elements on the vorticity line; to be more concise: vorticity lines are formed by the same fluid elements, they are material lines. This is the Second Helmholtz Theorem (1858)* (see also Oswatitsch 1959, p. 37, and Serrin 1959, p. 163). We mention that Eq. (5.24) can be directly derived from Eq. (5.34) by differentiation with respect to (Saffman 1992, p. 17). We apply this theorem to vorticity tubes. For these tubes the theorem holds that the integral over the product of the vorticity component normal to a cross-section times this area remains constant as the tube moves with the fluid. This is the Third Helmholtz Theorem. A simple proof will be given in Section 5.7. * The notation of First, Second, and Third Theorem is somewhat arbitrary. Helmholtz did not use this distinction,Going back to Helmholtz’s equation (5.24), we may mention that Friedmann in 1934 (see Kochin et al. 1964) proved for any vector field that Eq. (5.24) is the condition that the vector line is a material line and that the strength of a vector tube is preserved when itis carried away with the fluid. ‘Thus, by specifying the vector as the vorticity vector, we recover the second and third Helmboltz theorems. It appears that vorticity is the only known vector quantity in fluid flow which has this conservation property.* Some special cases will elucidate the subject matter discussed so far. As an example of Eq, (5.31) let us write the conservation law of vorticity for axisymmetric flows in cylindrical polar coordinates with vg = 0. From Eq. (5.12) we obtain A (Mey wor (5.37) Bq, (5.37) can be interpreted as the conservation law for an individual fluid element at 1, z in the meridional plane or as the conservation law for an individual ring of fluid elements at 2nr, z in the whole field. The amount of vorticity @p of a fluid element can change when r changes. In swirling motion with vg # 0, Eq, (5.37) is extended to 4M) 2. . Se Or (5.38) In addition, 0 itself satisfies the conservation law of angular momentum d(rv,)/dt = 0,as follows from Eq, (3.27) for v= 0. If we consider a two-dimensional flow in the (x, y) - plane, Eq. (5.34) reduces to az = = . 5. Oz = Oo 3, = Pd (5.39) Vorticity lines cannot be stretched in two-dimensional flows as can those in axi- symmetric flows, as demonstrated by Eq. (5.37). If in three-dimensional flows only the @, - component is nonzero, that is, @= @,k, the vorticity-transport equation simplifies to % An example from electromagnetics, in which vorticity is replaced by the magnetic field, is the “frozen-field”” equation for a magnetic field in a perfectly conducting fluid (Woltjer 1958, see also ‘Schmid 1966). - 100-do, ow. a ae (5.40) The vorticity line is stretched when w changes with 2. Finally, we may mention that Eq. (5.31) can be extended for a viscous fluid to (Oswatitsch 1959, p. 54) 4G: wa V3. (5.41) Friction changes the amount of generalized vorticity with time. A way to extend the theory of vortex dynamics to viscous fluids will be given in Section 20.2. 5.4. Vorticity Generation Since vorticity cannot occur in the flow of an incompressible homogeneous inviscid fluid if only conservative forces are acting, we must consider a viscous fluid. In such a fluid motion vorticity can be created only at the boundary of the flow domain. We have seen in Section 3.1 (see also Section 6.1) that in an irrotational flow only the velocity component normal to the boundary can be described; the tangential velocity component cannot. If we admit vorticity in a viscous fluid flow, the tangential velocity component can be prescribed, and this bouridary condition determines the amount of vorticity created. Boundary conditions for the vorticity-transport equation itself cannot be prescribed directly by the vorticity because vorticity on the boundary is usually part of the solution. We must incorporate the boundary conditions for the velocity into a vorticity solution. This is being done, for instance, in finite-difference schemes of numerical computations (Anderson 1989, see Exercise 5). ‘The physical explanation for the generation of vorticity causes conceptual difficulties (Lighthill 1963, Batchelor 1967, Morton 1984, Sarpkaya 1996), due to a large extent to the nature of nonslip. The struggle to understand nonslip at a solid wall has a long history (Goldstein 1965, p. 676) and must ultimately be resolved with a molecular theory (Koplik et al. 1989). Within the framework of continuum theory we accept this condition as given. Let us assume the following scenario: At t= 0a body is impulsively started from rest. Since pressure waves travel with infinite speed in an incompressible fluid, we can assume at this initial instant a potential flow everywhere and an infinitely thin shear layer along the body. This thin shear layer, which reduces the potential flow velocity to zero to satisfy the nonslip condition, has vorticity. It forms an infinitely -101-thin sheet of infinite amount.* The vorticity of this sheet immediately starts to diffuse into the fluid and is then carried away by a combination of diffusion and convection. Simultaneously vorticity is produced and redistributed at the wall in such a way that the nonslip condition and the solenoidality of vorticity are always fulfilled (Lighthill 1963, p. 59), whether the body is accelerated or in steady motion. The relation between surface vorticity and the amount of it diffused into the fluid with other flow quantities will give us physical insight into the generation process. For a flow along a nonslip surface z= 0, the surface vorticity can be expressed by velocity gradients according to the definition @ = curl? by a | p= FF Oy= Br (6.42) If we compare the terms of Eq, (5.42) with the viscous stress at the wall av. ou Tye WS Bae NSE Ty = 0, (5.43) we find P= -pRxB, (5.44) with? the vector of the wall-shear stress components ty, and ty,. The vorticity lines at the wall are perpendicular to the constant shear-stress lines or to the skin-friction lines which again are parallel to the streamlines.** For two-dimensional flows in the (x, z)-plane, with z= 0 representing the wall, the shear stress is ta = wo, = 2 (6.45) ‘Wall-shear stress thus has vorticity. Before we conclude that the occurrence of wall-shear stress is a mechanism for vorticity production at least at a nonslip surface, we discuss a simple example which seems to contradict that statement. In a-two- This model of unbounded vorticity generated at an instant is an artifice within the framework of continuum theory. However, on the molecular level, nonslip can be thought of as molecular reflection at a “‘nonspecular’” solid surface that justifies the picture of immediate vorticity generation in a continuum. ‘** In the sense of limiting streamlines at the nonslip surface. - 102-dimensional flow the natural coordinates for a rotating fluid element moving along a circular arc are the polar coordinates 7, . Vorticity and shear stress, given in Eqs. (3.31) and (3.32), respectively, are along this arc with radius of curvature 7, ay _ os o= Shee, (5.46) a 0 tp = 2UDrp = we. (5.47) ‘We notice that vorticity consists of two parts, the gradient dvg/dr and the solid-body rotation v9/r.* We recall from Section 1.5 that vorticity is only invariant with respect to rotation up to a constant, while the component D,» of the deformation tensor is invariant, as we can see immediately from Eqs. (5.46) and (5.47). Vorticity thus can originate from rotation and lateral deformation of a fluid element. In the case of a rotating rod of radius r, we obtain for the shear stress (divided by 1) ou 2a, = 2.2% eee SP (5.48) with nonslip but with zero vorticity.** If the rod is at rest, and the fluid is in solid- body rotation, we obtain fe _ 8 ie o. (5.49) A rotating fluid with 2 = Or yields @ = Q in agreement with Eq. (4.7). The torque on the rod is | M| = 4muQr? in both systems. We conclude that along a nonslip surface, which is in an inertial frame, vorticity is connected with the shear stress, and since shear stress is produced by nonslip, vorticity is generated by the shear stress. This vorticity is redistributed in a rotating frame which itself has constant vorticity. * Lindgren (1980) introduced the notion that vorticity is the sum of these two parts to obtain insight into the concept of vorticity within a fluid. Tritton (1982) rightly rejected this idea as not “‘useful.”” However, Lindgren’s idea becomes useful when applied to surface vorticity, as we will show in the text that follows. ** This situation is an exception because an irrotational flow in general cannot satisfy the nonslip condition (see the remark on d’Alembert’s paradox in the case of a laminar irrotational flow, Section 3.1). ~ 103 -We assume now a local free surface without surface tension in a steady two- dimensional flow. Then, the surface shear stress is zero, and the surface vorticity is o2 O= 7%. (5.50) In this case the origin of surface vorticity is solely due to local solid-body rotation (Longuet-Higgins 1992). Eq. (5.50) is an example of an explicit vorticity-boundary condition given in terms of the surface curvature and the surface tangential velocity.* The amount of vorticity that is diffused into the interior of the fluid next to the boundary is determitied by the flux of vorticity, which is equivalent to the vorticity source strength on the boundary (Lighthill 1963, p. 58). This flux of vorticity is invariant with respect to rotation. The flux density or local flux is, for simplicity in Cartesian coordinates with z= 0 the flat boundary, Op = vem wwe, (51) az (5.52) | Bey, \ ay : (5.53) | __ where the last equation comes from the solenoidal property of the vorticity vector. Q, is thus determined by the tangential components of div@. The vertical vorticity | component itself, is zero according to Eq. (5.42). At a nonslip surface Vu= Pufaz* and V0 = Fvfez". Eqs. (5.51) and (5.52) show that the expression for the vorticity flux is also the | expression for the viscous force (per unit mass, see Section 2.3) acting on the | boundary. We can also replace this viscous force with the other forces that balance the momentum equation. Thus, the Navier-Stokes equations give, for slip on the boundary (Lugt 1987), | * We have noticed in the definition of a free surface in Section 2.6 that the second fluid is neglected only in the sense that 2 < jz. Sarpkaya (personal communication) pointed out that, no matter how small Hz is, diffusion of surface vorticity will occur and that, therefore, a steady flow solution is not possible. All examples given in Chapter 15 are time-dependent and assume that the diffusion | of vorticity is slow compared to the changing flow field - 104-ayo Be Ly) : Ve xe 2 OE (5.54) 20, ay Bly vS sarncaa 4B (5.55) and for nonslip (Lighthill 1963, p. 54) a -v te 7 (5.56) aa, ~v oe (5.57) For the nonslip condition, the vorticity flux depends only on the pressure change along the wall.* Notice that a vorticity flux can exist on a flat free surface (perfect slip) while the surface vorticity itself is zero according to Eq. (5.50). We may emphasize that it does not follow from the interpretation of the vorticity flux as created by the pressure gradient (and the acceleration of the wall) that the vorticity flux requires the pressure concept for explaining vorticity production. In Section 12.1 flow separation will be explained by vorticity and not by the pressure gradient. For a two-dimensional fiow along a circular arc the vorticity flux is locally ee 12% ' (5.58) As mentioned before, the vorticity flux is invariant with respect to rotation in contrast to the surface vorticity. For two-dimensional flows a formal analogy exists to heat transfer (when neg- lecting dissipation) that is helpful in understanding the spreading of vorticity. From Eqs. (3.63) and (3.64) we obtain in dimensionless form * The terms du/dt + (u?/2)0x = dufdt in Eq. (5.54) and the corresponding terms in Eq. (5.55) can also be interpreted as an accelerated nonslip wall (Morton 1984). - 105 -1,20 , #o Du Dao = 2 at tae tay ~ Re ax? * ay? iy (5.59) ar, at ah _ 1 #r , eT Br HS tay Re ae * ay” (5.60) Like heat, vorticity is transported from the boundary into the fluid interior through molecular diffusion and convection. There ate, however, differences between heat and vorticity production. Whereas the (absolute) temperature can have only positive values, vorticity can be both positive and negative. Moreover, the boundary is always a source of positive and negative vorticity, but it is not a sink (Morton 1984). ‘Two examples may illustrate two different ways of bringing vorticity into the fluid. The first example is the plane, steady Couette flow: u= Uy/h with U the constant velocity of the upper wall y= h, zero velocity of the lower wall y= 0, and h the distance between the two walls. This linear velocity profile yields a constant vorticity field. Thus, the vorticity flux is zero at the walls. Vorticity is neither generated nor absorbed there, and vorticity can be generated and brought into the flow only during the transient phase of the flow. The second example is the plane steady Poiseuille flow with its parabolic velocity profile between two parallel walls. Here, the vorticity | flux is not zero at the wall, and vorticity is continuously produced there, positive vorticity at one wall and negative vorticity at the other. They average each other out. Ina steady two-dimensional flow, integration of the vorticity flux over the closed surface of a boundary (body or liquid surface) or over a closed streamline yields 9® ae ve =O, (5.61) which follows from Eq, (5.54) since the line integral over the first term to the right vanishes under the assumption of single-valuedness for pressure and velocity. This | result means that the total averaged vorticity flux from a closed surface is zero. ‘An interesting generalization of Eq. (5.61) for closed streamlines of a steady flow in three-dimensional space is, according to Prandtl (1904) (see also Batchelor 1956), vo culB- d= 0. (5.62) Prandtl argued that friction cannot do work in a steady circular motion (see Section 57). For almost inviscid fluids, that is, for v< 1, Prandtl obtained results for the | vorticity fields of two-dimensional (plane) and axisymmetric flows. In the first case | Helmholtz’s Second Theorem gives for a steady flow @ = k f(y) and hence - 106-curl@= curl & (5.63) oe ee oe = Fy R= FEV If we place this term into Eq. (5.62), we obtain voourl@d: a= VER ee 0, (5.64) which results in dffay = 0 or @ = const because $8" d3'# 0 (Section 5.7). We state: In an almost inviscid fluid the vorticity field of a steady two-dimensional flow is constant inside a closed streamline. Since the flow inside a closed streamline constitutes a vortex, this statement expresses an important property of vortices. The center of such vortices is defined by the center of nested streamlines. In the second case we can show that the field @/r of an axisymmetric steady flow is constant inside a closed streamline in the meridional plane. 5.5. Maximum Principles Maximum principles of partial differential equations can give information on general properties of vortices and vorticity fields independently of the solutions proper. They also confirm fundamental differences between two- and three- dimensional motions mentioned in the previous sections. In particular, they will reveal peculiar properties for steady, two-dimensional (plane and axisymmetric) flows. Before we discuss maximum principles of the vorticity equation, let us become familiar with this concept by looking at the one-dimensional maximum principle. We assume that a function f(x) exists in a certain interval of x with continuous derivatives f° and f’, and that g(x) is any bounded function. Then, we see immediately that the differential inequality f+ g(xif'>0 (6.65) does not permit a maximum of f inside the interval since such a maximum at x= ¢ requires f(c) = 0 and f(c)< 0. A quite trivial, but for our purpose important extension of Eq. (5.65) to a quasilinear case is that the inequality -107-f+ Glx, pf > 0 (5.66) also excludes a maximum. A corresponding argument can be made for the nonexistence of a minimum. According to Protter and Weinberger (1967) restriction (5.65) can be relaxed to f+ gf 2 0, (5.67) which does not permit a maximum in the interior of the interval unless f is a constant. The same proof is valid for the maximum principle associated with the quasilinear inequality (Lugt 1985) f° +GG, pf 2 0. (5.68) ‘The two-dimensional extension of the above maximum principle is a simplified version given by Hopf (Protter and Weinberger 1967) for the inequality 2 12 & + gata yt + g0(, oe 20. (5.69) | Again, the proof for this maximum principle is the same if we replace gy(x, y) and g2(%, y) by Gr (x, y, f) and Go(x, y, f).* | if we apply this maximum principle to the steady, two-dimensional vorticity equation (8.70) | We arrive at the important result that the vorticity @ has no extremum within the fluid. | This statement has interesting physical implications. It means that closed streamlines representing a vortex do not have a local extremum of vorticity. As an example, Fig. 5,2 shows the steady flow past a circular cylinder at a low Reynolds number. The vorticity is produced at the surface due to nonslip (or even perfect slip) and spreads out from there into the fluid. Hence, the maximum and minimum values | of the vorticity are at the surface. This result is physically plausible since an extremum of vorticity inside the fluid cannot be maintained without a source, and _ vorticity cannot be produced inside a fluid. This fact can also be expressed by Eq. | (6.61), which is but another formulation of the maximum principle (5.70). Eq. (5.61) |* A simple proof for the Laplace equation V?f= is given in Section 6.1 - 108-represents the conservation of the vorticity flux across a closed streamline and can be interpreted to mean that closed equivorticity lines are not possible in a steady flow (see Section 6.1). We must abandon the intuitive notion that a vortex is always accompanied by an extremum of vorticity. We shall call a vortex without an extremum of vorticity an attached vortex. In time-dependent two-dimensional flows, extrema of vorticity do occur. In fact, the occurrence of a local extremum of © within the fiuid signals the process of vortex shedding, and the vortex is called a detached, separated ot shed vortex. The vortex moves away from the body (Fig. 5.3). This process is inherently unsteady. So far we have defined the center of a vortex by the center of nested streamlines (Section 1.3). This definition is useful for an attached vortex, which is not moving relative to the reference frame fixed to the body. For a detached vortex moving relative to the reference frame the location of the center of nested streamlines is difficult to determine. Here, the location of the extremum of vorticity is an alternative definition for the vortex center. We must keep in mind, however, that the two locations do not coincide in many cases. () Fig, 5.2: Streamlines (a) and equivorticity lines (b) for a steady laminar flow past a circular cylinder at Reynolds number Re= 40 with Re= UDW (adapted from Takami and Keller 1969). The fact that an extremum of vorticity cannot occur inside a fluid indicates that at higher fiow velocities (or higher Reynolds numbers) other mechanisms may exist that can transport larger amounts of vorticity in a more efficient way than the steady-state flow. We shall see in Chapter 13 that shedding of vorticity pockets, that is shedding of vortices, will be that mechanism. - 109-Fig. 5.3: Two-dimensional vortex shedding behind an edge. Equivorticity lines at two consecutive times. Extrema of vorticity do occur in steady three-dimensional flows. One example will suffice to prove this statement. But first, let us consider an extension of the | maximum principle associated with Eq. (5.69). Protter and Weinberger (1967) have shown that, for f(x, y), if 2 oy + or arta, WE + gol, oe +h, yf20 (7) only a non-negative maximum at the boundary unless f is a constant in the interior of the domain. \ | | ima certain domain with h(x, y) < 0, and if g; and gp are bounded, then fcan assume An example that demonstrates the existence of extrema of vorticity in a steady | three-dimensional flow is Burgers’ vortex, Eqs. (4.70) through (4.72). For this special flow situation, @, = @» = 0, @, = ©, the vorticity equation reduces to FP yy (YO, BO , MWe | VG tg B)Ot EOP 0 SE + Si w= 0. (5.72) A comparison of Eqs. (5.71) and (5.72) shows that a non-negative maximum cannot exist if d0,/8z $ 0. However, such a maximum may exist if dv,/0z > 0. The term 3v,/éz is Burgers’ constant a. We know from the discussion of Burgers’ solution and its time-dependent extension (4.75) that the steady solution exists for a> 0 only. A hon-negative maximum of @ exists at r= 0. Physically, an influx of fluid toward the | axis occurs that convects vorticity with stretching toward r= 0 and thus balances the | outward diffusion of ©. This stretching cannot occur in two-dimensional flow and an extremum cannot be established. The extremum in three-dimensional flow is thus due | to a concentration process. Notice that the vorticity in Burgers’ equation is a scalar | and that, as in two-dimensional flows, all vorticity lines are parallel to each other, \ However, in the axisymmetric case, the occurrence of a local vorticity extremum (that lis, the appearance of nested closed equi-vorticity lines) is due to stretching of the icity lines. Another example of an axisymmetric flow illustrates the role of -110-stretching but reveals a disturbing complexity in defining a vortex by means of the vorticity: There seems to be an analogy between steady two-dimensional and axisymmetric flows ( @, @, = @, = 0, without vg) because, qualitatively, Fig. 5.2 can also represent the flow past a sphere. However, while for two-dimensional flows h(x, y) = 0 in Eq. (5.69), for axisymmetric flows h(r, z) = —I/r(v/r— ,). This means that in the axisymmetric case the maximum principle is not generally valid because stretching of the circular vorticity lines makes it possible for h to become positive. A direct analogy does exist between two-dimensional and axisymmetric flows, however, if in axisymmetric flows @, is replaced by w/t. Then, since 2. 2 Fe Fy sd oe 2 (Be) 2M ae a TN a Gar GIO TD v a globally valid maximum principle holds for w»y/. As a consequence, we must define vortex shedding as exemplified in Fig. 5.4 in the meridional plane of an axisymmetric flow with @, ©, = @, = 0s the development of an extremum in @4/r. The cognition that must be replaced by @y/r in axisymmetric flows with », ©, = @, = 0 in order to determine vortex shedding leads to a problem. Let us assume that we do not know the extremal properties of the @9/* - field. This situation, in fact, occurs in the general three-dimensional case, for which we do not know whether a function of @ exists that obeys a maximum principle, Then we have to rely on the @ - field itself and on the streamlines (Chapter 1), and we are left with the cumbersome analysis of the flow field over a certain time span. We will find more on this subject in Chapter 12. 5.6. Forces Acting on a Body Of great practical interest is knowledge of the forces and torques which act on a body in a flow field, The force pdF acting on a surface element dA is pdF= T-7?dA, (5.74) and the corresponding torque is pipxdP=7xT-?dA, (5.75) where 7%} is the distance from the center of rotation. Eq. (5.74) yields, for a nonslip surface according to Bgs. (2.31) and (5.44), pdF = (— p+ 2 D-WdA= (—pe+pSxw da. (5.76) -1ll-