Fluid Dynamics

1
Contents

1 Fluid Mechanics 12

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.1.1 Definition of a Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.1.2 Continuum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.1.3 Physical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.2 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.2.1 Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.2.2 Acceleration of Fluid Particle . . . . . . . . . . . . . . . . . . . . . 14

1.3 General Deformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.4 Rate of Strain Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.5 Rate of Rotation Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.5.1 Decomposition of ∂ui /∂xj . . . . . . . . . . . . . . . . . . . . . . . 20

1.6 Conservation of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.6.1 Continuity Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.6.2 Direct Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1.6.3 Simplifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

1.7 Pathlines and Streamlines . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

1.7.1 Pathlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

1.7.2 Streamlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

1.8 Streamfunction and the Velocity Potential . . . . . . . . . . . . . . . . . . 25

2
 1.8.1 Streamfunction in 2D . . . . . . . . . . . . . . . . . . . . . . . . . . 26

1.8.2 Stokes Streamfunction - 3D . . . . . . . . . . . . . . . . . . . . . . 29

1.8.3 The Velocity Potential . . . . . . . . . . . . . . . . . . . . . . . . . 29

1.9 Vortex Lines and Circulation . . . . . . . . . . . . . . . . . . . . . . . . . . 29

1.9.1 Vortex Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

1.9.2 Circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2 Fluid Dynamics - Euler Equations 31

2.1 Body Forces and Surface Forces . . . . . . . . . . . . . . . . . . . . . . . . 31

2.1.1 Body Forces (i.e. External Forces) . . . . . . . . . . . . . . . . . . . 31

2.1.2 Surface Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.2 Ideal (or Perfect) Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.3 Conservation of Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.4 Conservative Body Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.5 Potential Flow or Irrotational Flow . . . . . . . . . . . . . . . . . . . . . . 34

2.6 Irrotational Flow - Bernoulli Pressure . . . . . . . . . . . . . . . . . . . . . 35

3 Sources and Sinks 37

3.1 Spherical and Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . 37

3.1.1 Spherical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.1.2 Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . 40

3.2 Uniform Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.3 Point Sources and Sinks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.3.1 In Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . 43

3.4 Doublets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.5 Axi-Symmetric Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.5.1 Uniform Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3.5.2 Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3
 3.5.3 Doublet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.6 The Half-Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.7 Rankine Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.8 Flow Around a Shpere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.9 Line Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4 Viscosity - The Navier-Stokes Equation 54

4.1 Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.2 Newtonian Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.2.1 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.2.2 Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.2.3 Incompressible Newtonian Fluid . . . . . . . . . . . . . . . . . . . . 56

4.3 Navier-Stokes Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.3.1 Conservative Body Force . . . . . . . . . . . . . . . . . . . . . . . . 57

4.4 Alternative Form of the Navier-Stokes Equation and the Vorticity Equation 58

4.4.1 The Vorticity Equation in Two Dimensions . . . . . . . . . . . . . . 59

4.5 Vorticity and Viscocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

5 Streamfunction in 2-Dimensions 61

5.1 Cartesians . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

5.2 Cylindrical Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 62

5.2.1 Relating Velocity to Streamfunction by Solving the Continuity Equa-

tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5.2.2 Direct Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6 Plane Potential Flow 66

6.1 Potential Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

6.2 Complex Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

6.3 Simple Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4
 6.3.1 Uniform Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

6.3.2 Flow Near a Stagnation point . . . . . . . . . . . . . . . . . . . . . 70

6.3.3 Flow Around a Corner . . . . . . . . . . . . . . . . . . . . . . . . . 72

6.4 Singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

6.4.1 Source (and Sink) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

6.4.2 Vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

6.4.3 Dipole (or Doublet) . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

6.5 Flow Around a Circular Cylinder . . . . . . . . . . . . . . . . . . . . . . . 81

6.6 Flow Around a Rotating Cylinder . . . . . . . . . . . . . . . . . . . . . . . 84

6.6.1 New Complex Potential . . . . . . . . . . . . . . . . . . . . . . . . 84

6.6.2 Velocity Components . . . . . . . . . . . . . . . . . . . . . . . . . . 84

6.6.3 Stagnation Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

6.6.4 Force on a Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

6.7 Other Methods for Obtaining Potentials . . . . . . . . . . . . . . . . . . . 89

6.7.1 Method of Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

6.7.2 Continuous Distributions . . . . . . . . . . . . . . . . . . . . . . . . 90

6.7.3 Conformal Transformations . . . . . . . . . . . . . . . . . . . . . . 90

7 Boundary Conditions 91

7.0.4 Flow Past a Stationary Finite Solid Body . . . . . . . . . . . . . . . 91

8 Similarity, Scaling, Reynold’s Number 93

8.1 Dynamic Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

8.2 Stokes Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

9 Solutions of the Navier-Stoke’s Equations 96

9.1 Simplification of the Navier-Stokes Equations with Unidirectional Flow . . 97

9.2 Steady Unidirectional Velocity fields . . . . . . . . . . . . . . . . . . . . . 98

5
 9.2.1 2-D Poiseuille Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

9.2.2 Circular Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

9.2.3 Flow in an Annulus . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

9.2.4 Couette Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

9.3 Unsteady Unidirectional Velocity fields . . . . . . . . . . . . . . . . . . . . 101

9.3.1 Impulsively Started Plate . . . . . . . . . . . . . . . . . . . . . . . 101

9.3.2 General Solution for Infinite Plate . . . . . . . . . . . . . . . . . . . 103

9.3.3 Oscillating Plane Boundary (Stokes Boundary Layer) . . . . . . . . 105

9.3.4 Oscillating Pressure Gradiant - Circular Cylinder . . . . . . . . . . 107

9.4 Circular Streamlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

9.5 Two Dimensional Stagnation Point flow . . . . . . . . . . . . . . . . . . . . 114

10 Mathematical Theorems 118

10.1 Uniqueness Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

10.2 Minimum Dissipation Theorem . . . . . . . . . . . . . . . . . . . . . . . . 121

10.2.1 Energy Balance and Dissipation of Energy . . . . . . . . . . . . . . 121

10.2.2 Proof of Minimum Dissipation Theorem . . . . . . . . . . . . . . . 122

10.2.3 Vorticity Minimum Dissipation Theorem . . . . . . . . . . . . . . . 124

11 Stoke’s Streamfunction 125

11.1 Streamfunction in Cylindrical and Spherical Polar Coordinates . . . . . . . 125

11.1.1 Streamfunction in Cylindrical Polar Coordinates . . . . . . . . . . . 125

11.1.2 Streamfunction in Spherical Polar Coordinates . . . . . . . . . . . . 129

11.2 Flow Around a Corner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

11.3 Drag on Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

11.3.1 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 135

11.3.2 Calculating the Streamfunction . . . . . . . . . . . . . . . . . . . . 135

11.3.3 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

6
 11.3.4 Stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

11.3.5 Force on the sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

11.3.6 Calculating the force on the sphere by doing an integral at infinity . 140

A Glossary 142

B Vector Calculus, Divergence Theorem and Stoke’s Theorem 145

B.1 Vector Notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

B.2 Vector Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

B.2.1 Gradient of a Scalar . . . . . . . . . . . . . . . . . . . . . . . . . . 146

B.2.2 Divergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

B.2.3 Laplacian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

B.2.4 Curl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

B.2.5 Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

B.3 Definitions and Statement of Line Integral Theorem, Gauss’s Divergence

Theorem and Stoke’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 149

B.3.1 Defintion of a Line Integral . . . . . . . . . . . . . . . . . . . . . . 149

B.3.2 Theorem for a Line Integral . . . . . . . . . . . . . . . . . . . . . . 150

B.3.3 Definition of Flux and Statement Gauss’s Divergence Theorem . . . 151

B.3.4 Statement of Stokes’ Theorem . . . . . . . . . . . . . . . . . . . . . 151

B.3.5 Conservative Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

B.4 Proof of Theorem for Line Integral of ∇ψ . . . . . . . . . . . . . . . . . . . 152

B.5 Proof of Gauss’s Divergence Theorem . . . . . . . . . . . . . . . . . . . . . 153

B.6 Proof of Stokes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154

C Orthogonal Curvilinear Coordinates 157

C.1 Gradient of a Scalar Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

C.2 The Divergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

C.3 Laplacian of a Scalar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

7
 C.4 The Curl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

D Cylindrical Polar Coordinates 161

D.1 Vector Differential operators . . . . . . . . . . . . . . . . . . . . . . . . . . 161

D.2 Rate of strain tensor revisted . . . . . . . . . . . . . . . . . . . . . . . . . 163

D.3 Rate of strain tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

E Spherical Polar Coordinates 167

E.1 Vector Differential operators . . . . . . . . . . . . . . . . . . . . . . . . . . 167

E.2 Rate of strain tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

8
List of Figures

1.1 Rotation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.2 Pure translation of fluid element. . . . . . . . . . . . . . . . . . . . . . . . 15

1.3 An expansion of fluid element in the y−direction. . . . . . . . . . . . . . . 16

1.4 Rotation and shear. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.5 Rotation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1.6 Uniform flow over a flat plate n = 1. . . . . . . . . . . . . . . . . . . . . . 22

1.7 A pthline. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

1.8 Streamlines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

1.9 Stream tube. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

1.10 Stress. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

1.11 Say two points on a curve have tangent vector dx~ex + dy~ey then dy~ex −dx~ey
points ‘outward’ - this can be seen here following from the fact that dx and
dy are positive. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.2 Stress. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

6.1 Uniform motion in the x−direction.. . . . . . . . . . . . . . . . . . . . . . 69

6.2 Uniform motion at an angle α . . . . . . . . . . . . . . . . . . . . . . . . . 70

6.3 Flow at a stagnation point. . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

9
6.4 Flow at a stagnation point. . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

6.5 Flow at a corner. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

6.6 Extended solution from flow at a corner with walls at θ = 0 and θ = 2π/n. 74

6.7 Flow around a right-angled corner, n = 2. . . . . . . . . . . . . . . . . . . 75

6.8 Uniform flow over a flat plate n = 1. . . . . . . . . . . . . . . . . . . . . . 75

6.9 Stagnation point flow when n = 2. . . . . . . . . . . . . . . . . . . . . . . . 76

6.10 Flow over a wedge 1 ≤ n ≤ 2. . . . . . . . . . . . . . . . . . . . . . . . . . 76

6.11 Uniform flow over a flat plate n = 1. . . . . . . . . . . . . . . . . . . . . . 76

6.12 Doublet. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

6.13 Streamlines for a dipole. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

6.14 Superposition of uniform flow and a dipole results in ψ = 0 as a solid

object, a cylinder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

6.15 Flow around a cylinder. We have a dividing streamline for ψ = 0 as a circle

of radius a about the origin. Another solution for ψ = 0 is the line y = 0.
The points denoted by black circles denote stagnation points, namely where
dw/dz = 0. The x’s denote points of minimum pressure. Cavitation occurs
here first. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6.16 Flow around a clockwise rotating cylinder for different values of Γ. . . . . . 87

6.17 Flow around a rotating cylinder results in an upward force on the cylinder. 89

11.1 Streamlines for Stokes flow around a sphere. . . . . . . . . . . . . . . . . . 134

B.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

B.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

B.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

B.4 Finding the circulation around an infintesimal square. . . . . . . . . . . . . 155

B.5 When adding together the circulation of each loop the only remaing con-
tribution to the line integral comes from the outside edge. . . . . . . . . . 155

B.6 We have some surface bounded by the loop Γ. The surface is divided into
many small areas, each approximately a square. . . . . . . . . . . . . . . . 156

10
C.1 The gradient of ~v . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

C.2 The ~e1 component of the Curl of an infintesimal loop. . . . . . . . . . . . . 160

E.1 Stress. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

11
Chapter 1

Fluid Mechanics

1.1 Introduction

1.1.1 Definition of a Fluid

The defining property of a fluidis that it cannot withstand, shearing forces, however small,
without sustained motion. Since both gases and liquids have this property, both are fluids
and subject to a unified treatment as far as their macroscopic motion is concerned.

The rate at which the fluid deforms continuously depends not only on the magnitude of
the applied force but also on a property of the fluid called its viscosity - its resistance to
deformation and flow.

Solids also deform when sheared, but a position of equilibrium is soon reached in which
elestic forces induced by the deformation of the solid exactly counterbalance the applied
shear force, and further deformation ceases.

1.1.2 Continuum

There are a large number of atoms or molecules in any given volume. For example the
number of gas molecules in (1µcm)3 is 2.687 × 107 . Therefore gases and liquids can be
considered as continuum, for example water appears continuous, wind appears continuous
(even though we cant see it).

1.1.3 Physical Properties

What are the physical properties of a fluid? Density, viscocity and temperature and
pressure, velocity and acceration.

12
 Density ̺ = ̺(~x, t)
Voscosity µ = µ(~x, t)
Temperature T = T (~x, t)
Voscosity p = p(~x, t)
Temperature ~u = ~u(~x, t). (1.1)

1.2 Kinematics

Kinematics involves the description of a fluids motion without considering forces.

1.2.1 Velocity

Velocity is the rate of change of position with time, i.e.

d~x
~u = . (1.2)
dt

There are two possible descriptions:

Lagrangian

(a) In the Lagrangian description we follow a particle which starts at position ~x = ~x0 at
time t = t0 and we follow it and observe its velocity ~q(~x0 ; t, t0 ). In this case ~x and t are
dependent variables.

Newton’s laws and conservation of mass and energy, apply directly to each particle. How-
ever, fluid flow is a continuum phenomenon and it is not possibly to to track each “par-
ticle”. Let’s consider a second way of describing fluid motion.

Eulerian

(b) In the Eulerian description we define the velocity of the fluid particle which is situated
at ~x at time t to be ~u(~x, t).

~u(~x, t) = (U1 (x, y, z, t), U1 (x, y, z, t), U1 (x, y, z, t)) (1.3)

13
with ~x and t as independent variables.

Here one is not concerned about the location or velocity of any particular particle, but
rather about the velocity of whatever particle happens to be at a particular location of
interest at a particular time.

1.2.2 Acceleration of Fluid Particle

(a) Lagrangian description

∂~q
~a = (~x0 ; t, t0 ) = . (1.4)
∂t

(b) Eulerian.

The velocity field ~u gives the velocity of the fluid at someplace in the fluid. The accelera-
tion is not simply given by ∂~u/∂t as this represents the rate at which the velocity ~u(x, y, z)
changes at a fixed point in space. What we need is the rate at which the velocity changes
for a particular piece of fluid. After time dt the piece of fluid will have moved to a new
position. The particle at position (~x, t) moves to (~x + d~x, t + dt) where d~x = ~u(~x, t)dt

~u(~x + d~x, t + dt) − ~u(~x, t)

~a(~x, t) = (1.5)
dt

With use of the Taylor expansion we have

∂~u ∂~u
~u(~x + d~x, t + dt) = ~u(~x, t) + dt + dx (1.6)
∂t ∂xi i

so that

∂~u dxi ∂~u

~a(~x, t) = + (1.7)
∂t dt ∂xi

which can be written as

∂~u
~a(~x, t) = + ~|u ·{z
∇~u} . (1.8)
∂t
|{z} advective
local change

We define the substantial derivative by

14
 D ∂
= + ~u · ∇. (1.9)
Dt ∂t

1.3 General Deformations

Consider an infintesimal volume element of the fluid at time t0 . After an infintesimal time
later δt this volume element will have suffered a deformation that can be considered the
combination of a translation, linear expansion, rotation and shear (see fig 1.1).

= + + +

general translation expansion rotation shear

deformation

Figure 1.1: Rotation.

A translation is when the position of a fluid element is changed without changing its size
or shape, see fig 2. After a time δt the fluid element will be shifted in the x−direction an
amount ux δt and shifted in the y−direction an amount uy δt - see fig 1.2.

δy

O′
uy ux uy δt
O δx
ux δt

Figure 1.2: Pure translation of fluid element.

Expansion

Let us consider the case where we have linear expansion in the x−direction. At time t0
the x−component of the velocity at position (x + δx, y) where δx is small is given by the

15
first order Taylor expansion

∂u
u(x + δx, y, z, t0 ) = u(x, y, z, t0) + δx.
∂x

The situation we consider is shown in fig 1.3.

∂ux
ux ux + ∂y
δy
B C B C C′

δy δy

∂ux
ux ux + ∂y
δy
O δx A O δx A A′
 
∂ux
δl = ∂x
dx δt

Figure 1.3: An expansion of fluid element in the y−direction.

The change in volume after time δt will then be

δV = (δx + δl)δyδz − δxδyδz

= δlδyδz
= [u(x + δx, y, z)δt − u(x, y, z)δt]δyδz
 
∂ux
= δx δt(δyδz) (1.10)
∂x

and the rate of change in volume per unit volume is then

1 dδV ∂u
= x.
δV dt ∂x

If velocity gradients ∂uy /∂y and ∂uz /∂z are also present, it can easily be shown that

1 dδV ∂u ∂uy ∂uz

= x+ + = ∇ · ~u.
δV dt ∂x ∂y ∂z

16
Rotation and shear

We now consider the effects of rotation and shear. For simplicity we consider motion in
the x − y plane, however these considerations are easily extended to the more general
case. The situation is the one shown in fig 1.4. We have that the change in angle of the
line OA is

 
∂uy
δα = δx δt
∂x

and the change in angle of the line OB is

 
∂ux
δβ = δy δt.
∂y
 
∂ux
∂y
δy δt
∂ux
ux + ∂y
δy
B B B′
uy

δβ
δy δy
δα  
uy ux ux ∂uy A′ ∂uy
uy + ∂x
δx ∂x
δx δt
A
O δx A O δx

Figure 1.4: Rotation and shear.

Note that if ∂uy /∂x is positive, ωOA will be counterclockwise. If ∂ux /∂y is positive then
ωOB will be clockwise.

Rotation

We consider pure rotations first where after time δt the element has rotated by a small
angle θ in which case

∂uy
δxδt ∂uy
ϑ = δα = ∂x = δt
δx ∂x
∂ux
δxδt
∂y ∂u
= −δβ = − = − x δt. (1.11)
δx ∂y

17
Rotations are obviously volume preserving.

Shear

Shear type strains are where the initial fluid element is changed into a parallelogram.

∂uy
ϑ = δα = δt
∂x
∂u
= δβ = x δt. (1.12)
∂y

The area of the parallelogram is |OA||OB| sin(π − 2δα) or |OA||OB| cos(2δα). Taylor
expanding we get

 
1 2
|OA||OB| 1 − (2δα) + . . .
2!

so to first order the area is unchanged.

Extracting rotation and shear

When we have general ‘angular’ deformations we extract the part that corresponds to
rotation by taking the average of the angular velocities with counterclockwise rotation
considered positive,

 
1 δα δβ
ω = −
2 δt δt
 
1 ∂u y ∂ux
= − (1.13)
2 ∂x ∂y

We extract the shear part from

 
1 δα δβ
σ = +
2 δt δt
 
1 ∂uy ∂ux
= + (1.14)
2 ∂x ∂y

For pure rotation (1.14) is zero, but for pure shear we have δα = δβ and (1.13) is zero.

18
1.4 Rate of Strain Tensor
∂ui
ui (~x + ~r) = ui (~x) + rj + O(r 2 ) (1.15)
∂xj

!
1 ∂ui ∂uj
eij = + (1.16)
2 ∂xj ∂xi

 
ǫxx ǫxy ǫxz
 ǫyx ǫyy ǫyz  (1.17)
ǫzx ǫzy ǫzz

1.5 Rate of Rotation Tensor

Define

!
1 ∂ui ∂uj
ωij = − = −ωji . (1.18)
2 ∂xj ∂xi

This has three independent components. Behave like a vector. Only happens in three
dimensions.

We define the vorticity as

∂uk
ωi = ǫijk (1.19)
∂xj

which is written in vector notation as ∇ × ~ω .

The rate of rotation tensor can be written in terms of ~ω

ǫijk ǫklm = δil δjm − δim δjl (1.20)

We can now see clearly the meaning of the term ωij rj in the Taylor series expansion for
the displacements

1
ωij rj = ǫijk ωj rk (1.21)
2

which in vector notation is

19
 1
~ω × d~r. (1.22)
2

Say ~ω = ω~k and ~r = r~i

1 1
~ × d~r = ωr~j.
ω
2 2

z ~ω

Figure 1.5: Rotation.

1
ui (~x + ~r) = ui (~x) + eij rj + ǫijk ωj rk + O(r 2 ) (1.23)
2

1.5.1 Decomposition of ∂ui/∂xj

The tensor ∂ui /∂xj can be written

! !
∂ui 1 ∂ui ∂uj 1 ∂ui ∂uj
= + + −
∂xj 2 ∂xj ∂xi 2 ∂xj ∂xi
= eij + ωij
 
1 1 1
= ekk δij + eij − ekk δij + ǫijk ωk
3 3 2
1 1
= θδij + γij + ǫijk ωk (1.24)
3 2

where we have defined the expansion parameter θ as ekk . From (1.24)

X ∂u
i
=θ
i
∂xi

20
 P
where we have used i γii = 0. The term

1
θδ
3 ij

in (1.24) represents the isotropic expansion part, that is it is the volume change ‘averaged’
in each direct equally. Also from (1.24) we have

!
∂ui ∂uj
− = ǫijk ωk
∂xj ∂xi

If i 6= j then

∂ui ∂uj
+ .
∂xj ∂xi

represents pure shear. The non-isotropic part of the expansion is

1
γii = eii − θ (no summation implied)
3

as eii (no summation implied) is the volume change in the ith−direction and is minus the
ith part of isotropic expansion.

1.6 Conservation of Mass

1.6.1 Continuity Equation

Consider an arbitrary volume V bounded by a surface S so the mass in V is

Z
̺dV
V

and conservation of mass means the rate of change of mass equals the mass flux across
the boundary

Z Z
∂
̺dV = − ̺~u · ~ndS (1.25)
∂t V S

21
The minus sign is there as mass flux leaing the volume would reduce the mass. Using the
Divergence Theorem

Z  
∂̺
+ ∇ · (̺~u) dV = 0 (1.26)
V ∂t

since the volume is arbitrary the integrand must vanish

∂̺
+ ∇ · (̺~u) = 0 (1.27)
∂t
This is the differential form of the continuity equation.

1.6.2 Direct Derivation

∂(̺ux )
̺ux dydz [̺ux + ∂x
dx]dydz

Figure 1.6: Uniform flow over a flat plate n = 1.

Consider the mass flux through the x faces

 
∂
xf lux = ̺ux + (̺ux )dx dydz − ̺ux dydz
∂x
∂
= (̺ux )dxdydz.
∂x

similarly for the y and z faces

∂
yf lux = (̺uy )dxdydz
∂y
∂
zf lux = (̺uz )dxdydz (1.28)
∂z
22
The total net mass outflux must balance the rate of decrease of mass within the cubical
element which is

∂̺
− dxdydz
∂t

Combining the above equations, the balance of mass outflux with mass decrease is then
described by the equation

 
∂̺ ∂ ∂ ∂
+ (̺ux ) + (̺uy ) + (̺uz ) dxdydz = 0.
∂t ∂x ∂y ∂z

Or

∂̺ ∂ ∂ ∂
+ (̺ux ) + (̺uy ) + (̺uz ) = 0.
∂t ∂x ∂y ∂z

which is just (1.27).

1.6.3 Simplifications

Written out the continuity equation is

∂̺
+ ̺∇ · ~u + ~u · ∇̺ = 0 (1.29)
∂t

For constant ̺ the continuity equation linear.

This can be written in terms of the substantial derivative as

D̺
+ ̺∇ · ~u = 0
Dt

A simplification would be steady flow in which case the continuity equation becomes

∇ · (̺~u) = 0.

If the density is constant ̺ = ̺0 , that is, the fluid is incompressible the continuity equation
becomes

∇ · ~u = 0 (1.30)

23
 ~ such that
So ~u divergence free and hence solenoidal, that is, there is a vector A

~
~u = ∇ × A.

We exclusively deal with incompressible flows in these notes. Note that this equation can
be expressed in terms of the rate of strain tensor as ekk = 0.

1.7 Pathlines and Streamlines

1.7.1 Pathlines

~
The locus of an actual particle X(t) is called a pathline. Pathlines are associated with
the Lagrangian description. The corresponding acceleration of the particle is then

dX~
= ~q = ~u(~x, t) . (1.31)
dt |{z} | {z }
Lagrangian Eulerian

X(t)

Figure 1.7: A pthline.

Pathlines are associated with the Lagrangian description.

1.7.2 Streamlines

We can draw lines which are always tangent to the fluid velocity at an instant in time,
these are called streamlines. With steady flow (this is when the velocity of fluid doesn’t
change at any point with time) the streamlines dont change with time and in this case the
particle’s path coincides with the streamline. If we have unsteady flow the streamlines
vary from instant to instant and in general the particle’s path does not coindide with a
streamline as the particle can end up on another streamline. That is with time unsteady
flow pathlines and streamlines are not necessarily coincident.

Streamlines are defined by the three simulataneous equations written as

24
 Figure 1.8: Streamlines.

d~x
= λ~u(~x, t) (1.32)
ds
or

dx dy dz
~x + ~y + ~z = λ(ux~x + uy ~y + uz ~z) (1.33)
ds ds ds

where s is a parameter along the streamline. The parameter s should not be confused
with time, the above equations are integrated while keeping time fixed. The resulting
curves give us the streamlines at an instant in time.

The streamline passing through a closed curve which does not lie on a surface generated
by streamlines form a tubular surface. The fluid contained in such surface is called a
stream tube.

Figure 1.9: Stream tube.

1.8 Streamfunction and the Velocity Potential

For incompressible flow

∇ · ~u = 0 (1.34)

25
Here we breifly introduce the method using the so-called streamfucntion. We will go into
this in more detail in future chapters. The streamfunction method is introduced for 2D
motion. It was extended to axisymmetric flow in thre dimensions - then called Stoke’s
streamfucntion.

For steady flows pathlines and streamlines coincide.

1.8.1 Streamfunction in 2D

Imcompressibility condition

The imcompressibility condition

∇ · ~u = 0. (1.35)

In two dimensions (̺ = Const.) becomes

∂u ∂v
+ = 0. (1.36)
∂x ∂y

If we choose

∂ψ ∂ψ
u= , v=− (1.37)
∂y ∂x

then (1.35) is automatically satisfied:

   
∂ ∂ψ ∂ ∂ψ
− = 0.
∂y ∂y ∂x ∂y

Streamfunction giving Streamlines

We now prove that ψ = Const. is a streamline in steady flow. First note tat in general for
two infinitesimally close points a = (x, y) and b = (x + dx, y + dy) we have from calculus
that

∂ψ ∂ψ
ψ(x + dx, y + dy) − ψ(x, y) = dx + dy
∂x ∂y
= ∇ψ · d~r. (1.38)

26
where d~r is the vector frompoint a to point b. If a and b are points on a curve defined by
ψ = C where C is a constant, then d~r is tangent to the curve ψ = C at a and

0 = ψ(x + dx, y + dy) − ψ(x, y) = ∇ψ · d~r.

Implying that the vector ∇ψ is normal to the curve ψ = C. If we can show that everywhere
~u · ∇ψ = 0, using the formula for ~u in terms of ψ given by (1.37), then we will have proved
the result. This easily follows

 
∂ψ ∂ψ ∂ψ ∂ψ
~u · ∇ψ = + − = 0. (1.39)
∂y ∂x ∂x ∂y

Flux Through a curve

We have already defined the flux through a surface by summing over ~u · ~ndS where dS is
the area of an infintesimal area element with unit normal vector ~n. The two dimensional
analogue of the volume flux as the sum of terms

~u · ~ndl

along a given curve and where dl is the length of an infintesimal line element.

We first consider the simple case of the volume flux Φ through the curve x = Const. in
the x−direction from say a to b,

Z b Z b
∂ψ
Φ= udy = dy = ψ|b − ψ|a . (1.40)
a a ∂y

x
a

Figure 1.10: Stress.

Now consider any curve Γ from point P to point Q. The volume flux through a curve is
the integral of the dot product of the flow velocity vector (u, v) and the normal to the

27
curve. If (dx, dy) is the infinitesimal displacement from one point of the curve to another,
the unit normal ~n to the curve there multiplied by dl is (dy, −dx), i.e.

~ndl = dy~ex − dx~ey . (1.41)

This is easily seen from

(dx~ex + dy~ey ) · (dy~ex − dx~ey ) = 0

and

(dy~ex − dx~ey ) · (dy~ex − dx~ey ) = dx2 + dy 2 = dl2

= (~ndl) · (~ndl).

The vector (1.41) points ‘outward’ as can be understood from considering fig (1.8).
y
dx
dy

~ndl
x

Figure 1.11: Say two points on a curve have tangent vector dx~ex + dy~ey then dy~ex − dx~ey
points ‘outward’ - this can be seen here following from the fact that dx and dy are positive.

The volume flux through the curve Γ with end points P and Q is

Z Q
Φ = ~u · ~ndl
P
Z Q
= (u~ex + v~ey ) · (dy~ex − dx~ey )
P
Z Q
= (udy − vdx)
P
Z Q  
∂ψ ∂ψ
= dx + dy
P ∂x ∂y
Z Q
= ∇ψ · d~r (1.42)
P

28
where we have used (1.37). By the theorem proved in section B.4 we then have

Φ = ψ|P − ψ|Q . (1.43)

Note that the flux only depends on the values ψ take at the end points, and so the flux
is independent of which curve we choose between the end points.

1.8.2 Stokes Streamfunction - 3D

Since it is axisymmetric it can be written in terms of two coordinates. There are two
different choices of coordinates systems. This will be much expanded on in future chapters.

1.8.3 The Velocity Potential

The velocity potential arises in flow where the vorticity is zero.

Thus ~ω = ∇ × ~u = 0, that is, irrotational implies that we can write ~u as

~u = ∇φ (1.44)

where φ is the velocity potential. And is the fluid is incompressible

∇2 φ = 0 (1.45)

that is it satisfies the Laplace equation.

1.9 Vortex Lines and Circulation

1.9.1 Vortex Lines

Define vorticity ~ω = (ω1 , ω2 , ω3 )

A line to which vorticity vectors are tangent at all its points is called a vortex line. The
differential equations for a vortex line are,

d~x
= λ~ω . (1.46)
ds

29
Voretx lines have a direction of ~ω and have a density in any region proportional to the
magnitude of ~ω . As ~ω = ∇ × ~u, we have that ∇ · ~ω = 0. So vortex lines are like lines of a
magnetic force - they will form closed loops.

1.9.2 Circulation

The circulation Γ along any line is defined by the line integral

Z B Z B
Γ= ~u · d~x = uidxi . (1.47)
A A

There is a simple relationship between the circulation around a closed curve and the
vorticity of the fluid over any surface bounded by that curve.

Consider the following by Stokes teorem

Z Z
~u · d~x = ∇ × ~u · dS~
C ZS Z
= ~ω · ~ndS = ωn dS (1.48)
S S

normal component of vorticity through that surface.

30
Chapter 2

Fluid Dynamics - Euler Equations

2.1 Body Forces and Surface Forces

2.1.1 Body Forces (i.e. External Forces)

If Fi denotes the component of body forces per unit mass acting in the xi −direction, the
i − th component of the resulting body force acting on V is

Z
̺Fi dV. (2.1)
V

2.1.2 Surface Forces

Let Fi denote body forces per unit mass in the i − th direction acting upon a surface
whose normal is in the j − th direction and denoted by σij

Fi = σij dSj (2.2)

2.2 Ideal (or Perfect) Fluid

An ideal (or perfect) fluid is one which can exert no shearing stress across any surface

σij = −pδij (2.3)

that is the stress tensor for a perfect fluid is diagonal

31
  
−p 0 0
σ =  0 −p 0  (2.4)
0 0 −p
Z Z
σij dSj = − pdSi (2.5)
S S

this is done ignoring viscous effects, that is µ = 0 and the fluid is inviscid.

2.3 Conservation of Momentum

We can go through the same process for momentum as with did for mass. The total
momentum in the volume V is:

Z
Πi = ̺ui dV (2.6)
V

where i runs over the three components of the momentum. We allow the proof to include
compressible fluids however will only look at incompressible solution. It’s rate of change
is just

Z
∂Πi ∂(̺ui )
= dV (2.7)
∂t V ∂t

In the absense of any forces a momentum change within the volume V can occur by
momentum flowing across the boundary,

Z
∂Πi
=− (̺ui)~u · ~ndS (2.8)
∂t S

Using the Divergence theorem the surface integral can be written as a volume integral

Z  
∂(̺ui )
+ ∇ · (̺ui~u) dV = 0 (2.9)
V ∂t

As V is arbitrary this implies an equation for the local conservation of mommentum,

∂(̺ui )
+ ∇ · (̺ui~u) = 0. (2.10)
∂t

32
Note the analogy with the mass conservation equation (1.27).

Now we consider the effects of forces. From Newtons 2nd law:

d~p d
F~ = = (m~u).
dt dt

It must be beared in mind that the we write d/dt rather than ∂/∂t, and that the laws of
motion apply to particles, not to the control volume. We have “surface forces on S” plus
“body forces” in V but assume viscous forces are negligible so surface forces are purely
due to pressure. Then we can write for an ideal fluid

Z Z Z Z
∂
̺ui dV + ui ̺uk dSk = − pdSi + ̺Fi dV (2.11)
∂t V S S V

where the second term is the outlet momentum flux and the RHS is the sum of surface
and body forces. Converting again the flow contributions to a volume integral and then
deducing local equations one obtains

∂(̺ui )
+ ∇j (̺ui uj ) = −∇i p + ̺Fi . (2.12)
∂t

This d = 3 expression can readily be generalised to give an expression that is valid in

d + 1 spacetime.

Alternatively one often proceeds by writing the LHS of (2.12) as

∂(̺ui ) ∂u ∂̺
+ ∇ · (̺ui~u) = ̺ i + ui + [ui ∇ · (̺~u) + ̺(~u · ∇)ui]
∂t ∂t ∂t   
∂ui ∂̺
= ̺ + (~u · ∇)ui + ui + ∇ · (̺~u) .
∂t ∂t

Using the continuity equation, (1.27), this reduces to

 
∂ui Dui
̺ + (~u · ∇)ui ≡ ̺ . (2.13)
∂t Dt

Using this in (2.12) it can be written

Dui ∂p
̺ + − ̺Fi = 0 (2.14)
Dt ∂xi

33
These are known as the Euler equations of motion.

So in summary the Euler equations can be applied to compressible as well as incompress-

ible flow. In these notes we restrict ourselves to incompressible flow. For an incompress-
ible, ideal fluid the equations of motion are

D~u
̺ = −∇p + ̺F~ (2.15)
Dt
∇ · ~u = 0. (2.16)

2.4 Conservative Body Forces

F~ = −∇Ω Ω = scalar (2.17)

Thus we can redefine the pressure

p′ = p + ̺Ω

so the Euler equation becomes

D~u
̺ = −∇p′ . (2.18)
Dt

2.5 Potential Flow or Irrotational Flow

For vanishing vorticity, that is irrotational flow, ~ω = ∇ × ~u = 0 implies

~u = ∇φ (2.19)

and the continuity equation gives us

∇2 φ = 0. (2.20)

For potential flow the velocity field can thus be found by solving Laplace equation for
the velocity potential. Since the Laplace equation is linear we can superimpose solutions.
In particular we can take simple flows and superimpose them to form more complicated
flows of interest such as the flow around solid objects. In chapter 3 we find simple exact

34
solutions of the Laplace equation in three dimensions, simple flows, and construct more
complex flows. We do the same for plane flows in chapter 6.

In this chapter we have not included viscosity. An important aspect of viscous fluids is
the non-slip condition at the surface of a solid object. This states that the velocity of the
fluid at a solid boundary is zero relative to the boundary. As we shall see potential flows
often dont satisfy this non-slip condition - however all liquids have a non-zero viscosity.
Can potential flows be of physical interest?

As it turns out for small viscocity (more accurately large Reynolds number, see chapter 8)
the viscous effects are concentrated in a thin boundary layer around the surface. Outside
the bounary layer the fluid acts as an invisid fluid and so potential flow theory provides
both the velocity at the outer edge of the boundary layer and the pressure there (we
explain how to find the pressure in the next section). The pressure is not significantly
effected by the thin boundary layer, so the pressure calculated from the invisid flow gives
the pressure on the surace, and then for example this can be integrated over the body’s
surface to calculate a lift.

2.6 Irrotational Flow - Bernoulli Pressure

Having constructed potential flows, the Euler equation can then be used to find the
pressure. Substitute ~u = ∇φ into the Euler equation (2.18),

 
D~u ∂~u
̺ =̺ + ~u · ∇~u = −∇p′ . (2.21)
Dt ∂t

Using subscript notation rather than vectors

!
∂ui ∂u ∂p′
̺ + uj i =− (2.22)
∂t ∂xj ∂xi

∂φ
but ui = ∂xi

!
∂ ∂φ ∂φ ∂ 2 φ ∂p′
̺ + =− (2.23)
∂t ∂xi ∂xj ∂xi ∂xj ∂xi

or

  !2  
∂  ∂φ 1 ∂φ  ∂p′
̺ + =− (2.24)
∂xi  ∂t 2 ∂xj  ∂xi

35
or

 !2 
∂  ∂φ 1 ∂φ p′
+ + =0 (2.25)
∂xi ∂t 2 ∂xj ̺

and this implies that

!2
∂φ 1 ∂φ p′
+ + = Const. = p∞ (2.26)
∂t 2 ∂xj ̺

Therefore

∂φ 1
p′ = p0 − ̺ − ̺|~u|2 . (2.27)
|{z} |2 {z }
∂t
(1) (2)

This is the Bernoulli pressure, where (1) is the rate of change of potential with time and
(2) is the dependence quadratically on velocity.

We will prove later that potential flow is not influenced by viscous forces and vorticity.

36
Chapter 3

Sources and Sinks

As mentioned in the previous chapter, for potential flow the velocity field can be found
by solving Laplace equation for the velocity potential. Since the Laplace equation is
linear we can superimpose solutions. We can take simple flows and superimpose them
to form more complicated flows of interest such as the flow around solid objects. The
momentum conservation equation, that is the Euler’s equations of motion, can then be
used to calculate the pressure.

Exact solutions are easily found and described in spherical or cylindrical coordinates,
not just in this chapter but also chapters to come. In the next section we introduce the
mathematics for spherical and cylindrical coordinates with more details and derivations
given in the appendices D and E.

3.1 Spherical and Cylindrical Coordinates

3.1.1 Spherical Coordinates

Their relation to Cartesian coordinates is

x = r sin θ cos ϕ
y = r sin θ sin ϕ
z = r cos θ (3.1)

A general position vector is then

~r = r sin θ cos ϕ~ex + r sin θ sin ϕ~ey + r cos θ~ez (3.2)

37
 ~er
θ ~eϕ

r ~eθ
ϕ

Figure 3.1:

When we work with vectors in spherical polar coordinates, we use a new local vector
basis different from ~i, ~j and ~k of Cartesian coordinates. Instead we specify vectors as
components in the basis (~er , ~eθ , ~eϕ ) of unit vectors. These basis vectors may be visulalised
as follows. To see the direction of ~er , keep θ and ϕ and increase r. To see the direction
of ~eθ , keep r and ϕ fixed and increase θ. To see the direction of ~eϕ , keep r and θ fixed
and increase ϕ. If ~r corresponds to the position vector then mathematically these unit
vectors are written

1 ∂~r 1 ∂~r 1 ∂~r

~er = , ~eθ = , ~eϕ = (3.3)
∂~r ∂r ∂~r ∂θ ∂~r ∂ϕ
∂r ∂θ ∂ϕ

Any arbitrary vector ~v can be written

~v = vr~er + vθ ~eθ + vϕ~eϕ

When working in non-Cartesian coordinates, common notation is

∂~r
hi =
∂qi

Here q1 = r, q2 = θ and q3 = ϕ. We determine the hi with the use of (3.2):

∂~r
= sin θ cos ϕ~ex + sin θ sin ϕ~ey + cos θ~ez
∂r
∂~r
= r cos θ cos ϕ~ex + r cos θ sin ϕ~ey − r sin θ~ez
∂θ
∂~r
= −r sin θ sin ϕ~ex + r sin θ cos ϕ~ey (3.4)
∂ϕ

38
From these we can check that the basis vectors are indeed orthogonal to each other. We
calculate

h2r ~er · ~er = sin2 θ cos2 ϕ + sin2 θ sin2 ϕ + cos2 θ = 1

h2θ~eθ · ~eθ = r 2 cos2 θ cos2 ϕ + r 2 cos2 θ sin2 ϕ + r 2 sin2 θ = r 2
h2ϕ~eϕ · ~eϕ = r 2 sin2 θ sin2 ϕ + r 2 sin2 θ cos2 ϕ = r 2 sin2 θ. (3.5)

to obtain

hr = 1, hθ = r, hϕ = r sin θ (3.6)

We now look at what ∇φ is in spherical polar coordinates. Write

(∇φ)i = (∇φ) · ~ei (~r)

 
∂φ ∂φ ∂φ
= ~e + ~e + ~e · ~ei (~r) (3.7)
∂x x ∂y y ∂z z

We have

1 ∂~r
~ex · ~ei = ~ex ·
hi ∂qi
1 ∂  
= ~ex · x~ex + y~ey + z~ez
hi ∂qi
1 ∂x
= .
hi ∂qi

Substutuiting this and analogous results for ~ey and ~ez into (3.7) gives

1 ∂φ ∂x 1 ∂φ ∂y 1 ∂φ ∂z
(∇φ)i = + +
hi ∂x ∂qi hi ∂y ∂qi hi ∂z ∂qi
1 ∂φ
=
hi ∂qi

Therefore we have

39
 3
X 1 ∂φ
∇φ = ~ei (~r) . (3.8)
i=1
hi ∂qi

In the case of spherical polar coordinates,

∂φ 1 ∂φ 1 ∂φ
∇φ = ~er + ~eθ + ~eϕ . (3.9)
∂r r ∂θ r sin θ ∂ϕ

If ~u is the velocity and if ~u = ∇φ then we have,

∂φ 1 ∂φ 1 ∂φ
ur = , uθ = , uϕ =
∂r r ∂θ r sin θ ∂ϕ

3.1.2 Cylindrical Coordinates

Their relation to Cartesian coordinates is

x = ρ cos ϕ
y = ρ sin ϕ
z = z (3.10)

A general position vector is then

~r = ρ cos ϕ~ex + ρ sin ϕ~ey + z~ez (3.11)

These basis vectors may be visulalised as follows. To see the direction of ~eρ , keep ϕ and z
and increase ρ. To see the direction of ~eϕ , keep ρ and z fixed and increase ϕ. To see the
direction of ~ez , keep ρ and ϕ fixed and increase z. If ~r corresponds to the position vector
then mathematically these unit vectors are written

1 ∂~r 1 ∂~r 1 ∂~r

~eρ = , ~eϕ = , ~ez = (3.12)
∂~r ∂ρ ∂~r ∂ϕ ∂~r ∂z
∂ρ ∂ϕ ∂z

Any arbitrary vector ~v can be written

~v = vρ~eρ + vϕ~eϕ + vz ~ez

40
 ~ez
ρ ~eϕ
z
~eρ

Figure 3.2: Stress.

∂~r
= cos ϕ~ex + sin ϕ~ey
∂ρ
∂~r
= −ρ sin ϕ~ex + ρ cos ϕ~ey
∂ϕ
∂~r
= ~ez (3.13)
∂z

From these we can check that the basis vectors are indeed orthogonal to each other. We
calculate

h2ρ~eρ · ~eρ = cos2 ϕ + sin2 ϕ = 1

h2ϕ~eϕ · ~eϕ = ρ2 sin2 ϕ + ρ2 cos2 ϕ = ρ2
h2z ~ez · ~ez = 1. (3.14)

to obtain

hρ = 1, hϕ = ρ, hz = 1 (3.15)

As before we have the formula

3
X 1 ∂φ
∇φ = ~ei (~r) . (3.16)
i=1
hi ∂qi

In the case of cylindrical polar coordinates,

41
 ∂φ 1 ∂φ ∂φ
∇φ = ~eρ + ~eϕ + ~ez . (3.17)
∂ρ ρ ∂ϕ ∂z

If ~u is the velocity and if ~u = ∇φ then we have,

∂φ 1 ∂φ ∂φ
uρ = , uϕ = , uz =
∂ρ ρ ∂ϕ ∂z

3.2 Uniform Flow

It is easy to see that

φ = Ux (3.18)

represents uniform flow with constant velocity U in the x−direction.

The streamfunction for uniform flow is

ψ = Uy. (3.19)

3.3 Point Sources and Sinks

The simplest solution is a point source or sink.

Let us consider Laplace’s equation in spherical polar coordinates (r, θ, ϕ), that is

∇2 φ = 0 (3.20)

or

   
1 ∂ 2 ∂φ 1 ∂ ∂φ 1 ∂2φ
r + 2 sin θ + 2 2 =0 (3.21)
r 2 ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2

Consider the problem with radial symmetry, that is φ is a function of r only, that is
φ = φ(r). Then Laplace’s equation reduces to

 
1 ∂ 2 ∂φ
r =0 (3.22)
r 2 ∂r ∂r

42
so that

A
φ=− +B r 6= 0 (3.23)
r

Without loss of generality we choose B = 0 (since we always take the derivative of φ in

obtaining the velocities), so this solution satisfies Laplaqce’s equation except at the origin

m
φ=− (3.24)
4πr

the 4π is a normalising factor.

p p
Note r = x2 + y 2 + z 2 in Cartesian coordinates and r = ρ2 + z 2 in cylindrical coor-
dinates.

The potential (3.24) only has radial velocities

∂φ m
ur = = (3.25)
∂r 4πr 2

and this is outflow. We see that the minus sign in (3.24) corresponds to outward flow
from the origin, i.e. a source of material. So (3.24) is the velocity potential due to a
simple source of strength m at r = 0. If we wanted a sink would make the replacement
m → −m and we would have inflow.

The volume flux Φ through a spherical surface at r = a is radial velocity times the area
of the sphere,

m
Φ= × 4πa2 = m.
4πa2

3.3.1 In Cylindrical Coordinates

In cylindrical coordinates

m
φ=− p
4π x2 + ρ2

∂φ mx 1 ∂ψ
ux = = 2 2 3/2
=
∂x 4π(x + ρ ) ρ ∂ρ
∂φ mρ 1 ∂ψ
uρ = = 2 2 3/2
=− (3.26)
∂̺ 4π(x + ρ ) ̺ ∂x

43
so that ψ(x, ρ) is the streamfunction, that is

mx
ψ(x, ρ) = − p . (3.27)
4π x2 + ρ2

If the source is located at (x0 , y0 , z0 )

m
φ=− p (3.28)
4π (x − x0 )2 + (y − y0 )2 + (z − z0 )2

3.4 Doublets

If a source of strength m is situated at (d, 0, 0) and a sink of the same strength at (−d, 0, 0),
the potential is a superposition of the two, that is

!
m 1 1
φ= p −p (3.29)
4π (x + d)2 + y 2 + z 2 (x − d)2 + y 2 + z 2

If d is infintesimal then we have to first order in d in a Taylor expansion

1 1 ∂ 1
p = +d .
(x + d)2 + y 2 + z 2 r ∂x r

Thus to first order in d we have for φ

 
m ∂ 1 2md x
φ= 2d =−
4π ∂x r 4π r 3

limd→0 2md = µ

−µx
φ= (3.30)
4πr 3

This satisfies Laplaces’s equation and is called a doublet. The axis of a doublet is the
lines drawn from the sink to the source, in the case the x axis. Likewise with y and z
directions

µ2 y
φ = −
4πr 3
µ z
φ = − 33 (3.31)
4πr
44
In general

~µ · ~r
φ=− (3.32)
4πr 3

where (µ1 , µ2 , µ3 ) and doublet acting in the ~x direction.

3.5 Axi-Symmetric Flow

Spherical Polar coordinates (r, θ, ϕ)

∂φ 1 ∂ψ
ur = = 2
∂r r sin θ ∂θ

1 ∂φ 1 ∂ψ
uθ = =−
r ∂θ r sin θ ∂r

We have axi-symmetric flow if ψ = ψ(r, θ)

The formula for the velocities in terms of the streamfunctions will be given in chapter 10.

Cylindrical Polar coordinates (ρ, ϕ, z)

The velocity potential is

∂φ 1 ∂ψ
uρ = =−
∂ρ ρ ∂z

and the streamfunction is

∂φ 1 ∂ψ
uz = =
∂z ρ ∂ρ

Again the formula for the velocities in terms of the streamfunctions will be given in chapter
10.

45
3.5.1 Uniform Flow

In spherical Polar coordinates

φ = Ur cos θ

1
ψ = Ur 2 sin2 θ
2
as

∂φ 1 ∂φ
ur = = U cos θ and uθ = = −U sin θ
∂r r ∂θ

and

∂φ 1 ∂ψ 1 ∂ψ
ur = = 2 = U cos θ and uθ = − = −U sin θ.
∂r r sin θ ∂θ r sin θ ∂r

In cylindrical coordinates the velocity potential is

φ = Ux

The streamfunction is

1
ψ = Uρ2
2
as

∂φ ∂φ
uρ = =0 and ux = =U
∂ρ ∂x

and

1 ∂ψ 1 ∂ψ
uρ = − =0 and ux = =U
ρ ∂x ρ ∂ρ

46
3.5.2 Source

The velocity potential is

m
φ=−
4πr

The streamfunction is

m
ψ=− cos θ
4π

and the velocity potential in cylindrical polar coordinates is

m
φ=− p
4π z 2 + ρ2

which is the same function as given above but expressed in cylindrical polar coordinates.

The velocities in spherical coordinates are via the streamfunction

1 ∂ψ m 1 ∂ψ
ur = = and uθ = − = 0.
r 2 sin θ ∂θ 4πr 2 r sin θ ∂r

The velocities in cylinidrical coordinates via the velocity potential are

∂φ ∂ m mρ
uρ = =− p = p
∂ρ ∂ρ 4π z + ρ
2 2 4π z 2 + ρ2

∂φ mz
uz = = p
∂z 4π z 2 + ρ2

3.5.3 Doublet

The velocity potential in spherical polar coordinates is

µ cos θ
φ=−
4πr 2

The streamfunction in spherical polar coordinates is

47
 µ sin2 θ
ψ=
4πr

The velocity potential in spherical polar coordinates via the vector potential is

∂ µ cos θ µ cos θ 1 ∂ µ cos θ µ sin θ

ur = − 2
= and uθ = − =
∂r 4πr 2πr 3 r ∂θ 4πr 2 4πr 3

The velocity potential in spherical polar coordinates via the streamfunction is

1 ∂ µ sin2 θ µ cos θ 1 ∂ µ sin2 θ µ sin θ

ur = 2 = and uθ = − =
r sin θ ∂θ 4πr 2πr 3 r sin θ ∂r 4πr 4πr 3

The velocity potential in cylindrical polar coordinates is

µz
φ=−
4πρ3/2

The velocities in cylinidrical coordinates via the velocity potential are

∂φ ∂ µz 3 µz
uρ = =− 3/2
=
∂ρ ∂ρ 4πρ 2 4πρ5/2

∂φ µ
uz = =−
∂z 4πρ3/2

3.6 The Half-Body

We can use singular solutions to “build up” the fow around different shape bodies. First
example is the so-called “half-body”.

Consider what happens when a source is placed in free-stream.

1 m
ψ(r, θ) = Ur 2 sin2 θ − cos θ (3.33)
|2 {z } | 4π{z }
unif orm source

We have a dividing streamline for ψ(r, θ) = m/4π.

The equation of the “half-body” is given by

48
 m 1 m
= Ur 2 sin2 θ − cos θ
4π 2 4π

that is

m 1
(1 + cos θ) = Ur 2 sin2 θ
4π 2

In terms of half angles

 2
m 2 θ 1 2 θ θ
2 cos = Ur 2 sin cos
4π 2 2 2 2

which gives

 m 1/2 θ
r= cosec
4π 2

3.7 Rankine Bodies

A closed body is generated by superimposing a uniform stream plus sources and sinks
such that the outflow due to the sources is exactly balanced by the inflow due to the sinks,
with the line joining the sources and sinks along the stream direction.

A simply example is a source at x = −a and sink at x = a plus free-stream.

+ −
−d d

Figure 3.3: .

Potential (spherical polars)

m m
φ = Ur cos θ − + (3.34)
4πr1 4πr2

49
and streamfunction

1 m cos θ1 m cos θ2
ψ = Ur 2 sin2 θ − + (3.35)
2 4π 4π

3.8 Flow Around a Shpere

Important problem - many particles are spherical or nearly spherical. Ball games involve
spherical objects. Spherical inhomogeneties are used to change flow characteristics for
example to “trip” boundary layers or improve mixing in jet engines.

Consider a frame of reference such that the sphere is at rest and the fluid flow at ∞ is U
in the x−direction.

a
U

Figure 3.4: .

Consider the Rankine body - push the soure and sink together, so that we have a dipole
(source-doublet). The velocity potential φ and streamfunction are

µ cos θ
φ = Ur cos θ −
r2
1 2 2 µ sin2 θ
ψ = Ur sin θ + (3.36)
2 r

The first terms correspond to the uniform stream and the second to the dipole.

The radial velocity is

∂φ 2µ
vr = = U cos θ + 3 cos θ. (3.37)
∂r r

We require that this velocity is zero on the sphere (a physical consideration) so that the
1
boundary condition is ur = 0 on r = a so that µ = − Ua3 . Then
2
50
  
a3
φ = U r + 2 cos θ
2r
 
1 2 a3
ψ = U r − sin2 θ (3.38)
2 r

and the velocities are

 
a3
ur = U 1 − 3 cos θ
r
 
1 ∂φ a3
uθ = = −U 1 + 3 sin θ (3.39)
r ∂θ 2r

Notice that

(i) the disturbance to the freestream is proportional to 1/r 3 and

3
(ii) uθ (r = a) = − U sin θ and this is the slip velocity.
2
(iii) Stagnation points

Stagnation points at r = a, θ = 0 and θ = π, ur = uθ = 0

(iv) Pressure distribution on the sphere.

∂φ
As the flow is steady = 0 so that the Bernoulli pressure is given by
∂t

1
p = p0 − ̺|~u|2
2

so up stream of the sphere p = p∞ and ux = U

1 1
p∞ = p0 − ̺U 2 implying p0 = p∞ + ̺U 2
2 2

therefore

1 1
p = p∞ + ̺U 2 − ̺|~u|2 .
2 2

3
On r = a, ur = 0 and uθ − U sin θ therfore
2

51
  
p − p∞ 1 2 9 2
= U 1 − sin θ (3.40)
̺ 2 4

(v) Cavitation (boiling at ordinary temperature)

If the minimum pressure, pmin , is less than the saturated vapour pressure pc then liquid
π
will caviate. The minimum pressure for a sphere occurs at θ =
2

5
pmin = p∞ − ̺U 2 (3.41)
8

We have pmin ≤ pc if

5
p∞ − ̺U 2 ≤ pc
8
or

r  1/2
8 p∞ − pc
U≥ . (3.42)
5 ̺

Rearanged this condition becomes

5 p − pc
≥ ∞
1
4 2
̺U 2

In hydraulic engineering the cavitation number is defined by

p∞ − pc
κ= 1
2
̺U 2

for a sphere one has cavitation at κ = 54 , experimentally κ < 1

2
for torpedo shaped bodies.

3.9 Line Distributions

We have previously examined cases of a finite number of singularities representing a body

(for example oval shaped bodies or spheres).

The next step is to consider an infinite set of singularities each infinitesimally close to its
neighbour over a finite line segment.

52
For example uniform distribution of sources

Streamfunction for a point source at x = s is given in Cartesian coordinates by

m(x − s) m(x − s)
ψ=− =− p (3.43)
4πr 4π (x − s)2 + y 2 + z 2

Let the total strength of the sources be m, take the strength of the source in element δs
as mδs/a, with m constant for a uniform distribution. For a continuous distribution from
s = 0 to s = a if we let δs → 0

Z a
1 m(s)(x − s) m
ψ=− p ds = [R − R] (3.44)
4πa 0
2 2
4π (x − s) + y + z 2 4πa 2

where R2 is the distance from the point (a, 0, 0) and R the distance to the origin.

Airship forms source + line sink + uniform stream

1 m m
ψ = Ur 2 sin2 θ + (R − R2 ) − cos θ
2 4πa 4π

This technique is the basis of slender body theory.

53
Chapter 4

Viscosity - The Navier-Stokes

Equation

4.1 Equations of Motion

Z Z Z
d
̺ui dV = σij nj dS + ̺Fi dV (4.1)
dt V S V

Or

Z Z Z
D~u
̺ dV = σ · ~ndS + ̺f~dV (4.2)
V Dt S V

by taking the derivative inside the integral, so that it becomes a substantial dreivative.
On applying the Divergence theorem,

Z Z
~
∇ · AdV = ~ · ~ndS,
A
V S

and rearranging we obtain

Z  
D~u ~
̺ − ̺f − ∇ · σ dV = 0 (4.3)
V Dt

Since V is arbitrary the integrand must be indentically zero, that is

D~u
̺ = ̺f~ + ∇ · σ (4.4)
Dt

54
or in subscript notation

Dui ∂σij
̺ = ̺fi + (4.5)
Dt ∂xj

4.2 Newtonian Fluid

4.2.1 Pressure

For a real fluid (that is viscous, heat conducting, ...) in motion, we shall define the
(mechanical) pressure by

1 1
p = − T r(σij ) = − σii .
3 3
In Newtonian fluid there is only a linear dependence on the rate of strain tensor in which
case,

σij = −pδij + 2µeij + λekk δij (4.6)

If we let i = j and sum (4.6) we obtain

σii = −3p + (2µ + 3λ)eii

1
However by definition p = − σii , so wemust have the following relation between λ and µ
3

2
λ = − µ.
3
Therefore the constitutive relation for a Newtonianfluid isgiven by

 
1
σij = −pδij + 2µ eij − ekk δij
3

This relation also tells us that the principal axis for σij and eij coincide.

4.2.2 Viscosity

µ is known as the Newtonian viscocity.

55
4.2.3 Incompressible Newtonian Fluid

ekk = ∇ · ~u = 0 (4.7)

so

σij = −pδij + 2µeij . (4.8)

This is the constitutive relation on which is nearly all these nots are based.

4.3 Navier-Stokes Equation

Cauchy derived the equation

Dui ∂σij
̺ = ̺fi + (4.9)
Dt ∂xj

which together with the incompressibility condition for the continuity equation

∂uj
=0 (4.10)
∂xj

The stress tensor for a Newtonian fluid is,

σij = −pδij + 2µeij

which on substitution for eij leads to

!
∂ui ∂uj
σij = −pδij + µ +
∂xj ∂xi

This may now be substituted into the Cauchy’s equation to yield

" !#
Dui ∂p ∂ ui 2
∂ ∂uj
̺ = ̺fi − +µ 2
+ (4.11)
Dt ∂xi ∂xj ∂xi ∂xj

The last term on the RHS of this equation is identically zero because of the continuity
equation. The Navier-Stokes equations emerges

56
 Dui ∂p ∂2u
̺ = ̺fi − + µ 2i (4.12)
Dt ∂xi ∂xj

or in vector notation

D~u
̺ = −∇p + µ∇2~u + ̺f~ (4.13)
Dt

4.3.1 Conservative Body Force

Often the body force is conservative - that is

f~ = −∇Ω (4.14)

where Ω is the potential energy per unit mass. For example gravity

f~ = −g~k (4.15)

where ~k is the unit vertical upward vector, and hence

Ω = gz. (4.16)

Thus we can redefine the pressure as

p′ = p + ̺Ω (4.17)

which upon substitution into (4.13) leads to

D~u
̺ = −∇p′ + µ∇2~u (4.18)
Dt
It is convenient to write the equations as

D~u 1
= − ∇p′ + ν∇2 ~u (4.19)
Dt ̺

Writing out the substantial derivative

∂~u 1
+ ~u · ∇~u = − ∇p′ + ν∇2~u (4.20)
∂t ̺

57
4.4 Alternative Form of the Navier-Stokes Equation
and the Vorticity Equation

We first derive a couple of identities using ǫijk ǫki′ j ′ = (δii′ δjj ′ − δij ′ δji′ ):

[~u × ~ω ]i = ǫijk uj ωk
= ǫijk uj (ǫki′ j ′ ∂i′ uj ′ )
= ǫijk ǫki′ j ′ uj ∂i′ uj ′
= (δii′ δjj ′ − δij ′ δji′ )uj ∂i′ uj ′
= uj ∂i uj − uj ∂j ui
1 2
= ∂ u − ~u · ∇ui
2 i
1
= [ ∇~u2 − ~u · ∇~u]i
2

and

[∇ × ~ω ]i = ǫijk ∂j ωk
= ǫijk ∂j (ǫki′ j ′ ∂i′ uj ′ )
= ǫijk ǫki′ j ′ ∂j ∂i′ uj ′
= (δii′ δjj ′ − δij ′ δji′ )∂j ∂i′ uj ′
= ∂j ∂i uj − ∂j ∂j ui
= [−∇2 ~u + ∇(∇ · ~u)]i .

In vector notation these two identities read

1
~u · ∇~u = ∇|~u|2 − ~u × ~ω (4.21)
2

and

∇ × ~ω = −∇2~u + ∇(∇ · ~u). (4.22)

For incompressible flow (∇ · ~u = 0) this second identity becomes

∇ × ~ω = −∇2~u. (4.23)

58
Substitution of (4.21) and (4.23) into (4.20)

 
∂~u 1 2 p
− ~u × ω
~ = −∇ |~u| + − ν∇ × ~ω (4.24)
∂t 2 ̺

”This is the ‘best’ form of the equation to use when considering coordinate systems other
than Cartesian coordinates”.

We can take the curl of (4.20) using ∇ × (∇p′ ) = 0,

∂~ω
+ ~u · ∇~ω = ~ω · ∇~u + ν∇2 ~ω (4.25)
∂t

4.4.1 The Vorticity Equation in Two Dimensions

In two dimensions the first term on the RHS vanishes, that is

~ω · ∇~u = 0. (4.26)

Therefore in two dimensions the vorticity satisfies the diffusion equation,

D~ω
= ν∇2 ~ω . (4.27)
Dt

4.5 Vorticity and Viscocity

It is possible to show that without viscous forces there cannot be a vorticity fieled in the
flow which varies with coordinate directions. This is expressed in the following relationship

∂τij ∂ωk
= µ∇2 ui = −µǫijk
∂xj ∂xj

(recall σij = −pδij + τij where τij = 2µeij ). The equality of the last expression with the
central expression can be derived in the following manner

59
 ∂ωk ∂ ∂
−µǫijk = −µǫijk (ǫkmn u )
∂xj ∂xj ∂xm n
∂ ∂
= −µǫijk ǫkmn u
∂xj ∂xm n
∂ 2 un
= −µ(δim δjn − δin δjm )
∂xj ∂xm
∂ ∂uj ∂ 2 ui
= −µ( − )
∂xi ∂xj ∂xj ∂xj
= µ∇2 ui . (4.28)

where we used ∂uj /∂xj = 0.

If invisid flow starts with no vorticity then no vorticity will be produced. To understand
this intuitively we note that of the three types of force that can act on a cubic fluid
element, the pressure, body forces, and viscous forces, only the viscous shear forces are
able to give rotary motion. Hence if the viscous effects are nonexistent, vorticity cannot
be introduced.

60
Chapter 5

Streamfunction in 2-Dimensions

If

∇ · ~u = 0 (5.1)

~ x, t) such that
then it follows that there exists a vector field A(~

~
~u = ∇ × A. (5.2)

5.1 Cartesians

As noted in section 1.8.1 the continuity equation in 2D

∂u ∂v
+ =0
∂x ∂y

is solved by choosing

∂ψ ∂ψ
u= and v=− .
∂y ∂x

~ for A
The continuity equation is solved if we write ~u = ∇ × A ~ = ψ(x, y, t)~z, then

61
 ~ex ~ey ~ez
~ = ∂ ∂ ∂
~u = ∇ × A
∂x ∂y ∂z
0 0 ψ(x, y, t)
∂ψ ∂ψ
= ~ex − ~e (5.3)
∂y ∂x y

thus

∂ψ ∂ψ
u= , v=− (5.4)
∂y ∂x

5.2 Cylindrical Polar Coordinates

The relation to Cartesian coordinates is

x = r cos θ, y = r sin θ.

From section 3.1.2 we see that the formula for the velocity components ur and uθ in terms
of the velocity potential

∂φ 1 ∂φ
ur = and uθ = . (5.5)
∂r r ∂θ

5.2.1 Relating Velocity to Streamfunction by Solving the Con-

tinuity Equation

In cylindrical polar coordinates the continuity equation

1 ∂ 1 ∂uθ
(rur ) + =0 (5.6)
r ∂r r ∂θ

With the choice

1 ∂ψ ∂ψ
ur = and uθ = − (5.7)
r ∂θ ∂r

the continuity equation is solved.

62
We obtain these also from the curl in cylindrical polar coordinates

~er r~eθ ~ez

~ = 1 ∂ ∂ ∂
~u = ∇ × A
r ∂r ∂θ ∂z
0 0 ψ(r, θ, t)
1 ∂ψ ∂ψ
= ~er − ~e (5.8)
r ∂θ ∂r θ

We demonstrate that ψ is indeed constant along the streamlines using the formula (5.7)
that we have obtained relating the velocity components to this streamfunction. The
differential equation for streamlines are

d~x
= λ~u.
ds

and thus dr = λur ds and rdθ = λuθ ds. Using this find

∂ψ ∂ψ
dψ = dr + dθ
∂r ∂θ
= −uθ dr + rur dθ
= −uθ λur ds + ur λuθ ds = 0. (5.9)

5.2.2 Direct Calculation

Here we derive (5.7) directly from (5.4)

Since the streamfunction is defined completely by the geometry of the streamlines, it must
be the same in any coordinates system. So if ψC (x, y, t) is the streamfunction in Cartesian
coordinates, the stream function in polar coordinates will be

ψP (r, θ, t) = ψC (r cos θ, r sin θ, t). (5.10)

The position vector expressed in Cartesian basis vectors

~r = r cos θ~i + r sin θ~j.

Taking the derivative we obtain

63
 ∂~r
= cos θ~i + sin θ~j.
∂r

This already has unit length so

~er = cos θ~i + sin θ~j

∂~r
= −r sin θ~i + r cos θ~j
∂θ

This already has unit length so

~eθ = − sin θ~i + cos θ~j

~u = ur~er + uθ~eθ
= ur [cos θ~i + sin θ~j] + uθ [− sin θ~i + cos θ~j]
= [u cos θ − u sin θ]~i + [u sin θ + u cos θ]~j
r θ r θ (5.11)

having collected terms in ~i and ~j we see that

u = ur cos θ − uθ sin θ, v = ur sin θ + uθ cos θ.

Applying the formula for u and v in terms of the stream function

∂ψ
u = = ur cos θ − uθ sin θ
∂y
∂ψ
v = − = ur sin θ + uθ cos θ (5.12)
∂x

Solving for ur and uθ ,

∂ψC ∂ψ
ur = cos θ − sin θ C
∂y ∂x
∂ψ ∂ψ
uθ = − sin θ C − cos θ C (5.13)
∂y ∂x

64
Now using the chain rule we can write formula for the derivatives

∂ ∂x ∂ ∂y ∂
= +
∂r ∂r ∂x ∂r ∂y
∂ ∂
= cos θ + sin θ (5.14)
∂x ∂y

and

∂ ∂x ∂ ∂y ∂
= +
∂θ ∂θ ∂x ∂θ ∂y
∂ ∂
= −r sin θ + r cos θ . (5.15)
∂x ∂y

Solving for ∂/∂x and ∂/∂y we obtain

∂ ∂ 1 ∂
= cos θ − sin θ
∂x ∂r r ∂θ
∂ ∂ 1 ∂
= sin θ + cos θ (5.16)
∂y ∂r r ∂θ

Substituting these into (5.13) and using (5.10)

   
∂ 1 ∂ ∂ 1 ∂
ur = cos θ sin θ + cos θ ψ − sin θ cos θ − sin θ ψ
∂r r ∂θ P ∂r r ∂θ P
1 ∂ψ
= (5.17)
r ∂θ

   
∂ 1 ∂ ∂ 1 ∂
uθ = − sin θ sin θ + cos θ ψ − cos θ cos θ − sin θ ψ
∂r r ∂θ P ∂r r ∂θ P
∂ψ
= − (5.18)
∂r

In polar coordinates the formula for the velocity components in terms of the stream
function are

1 ∂ψ ∂ψ
ur = , uθ = − (5.19)
r ∂θ ∂r
where ~u = ur~er + uθ~eθ .

65
Chapter 6

Plane Potential Flow

6.1 Potential Flow

a) Potential flow has no vorticity

∂uk ∂2φ
ωi = ǫijk = =0
∂xj ∂xj ∂xk

b) Potential flow is not influenced by viscous forces. Recall the stress tensor can be
written: σij = −pδij + τij where τij is the viscous stress tensor - see (4.9). The viscous
term in the Navier-Stokes equation is

∂τij
= µ∇2 ui .
∂xj

From ∇ × ~ω = ∇ × (∇ × ~u) = ∇(∇ · ~u) − ∇2~u = −∇2 ~u (where we have used ∇ · ~u = 0),

∂ωk
µ∇2 ui = −µǫijk = 0.
∂xj

c) Potential flow satisfies Laplace equation,

0 = ∇ · ~u = ∇ · (∇φ) = ∇2 φ.

66
6.2 Complex Representation

We will show in the following that any analytic function represents a two dimensional
flow.
~ = ψ~k such that
Incompressible - ∇ · ~u = 0 implies that there exists a vector field A
~u = ∇ × ψ~k or

∂ψ ∂ψ
u= , v=− (6.1)
∂y ∂x

Irrotational - ~ω = ∇ × ~u = 0 implies there is φ such that ~u = ∇φ or

∂φ ∂φ
u= , v= . (6.2)
∂x ∂y

Using (6.2) in the continuity equation ∇~u = 0 gives

∂u ∂v
+ = ∇2 φ = 0.
∂x ∂y

Using that potential flow has no vorticity we find using (6.1)

∂v ∂v
−ω = − + = ∇2 ψ = 0.
∂x ∂y

Thus both the velocity potetial φ and the streamfrunction ψ the Laplace equation.

These imply

∂ψ ∂φ ∂ψ ∂φ
= , − = (6.3)
∂y ∂x ∂x ∂y

which are the Cauchy-Riemann equations. These are necessary condition for the complex
function φ(x, y) + iψ(x, y) to be analytic, as we shall now show. First we must note that

f ′ (z) exists and is unique ⇐⇒ f (z) ia an analytic function

requiring that a derivative of a complex function should be the same independent of the
direction we take the limit in the complex plane. We have

67
 1
f ′ (z) = lim [f (z + ∆z) − f (z)]
∆z→0 ∆z
1
= lim [φ(x + ∆x, y) − φ(x, y) + iψ(x + ∆x, y) − iψ(x, y)]
∆x→0 ∆x
1
= lim [φ(x, y + ∆y) − φ(x, y) + iψ(x, y + ∆y) − iψ(x, y)] (6.4)
i∆y→0 ∆x

Equating the real parts of thes last two give lines give

∂φ ∂ψ
=
∂x ∂y

and equating imaginary parts

∂ψ ∂φ
=− .
∂x ∂y

Thus we have that the Cauchy-Riemann equations as a necessary requirement for a com-
plex function to be analytic.

dw ∂φ ∂ψ
= +i
dz ∂x ∂x

dw
= u − iv (6.5)
dz

Note that

dw √
= u2 + v 2
dz

which is the magnitude of the velocity. dw dz

is the complex velocity and is analytic since
the derivative of an analytic function is analytic.

6.3 Simple Flows

(b) Uniform flows

(b) Flow near a stagnation point

68
(c) Flow around a corner

(i) Flows for 0 ≤ θ ≤ π/n

(ii) Flows for 0 ≤ θ ≤ 2π/n

6.3.1 Uniform Flows

Consider uniform flow in the x−direction,

ux = u = U, v=0

which implies that φ = Ux+k, ψ = Uy +k ′ where k and k ′ are arbitrary amd do not effect
the velocity or Cauchy-Riemann equations as these all involve derivatives, and hence can
be dropped without loss of generality. The complex potential is w(z) = Uz and

dw
= U = U + i0
dz

this is where the stream is aligned with the x−direction.

y

Figure 6.1: Uniform motion in the x−direction..

We now calculate the velocity field from the complex potential w = Ue−iα z.

dw
= Ue−iα = U(cos α − i sin α)
dz

u = U cos α
v = U sin α. (6.6)

69
 y

Figure 6.2: Uniform motion at an angle α

√
Obviously u2 + v 2 = U.

It is easy to show that multiplying any complex potential w(z) by e−iα rotates the velocity
vector anti-clockwise by an angle of α at each point in space: as

dw(z)
= u(x, y) − iv(x, y)
dz

then

d(w(z)e−iα )
= [u(x, y) − iv(x, y)][cos α − i sin α]
dz
= u(x, y) cos α − v(x, y) sin α − i[u(x, y) sin α + v(x, y) cos α]
  
cos α − sin α u(x, y)
= . (6.7)
sin α cos α v(x, y)

Note that we are rotating the velocity vector at each point in space rather than rotating
about the origin.

6.3.2 Flow Near a Stagnation point

A stagnation point zs has both velocity components zero, i.e. u = v = 0 at z = zs thus

dw
= 0. (6.8)
dz z=zs

For some potential w(z) (analytic about zs ) consider a Taylor expansion of w(z) about zs
to invetigate flow near a stagnation point.

70
 1
w(z) = w(zs ) + (z − zs )w ′ (zs ) + (z − zs )2 w ′′ (zs ) + O(|z − zs |3 )
2

so that

1
w(z) = w(zs ) + (z − zs )2 w ′′ (zs ) + O(|z − zs |3 ). (6.9)
2

Assuming that w ′′ (zs ) 6= 0 it may be written in polar coordinate form w ′′ (zs ) = Aeiα
A > 0 and α real. The term w(zs ) is a constant and so does not effect the resulting
velocity field and so can be neglected. The flow resulting from (6.9) is the same as the
flow resulting from (1/2)z 2 Aeiα + O(|z|3 ) but shifted by an amount xs in the x−direction
and an amount ys in the y−direction. We are considering the flow near the stagnation
point and so the terms represented by O(|z − zs |3 ) are to be neglected. We know that a
factor eiα rotates the velocity vector by an angle of α clock-wise at each point in space.

In the case of monomials in z − zs , we will find that we can obtain the flow by performing
a rotation about the point zs . First put z − zs = (x − xs ) + i(y − ys ). Consider the (x′ y ′ )
axis defined by x′ = x − xs and y ′ = y − ys . Then consider this axis rotated through −α/2
and becoming the X − Y axis so that

 ′   
x cos(α/2) sin(α/2) X
= . (6.10)
y ′
− sin(α/2) cos(α/2) Y

(see fig 6.3 and note for example that X = a and Y = 0 corresponds to x′ = a cos(α/2)
and y ′ = −a sin(α/2)) this is effected via

iα
(z − zs ) = (X + iY )e− 2
= [X cos(α/2) + Y sin(α/2)] + i[−X sin(α/2) + Y cos(α/2)] (6.11)

Now X + iY can also by written in polar form Reiθ

Then the complex potenetial in X − Y coordinates is

1
w = C + (X + iY )2
2
1
= Re(C) + (X 2 − Y 2 ) + i[Im(C) + AXY ]. (6.12)
2

Using

71
 y′ Y

y
α/2 x′
zs
X

Figure 6.3: Flow at a stagnation point.

w = φ̃(X, Y ) + iψ̃(X, Y )

ψ̃(X, Y ) = AXY + constant (6.13)

and

∂ ψ̃(X, Y ) ∂ ψ̃(X, Y )
uX = and uY = − . (6.14)
∂Y ∂X

So

uX = AX and uY = −AY. (6.15)

So the flow in the neighbourhood of zs in the X −Y frame is the same as the flow generated
1
by the complex potential z 2 .
2

6.3.3 Flow Around a Corner

A natural extension of the preceding analysis is to consider the complex potential

w(z) = Az n , (6.16)

In polar coordinates the complex potential becomes

72
 Y

Figure 6.4: Flow at a stagnation point.

Ar n einθ = Ar n cos (nθ) + iAr n sin (nθ) (6.17)

Thus the velocity potential and the streamfunction can be identified as

φ = Ar n cos (nθ)
ψ = Ar n sin (nθ) (6.18)

The complex velocity is then

dw
= Anz n−1 . (6.19)
dz

The streamfunction in spherical polar coordinates is given by

1 ∂ψ
ur =
r ∂θ

∂ψ
uθ = −
∂r

ur = Anr n−1 cos (nθ) , uθ = −Anr n−1 sin (nθ) (6.20)

Consider the streamlines Ψ = 0. This obviously implies that uθ = 0, which could concur
with there being a wall through which there is no flow. In this case 0 ≤ θ ≤ π/n.

73
At θ = 0 and θ = π/n (n 6= 1) we have uθ = 0, while at θ = π/2n ur = 0 and uθ has
its maximum value. Flow is parallel to the walls at the walls - the velocity component
perpendicular to the wall is zero as it should be. At the point r = 0 we have ur = uθ = 0
so the velocity of the fluid is zero there and we have a stagnation point.
Ψ=0

π
n
Ψ=0

Figure 6.5: Flow at a corner.

These solutions corresponing to integer value n can be extended by having the wall at
2π/n as shown in fig (5.4) while retaining the same velocity formula (6.20) but now with
the variable θ extended to the range 0 ≤ θ ≤ 2π/n. Where there was a wall there is
now a stagnation point streamline. The streamfunction is zero on the walls and on the
stagnation point streamline.
Ψ=0
Ψ=0

π
n
Ψ=0

Figure 6.6: Extended solution from flow at a corner with walls at θ = 0 and θ = 2π/n.

(i) Flows for 0 ≤ θ ≤ π/n

Flow for n = 2 and walls at θ = 0 and θ = π/2

Here we can consider Cartesian coordinates.

w(z) = Az 2 = A(x + iy)2 = A(x2 − y 2) + i2Axy

The complex velocity can then be obtained from

74
 dw(z)
= 2Az
dz
= 2A(x + iy).

Implying

u = 2Ax
v = −2Ay

Ψ=0

Figure 6.7: Flow around a right-angled corner, n = 2.

Figure 6.8: Uniform flow over a flat plate n = 1.

(ii) Flows for 0 ≤ θ ≤ 2π/n

Stagnation point flow when n = 2 and walls at θ = 0 and θ = π

Or if we allow

In chapter 7 we will reconsider this flow but for non-zero viscocity where we demand that
u = v = 0 at the wall but that the flow is still described by ψ = Axy as y → ∞.

Flow over a wedge 1 ≤ n ≤ 2

The wedge angle is

2π n−1
ϑ = 2π − = 2π
n n

75
 Figure 6.9: Stagnation point flow when n = 2.

Figure 6.10: Flow over a wedge 1 ≤ n ≤ 2.

Uniform flow over and below a plate for n = 1 with walls at θ = 0, θ = π and
θ = 2π

Figure 6.11: Uniform flow over a flat plate n = 1.

6.4 Singularities

(a) Source (and Sink)

(b) Vortex

(c) 3 (or Doublet)

76
6.4.1 Source (and Sink)

The complex potential for a two-dimensional source located at z = 0 is

w(z) = m ln z (6.21)

with m > 0, this gives us

dw m
= .
dz z

Recall that in polar coordinates m ln z = m(ln r + iθ), so that the streamfunction is given
by ψ = mθ

So the streamlines (ψ is a constant) are straight lines radiating from the source

dw m dw m
= e−iθ , =
dz r dz r

dw
From u − iv = dz
we have

m m
u= cos θ, v= sin θ (6.22)
r r

6.4.2 Vortex

Consider the potential

w(z) = Γi ln z, (6.23)

with derivative

dw iΓ
=
dz z

As

w(z) = Γi(ln(reiθ )) = Γ(i ln r − θ)

This implies that the velocity potential is φ = −Γθ and ψ = Γ ln r

77
 1 ∂ψ
ur = =0
r ∂θ

and

∂ψ Γ
uθ = − =−
∂r r

The circulation around a closed curve C given by

I
~u · d~r
C

let C be a circle of radius a so on this curve dr = 0

I Z 2π
Γ
~u · d~r = uθ adθ = − a2π = −Γ2π (6.24)
C 0 a

6.4.3 Dipole (or Doublet)

Place a source and sink on the x−axis a distance d from the origin. The complex potential
is

w(z) = m ln(z + d) − m ln(z − d)

which can be rewritten as

!
d
1+ z
w(z) = m ln d
.
1− z

m −m
x

−d d

Figure 6.12: Doublet.

78
We have the Taylor expansions

x2 x3
ln(1 + x) = x − + − ...
2 3
x2 x3
ln(1 − x) = −x − − − ...
2 3

Let d → 0 and use a Taylor series expansion

   
d d
w(z) = m ln 1 + − m ln 1 −
z z
 2 3
  
d 1d 1d d 1 d2 1 d3
= m − + − ... − m − − − + ...
z 2 z2 3 z3 z 2 z2 3 z3
 
2md 1 d2
= 1+ + ... (6.25)
z 3 z2

Now we consider the dual limit of d → 0 whilst m → ∞ and call the limit of 2md µ so
we have

µ
w(z) = (6.26)
z

the direction of a dipole is source to sink, that is plus to minus. Writting

µ µ µ(x − iy)
w(z) = = = 2
z x + iy x + y2

we can extract the imaginary part which gives us

µy
ψ=−
x2 + y2

Streamlines are given by ψ constant say ψ0 .

µy
x2 + y 2 = −
ψ0

or

 2
2 µ µ2
x + y+ =
2ψ0 4ψ02

79
 µ µ
which are circles with centre at (0, ) and radius .
2ψ0 2|ψ0 |

Figure 6.13: Streamlines for a dipole.

More generally the complex potential for a dipole at z0 = x0 + iy0 which makes an angle
α with the x−axis is given by

µeiα
w(z ∗ ) = . (6.27)
z ∗ − z0∗

First note that the complex potential

µ
z∗

invokes the same flow pattern as

µ
.
z

Then substituting

x 7→ x′ = x − x0 y 7→ y ′ = y − y0

shifts the origin to (x0 , y0 ) and this would be given by the complex potential

µ
.
z ∗ − z0∗

80
Then multiplying by eiα gives

µeiα µ(x′ + iy ′ )(cos α + i sin α)

=
x′ − iy ′ x′ 2 + y ′2
(x′ cos α − y ′ sin α) + i(x′ sin α + y ′ cos α)
= µ (6.28)
x′ 2 + y ′ 2

so if we write X = x′ cos α − y ′ sin α and Y = x′ sin α + y ′ sin α

     ′
X cos α − sin α x
= . (6.29)
Y sin α cos α y′

6.5 Flow Around a Circular Cylinder

From the choice of complex potential

µ
w(z) = Uz + (6.30)
z

we obatin the complex velocity

dw µ
(z) = U − 2 (6.31)
dz z

which is analytic except at the origin. We have stagnation points at z = ±a (that is

where x = ±a) if and only if µ = Ua2 . Using this value for µ and then writting out the
complex potential we obatin

   
a2 a2
w(z) = U z + = U x + iy + (6.32)
z x + iy

and for ψ, the imaginary part of w, we obatin

   
a2 y x2 + y 2 − a2
ψ=U y− = Uy (6.33)
x2 + y 2 x2 + y 2

The streamlines are given by ψ being a constant and a dividing streamline is when ψ = 0
for x2 + y 2 = a2 which is the equation of a circle of radius a, if we consider this dividing

81
 y
ψ=0

Figure 6.14: Superposition of uniform flow and a dipole results in ψ = 0 as a solid object,
a cylinder.

streamline to be solid, we then have a cylinder. In terms of polar coordinates the cylinder
is given by z = aeiθ so that

 
dw a2
= U 1− 2
dz z
−2iθ
= U(1 − e )
= Ue−iθ (eiθ − e−iθ )
= 2iU sin θe−iθ (6.34)

therefore

2
dw
= 4U 2 sin2 θ
dz

By Bernoulli’s equation,

p 1 2
+ |~u| = k,
̺ 2

at infinity

p∞ 1 2
+ U =k
̺ 2

and at cylinder

82
 p 1 2 2
+ 4U sin θ = k,
̺ 2

Equating the last two expressions (both equal to k) we obtain

p p 1
= ∞ + U 2 (1 − 4 sin2 θ). (6.35)
ρ ρ 2

ψ=0 y

ψ=0 a x ψ=0

w(z) = Uz

Figure 6.15: Flow around a cylinder. We have a dividing streamline for ψ = 0 as a circle
of radius a about the origin. Another solution for ψ = 0 is the line y = 0. The points
denoted by black circles denote stagnation points, namely where dw/dz = 0. The x’s
denote points of minimum pressure. Cavitation occurs here first.

Note that the pressure is the same atR diametrically

R opposite points on the cylinder. There-
fore the total force on the cylinder p~ndS = 0 since all the pressures cancel.

~n = cosθ~i + sin θ~j

Notice that minimum pressure occurs when θ = ±π/2 that is when sin2 θ = 1. Therefore
from (6.35) we have

3
pmin = p∞ − ̺U 2 .
2

We require the pressure p is always greater than zero, thus

3
p∞ ≥ ̺U 2
2

When p∞ is equal to 32 ̺U 2 cavitation will occur and the equations will no longer apply.

Note that pmax occurs at θ = 0 and at π i.e. at the stagnation points, this is always the
case for steady flow.

83
6.6 Flow Around a Rotating Cylinder

We can obtain flow around a rotating cylinder if we add a vortex to the previous solution.

6.6.1 New Complex Potential

We add a vortex to the previous complex potential (6.30) to obtain

 
a2 z
w(z) = U z+ + iΓ ln (6.36)
z a

which is equivalent to

 
a2
w(z) = U z + + iΓ ln z − iΓ ln a} .
z | {z
constant

The derivative is

 
dw a2 iΓ
=U 1− 2 + . (6.37)
dz z z

Now

 
a2 y
ψ = Im(w) = U y− 2 + Γ ln r − Γ ln a (6.38)
x + y2

so that the dividing streamline ψ = 0 is given by r = a, but y = 0 is no longer a streamline:

ψ(x, y = 0) = Γ ln x − Γ ln a 6= const.

and so the flow is no longer symmetric about the x−axis.

6.6.2 Velocity Components

The radial and angular velocity components can be obtained from the formula

1 ∂ψ ∂ψ
ur = , uθ = − .
r ∂θ ∂r

84
The streamfunction (6.38) written in terms of r and θ is

 
a2
ψ(r, θ) = Ur 1 − 2 sin θ + Γ ln r − Γ ln a
r

from which we obtain velocity components

 
a2
ur = U 1 − 2 cos θ
r
 
a2 Γ
uθ = −U 1 − 2 sin θ − (6.39)
r r

6.6.3 Stagnation Points

Finding the stagnation points via

dw
=0
dz

gives the quadratic equation

iΓ
z 2 − a2 + z=0
U

for z, with solution

s
iΓ Γ2
zs = − ± − + a2 . (6.40)
2U (2U)2

Obviously if Γ = 0 (non-rotating cylinder) then the stagnation points would be at zS = ±a

as was found before. There are three cases for Γ > 0 to consider here:
Γ2 Γ
(i) 0 < 2
< a2 (0 < < a). Here the square-root in (6.40) is real and the
(2U) 2U
mod-square of z is

 
2 Γ2 2 Γ2
|zs | = − 2
+a + 2
= a2
(2U) (2U)

and hence the stagnation points lie on the cylinder’s surface. As 0 < sin β < 1 for
0 < β < π/2, there is a β in this range such that Γ = 2aU sin β. We must solve

85
 Γ
a sin β =
2U

resulting in

q
zs = −ia sin β ± a2 − a2 sin2 β
= −ia sin β ± a cos β. (6.41)

The situation is symmetric about the y−axis.

Γ
(ii) = a. It is obvious from (6.40) that there is a single stagnation point at z = −ia.
2U
Γ
(iii) > a. Here the square-root in (6.40) becomes imaginary and we get one stagnation
2U
point inside r = a,

s
iΓ Γ2
zs = − +i − a2 .
2U (2U)2

and one outside,

s
iΓ Γ2
zs = − −i − a2
2U (2U)2

(both) lying on the negative imaginary axis.

6.6.4 Force on a Cylinder

At the cylinder, z = aeiθ

dw
iΓ
= U 1 − e−2iθ + e−iθ
dz z=aeiθ a

which can be rewritten

 
dw −iθ Γ
= ie 2U sin θ +
dz a

86
 Γ=0 0 < Γ < 2aU

Γ = 2aU Γ > 2aU

Figure 6.16: Flow around a clockwise rotating cylinder for different values of Γ.

and so

2  2
dw 2 Γ
= ~u = 2U sin θ +
dz a

This splits as

4ΓU Γ2
~u2 = 4U 2 sin2 θ + sin θ + 2
a a

The first parts are the previous velocity and the other is the addad circulation, from
Bernoulli’s equation ̺p + 21 U 2 = k
p 1 2
By Bernoulli’s equation, + |~u| = k first at infinity
̺ 2

p∞ 1 2
+ U =k
̺ 2

and then at cylinder

 
p 1 2 2 4ΓU Γ2
+ 4U sin θ + sin θ + 2 = k,
̺ 2 a a

Equating the last two expressions (both are equal to k) we obtain

87
  
p p 1 1 4ΓU Γ2
= ∞ + U2 − 2
4U sin θ + 2
sin θ + 2 (6.42)
̺ ̺ 2 2 a a

which gives the pressure. Now

Z
F~ = −p~ndS
S

for a cylinder

Z 2π
F~ = − p(θ)~er adθ (6.43)
0

where F~ is the force per unit length, and

~er = cos θ~ex + sin θ~ey .

R 2π R 2π R 2π R 2π
As 0 cos θdθ = 0 sin θdθ = 0 sin3 dθ = 0 sin2 θ cos θdθ = 0 because of symmetry,
the only term that contributes to F~ is

1 4ΓU
− ̺ sinθ~er
2 a
So that

Z 2π
F~ = 2̺ΓU sin θ(cos θ~ex + sin θ~ey )dθ (6.44)
0

R 2π R 2π
The integral 0
sin θ cos θdθ = 0 by symmetry. We are left with the integral 0
sin2 θdθ,

Z 2π Z 2π Z 2π
2 1 1
sin θdθ = (1 − cos 2θ) = dθ = π.
0 2 0 2 0

Therefore F~ is the force per unit length is

F~ = 2̺ΓUπ~ey . (6.45)

Notice that the force is upward.

We are familiar with the lift generated by a curving ball (backspin and top spin). The
flow associated with a rotating cylinder two dimensional and is more easy to understand.

88
Figure 6.17: Flow around a rotating cylinder results in an upward force on the cylinder.

6.7 Other Methods for Obtaining Potentials

(a) Method of images

(b) Continuous Distributions

(c) Conformal transformations

6.7.1 Method of Images

w(z) = m ln(z − a) + m ln(z + a) (6.46)

The image of a dipole at z = a is a dipole is a dipole at z = −a with axis reflected in the

y−axis.

µeiα µei(π−α)
w = − −
z−a z+a
iα
µe µe−iα
= − + (6.47)
z−a z+a

For z = iy the complex potential w becomes

µeiα µe−iα
w= +
a − iy a + iy

which is obviously real (being the sum of a complex number and its complex conjugate)
and hence corresponds to ψ = 0.

89
6.7.2 Continuous Distributions

So far we have only discussed discrete distributions of singularities. Let pregress to an

infinite set of singularities each infinitesimally close to its neighbour, over a finite line
segment.

Take strength of source in elements δs as m(s)δs/a (total strength is m)

Z a
m(s)
w(z) = ln(z − s)ds (6.48)
0 a

and

Z a
dw m(s) 1
= u − iv = ds (6.49)
dz 0 a z−s

we can consider the case we have already seen, which is m is a constant. The intragl is
easy to perform

Z a
1
ds = [ln(z − s)]a0 = ln(z − a) − ln z,
0 z−s

so we end up with

 
dw m z−a
= ln (6.50)
dz a z

If we write z = reiθ and z − a = r1 eiθ1 ,

 
dw m r1 eiθ1 m r m
= ln = ln( 1 ) + i (θ1 − θ)
dz a reiθ a r a

Therefore u = m
a
ln( rr1 ) and v = m
a
(θ1 − θ).

The technique is used in slendre body theory.

6.7.3 Conformal Transformations

Consider the angle preserving transformation ξ = ξ(z). We can use this technique to find
the complex potential for a large variety of geometries.

90
Chapter 7

Boundary Conditions

The Navier-Stokes equation

D~u 1
= − ∇p + ν∇2~u
Dt ̺

is a second order differential equation in space. This implies that each of the 3 velocity
components must have two boundary conditions for the problem to be well-posed.

7.0.4 Flow Past a Stationary Finite Solid Body

There are two sets of boundary conditions here, one on hte body and the other at ∞. At
∞ the flowtends to the free stream of constant velocity U∞ int the x−direction, that is

~ → (U , 0, 0)
U ∞

On the body surface S we have

1. kinematic condition, that is no flow through the surface

(~u · ~n)S = 0.

2. Non-Slip condition

In this case the velocities in tangental and binormal direction (the binormal is the cross
product of the tangential and normal vector and so is orthogonal to the normal and given
tangential vector. It also tangential to the surface but orthognal to the given tangential
vector) are zero (that is the fluid ‘sticks’ to the surface - hence the non-slip condition).

91
Thus the velocity on the surface can be set identically equal to zero because the velocity
in all three orthogonal directions (i.e. normal tangential and binormal) are zero, that is

~u = 0 on S.

~ and the fluid at infinity was

Note however if the surface were moving with a velocity U
at rest we would have

~u → at ∞
~u = U~ on S. (7.1)

3. Free-surface conditions.

(a)

Let F (~x, t) = 0 be the equation of any surface moving with the fluid (for example ocean
surface), then a prticle of fluid remains on the surface. This tells us that the total
(substantial) derivative of F must be indentically zero, that is

DF ∂F
= + ~u · ∇F = 0.
Dt ∂t

For example with two-dimensional water waves we might define the free surface height by
y = ζ(x, t)

FIG HERE

The kinematic condition for the free surface

F = ζ(x, t) − y = 0

becomes

∂ζ ∂ζ
+u − v = 0.
∂t ∂x

(b)

92
Chapter 8

Similarity, Scaling, Reynold’s

Number

Let us now scale the Navier-Stokes equation with respect to the following characteristic
scales of the flow field

U = Velocity scale
L = Characteristic length
ω = Characteristic frequency. (8.1)

The Navier-Stokes equation yet again is

∂~u 1 2
+ ~u · ∇~u = − ∇p + ν∇
| {z ~u} (8.2)
∂t
| {z } ̺
| {z } viscous
Inertial pressure

(ν = µ/̺). The order (size of) the respective terms is as follows

∂~u
1. = O(ωU)
∂t
 2
2. ~u · ∇~u = O UL

νU


3. ν∇2 ~u = O L2
.

1.

The ratio of (1) to (3) and (2) to (3) as follows

93
    
ωU ωL2
O =O
νU/L2 ν

2.

   
U 2 /L UL
O =O
νU/L2 ν

2
The quantity ωLν is often referred to as the oscillatory Reynolds number (which is related
to the Strouhal number) while,

UL UL
R= =̺ (8.3)
ν µ

is the famous Reynolds number which represents the ratio of inertial forces to viscous
forces. If R >> 1 then inertial forces dominate the flow field. Conversely, if R << 1 then
viscous forces dominate the flow field.

8.1 Dynamic Similarity

We can make the Navier-Stokes equation dimensionless if the units of time, distance and
velocity are scaled in accordance with

L ′
t = t
U
~x = L~x′
~u = U~u′ (8.4)

That is, distance is measured in multiples of L, time in multiples of U/T and velocity in
multiples of U. We also have

F U2 1 1
p= = (̺L3 ) F ′ 2 ′ = ̺U 2 p′ .
A L L A

With these new variables, the derivatives get changed from ∂/∂x to (1/L)∂/∂x′ , and so
on. We have for the terms of the Navier-Stokes equation

94
 ∂~u ̺U 2 ∂~u
̺ →
∂t L ∂t
̺U 2 ′
̺~u · ∇~u → ~u · ∇′~u′
L
̺U 2 ′ ′
∇p → ∇p
L
µU ′ 2 ′
µ∇2~u = ∇ ~u (8.5)
L2

Substituting these in gives

∂~u′ ′ ′ ′ 1 ′ ′ 1 ′2 ′
+ ~
u · ∇ ~
u = − ∇ p + ∇ ~u . (8.6)
∂t′ ̺ R

Whatever the scale, flows with the same R “look” the same - in terms of appropriate
scaled x′ , y ′ , z ′ and t′ .

The basis of tests on scaled down models in a wind or water tunnel based on the concept
of dynamical similarity that is experiments should be conducted at the same Reynolds
number.

We are able to predict the values quantities to be expected on a prototype from measure-
ments on a model.

For example: for a ship the Reynolds number R is the same in a water channel as an
ocean. As ν is the same because the fluid is the same, then if L decreases, U must increase
to keep R constant.

Usefull in experiments to use the same liquid for example but to change the Reynolds
number by varying other characteristic quantities.

8.2 Stokes Flow

These equations

∇p = µ∇2~u (8.7)

are called the Stokes flow equations where viscous effects totally dominate over inertial
forces. The flow is characterised by either extreme viscosities or microscopic length scales
and velodcities.

95
Chapter 9

Solutions of the Navier-Stoke’s

Equations

(a) Unidirectional (or rectilinear) flow.

As we shall see here the nonlinear term (~u · ∇)~u of the Navier-Stoke’s equations vanishes.
This has the significant advantage that the equation is linear allowing the use of powerfull
mathematical techniques for linear analysis.

(b) Circular streamlines. Example of where the nonlinear term decouples from the equa-
tion which determines the velocity vector field.

(c) Two-dimensional stagnation point flow of viscous fluid.

Recall the continuity equation in two dimensions

∂u ∂v
+ = 0.
∂x ∂y

D

∇2 Ψ = ν∇4 Ψ (9.1)
Dt
expanding the substantial derivative

D
∂
∂
∂

∇2 Ψ = ∇2 Ψ + u ∇2 Ψ + v ∇2 Ψ
Dt ∂t ∂x ∂y
∂
∂Ψ ∂
∂Ψ ∂

= ∇2 Ψ + ∇2 Ψ − ∇2 Ψ
∂t ∂y ∂x ∂x ∂y
2
∂(∇ Ψ, Ψ)
=
∂(x, y)
= ν∇4 Ψ (9.2)

96
where the Jacobian is defined by

   
∂(∇2 Ψ, Ψ) ∂Ψ 2 ∂Ψ ∂Ψ 2 ∂Ψ
= ∇ − ∇ . (9.3)
∂(x, y) ∂y ∂x ∂x ∂y

The RHS of this is nonlinear and again is a majour source of difficulty when trying to
solve the Navier-Stokes equations.

9.1 Simplification of the Navier-Stokes Equations with

Unidirectional Flow

Unidirectional (or rectilinear) flow is when the velocity vector field is given by ~u = (u, 0, 0),
that is

~u = u(~x, t)

The continuity equation ∇ · ~u = 0 implies that

∂u
=0 (9.4)
∂x

Thus u can only be a function of y, z and t, that is,

u = u(y, z, t).

The advective nonlinear term in the Navier-Stokes equation ~u · ∇~u is then

∂u
~u · ∇~u = u =0
∂x

because (9.4).

As there are no y and z components of velocity there can be no forces on the fluid in the
the y or z direction, implying

∂p ∂p
= 0, = 0. (9.5)
∂y ∂z

Therefore the pressure p can only be a function of x and t, that is,

97
 p = p(x, t).

The Navier-Stokes equation reduces to

∂u 1 ∂p
=− + ν∇2 u (9.6)
∂y ̺ ∂x | {z }
|{z} | {z } (3)
(1) (2)

for unidirectional flows. It has the significant advantage that the equation is linear al-
lowing the use of powerfull mathematical techniques for linear analysis. If we look more
closely at (9.6), we see that parts (1) and (3) are functions of y, z and t whereas (2) is
a function of x and t. As x is an independent variable ∂p/∂x cannot be a function of x
because neither of the other two terms are. It can be at most a function of t. Therefore
the pressure term can be at most be a function of t, that is

1 ∂p
− = G(t). (9.7)
̺ ∂x

This futher simplification now allows us to write (9.6) as follows

∂u
= G(t) + ν∇2 u. (9.8)
∂t

In practice the pressure gradient is usually prescribed and the problem is to determine u,
subject to specified boundary conditions.

9.2 Steady Unidirectional Velocity fields

(i) 2-D Poiseuille flow - the flow between two flat plates a distance b apart.

(ii) Circular cylinder

(iii) Flow in an annulus

(iv) Flow due to motion of boundaries (Couette flow)

98
9.2.1 2-D Poiseuille Flow

9.2.2 Circular Cylinder

 
1 ∂ ∂u G
r =− . (9.9)
r ∂r ∂r ν

The no-slip boundary condition at the surface of the cylinder r = a provides us with one
boundary condition

u = 0 on r = a.

A “smoothness” condition at r = 0 of

∂u
= 0 on r = 0.
∂r

This says that there is no “stress jump” along the axis of symmetry.

Integrating (9.9) once

∂u G1 A
=− r+
∂r ν2 r

Using the boundary condition at r = 0 implies that A = 0. Integrating again

G 2
u(r) = − r +B (9.10)
4ν

The non-slip condition implies B = Ga2 /4, so that we finally have

G 2
u(r) = (a − r 2 ) (9.11)
4ν

9.2.3 Flow in an Annulus

Same as with (9.9) but now the non-slip conditions aplly to both cylinders, that is

u = 0 on r = a, b.

99
 G 2
u(r) = − r + A ln r + B (9.12)
4ν

Using the non-slip boundary conditions (where a is the radius of the outer cylinder and
b is the radius of the inner cylinder)

G 2
0 = − a + A ln a + B (9.13)
4ν
G
0 = − b2 + A ln b + B (9.14)
4ν

Subtracting we get

G (a2 − b2 )
A=
4ν ln a/b

Substituting this into (9.13) gives

G 2 G (a2 − b2 )
B= a − ln a
4ν 4ν ln a/b

The final solution is

 
G 2 2 (a2 − b2 ) ln r/a
u(r) = a −r + . (9.15)
4ν ln a/b

In the limit b → 0, ln a/b → −∞ and we recover the parabolic profile given in the previous
section.

9.2.4 Couette Flow

Here there is no pressure gradient so G = 0. The problem reduces to solving

∂2u
=0
∂y 2

subject to the no-slip boundary conditions

100
 u = 0 on y = 0
u = U on y = b

The solution is

Uy
u= (9.16)
b

a linear velocity profile.

The vorticity in this case is

~i ~j ~k
U
~ω = ∇ × ~u = ∂
∂x
∂
∂y
∂
∂z = − ~k. (9.17)
Uy b
b
0 0

This simple flow is exploited in the design of the Couette viscometer which is used to
measure the viscocity of a liquid.

9.3 Unsteady Unidirectional Velocity fields

time dependence of flow have two sources

(a) through the pressure gradient G(t) or

(b) through time dependence of the boundary conditions.

9.3.1 Impulsively Started Plate


0 t<0
u= (9.18)
U0 t > 0

As no pressure gradient is acting, the equations to be solved are

∂u ∂2u
=ν 2
∂t ∂y

with boundary conditions on y = 0. There is no flow at infinity so

101
 u(∞, t) = 0

while the no slip boundary condition for flow at the plate for t > 0 implies

u(0, t > 0) = U0 (9.19)

The equation is the diffusion equation which is known to have a similarity solution.

u(y, t) = U0 f (η) where η = y/(2(νt)1/2 )

It is easy

 
∂u ′ ∂η ′ 1
= U0 f (η) = U0 f (η) − η
∂t ∂t 2t

and

∂2u 1
2
= U0 f ′′ (η)
∂y 4νt

We obtain the ODE

f ′′ + 2ηf ′ = 0 (9.20)

or

2
(e−η f )′ = 0

so

2
f ′ = Ae−η

and when integrated again we obtain

f (η) = Cerf (η) + D (9.21)

where erf (η) is defined as follows

102
 Z η
2 2
erf (η) = √ e−ξ dξ. (9.22)
π 0

We now find C and D from the boundary conditions, recalling that u(y, t) = U0 f (η) and
1/2
η = y/(2(νt)
R ∞ −ξ2 . From
√ no flow at infinity u(∞, t) = 0 we get C = −D (where we have
used 0 e dξ = π/2). From u(0, t > 0) = U0 we get D = U0 . Finally we have

u(y, t) = U0 (1 − erf (η))Θ(t) (9.23)

9.3.2 General Solution for Infinite Plate

The solution can be generalised to a general velocity U(t), t > 0, of the infinite plate. We
begin with a couple of trivial cases that will instructive in devloping the general solution.

Impulsively started plate at time t1

First consier the case where we have the impulsively started plate that moves at speed
(U1 − U0 ) after t = t1 , then write

u(y, t) = (U1 − U0 )f (η) where η = y/(2(t − t1 )1/2 )

then

 
∂u ′ ∂η ′ 1
= (U1 − U0 )f (η) = (U1 − U0 )f (η) − η
∂t ∂t 2(t − t1 )

and

∂2u 1
2
= (U1 − U0 )f ′′ (η)
∂y 4ν(t − t1 )

We obtain f (η) = Cerf (η) + D where η = y/(2(t − t1 )1/2 ). No flow at infinity implies
again that C = −D. And u(0, t > t1 ) = U1 − U0 implies D = U1 − U0

103
Twice impulsively motioned plate

Second consider the case where we have the impulsively started plate that moves at speed
U0 after t = 0 and then later at speed U1 for t > t1 then the we have for the plate

u(t) = U0 Θ(t) + (U1 − U0 )Θ(t − t1 ) (9.24)

As the differential equation is linear we may add the solutions to obtain

u(y, t) = U0 (1 − erf (η 0 ))Θ(t) + (U1 − U0 )(1 − erf (η t1 ))Θ(t − t1 ) (9.25)

p
where η t1 = y/(2 ν(t − t1 ))

Let t > t1

General Solution for Infinite Plate

The motion of the plate can be written

X U(τ + dτ ) − U(τ )
u(t) = U(0)Θ(t) + dτ ( )Θ(t − τ − dτ )
dτ
Zτ =0∞
dU
= U(0)Θ(t) + dτ Θ(t − τ ) (9.26)
0 dτ

By linearity we can write for the solution

Z ∞
0 dU(τ )
u(y, t) = U(0)(1 − erf (η ))Θ(t) + Θ(t − τ )(1 − erf (η τ ))dτ
0 dτ
= U(0)(1 − erf (η 0 ))Θ(t) + [U(τ )Θ(t − τ )(1 − erf (η τ ))]∞
0
Z ∞
d
− U(τ ) [Θ(t − τ )(1 − erf (η τ ))]dτ
0 dτ
Z ∞
∂
= U(τ ) [Θ(t − τ )(1 − erf (η τ ))]dτ (9.27)
0 ∂t

where

p
η τ = y/(2 ν(t − τ )).

104
The final solution for an infinite plate.

Z ∞
∂
u(y, t) = U(τ ) [Θ(t − τ )(1 − erf (η τ ))]dτ. (9.28)
0 ∂t

9.3.3 Oscillating Plane Boundary (Stokes Boundary Layer)

Consider the sinusoidal oscillations of an infinite plate.

u(0, t) = V cos σt (9.29)

Assume the solution is of the form

u(y, t) = U1 (y) cos σt + U2 (y) sin σt. (9.30)

Substituting this into

∂u ∂2u
=ν 2
∂t ∂y

we obtain

 
d2 U1 d2 U2
−U1 (y)σ sin σt + U2 (y)σ cos σt = ν cos σt 2 + sin σt 2 . (9.31)
dy dy

Implying

d2 U1 σ d2 U2 σ
2
= U2 (y), 2
= − U1 (y) (9.32)
dy ν dy ν

Or in matrix form

    
d2 U1 (y) 0 σ/ν U1 (y)
= . (9.33)
dy 2 U2 (y) −σ/ν 0 U2 (y)

Use the trial solution

 
a
eαy (9.34)
b

105
in (9.33)

   
a 2 bν/σ
α = (9.35)
b −aν/σ

So aα2 = bν/σ = −a(ν/σ)2 /α2 . Or α4 = −(ν/σ)2 , implying α2 = ±iν/σ, implying that

       
a 1 2 a 1
= (for α = +iν/σ), = (for α2 = −iν/σ)
b i b −i

and

1+i 1−i
α=± √ , α=± √ .
2 2

The general solution is then

    
   r  r 
U1 (y) 1 σ 1 σ
= A exp (1 + i) y +B exp −(1 + i) y
U2 (y) i 2ν i 2ν
   r     r 
1 σ 1 σ
+ C exp (1 − i) y +D exp −(1 − i) y
−i 2ν −i 2ν
(9.36)

At this point we can use the boundary condition at infinity which is that there is no flow
ther to get A = C = 0, so that

     r     r 
U1 (y) 1 σ 1 σ
=B exp −(1 + i) y +D exp −(1 − i) y (9.37)
U2 (y) i 2ν −i 2ν

We get

106
      √   √  √
U1 (y) 1 σ
y 1 σ σ
= B e−i 2ν
+D ei 2ν y e− 2ν y
U2 (y) i −i
  r  r 
1 σ σ
= [B (cos y − i sin y )
i 2ν 2ν

1
 r
σ
 r
σ
 √σ
+D (cos y + i sin y )]e− 2ν y
−i 2ν 2ν
  r    r  √
B+D σ 0 σ σ
= [ cos y + sin y ]e− 2ν y
0 2ν B+D 2ν
  r    r  √
0 σ −B + D σ σ
+ i[ cos y + sin y ]e− 2ν y .
B−D 2ν 0 2ν
(9.38)

For the solution to be real we must have B = D, and then we have

    r    r   r 
U1 (y) 2B σ 0 σ σ
=[ cos y + sin y ] exp − y (9.39)
U2 (y) 0 2ν 2B 2ν 2ν

Substituting this result into (9.30) we find for u(y, t) we obtain

r  r   r 
σ σ σ
u(y, t) = [2B cos y cos σt + 2B sin y sin σt] exp − y (9.40)
2ν 2ν 2ν

Using the boundary condition that u(0, 0) = V implies 2B = V . So that

r  r   r 
σ σ σ
u(y, t) = V (cos y sin σt + sin y sin σt) exp − y (9.41)
2ν 2ν 2ν

Using cos(α + β) = cos α cos β − sin α sin β, the solution can be written in the final form

 r   r 
σ σ
u(y, t) = V cos σt − y exp − y . (9.42)
2ν 2ν

9.3.4 Oscillating Pressure Gradiant - Circular Cylinder

We define the oscillatory gradient by

107
 G(t) = G cos σt (9.43)

while requiring the non-slip condition to be satisfied on the surface of the cylinder r = a.

The equation to be solved becomes

 
∂u ∂ 2 u 1 ∂u
= G cos σt + ν + (9.44)
∂t ∂r 2 r ∂r

Assume the solution has the form

u(r, t) = U1 (r) cos σt + U2 (r) sin σt (9.45)

which leads to

 
d2 U1 (r) 1 dU1 (r)
−U1 (r)σ sin σt + U2 (r)σ cos σt = G cos σt + ν + cos σt
dr 2 r dr
 2 
d U2 (r) 1 dU2 (r)
+ν + sin σt (9.46)
dr 2 r dr

which gives the coupled differential equations

d2 U1 (r) dU (r) σ 2 G
r2 2
+r 1 − r U2 (r) = −r 2
dr dr ν ν
2
d U2 (r) dU (r) σ 2
r2 2
+r 2 + r U1 (r) = 0 (9.47)
dr dr ν

The non-slip condition satisfied at the surface of the cylinder r = a reads

u(a, t) = U1 (a) cos σt + U2 (a) sin σt = 0. (9.48)

which implies

U1 (a) = U2 (a) = 0 (9.49)

Defining Ũ2 (r) = U2 (r) − G/σ, then we can write (9.47) as

108
 d2 U1 (r) dU1 (r) σ 2
r2 + r − r Ũ2 (r) = 0
dr 2 dr ν
2
d Ũ2 (r) dŨ (r) σ 2
r2 2
+r 2 + r U1 (r) = 0 (9.50)
dr dr ν

If we write

z(r) = U1 (r) − iŨ2 (r) (9.51)

Then equations (9.50) can be rewritten

d2 z(r) dz(r) σ
r2 2
+r − i r 2 z(r) = 0 (9.52)
dr dr ν

and has the solution

r
1/2 σ
z(r) = CI0 (i r) (9.53)
ν

where I0 is the modified Bessel function of the first kind which satisfies the differential
equation

d2 y(x) dy(x)
x2 2
+ x − x2 y(x) = 0.
dx dr

From I0 (x) = J0 (ix) we have I0 (i1/2 x) = J0 (i3/2 x) (where J0 is the Bessel function of the
first kind). The real and imaginary parts of J0 (i3/2 x) are the so-called Kelvin functions,

ber0 (x) = Re J0 (i3/2 x), bei0 (x) = Im J0 (i3/2 x) (9.54)

The solution is then

r r
σ σ
z(r) = C[ber0 ( r) + ibei0 ( r)]
ν ν

or

r r
σ σ
U1 (r) − iŨ2 (r) = (CR + iCI )[ber0 ( r) + ibei0 ( r)].
ν ν

109
Written out

G
U1 (r) − i(U2 (r) − )
rσ r
σ σ
= (CR + iCI )(ber0 ( r) + i bei0 ( r))
ν ν
r r r r
σ σ σ σ
= CR ber0 ( r) − CI bei0 ( r) + i[CI ber0 ( r) + CR bei0 ( r)]
ν ν ν ν
(9.55)

The boundary conditions (9.49) imply

r r
σ σ
U1 (a) = CR ber0 ( a) − CI bei0 ( a) = 0
ν ν
r r
G σ σ G
U2 (a) − = −CI ber0 ( a) − CR bei0 ( a) = − (9.56)
σ ν ν σ

The first condition is solved by

r r
σ σ
CR = A bei0 ( a), CI = A ber0 ( a)
ν ν

inserting this into the second condition gives

 r r 
2 σ 2 σ G
A ber0 ( a) + bei0 ( a) = .
ν ν σ

Therefore we have:

pσ pσ p p
G bei0 ( ν
a)ber0 ( r) − ber0 ( σν a)bei0 ( σν r)
U1 (r) = pν p (9.57)
σ ber02 ( σν a) + bei20 ( σν a)

and

pσ p p p
G [ber0 ( ν
r)ber0 ( σν a) + bei0 ( σν r)bei0 ( σν a)] G
U2 (r) = − p p + (9.58)
σ ber02 ( σν a) + bei20 ( σν a) σ

Let us introduce

110
 r
σ
α= a (9.59)
ν

the well known Womersley parameter of blood control. If we write

ber0 (α) bei0 (α)

Br = and Bi =
ber02 (α)+ bei20 (α) ber02 (α)+ bei20 (α)

then (9.45) can be written as

G
u(r, t) = [(Bi ber0 (αr/a) − Br bei0 (αr/a)] cos σt +
σ
G
+ [1 − Br ber0 (αr/a) + Bi bei0 (αr/a)] sin σt. (9.60)
σ

This can be written as

" √ ! #
G I0 ( iαr/a) iσt
u(r, t) = Re 1− √ e (9.61)
iσ I0 ( iα)

as can be seen from writting out

√ !
G1 I ( iαr/a)
1− 0 √
σ i I0 ( iα)
 
G1 ber0 (αr/a) + ibei0 (αr/a)
= 1−
σ i ber0 (α) + ibei0 (α)
 
G1 [ber0 (αr/a) + ibei0 (αr/a)][ber0 (α) − ibei0 (α)]
= 1−
σ i ber02 (α) + bei20 (α)
G1
= (1 − [ber0 (αr/a) + ibei0 (αr/a)][Br − iBi ])
σ i
G1
= (1 − Br ber0 (αr/a) − Bi bei0 (αr/a) + i [−Br bei0 (αr/a) + Bi ber0 (αr/a)])
σ i
G G
= [Bi ber0 (αr/a) − Br bei0 (αr/a)] − i [1 − Br ber0 (αr/a) + Bi bei0 (αr/a)] .
σ σ
(9.62)

111
9.4 Circular Streamlines

Let us consider steady flow between two rotating cylinders. The inner cylinder has radius
r1 and angular velocity Ω1 whereas the outer cylinder has radius r2 and angular velocity
Ω2 .

The boundary conditions for the velocity are

v = Ω1 r1 on r = r1
v = Ω2 r2 on r = r2 . (9.63)

~u = ur~er + uφ~eφ + uz~ez (9.64)

∂ ∂
~e = −~er , ~e = ~eφ (9.65)
∂φ φ ∂φ r

The Laplace of a scalar in cylindrical polar coordinates (with no z dependence and uz = 0)

 
2 1 ∂ ∂F 1 ∂2F
∇F = r + 2 2.
r ∂r ∂r r ∂φ

The advective term becomes

∂ 1 ∂
~u · ∇~u = (uφ~eφ ) · (~r + ~eφ )(u ~e )
∂r r ∂φ φ φ
v(r) ∂
= (v(r)~eφ )
r ∂φ
v2
= − ~er (9.66)
r

We inspect the effect of the Laplacian on the vector v~eφ employing (9.65) and using
v = v(r),

      
1 ∂ ∂ 1 ∂2 ~ = ~e 1 ∂ ∂ 1 ∂
r + 2 2 (v φ) φ r v+ (−v~er )
r ∂r ∂r r ∂φ r ∂r ∂r r 2 ∂φ
  
1 ∂ ∂ v
= ~eφ r v− ~e
r ∂r ∂r r2 φ
 
2 1
= ~eφ ∇ − 2 v (9.67)
r

112
(this illustrates the fact that in non-Cartesian coordinates the vector Laplacian resolved
in these non-Cartesian coordinates does not coincide with the Laplacian of a scalar in the
non-Cartesian coordinates).

The Navier-Stokes equation

1 dp
~u · ∇~u = − + ν∇2~u
̺ dr

becomes

 
v2 1 dp 2 1
− ~er = − ~e + ν~eφ ∇ − 2 v
r ̺ dr r r

and decouple to yield

1 dp v2
= (9.68)
̺ dr r

and

 
2 1
∇ − 2 v(r). (9.69)
r

Rewriting (9.69)

 2

2 d d
r + r − 1 v(r) = 0
dr 2 dr

and trying the trial solution r λ gives

(λ(λ − 1) + λ − 1)r λ = 0

or (λ − 1)(λ + 1) = 0. The general solution is then

B
v(r) = Ar + (9.70)
r

The boundary conditions (9.63) read

B B
Ω1 r1 = Ar1 + and Ω2 r2 = Ar2 +
r1 r2

113
implying Ω2 r22 − Ω1 r12 = A(r22 − r12 ) and Ω1 r1 (r1 r22 ) − Ω2 r2 (r2 r12 ) = B(r22 − r12 ), so that

Ω2 r22 − Ω1 r12
A =
r22 − r12
(Ω − Ω )r 2 r 2
B = − 2 2 1 2 1 2. (9.71)
r2 − r1

Special cases:

(i) B = 0 is obtained if Ω2 = Ω1 = Ω. The solution is then v = Ωr and this corresponds

to ridgid body rotation.

(ii) A = 0 occurs when Ω2 = 0 and r2 → ∞. In this case v = Ω1 r12 /r which corresponds

to the motion due to an irrotational vortex: the circulation around a circle centered at
the origin of radius r ≥ r1 is

Z 2π
Ω1 r12
Γ= (rdθ) = 2πΩ1 r12 (9.72)
0 r

so no-zero, yet the vorticity vanises everywhere where in the fluid,

~er r~eφ ~ez

1 ∂ ∂ ∂
~ω = ∇ × ~u = ∂r ∂φ ∂z = 0. (9.73)
r
0 Ω1 r12 0

Consider the situation when r1 → 0 while Ω1 r12 = Const. and invoke Stokes theorem

I Z
~u · d~x = ~ω · ~ndS,
C S

this implies a delta function of vorticity at the origin.

9.5 Two Dimensional Stagnation Point flow

The streamfunction was

ψ = Axy

The velocities were

114
 ∂ψ
u= = Ax
∂y

∂ψ
v=− = −Ay
∂x

This will not be a solution for viscous flow because u 6= 0 on the boundary y = 0.

We require the far-field to look like the above solution whereas on the plane y = 0,
u = v = 0. In summary

u = v = 0 on y = 0
ψ = Axy as y → ∞. (9.74)

The steady, two-dimensional Navier-Stokes equations become

∂(∇2 ψ, ψ)
= ν∇4 ψ (9.75)
∂(x, y)

Let us consider a solution of the form

ψ = xF (y) (9.76)

The velocities are then

∂ψ
u = = xF ′ (y)
∂y
∂ψ
v = − = F (y). (9.77)
∂x

The assumption (9.76) leads to

∇2 ψ = xF ′′ ∇4 ψ = xF ′′′′
∂(∇2 ψ, ψ)
= x(F ′′ F ′ − F ′′′ F ). (9.78)
∂(x, y)

where we used (9.3). The equation reduces to

115
 F ′′ F ′ − F ′′′ F = νF ′′′′

This can be written

((F ′ )2 − F F ′′ )′ = νF ′′′′

upon integrating once gives

(F ′ )2 − F F ′′ = νF ′′′ + c

where c is the constant of integration. It is clear that as y → ∞ that F → Ay, F ′ → A

etc so that c = A2 . Then

(F ′ )2 − F F ′′ = νF ′′′ + A2

This equation has two parameters ν and A. If we apply the following scaling

r
A √
η= y, F (y) = AνG(η)
ν

the equation reduces to

(G′ )2 − GG′′ = G′′′ + 1

with boundary conditions u = v = 0 at y = 0 which imply

G(0) = G′ (0) = 0

and boundary condition ψ = Axy as y → ∞ which implies

r √
dF A d AνG)
(∞) = A or (∞) = A
dy ν dη

or

G′ (∞) = 1.

This may be solved numerically. The solution of the equation is now of the form

116
 r !
√ A
ψ= AνxG y
ν

117
Chapter 10

Mathematical Theorems

The theorems derived here pertain to the approximations to the Navier-Stokes equation
given by

∇p = µ∇2~u

and the continuity equationd

∇ · ~u = 0.

These are the Stokes flow equations for slow viscous flow of an incompressible fluid.

Recall the rate of strain tensor

!
1 ∂ui ∂uj
eij = +
2 ∂xj ∂xi

and that the incompressibilty condition ∇ · ~u = 0 is equivalent to ekk = 0. Also that the
constituative relation then for incompressible flow is

σij = −pδij + 2µeij

It is easy to see that an alternative form of Stokes flow equations are

∂σij
=0 (10.1)
∂xj

118
using the continuity equation in component form

∂uj
= 0.
∂xj

10.1 Uniqueness Theorem

T heorem Let (~u, p) satisfy

∇p = µ∇2~u ∇ · ~u = 0 (10.2)

in a volume V with boundary conditions

~ (x) on S
~u = U

(x ∈ S, S may consist of interior as well as exterior boundaries). Let (~u′ , p′ ) also satisfy
(10.2); then

~u ≡ ~u′ in V. (10.3)

P roof

Let

~u˜ = ~u − ~u′
ẽij = eij − e′ij
σ̃ij = σij − σij′ (10.4)

Consider the following integral,

Z Z !
1 ∂ ũi ∂ ũj
ẽij σ̃ij dV = + σ̃ij dV
V V 2 ∂xj ∂xi
Z
∂ ũi
= σ̃ dV (10.5)
V ∂xj ij

119
The last line on the RHS is obtained because of the symmetry of σ̃ij namely σ̃ij = σ̃ji .
Using the alternative form of Stokes flow equation, (10.1), we can write (10.5) as

Z
∂
(ũ σ̃ )dV
V ∂xj i ij

which becomes

Z
ũiσ̃ij nj dS = 0
S

after employing the divergence theorem and noting that

~u˜ = ~u − ~u′ = 0

on S, because both the velocity fields satisfy the boundary conditions. We have thus
shown that (10.5) vanishes.

Now because

σ̃ij = −p̃δij + 2µẽij

and

ẽij δij = ẽkk = 0

it is easy to show that the LHS of (10.5) is equivalent to

Z
2µ ẽij ẽij dV
V

which we have shown is zero. Clearly the integrand must be zero, that is

ẽij = 0

everywhere in V which tells us that

~u˜ := ~u − ~u′

120
induces at most a uniform translation or rotation. However since ~u = ~u′ on S it implies
that

~u = ~u′ = 0 in V.

This proves the theorem.

10.2 Minimum Dissipation Theorem

10.2.1 Energy Balance and Dissipation of Energy

Z
1
Ekin = ̺~u · ~udV
2 V

The rate of change of Kinetic Energy is

Z
∂Ekin 1 ∂(̺~u · ~u)
= (10.6)
∂t V 2 ∂t

In the absense of any forces an kinetic Energy change within the volume V can occur by
Energy flowing across the boundary,

Z
∂Ekin 1
=− ( ̺~u · ~u)~u · ~ndS (10.7)
∂t S 2

In the pressence of viscous and body forces the rate of change of energy is

Z Z Z
∂Ekin 1
=− ( ̺~u · ~u)~u · ~ndS = ui σij nj dS + ̺ui fi dV (10.8)
∂t S 2 S V

Using the divergence theorem

121
Z   Z " #
∂E ∂
+ ∇ · (E ~u) dV = (u σ ) + ̺ui fi dV
V ∂t V ∂xj i ij
Z " #
∂ui ∂σij
= σ + ui + ̺ui fi dV
V ∂xj ij ∂xj
Z " #
∂σij
= eij σij + ui + ̺ui fi dV
V ∂xj
Z " #
∂
= eij (−pδij + 2µeij ) + ui (−pδij + 2µeij ) + ̺ui fi dV
V ∂xj
Z h i
= 2µeij eij − ~u · ∇p + µui ∇2 ui + ̺ui fi dV.
V
(10.9)

The LHS can be rewritten

1∂ 1 ∂ 1 1 ∂̺ 1 1
(̺u2 ) + ∇ · (̺u2~u) = ̺ ( u2 ) + u2 + ̺~u · ∇( u2 ) + u2 ∇ · (̺~u)
2 ∂t 2 ∂t
 2 2 ∂t  2  2 
∂ 1 2 1 2 1 2 ∂̺
= ̺ ( u ) + ~u · ∇( u ) + u + ∇ · (̺~u)
∂t 2 2 2 ∂t
 
∂ 1 2 1 2
= ̺ ( u ) + ~u · ∇( u )
∂t 2 2
D 1 2
= ̺ ( u ). (10.10)
Dt 2

As we are assuming an incompressible fluid we can bring the density ̺ inside the sub-
stantial derivative. We obtain an Energy balance equation, Energy is not conserved and
there is a viscous dissipation of Energy on the RHS

D 1 2
( ̺u ) + ~u · ∇p − ̺~u · f~ − µ~u · ∇2~u = 2µeij eij . (10.11)
Dt 2

10.2.2 Proof of Minimum Dissipation Theorem

Viscous dissipation is given

Z
2µ eij eij dV. (10.12)
V

122
T heorem Let (~u, p) be the unique flow satisfying

∇p = µ∇2~u, ∇ · ~u = 0
~
~u = U on S (10.13)

and let ~u′ be any other vector field satisfying

∇ · ~u′ = 0 and ~
~u′ = U on S (10.14)

then the rate of dissipation of energy is least in the flow satisfying (10.13).

P roof

Let

~u˜ = ~u − ~u′
ẽij = eij − e′ij (10.15)

where the ~u′ is the one defined in this theorem and e′ij corresponds to this ~u′ . We do a
caluculation similar to one done in the previous theorem,

Z Z
∂ ũi
ẽij σij dV = σij dV
V V ∂xj
Z
∂
= (ũi σij )dV
V ∂xj
Z
= ũi σij nj dS = 0. (10.16)
S

As ~u − ~u′ = 0 on S. Substituiting σij = −pδij + 2µeij into the first integral implies

Z
2µ ẽij eij dV = 0. (10.17)

Z Z
2µ e′ij e′ij dV = 2µ (eij − ẽij )2 dV
V
ZV Z Z
= 2µ eij eij dV + 2µ ẽij ẽij dV − 4µ eij ẽij dV (10.18)
V V V

123
Since the last term is zero by (10.17) and ẽij ẽij ≥ 0 we must have

Z Z
2µ e′ij e′ij dV ≥ 2µ eij eij dV (10.19)
V V

which proves the theorem. That is the Stokes flow is the one of minimum dissipation for
given boundary conditions.

10.2.3 Vorticity Minimum Dissipation Theorem

Assume ~u is an irrotaional velocity field field described by a velocity potential, φ, and ~u′
is an arbitrary solenoidal rotational velocity field. The velocity field corresponding to an
irrotational flow has a least amount of kinetic energy. We impose the boundary conitions
on these velocity fields

~u · ~n = ~u′ · ~n. (10.20)

Assuming, the same, constant density ̺ the continuity equation for both these velocity
fields are ∇ · ~u = ∇ · ~u′ = 0. The difference in kinetic energies of the two flows, ∆Ekin =
Ekin (~u′) − Ekin (~u), is

Z
1
∆Ekin = ̺ (~u′ · ~u′ − ~u · ~u)dV
2 V
Z Z
1
= ̺ (~u − ~u) · (~u − ~u) + ̺ (~u′ − ~u) · ~udV.
′ ′
(10.21)
2 V V

Using ~u = ∇φ and the divergence theorem on the last term on the RHS gives

Z Z
′
(~u − ~u) · ∇φdV = ∇ · [(~u′ − ~u)φ]dV
V ZV
= φ(~u′ − ~u) · ~ndS, (10.22)
S

which is zero by the boundary condition (10.20). This implies that the RHS of (10.21)
is positive and therefore the kinetic energy of rotational flow with velocity ~u′ is greater
than the kinetic energy of the corresponding irrotational flow with velocity ~u.

124
Chapter 11

Stoke’s Streamfunction

Three-dimensional axisymmetric flows.

∇ · ~u = 0

implies

~
~u = ∇ × A (11.1)

11.1 Streamfunction in Cylindrical and Spherical Po-

lar Coordinates

11.1.1 Streamfunction in Cylindrical Polar Coordinates

The velocity field is independent of φ and only have

~u = uρ~eρ + uz~ez

~ = Ψ(ρ, z) ~e
A (11.2)
ϕ
ρ

125
 ~u = ∇ × (Aρ~eρ + Aϕ~eϕ + Az ~ez )
~eρ ρ~eϕ ~ez
1 ∂ ∂ ∂
= ∂r ∂ϕ ∂z
ρ
Aρ ρAϕ Az
~eρ ρ~eϕ ~ez
1 ∂ ∂ ∂
= ∂ρ ∂ϕ ∂z
ρ
0 Ψ(ρ, z) 0
1 ∂Ψ 1 ∂Ψ
= − ~eρ + ~e (11.3)
ρ ∂z ρ ∂ρ z

Reading off the velocity components,

1 ∂Ψ 1 ∂Ψ
ur = − , uz = (11.4)
ρ ∂z ρ ∂ρ

We confirm that Ψ is indeed constant along streamlines, which are defined by

d~x
= λ~u
ds

Streamlines given by dρ = λuρds and dz = λuz ds, so

∂Ψ ∂Ψ
dΨ = dρ + dz
∂r ∂z
= ρuz dρ − ρuρ dz
= ρuz λuρ ds − ρuρ λuz ds = 0. (11.5)

Therefore Ψ = Const. on streamlines (actually stream surfaces in 3D flow).

By using the formula for the curl in cylindrical polar coordinates (an example of an
orthogonal curvilinear coordinate system)

~ = ∇(∇ · A)
∇ × ~u = ∇ × (∇ × A) ~ − ∇2 A.
~

~ vanishes under certain circumstances e.g., when either

It turns out that the term ∇(∇ · A)
∇ × ~u is bounded in space or vanishes more rapidily than 1/ρ for large ρ. Then

~
∇ × ~u = −∇2 A

126
 ∇ × ~u = ∇ × (uρ~eρ + uθ~eθ + uϕ~eϕ )
~eρ ρ~eϕ ~ez
1 ∂ ∂ ∂
= ∂ρ ∂ϕ ∂z
ρ
uρ ρuϕ uz
~eρ ρ~eϕ ~ez
1 ∂ ∂ ∂
= ∂ρ ∂ϕ ∂z
ρ
− ρ1 ∂Ψ
∂z
0 ρ1 ∂Ψ ∂ρ
   
∂ 1 ∂Ψ 1 ∂2Ψ
= − − ~eϕ
∂ρ ρ ∂ρ ρ ∂z 2
 
1 ∂ 2 Ψ 1 ∂Ψ ∂ 2 Ψ
= − − + ~eϕ (11.6)
ρ ∂ρ2 ρ ∂ρ ∂z 2

We define the operator D via

1
∇ × ~u = − D 2 Ψ~eϕ . (11.7)
ρ

hence

∂2 1 ∂ ∂2
D2 = − + (11.8)
∂ρ2 ρ ∂ρ ∂z 2

That ∇p = −µ∇2~u and from the identity ∇×(∇p) = 0, we have that ∇×(∇×(∇×~u)) = 0
but

∇ × (∇ × (∇ × ~u)) = ∇(∇ · (∇ × ~u)) − ∇2 (∇ × ~u)

= −∇2 (∇ × ~u) (11.9)

where we have used the identity ∇ · (∇ × ~u) = 0. We have already deomstrated that

~ = ∇ × (∇ × Ψ (−D 2 Ψ)
∇ × (∇ × A) ~eϕ ) = ~eϕ .
ρ ρ

Following the same steps as before we find

(−D 2 Ψ) (−D 4 Ψ)
∇ × (∇ × (∇ × ~u)) = ∇ × (∇ × ~eϕ ) = ~eϕ .
ρ ρ

127
The streamfucntion equation is then

D 4 Ψ = 0. (11.10)

Pressure

From ∇ × (∇ × ~u) = ∇(∇ · ~u) − ∇2~u and using ∇ · ~u = 0 we can rewrite ∇p = −µ∇2~u as

∇p = −µ∇ × (∇ × ~u) (11.11)

The usual formula for the grad operator in cylindrical polar coordinate basis (here no φ
component)

∂p ∂p
∇p = ~eρ + ~ez (11.12)
∂ρ ∂z

and

1
∇ × (∇ × ~u) = ∇ × (− D 2 Ψ~eϕ )
r
~eρ ρ~eϕ ~ez
1 ∂ ∂ ∂
=
ρ ∂ρ ∂ϕ ∂z
0 −D 2 Ψ 0
 
1 ∂ 2 ∂ 2
= − (−D Ψ)~eρ + (−D Ψ)~ez (11.13)
ρ ∂z ∂ρ

Using (11.11), (11.12) and (11.13) we obtain equations for the pressure in terms of the
stream function,

∂p ∂ −D 2 Ψ
=µ ( ) (11.14)
∂ρ ∂z ρ

and

∂p 1 ∂ −D 2 Ψ
= −µ ρ( ). (11.15)
∂z ρ ∂ρ ρ

128
11.1.2 Streamfunction in Spherical Polar Coordinates

The velocity field is independent of φ and only have

~u = ur~er + uθ~eθ

~ must be int the ϕ

implying that A ~ -direction:

~ = Ψ(r, θ) ~e
A (11.16)
r sin θ ϕ

~ we find
Computing ∇ × A

~u = ∇ × (Ar~er + Aθ~eθ + Aϕ~eϕ )

~er r~eθ r sin θ~eϕ
1 ∂ ∂ ∂
= 2 ∂r ∂θ ∂ϕ
r sin θ
Ar rAθ r sin θAϕ
~er r~eθ r sin θ~eϕ
1 ∂ ∂ ∂
= 2 ∂r ∂θ ∂ϕ
r sin θ
0 0 Ψ(r, θ)
1 ∂Ψ 1 ∂Ψ
= 2 ~er − ~e . (11.17)
r sin θ ∂θ r sin θ ∂r θ

Reading off the velocity components,

1 ∂Ψ 1 ∂Ψ
ur = , uθ = − (11.18)
r2 sin θ ∂θ r sin θ ∂r

We confirm that Ψ is indeed constant along streamlines, which are defined by

d~x
= λ~u
ds

Streamlines given by dr = λur ds and rdθ = λuθ ds, so

∂Ψ ∂Ψ
dΨ = dr + dθ
∂r ∂θ
= −r sin θuθ dr + r 2 sin θur dθ
= −r sin θuθ λur ds + r sin θur λuθ ds = 0. (11.19)

129
Therefore Ψ = Const. on streamlines (again stream surfaces as it is 3D flow).

By using the formula for the curl in spherical polar coordinates

∇ × ~u = ∇ × (ur~er + uθ~eθ + uϕ~eϕ )

~er r~eθ r sin θ~eϕ
1 ∂ ∂ ∂
= 2 ∂r ∂θ ∂ϕ
r sin θ
ur ruθ r sin θuϕ
~er r~eθ r sin θ~eϕ
1 ∂ ∂ ∂
= 2 ∂r ∂θ ∂ϕ
r sin θ 1 ∂Ψ
r 2 sin θ ∂θ
− sin1 θ ∂Ψ 0
 2
 ∂r 
1 1 ∂ Ψ 1 ∂ 1 ∂Ψ
= − − 2 ~eϕ
r sin θ ∂r 2 r ∂θ sin θ ∂θ
 2  
1 ∂ Ψ sin θ ∂ 1 ∂Ψ
= − + 2 ~eϕ (11.20)
r sin θ ∂r 2 r ∂θ sin θ ∂θ

We define an operator D via

1
∇ × ~u = − D 2 Ψ~eϕ . (11.21)
r sin θ

hence

 
∂2
2 sin θ ∂ 1 ∂
D = 2+ 2 (11.22)
∂r r ∂θ sin θ ∂θ

The streamfucntion equation is

D 4 Ψ = 0. (11.23)

Pressure

Again from ∇×(∇×~u) = ∇(∇·~u)−∇2~u and using ∇·~u = 0 we can rewrite ∇p = −µ∇2~u
as

∇p = −µ∇ × (∇ × ~u) (11.24)

The usual formula for the grad operator in spherical polar coordinate basis (here no ϕ
component)

130
 ∂p 1 ∂p
∇p = ~er + ~e (11.25)
∂r r ∂θ θ

and

1
∇ × (∇ × ~u) = ∇ × (− D 2 Ψ~eϕ )
r sin2 θ
~er r~eθ r sin θ~eϕ
1 ∂ ∂ ∂
= ∂r ∂θ ∂ϕ
r 2 sin θ
0 0 − sin1 θ D 2 Ψ
 
1 ∂ 1 2 ∂ 1 2
= 2 − D Ψ~er + r D Ψ~eθ (11.26)
r sin θ ∂θ sin θ ∂r sin θ

Using (11.24), (11.25) and (11.26) we obtain equations for the pressure in terms of the
stream function,

∂p 1 ∂ −D 2 Ψ
= −µ ( ) sin2 θ (11.27)
∂r r sin θ ∂θ r sin2 θ

and

1 ∂p 1 ∂ −D 2 Ψ
=µ r( ) sin θ. (11.28)
r ∂θ r ∂r r sin2 θ

11.2 Flow Around a Corner

Suppose the rigid boundary OA is scraped along the plane OB at a constant angle α with
velocity V Relative to O the flow is steady.

Near the origin the equation of motion will be

   2
4 1 ∂ ∂ 1 ∂2
∇ ψ= r + 2 2 ψ=0
r ∂r ∂r r ∂θ

The boundary conditions on the rigid boundaries are ψ = c, a constant on θ = 0, α,

1 ∂ψ
ur = = −V on θ = 0 (11.29)
r ∂θ

and

131
 1 ∂ψ
ur = = 0 on θ = α. (11.30)
r ∂θ

From the first of the ur boundary conditions it is clear that the solution is of the form

ψ = rV f (θ) (11.31)

where

f (0) = f (α) = 0 (11.32)

and the boundary conditions (11.29) and (11.30) become

f ′ (0) = −1 and f ′ (α) = 0. (11.33)

Now we have


  
2 ∂
1 ∂ 1 ∂2
∇ ψ = r + 2 2 rV f (θ)
r ∂r
∂r r ∂θ
V V F (θ)
= (f + f ′′ ) =: (11.34)
r r

and

   
4 1 ∂ ∂ 1 ∂2 V F (θ)
∇ψ = r + 2 2
r ∂r ∂r r ∂θ r
V
= 3
(F + F ′′ ) = 0. (11.35)
r

The solution to (11.35) is

F = Ã cos θ + B̃ sin θ

which substituted into (11.34) gives the differential equation

f + f ′′ = Ã cos θ + B̃ sin θ

132
The remainder of the solution is found from considering f + f ′′ = 0, this as we already
know is of the form A cos θ + B sin θ. This leads to the general solution

f (θ) = A cos θ + B sin θ + Cθ cos θ + Dθ sin θ. (11.36)

Substitution of the boundary conditions (11.32) into this gives

A = 0, and B sin α + Cα cos α + Dα sin α = 0.

Substitution of the boundary conditions (11.33) gives

B + C = −1, and B cos α + C cos α + D sin α − Cα sin α + Dα cos α = 0.

Eliminating B using B = −C − 1 from the equations above we obtain,

(−α sin α)C + (α cos α + sin α)D = cos α

(α cos α − sin α)C + (α sin α)D = sin α. (11.37)

Multiplying the first by α sin α and the second by α cos α + sin α gives

(−α2 sin2 α)C + (α cos α + sin α)(α sin α)D = α cos α sin α
(α2 cos2 α − sin2 α)C + (α sin α)(α cos α + sin α)D = (α cos α sin α + sin2 α)

subtracting the first from the second and simplifying gives

− sin2 α
C= .
sin2 α − α2

Substituting this into the second equation of (11.37) and dividing both sides by − sin α
gives

(α cos α sin α − sin2 α)

− D = −1.
sin2 α − α2

From which we easily get the expression for D. We easily find the expression for B by
substituting the expression for C into B = −C − 1. Altogather we have,

133
 α2 − sin2 α cos α sin α − α
B= C= D=
sin2 α − α2 sin2 α − α2 sin2 α − α2

Substituting these into (11.36) gives

(α2 sin θ + [− sin2 α]θ cos θ + [cos α sin α − α]θ sin θ)

f (θ) = .
sin2 α − α2

Simplifying we obtain

(θ sin(θ − α) sin α − α(θ − α) sin θ)

f (θ) = .
sin2 α − α2

Substituting this into (11.31) gives

(θ sin(θ − α) sin α − α(θ − α) sin θ)

ψ = rV (11.38)
sin2 α − α2

11.3 Drag on Sphere

Let us consider Stokes flow around a sphere

Figure 11.1: Streamlines for Stokes flow around a sphere.

Here we present the classic solution of the Stokes equations representing uniform motion
of a sphere of radius a in an infinitge expanse of fluid.

The force on a sphere is

Z
Fi = σij nj dS (11.39)
S r=a

134
 The net force FD on a sphere is in the z−direction and given by F~ · ~z . Using have
~er · ~ez = cos θ and ~eθ · ~ez = − sin θ we have

Z 2π Z π
2
FD = a [σrr cos θ − σrθ sin θ] sin θdθ (11.40)
0 0

11.3.1 Boundary conditions

At infinity

ur = U cos θ, uθ = −U sin θ (11.41)

Non-slip conditions on the shpere

ur = 0, uθ = 0 (11.42)

We seek a separable solution of the form

Ψ = f (r) sin2 θ. (11.43)

11.3.2 Calculating the Streamfunction

Using separation of variables and from the behaviour at infinity we try the solution of the
form Ψ(r, θ) = f (r) sin2 θ. Calculating D 2 Ψ first,

  
2 2 ∂2 sin θ ∂ 1 ∂
D [f (r) sin θ] = + 2 f (r) sin2 θ
∂r 2 r ∂θ sin θ ∂θ
 
2 ∂2 f (r) ∂ 1 ∂
= sin θ 2 f (r) + 2 sin θ sin2 θ
∂r r ∂θ sin θ ∂θ
∂2 f (r)
= sin2 θ 2 f (r) − sin2 θ 2
 2 ∂r  r
d 2
= 2
− 2 f (r) sin2 θ. (11.44)
dr r

It is obvious then that

135
  2
4 2 2 d2 2
D [f (r) sin θ] = sin θ − f (r) (11.45)
dr 2 r 2

The equation for f (r) is then

 2
d2 2
2
− 2 f (r) = 0. (11.46)
dr r

Trying f = r λ we get (λ2 − 1)(λ − 2)(λ − 4) = 0 and therefore

A
f (r) = + Br + Cr 2 + Dr 4 . (11.47)
r

We can express the velocity components in terms of Ψ via (11.4). Using the above
boundary conditions on the velocity components allows us to determine the constants
A, B, C and D.

1 ∂Ψ
ur =
r2
sin θ∂θ 
2 cos θ A 2 4
= + Br + Cr + Dr
r2 r
 
2A 2B 2
= cos θ + + 2C + 2Dr (11.48)
r3 r

1 ∂Ψ
uθ = −
r sin θ∂r 
A B 2
= − sin θ − 3 + + 2C + 4Dr (11.49)
r r

D = 0 otherwise flow at infinity diverges. From (11.41) we have 2C = U. Non-slip

conditions at r = a give

 
2A 2B
ur |r=a = cos θ + +U =0
a3 a
 
A B
uθ |r=a = − sin θ − 3 + + U = 0. (11.50)
a a

136
Solving for A and B gives

1 3
A = Ua3 , B = − Ua.
4 4

So the streamline functions is

1 1 a3 3
Ψ = U( r 2 + − ar) sin2 θ (11.51)
2 4 r 4

Velocities

The velocities are then

 
1  a 3 3  a 
ur = U cos θ 1 + − (11.52)
2 r 2 r

and

 
1  a 3 3  a 
uθ = −U sin θ 1 − − . (11.53)
4 r 4 r

Rate of strain tensor

Expressions for the rate of strain tensor in spherical polar coordinates are given in the
appendix. We will need the components ǫrr and ǫrθ .

∂ur
ǫrr = (11.54)
∂r

 
∂ 1  a 3 3  a 
ǫrr = U cos θ 1+ −
∂r 2 r 2 r
 3

U −3a a
= cos θ 4
+3 2 (11.55)
2 r r

r ∂  uθ  1 ∂ur
ǫrθ = + (11.56)
2 ∂r r 2r ∂θ

137
  
r ∂ 1 1 a3 3 a 1
ǫrθ = −U sin θ − 4
− 2 − U sin θ ur
2 ∂r r 4 r 4r 2r
 3
  3

U sin θ 1 a 3a 1 1a 3a
= − − + 4+ 2 + + −
2 r r 2r r 2 r4 2 r2
3U sin θ a3
= − (11.57)
4 r4

11.3.3 Pressure

Differential equations for the pressure p come from (11.27) and (11.28). We need to
2
calculate rDsin[Ψ]
2 θ from (11.51). Using (11.44) we can write

 
D 2 [Ψ] 1 d2 2
2 = − f (r)
r sin θ r dr 2 r 2
 2 
U d 2 1 2 1 a3 3
= − ( r + − ar)
r dr 2 r 2 2 4 r 4
 3 3

U 1a 1a 3a
= (1 + ) − (1 + − )
r 2 r3 2 r3 2 r
3a
= U 2 (11.58)
2r

From (11.27) we can write

∂p 1 ∂ 3a
= −µ (−U 2 ) sin2 θ
∂r r sin θ ∂θ 2r
1
= 3µUa 3 cos θ. (11.59)
r

This implies

3 cos θ
p = − µUa 2 + g(θ) (11.60)
2 r

where g(θ) is an arbitray function of θ. From (11.28) we can write

1 ∂p 1 ∂ 3a
= µ r(−U 2 ) sin θ.
r ∂θ r ∂r 2r
3 1
= µUa 3 sin θ. (11.61)
2 r
138
This implies

3 cos θ
p = − µUa 2 + h(r) (11.62)
2 r

where h(r) is an arbitray function of r. Comparing (11.60) and (11.62) noting that we
must have that g(θ) = h(r) = Const. = p∞ we obtain the final result

3 cos θ
p = − µUa 2 + p∞ (11.63)
2 r

11.3.4 Stress tensor

The stress tensor is related to the strain tensor.

σij = −pδij + 2µeij (11.64)

σrr = −p + 2µǫrr (11.65)

σrθ = 2µǫrθ (11.66)

 
−3a3 a
σrr = −p + µU cos θ 4
+3 2 (11.67)
r r

3 a3
σrθ = −µU sin θ (11.68)
2 r4

11.3.5 Force on the sphere

The force on a sphere is

Z 2π Z π
2
FD = a [σrr cos θ − σrθ sin θ] sin θdθ (11.69)
0 0

Z π   
2 3 cos θ −3a3 a
FD = 2πa µUa 2 − p∞ + µU cos θ +3 2 cos θ
0 2 r r4 r r=a
  
3 a3
+ µU sin θ 4 sin θ sin θdθ (11.70)
2 r r=a

139
The p∞ term vanishes because of the integral

Z π Z  −1
−1
u2
cos θ sin θdθ = − udu = = 0.
0 1 2 1

where we used substitution u = cos θ. The drag force is then given by

Z π Z π
2
FD = 3πµaU cos θ sin θdθ + 3πµaU sin3 θdθ (11.71)
0 0
| {z } | {z }
pressure viscous

The integrals are easy

Z π Z  −1
2
−1
2 u3 2
cos θ sin θdθ = − u du = =
0 1 3 1 3

Z π Z π
3 2 4
sin θdθ = sin θ(1 − cos2 θ)dθ = [− cos θ]π0 − = .
0 0 3 3

The final result is

FD = 2πµaU + 4πµaU = 6πµaU. (11.72)

| {z } | {z }
pressure viscous

That is, one third of the drag is due to pressure forces, two-thirds are due to viscous
forces.

11.3.6 Calculating the force on the sphere by doing an integral

at infinity
Z
Fi = σij nj dS (11.73)
S r=a

!
∂ui ∂uj
σij = −pδij + µ +
∂xj ∂xi

Taking the derivative

140
 !
∂σij ∂ ∂ 2 ui ∂ ∂uj
=− p+µ +
∂xj ∂xi ∂x2j ∂xi ∂xj

∂uj
From the continuity equation ∂xj
= 0, so

∂σij ∂ ∂2u
=− p + µ 2i
∂xj ∂xi ∂xj

But as ∇p = µ∇2~u, we have

∂σij
= 0.
∂xj

By Gauss theorem

Z Z Z
∂σij
σij nj dS + σij nj dS = dV = 0.
S r=a S r=∞ V ∂xj

Implying that

Z
Fi = − σij nj dS. (11.74)
S r=∞

141
Appendix A

Glossary

• Body force: Force acting on an infintesimal volume of fluid per unit mass.

• Boundary layer flow: There will be a layer close to the solid wall where the viscous
terms are important even for very high Reynolds number flow. The non-slip condition
holds at the surface while a height above the fluid motion is uneffected. This layer will
be very thin and the flow in that region is called the boundary layer flow.
∂ui ∂σij
• Cauchy’s equation: ̺ + (~u · ∇)ui = ̺fi + where σij is the stress tensor.
∂t ∂xj

• Cavitation: Cavitation is when a liquid vaporizes because the pressure is sufficiently

low. It occurs whenever the cavitation number σ, defined by

p∞ − pv
σ= 1
2
̺V 2

is less than the critical cavitation number σcrit , which depends on the geometry and the
Reynolds number.
∂̺
• Continuity equation: The continuity equation is + ∇ · (̺~u) = 0 where ̺ is the density
∂t
and ~u is the fluid velocity. The continuity equation expressing the conservation of mass.
For constant density, or incompressible flow the continuity equation reduces to ∇ · ~u = 0.

• Conservative force: F~ = −∇Ω

• Density: Denoted ̺ here, is the mass of fluid per unit volume. It can be a function fo
space and time, but oftent the assumption that it is constant is made.

• Euler equations of motion: For flows with high Reynolds number R, where the viscous
effects are small, most of the flow can be considered inviscid and a simpler set of equations

142
can be solved, which corresponds to the R → ∞ limit in the Navier-Stokes equations.
∂
They read ρ ~u + (~u · ∇)~u = −∇p + ρf~.
∂t
R H
• Guass Divergence Theorem: V ∇ · AdV ~ = SA ~ · ~ndS where V is a volume, S is the
surface bounding the volume V , and ~n is an outward pointing vector normal to the surface
S.

• Ideal (or perfect) fluid: This is a fluid which can exert no shearing stress across any
surface σij = −pδij that is the stress tensor for a perfect fluid is diagonal. This is done
by ignoring viscous effects.

• Incompressible fluid is a fluid which always has constant density. The continuity equa-
tion is then ∇ · ~u = 0.

• Inviscid: An inviscid fluid has zero viscocity.

• Kinematic viscocity: The kinematic viscocity is the ratio of viscocity and the density:
ν = µ/ρ.

• Lagrangian description:
∂~u
• Navier-Stokes equation: The Navier-Stokes equation is ρ +(~u·∇)~u = −∇p+µ∇2~u+ρf~
∂t
together with the continuity equation
 ∇ ·~u = 0. Another form of the Navier-Stokes
∂~u 1 2 p
equation is − (~u × ω) = −∇ |~u| + ~ + f~ where ~ω is the vorticity.
− ν∇ × ω
∂t 2 ρ
• Non-slip condition: Dictates that the speed of the fluid at a solid boundary is zero
relative to the boundary.

• Newtonian fluid:

• Pathline: A pathline is the locus of a particle.

!
1 ∂ui ∂uj
• Rate of strain tensor: eij = +
2 ∂xj ∂xi

• Reynolds number: R = UνL where U is the characterisic velocity, L the characteristic

length sclae and ν the viscocity. The smallness of R can be achieved by considering the
extremely small length scales, or dealing with a very viscous liquid, or treating flows with
very small velocity, so-called creeping flows.

• Slender body theory: Methodology that can be used to take advantage of the slenderness
of a body to obtain an approximation to the fluid field surrounding it and/or the net effect
of the field on the body.

• Stagnation point: A point in the flow field where the local velocity of the fluid is zero.

143
• Stokes equation: For Stoke’s flow the Navier-Stokes equation reduces to Stoke’s equation
∇p = µ∇2~u.

• Stokes flow: A type of flow where advective inertial forces are small compared with
viscous forces. The Reynolds number is low.

•R Stokes Theorem:
H Stayes that for any surface S encloed by the the closed curve C
∇×A ~= A~ · d~l holds.
S C

• Streamline: A line that is everywhere tangent to the velocity vector of a flow.

• Stress tensor: σij .

D ∂
• Substantial derivative: = + (~u · ∇)
Dt ∂t
• Surface forces:

• Viscocity: Viscocity is a measure of the fluid’s resistance to deformation by shear stress.

• Vorticity: ~ω = ∇ × ~u

144
Appendix B

Vector Calculus, Divergence

Theorem and Stoke’s Theorem

B.1 Vector Notions

An important vector operation is the cross product defined by

~ex ~ey ~ez

~
~a × b = a1 a2 a3 (B.1)
b1 b2 b3

which means

~a × ~b = (a2 b3 − b2 a3 )~ex + (a1 b3 − b1 a3 )~ey + (a1 b2 − b1 a2 )~ez (B.2)

The dot product of ~a and ~a × ~b is

a1 a2 a3
~a · (~a × ~b) = a1 a2 a3 = 0.
b1 b2 b3

Similarly ~b · (~a × ~b) = 0 therefore ~a × ~b is a vector perpendicular to the plane containing

~a and ~b. Also note

~a × ~b = −~b × ~a.

145
We consider the length of a cross product. We can simplify things by writing ~a = |~a|~ex
and ~b = |~b| cos θ~ex + |~b| sin θ~ey , then

~ex ~ey ~ez

|~a × ~b| = |~a| 0 0 = |~a||~b| sin θ~ez .
|~b| cos θ |~b| sin θ 0

Thus ~a × ~b has length |~a||~b| sin θ and a direction ~ez perpedicular to the plane containing
~a and ~b.

B.2 Vector Calculus

B.2.1 Gradient of a Scalar

The difference in φ is a scalar. However

∂φ ∂φ ∂φ
dφ = dx + dy + dz
∂x ∂y ∂z

As the the LHS is a scalr and dx, dy and dz are components of a vector the three numbers

∂φ ∂φ ∂φ
, ,
∂x ∂y ∂z

are aslo components of a vector. This vector is often denoted as ∇φ and so

∂φ ∂φ ∂φ
∇φ = ~ex + ~ey + ~ez
∂x ∂y ∂z

We call ∇φ the gradient of a scalar. Now the argument that ∇φ is a vector does not depend
on what scalar field we differentiated. As the trasformation as a vector is independent of
what scalar field we are differentiatinf we could just as well omit the φ.

Since the differential operators themselves transform as the components of a vector, we

call them components of a vector operator. We can write

∂ ∂ ∂
∇ = ~ex + ~ey + ~ez (B.3)
∂x ∂y ∂z

The gradient ∇ is an operator as it must act upon something.

146
B.2.2 Divergence

An obvious object the gradient ∇ could operate on would be a vector field ~v ,

∇ · ~v = ∇x vx + ∇y vy + ∇z vz
∂vx ∂vy ∂vz
= + + (B.4)
∂x ∂y ∂z

This is called the divergence and obviously transforms as a scalar.

Another way to combine a vector ~v with ∇ via a dot product is

~v · ∇

This is a scalar operator and does appear in fluid mechanics, in particular the advective
term of the substantial derivative.

B.2.3 Laplacian

It is possibly to combine two ∇? Take ∇ · (∇φ),

∇ · (∇φ) = (∇ · ∇)φ = ∇2 φ.

The ∇2 is a scalar operator that often appears in physics and is called the Laplacian,

2 ∂2 ∂2 ∂2
∇ = 2+ 2+ 2 (B.5)
∂x ∂y ∂z

How about acting ∇2 on a vector (we will at times need to do this as this terms appears
in the /navier-Stokes equations and the Stokes flow equations). In Cartesian coordinates
it is straightforward. For example the x-component of ∇2~v is simply

 
2 ∂2 ∂2 ∂2
(∇ ~v )x = + + vx = ∇2 vx . (B.6)
∂x2 ∂y 2 ∂z 2

The situation is not so simply in coordinates systems other than Cartesian coordinates.
We will later derive formula for spherical polar and cylindrical polar coordinates.

147
B.2.4 Curl

There is another object we can come uo with from the operator ∇ and a vector field ~v
using the cross product. The curl denoted ∇ × ~v can be defined via the cross product,

~ex ~ey ~ez

∂ ∂ ∂
∇ × ~v = (B.7)
∂x ∂y ∂y
~vx ~vy ~vz

The curl can be written in component form

∂
[∇ × ~v]i = ǫijk v (B.8)
∂xj k

B.2.5 Identities
∂ ∂
∇ × ∇φ = ǫijk φ=0
∂xj ∂xk

as partial derivatives commute, hence

∇ × ∇φ = 0. (B.9)

∂ ∂
∇ · (∇ × ~v ) = ǫijk ~v = 0
∂xi ∂xj k

again as partial derivatives commute, hence

∇ · (∇ × ~v) = 0. (B.10)

We can derive an identity for

∇ × (∇ × ~v )

with the usefull identity

ǫijk ǫki′ j ′ = (δii′ δjj ′ − δij ′ δji′ ) (B.11)

148
 [∇ × (∇ × ~v )]i = ǫijk ∂j [∇ × ~v ]k
= ǫijk ∂j (ǫki′ j ′ ∂i′ vj ′ )
= ǫijk ǫki′ j ′ ∂j ∂i′ vj ′
= (δii′ δjj ′ − δij ′ δji′ )∂j ∂i′ vj ′
= ∂i (∂j vj ) − ∂j ∂j vi . (B.12)

In vector notation this reads

∇ × (∇ × ~v ) = ∇(∇ · ~v ) − ∇2~v . (B.13)

B.3 Definitions and Statement of Line Integral The-

orem, Gauss’s Divergence Theorem and Stoke’s
Theorem

B.3.1 Defintion of a Line Integral

The first type of integral will need to introduce is a line integral of a scalar function f ,
denoted

Z Q
f dl (B.14)
P
|{z}
along Γ

We define the line integral as the limit of a sum,

X
fn ∆ln (B.15)
n

where fn is the value of the fucntion at the nth segment. The integral (B.14) comes from
increasing the number of segments so that the largest ∆ln → 0

The particular type of line integral that we will be interested in is of the form

Z Q
~v · d~l (B.16)
P
|{z}
along Γ

149
 Q

∆ln
∆l3
∆l1
P ∆l2

Figure B.1: .

where ~v is a vector field and d~l is an infintesimal displacement vector along the line Γ.
Hence it is of the form (B.14) where instead of f we have another scalar - the component
of ~v in the direction of d~l,

(~v · ~t) dl = ~v · d~l.

where ~t is the unit tangent vector to the curve (~t is the unit vector parallel to d~l). The
integral (B.16) is the sum of such terms.
~v
Q

d~l
P

Figure B.2: .

B.3.2 Theorem for a Line Integral

Our first theorem comes from using ~v = ∇ψ in the line integral.

Theorem: If Γ is any curve joining point P and point Q, the following is true,

Z Q
ψ(Q) − ψ(P ) = (∇ψ) · d~l. (B.17)
P
|{z}
any curve f rom
P to Q

This will be proved in section B.4.

150
B.3.3 Definition of Flux and Statement Gauss’s Divergence The-
orem

Given a surface S with outward unit-normal vector field ~n and a vector field ~u, the flux
is defined as

Z
~u · ~ndS.
S

Gauss’s theorem pertains to a closed surface S that encloses a volume V , and relates the
flux through the surface to the behaviour of a vector field inside the surface.

Gauss’s Divergence theorem:

Let V be a closed volume enclosed by the surface S, then

Z I
~
∇ · AdV = ~ · ~ndS
A (B.18)
V S

This will be proved in section B.5.

B.3.4 Statement of Stokes’ Theorem

Stokes’ theorem Let C be any simple closed path (i.e.,a path that starts and ends at
the same point and has no intersections), and consider any surface S of which C is the
boundary.Then stokes’theorem says that

Z I
~ · ~ndS =
∇×A ~ · d~l
A (B.19)
S C

This will be proved in section B.6.

B.3.5 Conservative Forces

A conservative force is a force with the property that the work done, W , in moving a
particle between two points

Z Q
W = F~ · d~l
P

151
is independent of the path taken. Equivalently, if a particle travels around a closed loop,
the net work done by a conservative force is zero. If a force F~ can be written in the form

F~ = −∇Φ

where Φ is a scalar field, the potential, the above theorem guarantees the force is conser-
vative. An obvious example of a conservative force is the gravitational force.

B.4 Proof of Theorem for Line Integral of ∇ψ

We now prove the theorem staed in section B.3.2. Consider the curve Γ split up into
segments as in fig (B1) where

∆~ln = ~pn+1 − p~n .

Then from multi-variable calculus

∂ψ ∂ψ ∂ψ
(∇ψ) · ∆~ln = ∆xn + ∆yn + ∆z
∂x ∂y ∂y n
= (∆ψ)n
= ψ(~pn+1 ) − ψ(~pn ). (B.20)

Now with ψ(~p0 ) = ψ(P ) and ψ(~pN +1 ) = ψ(Q), the approximation to the line integral is
the summation

N
X
(∇ψ) · ∆~ln = [ψ(~p1 ) − ψ(~p0 )] + [ψ(~p2 ) − ψ(~p1 )] + [ψ(~p3 ) − ψ(~p2 )] + · · · +
n=0
+ · · · + [ψ(~pN ) − ψ(~pN −1 )] + [ψ(~pN +1 ) − ψ(~pN )]
= ψ(~pN +1 ) − ψ(~p0 )
= ψ(Q) − ψ(P ). (B.21)

where all but two terms cancel. Taking the continuum limit does not change the last line
on the RHS. Also it is obvious that the answer will be the same regardless which curve
we use between P and Q.

152
B.5 Proof of Gauss’s Divergence Theorem

Say we have a closed surface S that encloses a volume V . Separate the volume into two
parts by taking a slice resulting it two closed surfaces and volumes as in fig .. The volume
V1 is enclosed by the closed surface S1 , which is made up of part of the original surface S1′
and of the surface of the slice S12 . The volume V2 is enclosed by S2 , which is made up of
the rest of the original surface S2′ and the surface of the slice S12 . The flux through S1 is
~u
S1′ S2′
~n

~n2 ~n1

V1 S12 V2

Figure B.3: .

We then have S1 = S1′ + S12 and S2 = S2′ + S12 . The flux through S1 can be written as
the sum of two parts

Z Z
~ · ~ndS +
A ~ · ~n dS
A (B.22)
1
S1′ S12

and similarly for the flux through S2 ,

Z Z
~ · ~ndS +
A ~ · ~n dS
A (B.23)
2
S2′ S12

As ~n1 = −~n2 , adding the fluxes of each of these surfaces gives

Z Z
~ · ~ndS +
A ~ · ~ndS
A (B.24)
S1′ S2′

which is just to the flux through the original surface S = S1 + S2 . We can subdivide
the volume again and again and it will generally be true that the flux through the outer
surface will be equal to the sum of the fluxes out of all the smaller interior pieces.

We now consider the special case of the flux out of a small cube.

153
  
∂vx ∂v
vx + dx dydz − vx dydz = x dxdydz (B.25)
∂x ∂x

Similar contributions come from the other two pairs of faces, adding together their con-
tributions the total flux through all faces is

Z  
∂v1 ∂v2 ∂v3
~v · ~ndS = + + dxdydz (B.26)
cube ∂x ∂y ∂z

or

Z
~v · ~ndS = (∇ · ~v )dxdydz (B.27)
cube

Splitting the volume V enclosed by a closed surface S into infintesimally small cubes,
summing the LHS of the above equation gives the total flux out of the closed surface and
summing over the RHS gives the volume integral

Z
∇ · ~vdV
V

resulting in

Z Z
~v · ~ndS = ∇ · ~v dV (B.28)
S V

B.6 Proof of Stokes’ Theorem

Let us find the circulation around an infintesimal square.

The circulation around the square is then

I
∂vy ∂v
~v · d~l = vx dx + (vy + )dy − (vx + x )dx − vy dy
∂x ∂y
 
∂vy ∂v
= − x dxdy (B.29)
∂x ∂y

Now say we had a collection of such squares as in fig (B.2) and wished to add up the
circulation from each individual square. Interior paths are transversed in opposite di-
rections, thus their contributions to each line integral cancel pairwise. Therefore only

154
 y

vy
dy v
v
(x, y) vx

dx

Figure B.4: Finding the circulation around an infintesimal square.

the outside edge contributes. This observation is an underlying principle in the proof of
Stoke’s theorem.

Figure B.5: When adding together the circulation of each loop the only remaing contri-
bution to the line integral comes from the outside edge.

We can write (B.6)

(∇ × ~v)z dS (B.30)

Here the z-component is the normal to the surface. We can therefore write the circulation
around a infintesimal square in an invariant vector form,

I
~v · d~l = (∇ × ~v ) · ~ndS (B.31)
C

So we have that the circulation of any vector ~v around an infintesimal square is the
component of the curl of ~v normal to the surface, times the area of the square. This result
is independent on the orientation of the square.

155
 Γ

Figure B.6: We have some surface bounded by the loop Γ. The surface is divided into
many small areas, each approximately a square.

Now suppose that we have a loop which is the boundary of some surface. There are
of course an infinite number of surfaces that have all have this loop as their boundary.
However the result does not depend on the particular surface choosen. Let the choosen
surface be divided into many small loops. If we take the loops small enough, we can
assume that each of the small loops enclose an area which is essentially flat. Also we
can choose our small loops so that each is very nearly a square. Combining (B.31) with
the fact that when you add up the circulation of each individual loop the only remaining
contribution comes from the outside edge, we have

Z I
∇ × ~v · ~ndS = ~v · d~l. (B.32)
S Γ

156
Appendix C

Orthogonal Curvilinear Coordinates

Curvilnear coordinates are a coordinate system for Euclidean space in which the coor-
dinate lines may be curved. We have already met such coordinates - cylindrical and
spherical polar coordinates.

We will use Guass’s theroem and Stoke’s theorem to find formula for the Divergence and
Curl respectively in orthogonal curvilinear coordinates. We have already derived formula
for the Gradient in the main text,combining this with formula for the Divergence we will
obtain a formula for the Laplacian of a scalar field.

The vector Laplacian and the advective term, both of which appear in the Navier-Stokes
equations, will be resolved in circular cylindrical and spherical polar coordinates respec-
tively in appendices D and E.

C.1 Gradient of a Scalar Field

We have already derived formula for the Gradient in the main text in section 3.1. The
formuala for the grad of a scalar is

3
X 1 ∂φ
∇φ = ~ei . (C.1)
i=1
hi ∂qi

The hi is defined by

∂~r
hi =
∂qi

where ~r is the position vector.

157
In cylindrical coordinates we have

q1 = ρ, q2 = ϕ, q3 = z
h1 = 1, h2 = ρ, h3 = 1 (C.2)

and in spherical coordinates

q1 = r, q2 = θ, q3 = ϕ
h1 = 1, h2 = r, h3 = r sin θ. (C.3)

From these examples it is clear that hi dqi (no summation over i implied) is the distance
change brought about by a coordinate change dqi .

C.2 The Divergence

We can use the Gauss’s Divergence theorem to find a formula for the divergence.

R
~
~v · dS
∇ · ~v (q1 , q2 , q3 ) = lim R (C.4)
dV →0 dV

with differencial volume

dV = h1 h2 h3 dq1 dq2 dq3 .

Define

v1 = ~v · ~e1 , v2 = ~v · ~e2 , v1 = ~v · ~e3 .

The area integral for the two faces q1 = Const. is given by

 
∂ ∂
v1 h2 h3 + (v1 h2 h3 )dq1 dq2 dq3 − v1 h2 h3 dq2 dq3 = (v h h )dq dq dq . (C.5)
∂q1 ∂q1 1 2 3 1 2 3

Adding in similar results for the other two pair of surfaces, we obtain

158
 ~e3
~e2 h3 dq3
h2 dq2

~e1 h1 dq1

Figure C.1: The gradient of ~v .

Z  
~ ∂(v1 h2 h3 ) ∂(v2 h1 h3 ) ∂(v3 h1 h2 )
~v (q1 , q2 , q3 ) · dS = + + dq1 dq2 dq3 . (C.6)
∂q1 ∂q2 ∂q3

And division by the diffrential volume (C.4) yields

 
1 ∂(v1 h2 h3 ) ∂(v2 h1 h3 ) ∂(v3 h1 h2 )
∇ · ~v (q1 , q2 , q3 ) = + + (C.7)
h1 h2 h3 ∂q1 ∂q2 ∂q3

C.3 Laplacian of a Scalar

Combining (C.1) and (C.7) we obtain the formula for the Laplacian of a scalar field,

      
2 1 ∂ h2 h3 ∂φ ∂ h1 h3 ∂φ ∂φ h1 h2 ∂φ
∇ φ(q1 , q2 , q3 ) = + +
h1 h2 h3 ∂q1 h1 ∂q1 ∂q2 h2 ∂q2 ∂q3 h3 ∂q3
(C.8)

C.4 The Curl

Using Stoke’s theorem

Z I
~=
∇ × ~v · dS ~v · d~l
C

in differential form we calculate the componet ~e1 · ∇ × ~v from

159
 h3 dq3

h2 dq2

~e1

Figure C.2: The ~e1 component of the Curl of an infintesimal loop.

  
∂
~e1 · ∇ × ~v dq2 dq3 h2 h3 = h3 v3 + (h v )dq − h3 v3 dq3
∂q2 3 3 2
  
∂
− h2 v2 + (h v )dq − h2 v2 dq2 (C.9)
∂q3 2 2 3

and so

 
1 ∂ ∂
~e1 · ∇ × ~v = (h v ) − (h v ) (C.10)
h2 h3 ∂q2 3 3 ∂q3 2 2

Similar results for the other two rectangles, those orthogonal to ~e2 and ~e3 , we obtain

 
1 ∂ ∂
∇ × ~v = (h v ) − (h v ) ~e1
h2 h3 ∂q2 3 3 ∂q3 2 2
 
1 ∂ ∂
+ (h v ) − (h v ) ~e2
h1 h3 ∂q3 1 1 ∂q1 3 3
 
1 ∂ ∂
+ (h v ) − (h v ) ~e3 (C.11)
h1 h2 ∂q1 3 3 ∂q3 1 1

which is compactly written as

∇ × ~v = ∇ × (v1~e1 + v2~e2 + v3~e3 )

~e1 h1 ~e2 h2 ~e3 h3
1 ∂ ∂ ∂
= . (C.12)
h1 h2 h3 ∂q1 ∂q2 ∂q3
h1 v1 h2 v2 h3 v3

160
Appendix D

Cylindrical Polar Coordinates

The formuala for

Relation to Cartesian coordinates:

x = ρ cos ϕ
y = ρ sin ϕ
z = z (D.1)

p
ρ = x2 + y 2
ϕ = tan−1 (y/x)
z = z (D.2)

D.1 Vector Differential operators

The non-zero derivatives of the basis vectors are

∂~eρ ∂~eϕ
= ~eϕ , = −~eρ . (D.3)
∂ϕ ∂ϕ

∂φ 1 ∂φ ∂φ
∇φ = ~eρ + ~eϕ + ~e (D.4)
∂r ρ ∂ϕ ∂z z

The Laplace operator is obtained from (C.2) and (C.8),

161
  
2 1 ∂ ∂ 1 ∂2 ∂2
∇ = ρ + 2 2 + 2. (D.5)
ρ ∂ρ ∂ρ ρ ∂ϕ ∂z

The curl of a vector field is obtained from (C.2) and (C.12),

∇ × ~u = ∇ × (uρ~eρ + uϕ~eϕ + uz ~ez )

~eρ ρ~eϕ ~ez
1 ∂ ∂ ∂
= ∂ρ ∂ϕ ∂z (D.6)
ρ
uρ ρuϕ uz

The Laplacian of a vector field ∇2~u resolved in cylindrical polar coordinates is

1 2 ∂uϕ
∇2~u ρ
= ∇2 u ρ − 2
uρ − 2
ρ ρ ∂ϕ
1 2 ∂uρ
∇2~u ϕ
= ∇2 u ϕ − 2 u ϕ + 2
ρ ρ ∂ϕ
∇2~u z
2
= ∇ uz (D.7)

where we have applied (D.5) to ~u = uρ~eρ + uϕ~eϕ + uz ~ez and used (D.3). These would
be used in writing the µ∇2~u term in the Navier-Stokes equations in cylindrical polar
coordinates.

If ~u is the velocity field and if we had incompressible flow, ∇ · ~u = 0. Then from the
identity ∇ × (∇ × ~u) = ∇(∇ · ~u) − ∇2~u and using ∇ · ~u = 0, then we can use the
alternative formula ∇2~u = −∇ × (∇ × ~u).

The non-linear term (~u · ∇)~u of the Navier-Stokes equations using (D.3) and (D.4) is

∂ 1 ∂ ∂
(~u · ∇)~u = (uρ + uϕ + uz )(uρ~eρ + uϕ~eϕ + uz~ez )
∂ρ ρ ∂ϕ ∂z
!
∂uρ 1 ∂uρ ∂uρ u2ϕ
= uρ + uϕ + uz − ~eρ
∂ρ ρ ∂ϕ ∂z ρ
 
∂uϕ 1 ∂uϕ ∂uϕ uρ uϕ
+ uρ + uϕ + uz + ~eϕ
∂ρ ρ ∂ϕ ∂z ρ
 
∂uz 1 ∂uz ∂uz
+ uρ + uϕ + uz ~ez (D.8)
∂ρ ρ ∂ϕ ∂z

Or

162
 u2ϕ
(~u · ∇)~u|ρ = (~u · ∇)uρ −
ρ
uρ uϕ
(~u · ∇)~u|ϕ = (~u · ∇)uϕ +
ρ
(~u · ∇)~u|z = (~u · ∇)uz . (D.9)

D.2 Rate of strain tensor revisted

Recall the rate of strain tensor in cartesian coordinates

!
1 ∂ui ∂uj
ǫij = + . (D.10)
2 ∂xj ∂xi

Take

 
∂ ∂ ∂
∇~u = ~ex + ~ey + ~ez ~u
∂x ∂y ∂z

where

~u = ux~ex + uy~ey + uz~ez

Write

 
∂ ∂ ∂
∇ ⊗ ~u = ~ex + ~ey + ~ez ⊗ ~u
∂x ∂y ∂z
∂ ∂ ∂
= (ux~ex ) ⊗ ~ex + (ux~ex ) ⊗ ~ey + (ux~ex ) ⊗ ~ez
∂x ∂y ∂z
∂ ∂ ∂
(uy~ey ) ⊗ ~ex + (uy~ey ) ⊗ ~ey + (uy~ey ) ⊗ ~ez
∂x ∂y ∂z
∂ ∂ ∂
(uz~ez ) ⊗ ~ex + (uz~ez ) ⊗ ~ey + (uz ~ez ) ⊗ ~ez
∂x ∂y ∂z
X ∂u
i
= ~ei ⊗ ~ej (D.11)
i,j
∂x j

where we have used that the derivatives do not effect the basis vectors. This gives twice
the second part of (D.10). It can be seen that

163
  
T ∂ ∂ ∂
(∇ ⊗ ~u) = ~u ⊗ ~ex + ~ey + ~ez
∂x ∂y ∂z
∂ ∂ ∂
= (ux~ex ) ⊗ ~ex + (uy~ey ) ⊗ ~ex + (u ~e ) ⊗ ~ex
∂x ∂x ∂x z z
∂ ∂ ∂
(ux~ex ) ⊗ ~ey + (uy~ey ) ⊗ ~ey + (u ~e ) ⊗ ~ey
∂y ∂y ∂y z z
∂ ∂ ∂
(ux~ex ) ⊗ ~ez + (uy~ey ) ⊗ ~ez + (uz~ez ) ⊗ ~ez
∂z ∂z ∂z
X ∂uj
= ~ei ⊗ ~ej (D.12)
i,j
∂x i

gives twice the first part of (D.10).

So that

1

ǫ= (∇ ⊗ ~u)T + ∇ ⊗ ~u (D.13)
2

with matrix components

 
ǫxx ǫxy ǫxz
ǫ = ǫyx ǫyy ǫyz  (D.14)
ǫzx ǫzy ǫzz

which are

∂ux ∂uy
ǫxx = ǫyy =
∂x ∂y
 
∂uz 1 ∂ux ∂uy
ǫzz = ǫxy = +
∂z 2 ∂y ∂x
   
1 ∂uy ∂uz 1 ∂ux ∂uz
ǫyz = + ǫxz = + (D.15)
2 ∂z ∂y 2 ∂z ∂x

D.3 Rate of strain tensor

We now consider the rate of strain tensor in cylindrical polar coordinates. Take

 
T ∂ 1 ∂ ∂
(∇ ⊗ ~u) = ~u ⊗ ~eρ + ~eϕ + ~ez
∂ρ ρ ∂ϕ ∂z

164
where

~u = uρ~eρ + uϕ~eϕ + uz~ez .

Non-trival terms of this are

1 ∂ 1 ∂uρ uρ ∂~eρ
(uρ~eρ ) ⊗ ~eϕ = ~e ⊗ ~eϕ + ⊗ ~eϕ
ρ ∂ϕ ρ ∂ϕ ρ ρ ∂ϕ
1 ∂uρ uρ
= ~e ⊗ ~eϕ + ~e ⊗ ~eϕ (D.16)
ρ ∂ϕ ρ ρ ϕ

and

1 ∂ 1 ∂uϕ uϕ ∂~eϕ
(uϕ~eϕ ) ⊗ ~eϕ = ~e ⊗ ~eϕ + ⊗ ~eϕ
ρ ∂ϕ ρ ∂ϕ ϕ ρ ∂ϕ
1 ∂uϕ uϕ
= ~e ⊗ ~eϕ − ~e ⊗ ~eϕ (D.17)
ρ ∂ϕ ϕ ρ ρ

All the rest are trivial as the derivatives dont effect the basis vectors. For example

∂ ∂uρ 1 ∂ 1 ∂uz
(uρ~eρ ) ⊗ ~eρ = ~e ⊗ ~eρ , (uz~ez ) ⊗ ~eϕ = ~e ⊗ ~eϕ .
∂ρ ∂ρ ρ ρ ∂ϕ ρ ∂ϕ z

All these terms sum up to give an expression of the form

(∇ ⊗ ~u)T = Mij ~ei ⊗ ~ej . (D.18)

It is then seen that the matrix representing the components is

 ∂u 
ρ 1 ∂uρ uϕ ∂uρ
∂ρ ρ ∂ϕ
− ρ ∂z
 ∂u 1 ∂uϕ uρ ∂uϕ 
 ∂ρϕ ρ ∂ϕ
+ ρ ∂z 
(D.19)
∂uz 1 ∂uz ∂uz
∂r ρ ∂ϕ ∂z

The rate of strain tensor is obtained by adding the transpose of the velocity gradient
tensor to the velocity gradient tensor and dividing by 2:

1

ǫ= (∇ ⊗ ~u)T + ∇ ⊗ ~u (D.20)
2
165
with components

 
ǫρρ ǫρϕ ǫρz
 ǫϕρ ǫϕϕ ǫϕz  (D.21)
ǫzρ ǫzϕ ǫzz

which are

∂uρ 1 ∂uϕ uρ
ǫρρ = ǫϕϕ = +
∂ρ ρ ∂ϕ ρ
 
∂u ρ ∂ u ϕ 1 ∂uρ
ǫzz = z ǫρϕ = +
∂z 2 ∂ρ ρ 2ρ ∂ϕ
   
1 1 ∂uz ∂uϕ 1 ∂uρ ∂uz
ǫϕz = + ǫρz = + . (D.22)
2 ρ ∂ϕ ∂z 2 ∂z ∂ρ

166
Appendix E

Spherical Polar Coordinates

Relation to Cartesian coordinates:

x = r sin θ cos ϕ
y = r sin θ sin ϕ
z = r cos θ (E.1)

p
r = x2 + y 2 + z 2
θ = cos−1 (z/(x2 + y 2 + z 2 )1/2 )
ϕ = tan−1 (y/x) (E.2)

E.1 Vector Differential operators

The derivatives of the basis vectors are

∂~er ∂~e ∂~eϕ ∂~er ∂~eθ ∂~eϕ

= θ = =0 = ~eθ = −~er =0
∂r ∂r ∂r ∂θ ∂θ ∂θ
∂~er ∂~eθ ∂~eϕ
= sin θ~eϕ = cos θ~eϕ = − sin θ~er − cos θ~eθ (E.3)
∂ϕ ∂ϕ ∂ϕ

These can be derived by differentiating

167
 ~er = sin θ cos ϕ~ex + sin θ sin ϕ~ey + cos θ~ez
~eθ = cos θ cos ϕ~ex + cos θ sin ϕ~ey − sin θ~ez
~eϕ = − sin θ~er − cos θ~eθ (E.4)

∂φ 1 ∂φ 1 ∂φ
∇φ = ~er + ~eθ + ~e (E.5)
∂r r ∂θ r sin θ ∂ϕ ϕ

The divergence of a vector field ~u is

1 ∂ 2 1 ∂ 1 ∂uϕ
∇ · ~u = 2
(r ur ) + (sin θuθ ) + (E.6)
r ∂r r sin θ ∂θ r sin θ ∂ϕ

The Laplace operator is (C.3) and (C.8),

   
2 1 ∂ 2 ∂ 1 ∂ ∂ 1 ∂2
∇ = 2 r + 2 sin θ + 2 2 (E.7)
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2

The curl of a vector field is given by (C.3) and (C.12),

∇ × ~u = ∇ × (ur~er + uθ~eθ + uϕ~eϕ )

~er r~eθ r sin θ~eϕ
1 ∂ ∂ ∂
= 2 ∂r ∂θ ∂ϕ (E.8)
r sin θ
ur ruθ r sin θuϕ

The Laplacian of a vector field ∇2~u resolved in spherical polar coordinates is

2 2 ∂uθ 2 cos θ 2 ∂uϕ

∇2~u = ∇2 u r − u − − u −
r r 2 r r 2 ∂θ r 2 sin θ θ r 2 sin θ ∂ϕ
1 2 ∂u 2 cos θ ∂uϕ
∇2~u θ
= ∇2 u θ − 2 2 u θ + 2 r − 2 2
r sin θ r ∂θ r sin θ ∂ϕ
1 2 ∂ur 2 cos θ ∂uθ
∇2~u ϕ
= ∇2 u ϕ − 2 2 u ϕ + 2 + 2 2 . (E.9)
r sin θ r sin θ ∂ϕ r sin θ ∂ϕ

where we have applied (E.7) to ~u = ur~er +uθ~eθ +uϕ~eϕ and used (E.3). These would be used
in writing the µ∇2~u term in the Navier-Stokes equations in spherical polar coordinates.

168
If ~u is the velocity field and if we had incompressible flow, ∇ · ~u = 0. Then from the
identity ∇ × (∇ × ~u) = ∇(∇ · ~u) − ∇2~u and using ∇ · ~u = 0, then we can use the
alternative formula ∇2~u = −∇ × (∇ × ~u).

The non-linear term (~u · ∇)~u of the Navier-Stokes equations

u2θ + u2ϕ
(~u · ∇)~u|r = (~u · ∇)ur −
r
uu u2ϕ cot θ
(~u · ∇)~u|θ = (~u · ∇)uθ + r θ −
r r
ur uϕ uθ uϕ cot θ
(~u · ∇)~u|ϕ = (~u · ∇)uϕ + − . (E.10)
r r

E.2 Rate of strain tensor

We now consider the rate of strain tensor in spherical polar coordinates. Take

 
T ∂ 1 ∂ 1 ∂
(∇ ⊗ ~u) = ~u ⊗ ~er + ~eθ + ~eφ
∂r r ∂θ r sin θ ∂φ

where

~u = ur~er + uθ~eθ + uφ~eφ .

Non-trivial terms are

1 ∂ 1 ∂ur ur ∂~er
(ur~er ) ⊗ ~eθ = ~e ⊗ ~eθ + ⊗ ~eθ
r ∂θ r ∂θ r r ∂θ
1 ∂ur ur
= ~e ⊗ ~eθ + ~e ⊗ ~eθ (E.11)
r ∂θ r r θ

The matrix elements of ∇~u are

 ∂ur 1 ∂ur uθ 1 ∂vr u 

∂r r ∂θ
− r r sin θ ∂φ
− rφ
 ∂θ 1 ∂uθ ur 1 ∂uθ u 
 ∂r r ∂θ
+ r r sin θ ∂φ
− cot θ rφ . (E.12)
∂φ ∂u 1 ∂uφ
∂r
1
r ∂θ
φ
r sin θ ∂φ
+ cot θ urθ + urr

The rate of strain tensor is obtained by adding the transpose of the velocity gradient
tensor and the velocity gradient tensor and dividing by 2:

169
 1

ǫ= (∇ ⊗ ~u)T + ∇ ⊗ ~u (E.13)
2

whose components are the matrix elements are

 
ǫrr ǫrθ ǫrφ
 ǫθr ǫθθ ǫθφ  (E.14)
ǫφr ǫφθ ǫφφ

which are

∂ur 1 ∂uθ ur
ǫrr = ǫθθ = +
∂r r ∂r r
1 ∂uφ ur uθ cot θ r ∂  uθ  1 ∂ur
ǫφφ = + + ǫrθ = +
r sin θ ∂φ r r 2 ∂r r 2r ∂θ
   
sin θ ∂ uφ 1 ∂uθ 1 ∂ur r ∂ uφ
ǫθφ = + ǫφr = + .(E.15)
2r ∂θ sin θ 2r sin θ ∂φ 2r sin θ ∂φ 2 ∂r r

ǫϕϕ

ǫθθ
ǫrr

Figure E.1: Stress.

170