Cantwell - Fundamentals of Compressible Flow PDF

AA210A
Course Reader
Fundamentals of Compressible Flow
by
Brian J. Cantwell
Department of Aeronautics and Astronautics

Stanford University, Stanford, California 94305
All rights reserved. These AA210 course notes are subject to copyright
and cannot be reproduced without the written permission of the author.
October 3, 2018
Contents
1 Introduction to fluid flow 1-1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1
1.2 Conservation of mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1
1.2.1 Incompressible flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3
1.2.2 Index notation and the Einstein convention . . . . . . . . . . . . . . 1-3
1.3 Particle paths, streamlines and streaklines in 2-D steady flow . . . . . . . . 1-5
1.3.1 Analysis of particle paths and streamlines . . . . . . . . . . . . . . . 1-8
1.3.2 The integrating factor . . . . . . . . . . . . . . . . . . . . . . . . . . 1-11
1.3.3 Sample problem - the integrating factor and reconstruction of a func-
tion from its perfect differential . . . . . . . . . . . . . . . . . . . . . 1-12
1.3.4 Incompressible flow in two dimensions . . . . . . . . . . . . . . . . . 1-14
1.3.5 Compressible flow in two dimensions . . . . . . . . . . . . . . . . . . 1-15
1.4 Particle paths in three dimensions . . . . . . . . . . . . . . . . . . . . . . . 1-16
1.5 The substantial derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-17
1.5.1 Frames of reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-17
1.6 Momentum transport due to convection . . . . . . . . . . . . . . . . . . . . 1-19
1.7 Momentum transport due to molecular motion at microscopic scales . . . . 1-23
1.7.1 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-23
1.7.2 Viscous friction - plane Couette flow . . . . . . . . . . . . . . . . . . 1-24
1.7.3 A question of signs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-25
1.7.4 Newtonian fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-26
1.7.5 Forces acting on a fluid element . . . . . . . . . . . . . . . . . . . . . 1-26
1.8 Conservation of energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-29
1.8.1 Pressure and viscous stress work . . . . . . . . . . . . . . . . . . . . 1-31
1.9 Summary - the equations of motion . . . . . . . . . . . . . . . . . . . . . . . 1-34
1.10 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-36
2 Thermodynamics of dilute gases 2-1

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-1
2.2 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-1
1
CONTENTS 2
2.2.1 Temperature and the zeroth law . . . . . . . . . . . . . . . . . . . . 2-1

2.2.2 The first law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
2.2.3 The second law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4
2.3 The Carnot cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5
2.3.1 The absolute scale of temperature . . . . . . . . . . . . . . . . . . . 2-8
2.4 Enthalpy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-9
2.4.1 Gibbs equation on a fluid element . . . . . . . . . . . . . . . . . . . 2-10
2.5 Heat capacities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-12
2.6 Ideal gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-13
2.7 Constant specific heat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-16
2.8 The entropy of mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-18
2.8.1 Sample problem - thermal mixing . . . . . . . . . . . . . . . . . . . . 2-18
2.8.2 Entropy change due to mixing of distinct gases . . . . . . . . . . . . 2-20
2.8.3 Gibbs paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-22
2.9 Isentropic expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-24
2.9.1 Blowdown of a pressure vessel . . . . . . . . . . . . . . . . . . . . . . 2-24
2.9.2 Work done by an expanding gas . . . . . . . . . . . . . . . . . . . . 2-25
2.9.3 Example - Helium gas gun . . . . . . . . . . . . . . . . . . . . . . . . 2-27
2.9.4 Entropy increase due to viscous friction . . . . . . . . . . . . . . . . 2-28
2.10 Some results from statistical mechanics . . . . . . . . . . . . . . . . . . . . 2-29
2.10.1 Diatomic gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-30
2.10.2 Characteristic vibrational temperature . . . . . . . . . . . . . . . . . 2-30
2.11 Enthalpy - diatomic gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-31
2.12 Speed of sound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-31
2.13 Atmospheric models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-32
2.14 The third law of thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . 2-33
2.15 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-36
3 Control volumes, vector calculus 3-1

3.1 Control volume definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1
3.2 Vector calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2
3.2.1 Useful vector identities . . . . . . . . . . . . . . . . . . . . . . . . . . 3-5
3.3 The Gauss theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-6
3.4 Stokes’ theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-7
3.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-8
4 Kinematics of fluid motion 4-1

4.1 Elementary flow patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1
4.1.1 Linear flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1
4.1.2 Linear flows in two dimensions . . . . . . . . . . . . . . . . . . . . . 4-3
4.1.3 Linear flows in three dimensions . . . . . . . . . . . . . . . . . . . . 4-4
CONTENTS 3
4.1.4 Incompressible flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-6

4.2 Rate-of-strain and rate-of-rotation tensors . . . . . . . . . . . . . . . . . . . 4-8
4.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-10
5 The conservation equations 5-1

5.1 Leibniz’ rule for differentiation of integrals . . . . . . . . . . . . . . . . . . . 5-1
5.1.1 Differentiation under the integral sign . . . . . . . . . . . . . . . . . 5-1
5.1.2 Extension to three dimensions . . . . . . . . . . . . . . . . . . . . . 5-3
5.2 Conservation of mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-5
5.3 Conservation of momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-6
5.4 Conservation of energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-8
5.5 Summary - differential form of the equations of motion . . . . . . . . . . . . 5-10
5.6 Integral form of the equations of motion . . . . . . . . . . . . . . . . . . . . 5-11
5.6.1 Integral equations on an Eulerian control volume . . . . . . . . . . . 5-11
5.6.2 Mixed Eulerian-Lagrangian control volumes . . . . . . . . . . . . . . 5-12
5.7 Applications of control volume analysis . . . . . . . . . . . . . . . . . . . . . 5-13
5.7.1 Example 1 - Solid body at rest in a steady flow . . . . . . . . . . . . 5-13
5.7.2 Example 2 - Channel flow with heat addition . . . . . . . . . . . . . 5-14
5.7.3 Example 3 - Stationary flow about a rotating fan . . . . . . . . . . . 5-16
5.7.4 Example 4 - Combined heat transfer and work . . . . . . . . . . . . 5-18
5.8 Stagnation enthalpy, temperature and pressure . . . . . . . . . . . . . . . . 5-18
5.8.1 Stagnation enthalpy of a fluid element . . . . . . . . . . . . . . . . . 5-18
5.8.2 Blowdown from a pressure vessel revisited . . . . . . . . . . . . . . . 5-19
5.8.3 Stagnation enthalpy and temperature in steady flow . . . . . . . . . 5-21
5.8.5 Stagnation pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-23
5.8.6 Transforming the stagnation pressure between fixed and moving frames5-25
5.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-25
6 Several forms of the equations of motion 6-1

6.1 The Navier-Stokes equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-1
6.1.1 Incompressible Navier-Stokes equations . . . . . . . . . . . . . . . . 6-2
6.2 The momentum equation expressed in terms of vorticity . . . . . . . . . . . 6-2
6.3 The momentum equation expressed in terms of the entropy and vorticity . . 6-3
6.3.1 Crocco’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-4
6.3.2 The energy equation for inviscid, non-heat conducting flow . . . . . 6-4
6.3.3 Steady flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-5
6.4 Inviscid, irrotational, homentropic flow . . . . . . . . . . . . . . . . . . . . . 6-6
6.4.1 Steady, inviscid, irrotational flow . . . . . . . . . . . . . . . . . . . . 6-6
6.5 The velocity potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-8
CONTENTS 4
6.5.1 Unsteady potential flow, the unsteady Bernoulli integral . . . . . . . 6-8

6.5.2 Incompressible, irrotational flow . . . . . . . . . . . . . . . . . . . . 6-11
6.6 The vorticity equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-11
6.7 Fluid flow in three dimensions, the dual stream function . . . . . . . . . . . 6-13
6.8 The vector potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-15
6.8.1 Selection of a Coulomb gauge . . . . . . . . . . . . . . . . . . . . . . 6-15
6.9 Incompressible flow with mass and vorticity sources . . . . . . . . . . . . . 6-16
6.10 Turbulent flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-17
6.10.1 Turbulent, incompressible, isothermal flow . . . . . . . . . . . . . . . 6-20
6.11 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-21
7 Entropy generation and transport 7-1

7.1 Convective form of the Gibbs equation . . . . . . . . . . . . . . . . . . . . . 7-1
7.2 The kinetic energy equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-2
7.3 Internal energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-3
7.3.1 Viscous dissipation of kinetic energy . . . . . . . . . . . . . . . . . . 7-4
7.4 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-5
7.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-8
8 Viscous flow along a wall 8-1

8.1 The no-slip condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-1
8.2 Equations of motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-3
8.3 Plane, compressible Couette flow . . . . . . . . . . . . . . . . . . . . . . . . 8-4
8.3.1 The energy integral in plane Couette flow . . . . . . . . . . . . . . . 8-6
8.3.2 The adiabatic wall recovery temperature . . . . . . . . . . . . . . . . 8-7
8.3.3 Velocity distribution in Couette flow . . . . . . . . . . . . . . . . . . 8-9
8.4 The viscous boundary layer on a wall . . . . . . . . . . . . . . . . . . . . . . 8-11
8.4.1 Measures of boundary layer thicknessl . . . . . . . . . . . . . . . . . 8-17
8.5 The Von Karman integral momentum equation . . . . . . . . . . . . . . . . 8-17
8.6 The laminar boundary layer in the limit M 2 → 0. . . . . . . . . . . . . . . . 8-21
8.6.1 The zero pressure gradient, incompressible boundary layer . . . . . . 8-22
8.7 The Falkner-Skan boundary layers . . . . . . . . . . . . . . . . . . . . . . . 8-27
8.7.1 The case β = −1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-31
8.7.2 Falkner-Skan sink flow. . . . . . . . . . . . . . . . . . . . . . . . . . 8-33
8.8 Thwaites’ method for approximate calculation of laminar boundary layer
characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-36
8.8.1 Example - free stream velocity from the potential flow over a circular
cylinder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-42
8.9 Compressible laminar boundary layers . . . . . . . . . . . . . . . . . . . . . 8-45
8.9.1 The energy integral for a compressible boundary layer with an adia-
batic wall and Pr = 1. . . . . . . . . . . . . . . . . . . . . . . . . . . 8-46
CONTENTS 5
8.9.2 Non-adiabatic wall with dP/dx = 0 and Pr = 1 . . . . . . . . . . . . 8-47

8.10 Mapping a compressible to an incompressible boundary layer . . . . . . . . 8-49
8.11 Turbulent boundary layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-61
8.11.1 The incompressible turbulent boundary layer velocity profile . . . . 8-63
8.12 Transformation between flat plate and curved wall boundary layers . . . . . 8-67
8.13 Head’s method for approximate calculation of turbulent boundary layer
characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-71
8.13.1 Head’s method applied to the zero pressure gradient flat plate . . . 8-77
8.13.2 Head’s method used to study the effect of Reynolds number on flow
separation on a circular cylinder . . . . . . . . . . . . . . . . . . . . 8-78
8.14 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-81
9 Quasi-one-dimensional flow 9-1

9.1 Control volume and integral conservation equations . . . . . . . . . . . . . . 9-1
9.1.1 Conservation of mass . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-2
9.1.2 Conservation of x -momentum . . . . . . . . . . . . . . . . . . . . . . 9-3
9.1.3 Conservation of energy . . . . . . . . . . . . . . . . . . . . . . . . . . 9-4
9.2 Area averaged flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-5
9.2.1 The traction vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-7
9.2.2 Example - steady, gravity-free, adiabatic flow of a compressible fluid
in a channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-11
9.3 Normal shock waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-12
9.3.1 The Rankine-Hugoniot relations . . . . . . . . . . . . . . . . . . . . 9-14
9.3.2 Shock property ratios in a calorically perfect ideal gas . . . . . . . . 9-15
9.3.3 Stagnation pressure ratio across a normal shock wave . . . . . . . . 9-18
9.3.4 Example - stagnation at a leading edge in supersonic flow . . . . . . 9-20
9.4 Shock wave thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-24
9.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-30
10 Gasdynamics of nozzle flow 10-1

10.1 Area-Mach number function . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-1
10.1.1 Mass conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-4
10.2 A simple convergent nozzle . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-6
10.2.1 The phenomenon of choking . . . . . . . . . . . . . . . . . . . . . . . 10-7
10.3 The converging-diverging nozzle . . . . . . . . . . . . . . . . . . . . . . . . . 10-8
10.3.1 Case 1 - Isentropic subsonic flow in the nozzle . . . . . . . . . . . . . 10-9
10.3.2 Case 2 - Non-isentropic flow - shock in the nozzle . . . . . . . . . . . 10-10
10.3.3 Case 3 - Isentropic supersonic flow in the nozzle . . . . . . . . . . . 10-12
10.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-13
10.4.1 Shock in a nozzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-13
10.4.2 Cold gas thruster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-16
CONTENTS 6
10.4.3 Gasdynamics of a double throat, starting and unstarting supersonic

flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-17
10.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-20
11 Area change, wall friction and heat transfer 11-1

11.1 Control volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-1
11.2 Entropy and stagnation pressure . . . . . . . . . . . . . . . . . . . . . . . . 11-2
11.3 Velocity, density, temperature and pressure . . . . . . . . . . . . . . . . . . 11-4
11.4 The Mach number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-5
11.5 Mass flow, area-Mach-number function . . . . . . . . . . . . . . . . . . . . . 11-6
11.6 Integrated relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-7
11.6.1 1-D, adiabatic, constant area flow with friction (Fanno line flow) . . 11-7
11.6.2 Example - frictional flow in a long pipe . . . . . . . . . . . . . . . . 11-8
11.6.3 Integrated relations - 1-D frictionless, constant area flow with heat
transfer (Rayleigh line flow) . . . . . . . . . . . . . . . . . . . . . . . 11-11
11.6.4 Example - double throat with heat addition . . . . . . . . . . . . . . 11-13
11.7 Detonations and deflagrations . . . . . . . . . . . . . . . . . . . . . . . . . . 11-15
11.7.1 Example - detonation in a mixture of fluorine and hydrogen diluted
by nitrogen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-20
11.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-20
12 Steady waves in compressible flow 12-1

12.1 Oblique shock waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-1
12.1.1 Exceptional relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-4
12.1.2 Flow deflection versus shock angle . . . . . . . . . . . . . . . . . . . 12-5
12.2 Weak oblique waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-7
12.3 The Prandtl-Meyer expansion . . . . . . . . . . . . . . . . . . . . . . . . . . 12-10
12.3.1 Example - supersonic flow over a bump . . . . . . . . . . . . . . . . 12-12
12.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-13
13 Unsteady waves in compressible flow 13-1

13.1 Governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-1
13.2 The acoustic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-1
13.3 Propagation of acoustic waves in one space dimension . . . . . . . . . . . . 13-4
13.4 Isentropic, finite amplitude waves . . . . . . . . . . . . . . . . . . . . . . . . 13-7
13.5 Centered expansion wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-10
13.6 Compression wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-12
13.7 The shock tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-12
13.7.1 Example - flow induced by the shock in a shock tube . . . . . . . . . 13-16
13.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-18
CONTENTS 7
14 Thin airfoil theory 14-1

14.1 Compressible potential flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-1
14.1.1 The full potential equation . . . . . . . . . . . . . . . . . . . . . . . 14-1
14.1.2 The nonlinear small disturbance approximation . . . . . . . . . . . . 14-3
14.1.3 Linearized potential flow . . . . . . . . . . . . . . . . . . . . . . . . . 14-7
14.1.4 The pressure coefficient . . . . . . . . . . . . . . . . . . . . . . . . . 14-9
14.1.5 Drag coefficient of a thin symmetric airfoil . . . . . . . . . . . . . . . 14-11
14.1.6 Thin airfoil with lift and camber at a small angle of attack . . . . . 14-12
14.2 Similarity rules for high speed flight . . . . . . . . . . . . . . . . . . . . . . 14-16
14.2.1 Subsonic flow M∞ < 1 . . . . . . . . . . . . . . . . . . . . . . . . . . 14-18
14.2.2 Supersonic similarity M∞ > 1 . . . . . . . . . . . . . . . . . . . . . . 14-22
14.2.3 Transonic similarity M∞ ∼ =1 . . . . . . . . . . . . . . . . . . . . . . 14-23
14.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-24
A Some results from the kinetic theory of gases A-1

A.1 Distribution of molecular velocities in a gas . . . . . . . . . . . . . . . . . . A-1
A.1.1 The distribution derived from the barometric formula . . . . . . . . A-1
A.1.2 The Maxwell velocity distribution function . . . . . . . . . . . . . . A-5
A.2 Mean molecular velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-7
A.3 Distribution of molecular speeds . . . . . . . . . . . . . . . . . . . . . . . . A-9
A.4 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-10
A.4.1 Kinetic model of pressure . . . . . . . . . . . . . . . . . . . . . . . . A-10
A.4.2 Pressure directly from the Maxwellian pdf . . . . . . . . . . . . . . . A-12
A.5 The mean free path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-13
A.6 Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-13
A.7 Heat conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-15
A.8 Specific heats, the law of equipartition . . . . . . . . . . . . . . . . . . . . . A-15
A.9 Diatomic gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-17
A.9.1 Rotational degrees of freedom . . . . . . . . . . . . . . . . . . . . . . A-17
A.9.2 Why are only two rotational degrees of freedom excited? . . . . . . . A-19
A.9.3 Vibrational degrees of freedom . . . . . . . . . . . . . . . . . . . . . A-21
A.10 Energy levels in a box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-24
A.10.1 Counting energy states . . . . . . . . . . . . . . . . . . . . . . . . . . A-26
A.10.2 Entropy of a monatomic gas in terms of the number of states . . . . A-28
B Equations in cylindrical and spherical coords B-1

B.1 Coordinate systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-1
B.1.1 Cylindrical coordinates . . . . . . . . . . . . . . . . . . . . . . . . . B-1
B.1.2 Spherical polar coordinates . . . . . . . . . . . . . . . . . . . . . . . B-2
B.2 Transformation of vector components . . . . . . . . . . . . . . . . . . . . . . B-3
CONTENTS 8

B.3 Summary of differential operations . . . . . . . . . . . . . . . . . . . . . . . B-4
B.4 Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-7
B.5 Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-8
B.6 Energy equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-9
B.7 Components of the stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . B-12
B.8 Energy dissipation function Φ . . . . . . . . . . . . . . . . . . . . . . . . . . 13
B.8.1 Cylindrical coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 13
B.8.2 Spherical polar coordinates . . . . . . . . . . . . . . . . . . . . . . . 13
Chapter 1
Introduction to fluid flow
1.1 Introduction
Compressible flows play a crucial role in a vast variety of natural phenomena and man-
made systems. The life-cycles of stars, the creation of atmospheres, the sounds we hear,
the vehicles we ride and the systems we build for generating energy and propulsion all
depend in an important way on the mechanics and thermodynamics of compressible flow.
The purpose of this course is to introduce students in Aeronautics and Astronautics to
the fundamental principles of compressible flow with emphasis on the development of the
equations of motion, as well as some of the analytical tools from calculus needed to solve
practically important problems involving nozzles, channels and lifting bodies.
1.2 Conservation of mass
Mass is neither created nor destroyed. This basic principle of classical physics is one of the
fundamental laws governing fluid motion and is a good departure point for our introductory
discussion. Figure 1.1 below shows an infinitesimally small stationary, rectangular control
volume ∆x∆y∆z through which a fluid is assumed to be moving. A control volume of
this type, with its surface fixed in space, is called an Eulerian control volume. The fluid
velocity vector has vector components Ū = (U, V, W ) in the x̄ = (x, y, z) directions and the
fluid density is ρ. In a general, unsteady, compressible flow, all flow variables may depend
on position and time. The law of conservation of mass over this control volume is stated
as
1-1
CHAPTER 1. INTRODUCTION TO FLUID FLOW 1-2
     

 rate of mass 
 
 

accumulation f low f low
     
= − . (1.1)

 inside the control 
  into the control   out of the control 
  
  
volume volume volume
   
Figure 1.1: Fixed control volume in a moving fluid. The arrows denote fluxes of mass on
the various faces of the control volume.
Consider a pair of faces perpendicular to the x-axis. The vector mass flux is ρŪ with
units mass/ (area − time) . The rate of mass flow in through the face at x is the flux in
the x-direction times the area ρ U |x ∆y∆z . The mass flow rate out through the face at
x + ∆x is ρ U |x+∆x ∆y∆z . Similar expressions apply to the other two pairs of faces. The
rate of mass accumulation inside the volume is ∆x∆y∆z (∂ρ/∂t) and this is equal to the
sum of the mass fluxes over the six faces of the control volume. The mass balance (1.1) is
expressed mathematically as

∂ρ
∆x∆y∆z = ∆y∆zρ U |x − ∆y∆zρ U |x+∆x +
∂t (1.2)
∆x∆zρ V |y − ∆x∆zρ V |y+∆y + ∆x∆yρ W |z − ∆x∆yρ W |z+∆z
which can be rearranged to read


∂ρ
∆x∆y∆z + ∆y∆z ρ U |x+∆x − ρ U |x +
∂t . (1.3)

∆x∆z ρ V |y+∆y − ρ V |y + ∆x∆y ρ W |z+∆z − ρ W |z = 0
Divide (1.3) through by the infinitesimal volume ∆x∆y∆z .
∂ρ ρ U |x+∆x − ρ U |x ρ V |y+∆y − ρ V |y ρ W |z+∆z − ρ W |z

+ + + =0 (1.4)
∂t ∆x ∆y ∆z
Let ∆x → 0, ∆y → 0, ∆z → 0. In this limit (1.4) becomes
∂ρ ∂ρU ∂ρV ∂ρW

+ + + = 0. (1.5)
∂t ∂x ∂y ∂z
Equation (1.5) is called the continuity equation and is the general partial differential equa-
tion for conservation of mass for any moving, continuous medium (continuum). The con-
tinuum might be a compressible gas, a liquid or a moving solid such as glacier ice or the
rock crust of the Earth.
1.2.1 Incompressible flow
The compressibility of a medium becomes important when the speed of a body begins to
approach the speed of sound in the medium. At speeds much lower than the speed of sound
the disturbance created by a body pushing aside the fluid is too small to significantly change
the thermal energy of the fluid and fluid behaves as if it is incompressible ρ = constant .
In this limit (1.5) reduces to
∂U ∂V ∂W
+ + = 0. (1.6)
∂x ∂y ∂z
Note that equation (1.6) applies to both steady and unsteady incompressible flow.
1.2.2 Index notation and the Einstein convention
For convenience, vector components are often written with subscripts. This is called index
notation and one makes the following replacements.
(x, y, z) → (x1 , x2 , x3 )
(1.7)
(U, V, W ) → (U1 , U2 , U3 )
In index notation equation (1.5) is concisely written in the form
3
∂ρ X ∂ (ρUi )
+ =0 (1.8)
∂t ∂xi
i=1
where the subscript refers to the ith vector component.

Vector calculus is an essential tool for developing the equations that govern compressible
flow and summed products such as (1.8) arise often. Notice that the sum in (1.8) involves
a repeated index. The theory of relativity is another area where such sums arise often
and when Albert Einstein was developing the special and general theory he too recognized
that such sums always involve an index that is repeated twice but never three times or
more. In effect the presence of repeated indices implies a summation process and the
summation symbol can be dropped. To save effort and space Einstein did just that and
the understanding that a repeated index denotes a sum has been known as the Einstein
convention ever since. Using the Einstein convention (1.8) becomes
∂ρ ∂ (ρUi )
+ = 0. (1.9)
∂t ∂xi
Remember, the rule of thumb is that a single index denotes a vector component and a
repeated index represents a sum. Three or more indices the same means that there is a
mistake in the equation somewhere! The upper limit of the sum is 1, 2 or 3 depending on
the number of space dimensions in the problem. In the notation of vector calculus (1.9) is
written
∂ρ
+ ∇ · ρŪ = 0 (1.10)
∂t
and (1.6) is ∇· Ū = 0 . Vector notation has the advantage of being concise and independent
of the choice of coordinates but is somewhat abstract. The main advantage of index
notation is that it expresses precisely what differentiation and summation processes are
being done in a particular coordinate system. In Cartesian coordinates, the gradient vector
operator is

∂ ∂ ∂
∇≡ , , . (1.11)
∂x ∂y ∂z
The continuity equation as well as the rest of the equations of fluid flow are presented in
cylindrical and spherical polar coordinates in Appendix B.
1.3 Particle paths, streamlines and streaklines in 2-D steady

flow
Let’s begin with a study of fluid flow in two dimensions. Figure 1.2 shows the theoretically
computed flow over a planar, lifting airfoil in steady, inviscid (non-viscous) flow. The
theory used to determine the flow assumes that the flow is irrotational
∇ × Ū = 0 (1.12)
and that the flow speed is very low so that (1.6) holds.
∇ · Ū = 0 (1.13)
A vector field that satisfies (1.12) can always be represented as the gradient of a scalar
potential function φ = Φ (x, y) therefore
U = ∇Φ (1.14)
or, in terms of components

∂Φ ∂Φ
(U, V ) = , (1.15)
∂x ∂y
When (1.14) is substituted into (1.13) the result is Laplace’s equation.
∇ · ∇Φ = ∇2 Φ = 0 (1.16)
Figure 1.2 depicts the solution of equation (1.16) over the airfoil shown.
The boundary conditions on the airfoil are that the flow is allowed to slip tangent to the
surface but cannot penetrate the surface and the flow is required to leave the sharp trailing
edge of the airfoil smoothly. The latter requirement is the famous Kutta condition. In a
real airfoil viscous friction prevents the flow from going around the sharp trailing edge. By
enforcing the Kutta condition, the solution of (1.16) is able to mimic the effects of viscosity
in real airfoils thus allowing the lift to be determined. Once the scalar potential is found
Figure 1.2: Flow over a 2-D lifting wing, (a) streamlines, (b) streaklines.
the velocity field is generated using (1.14). The streamlines shown in Figure 1.2(a) are
simply lines that are everywhere parallel to the velocity vector field.
A careful examination of the picture reveals quite a bit of information. First, note that
the mass flow rate between any two streamlines is fixed (fluid cannot cross streamlines).
As the streamlines approach the leading edge of the airfoil, the separation between lines
increases indicating a deceleration of the fluid. Near the point of maximum wing thickness
on the upper surface, the streamlines come closer together indicating a speed-up of the
fluid to greater than the free stream speed. Below the airfoil the streamlines move apart
indicating that the fluid slows down to less than the free stream speed. Downstream of
the airfoil trailing edge the flow speed recovers to the free stream value. Notice that the
largest velocity changes occur near the wing. In the upper and lower parts of the picture,
far away from the wing, the flow is deflected upward by the wing but the distance between
streamlines changes little and the corresponding flow speed change is relatively small.
Figure 1.2 (b) depicts streaklines in the flow over the airfoil. These are produced numer-
ically the same way one would produce dye lines in a real flow. Fluid elements that pass
through a given point upstream of the airfoil are marked forming a streak. In the figure,
alternating bands of fluid are marked light and dark. The flow pattern produced by the
streaklines is identical to the streamline pattern but there is one added piece of informa-
tion. The streakline pattern depicts the time-integrated effect of the flow velocity on the
position of the fluid elements that constitute the streak.
One of the popular explanations of how an airfoil produces lift is that adjacent fluid el-
ements on either side of the stagnation streamline ahead of the wing must meet at the
trailing edge at the same time. The reasoning then goes as follows; since a particle that
travels above the wing must travel farther than the one below, it must travel faster thus
reducing the pressure over the wing and producing lift. Figure 1.2 (b) clearly shows that
this explanation is completely erroneous. The fluid below the wing is substantially retarded
compared to the fluid that passes above the wing.
The figures below show two more streamline examples in a slightly compressible situation
with the effects of viscosity included. These are computations of the flow over a wing flap
at a free stream speed of approximately 30 % of the speed of sound. The flow conditions
are the same in both cases except that in the right-hand picture the trailing edge of the
main wing (visible in the upper left corner of the picture) has attached to it a small vertical
flap called a Gurney flap.
Figure 1.3: Computed streamlines over a wing flap.
The Gurney flap was developed by Dan Gurney in an effort to improve the speed of his
new IRL race car prior to the start of the 1971 racing season in Phoenix. The car was
disappointingly slow and after several days of testing, the driver Bobby Unser challenged
Gurney to find a solution. Gurney decided to add a spoiler to the rear wing by riveting a
length of aluminum right-angle to the trailing edge. When Unser tested the car it was just
as slow as before and so everyone felt that the idea was a failure. Later Unser confided
to Gurney that the reason the car was slow was that the down-force from the flap on the
rear wheels was so large that the car was understeering badly. It was clear that with some
adjustments to increase the down-force on the front wings the car could be made much
faster. For a while the Gurney team deflected questions with the fiction that the purpose
of the flap was purely to increase the structural strength of the wing, but eventually the
competition got wise and the Gurney flap was widely adopted.
While the Gurney flap increases both the lift and drag of a single wing the effect on a two-
element airfoil is to cause the flow over the flap to reattach as shown in Figure 1.3 thereby
increasing lift and reducing the overall drag of the wing-flap system. This allows the system
to be effective at much higher flap angles. Figure 1.3 was produced using the computed
velocity field. Once the velocity field is known, the user places computational particles
at selected grid points and integrates the trajectory of the particles forward in time. The
choice of initial points depends on the amount of detail one wants to visualize and this
accounts for the uneven distribution of streamlines in these figures. The integration traces
out the trajectories of the particles. Since the flow is steady, this process effectively traces
out the streamlines. Changes in flow speed can be seen in terms of the convergence and
divergence of streamlines, just as in the incompressible case, although the variation of fluid
density over the field complicates this interpretation slightly. We shall return to this last
point shortly.
1.3.1 Analysis of particle paths and streamlines
Let’s learn how to analyze particle paths and streamlines theoretically. Figure 1.4 shows a
typical trajectory in space of a fluid element moving under the action of a two-dimensional
steady velocity field.
Figure 1.4: Particle trajectory in a 2-D steady flow field.
The equations that determine the trajectory are:
dx (t)
= U (x (t) , y (t))
dt
(1.17)
dy (t)
= V (x (t) , y (t))
dt
where U and V are the flow velocity components in the x and y directions respectively.
The velocity field is assumed to be a smooth function of position. Formally, these equations
are solved by integrating the velocity field forward in time.
Z t
x (t) = x0 + U (x (t) , y (t))dt
0
(1.18)
Z t
y (t) = y0 + V (x (t) , y (t))dt
0
The result is a set of parametric functions for the particle coordinates x and y in terms of
the time, t , along a particle path.
x (t) = F (x0 , y0 , t)
(1.19)
y (t) = G (x0 , y0 , t)
The solution of (1.17) can also be expressed as a family of lines derived by eliminating t
between the functions F and G in (1.19).
ψ = Ψ (x, y) (1.20)
Figure 1.5: Streamlines in steady flow. The value of a particular streamline is determined
by the coordinates of any point on the streamline.
This is essentially how the streamlines observed in Figure 1.2(a) and Figure 1.3 are gener-
ated. The value of a particular streamline is determined by the initial conditions.
ψ0 = Ψ (x0 , y0 ) (1.21)
This is the situation depicted schematically in Figure 1.5 .

The streaklines in Figure 1.2(b) were generated by shading a segment of fluid elements that
pass through an initial point (x0 , y0 ) during a fixed interval in time. Selecting a vertical
line of initial points well upstream of the airfoil leads to the bands shown in the figure.
The length of a segment is directly related to the velocity history of the fluid particles
that make up the segment. The stream function can also be determined as the solution of
the first order ordinary differential equation obtained by eliminating dt between the two
particle path equations in (1.17).
dy V (x, y)
= (1.22)
dx U (x, y)
The differential of Ψ (x, y) is
∂Ψ ∂Ψ
dψ = dx + dy. (1.23)
∂x ∂y
If we use (1.17) to replace the differentials dx and dy in (1.23) the result is

∂Ψ ∂Ψ
dψ = U (x, y) + V (x, y) dt. (1.24)
∂x ∂y
On a line of constant ψ = ψ0 the differential dψ = 0 and so for nonzero dt , the right hand
side of (1.24) can be zero only if the expression in parentheses is zero. Thus the stream
function, Ψ (x, y), can be determined in two ways; either as the solution of a linear, first
order PDE (1.25)
∂Ψ ∂Ψ
U · ∇Ψ = U (x, y) + V (x, y) =0 (1.25)
∂x ∂y
or as the solution of the ODE (1.22) which we can write in the form,
−V (x, y) dx + U (x, y) dy = 0. (1.26)
Equation (1.25) is the mathematical expression of the statement that streamlines are par-
allel to the velocity vector field.
1.3.2 The integrating factor
On a line of constant ψ = ψ0 the differential dψ = 0 and
∂Ψ ∂Ψ
dx + dy = 0. (1.27)
∂x ∂y
A very important question concerns the relationship between the differential expressions,
(1.26) and (1.27). Are they the same? In particular, can (1.26) be regarded as a perfect
differential? The test for a perfect differential is to compare cross derivatives. Equation
(1.26) is a perfect differential if and only if,
∂V ∂U
− = . (1.28)
∂y ∂x
If the integrability condition (1.28) is satisfied then there must exist some function Ψ (x, y)
such that V = −∂Ψ/∂x and U = −∂Ψ/∂y and the expression (1.26) can be equated to
the differential of this function dψ. In this case it is easy to solve for Ψ (x, y) by direct
integration of (1.26).
But for an arbitrary choice of the functions U (x, y) and V (x, y), the answer to the above
question is in general no! In other words, for general V and U
V ∂Ψ/∂x
=− (1.29)
U ∂Ψ/∂y
but
∂Ψ ∂Ψ
V 6= − U 6= . (1.30)
∂x ∂y
In order to convert equation (1.26) to a perfect differential it must be multiplied by an

integrating factor. In general there is no systematic method for finding the integrating
factor and so the analytic solution of (1.26) for general functions U and V remains a
difficult unsolved problem in mathematics. However we shall see later that in the case of
fluid flow the integrating factor can be identified using the equation for conservation of
mass.
It was shown by the German mathematician Johann Pfaff in the early 1800’s that an
integrating factor for the expression −V dx + U dy always exists. That is, for any choice of
smooth functions U and V , there always exists a function M (x, y) such that,
dψ = −M (x, y) V (x, y) dx + M (x, y) U (x, y) dy (1.31)
and the partial derivatives are
∂Ψ
= −M (x, y) V (x, y)
∂x
(1.32)
∂Ψ
= M (x, y) U (x, y)
∂y
A differential expression like −V dx+U dy is often called a Pfaffian form. In the language of
differential geometry it is called a differential 1-form. Pfaffian forms of higher dimension,
say Adx + Bdy + Cdz are often encountered in physics and with certain restrictions on
A (x, y, z), B (x, y, z) and C (x, y, z) an integrating factor can be found. But, the existence
of the integrating factor is only assured unconditionally in two dimensions.
Pfaff’s theorem can be understood by noting that the flow patterns formed by the vector
field with local slope V /U and the flow pattern with slope (M V ) / (M U ) are identical
whereas the local flow speed differs by the scale factor M (x, y) . Since the solution of
(1.17) clearly exists (we could just integrate the equations forward in time numerically)
then the integrating factor must also exist.
Perfect differentials and integrating factors are generally covered in an upper level under-
graduate course in calculus. Unfortunately the presentation is often cursory and tends to
get passed over fairly quickly, often without making the deep connection that exists to
the analysis of physical problems. In this chapter we see the connection between perfect
differentials and the theory of fluid flow. In Chapter 2 we will find that the concept of a
perfect differential is one of the fundamental building blocks needed to develop the laws of
thermodynamics. The following sample problem is worked out in quite a bit of detail to
help strengthen your understanding.
1.3.3 Sample problem - the integrating factor and reconstruction of a

function from its perfect differential
Solve the first order ordinary differential equation
dy y
= H (xy) (1.33)
dx x
where H (x, y) is an arbitrary smooth function. Rearrange (1.33) to read

−yH (xy) dx + xdy = 0 (1.34)
This differential expression fails the cross derivative test
∂ (−yH (xy)) ∂ (x)

6= (1.35)
∂y ∂x
and is not a perfect differential. The integrating factor for (1.35) is known to be
1
M= (1.36)
xy + xyH (xy)
Multiplying (1.34) by (1.36) converts it to a perfect differential.
yH (xy) x
dψ = − dx + dy (1.37)
xy + xyH (xy) xy + xyH (xy)
We now know the partial derivatives of the solution.
∂Ψ yH (xy)
=−
∂x xy + xyH (xy)
(1.38)
∂Ψ x
=
∂y xy + xyH (xy)
The expression (1.37) can now be integrated by integrating either partial derivative with
the appropriate variable held fixed. Let’s start by integrating the partial derivative with
respect to y.
Z
xdy
ψ= + f (x) (1.39)
xy + xyH (xy) x−constant
Note that we have to include a constant of integration that can be at most a function of
x. Let α = xy . The solution can now be expressed as
Z
dα
ψ= + f (x) (1.40)
α + αH (α)
Now differentiate (1.40) with respect to x and equate to the known partial derivative with
respect to x in (1.38).
Z
∂ dα y df
+ f (x) = + =
∂x α + αH (xy) xy + xyH (xy) dx
(1.41)
yH (xy) df 1
− → = − → f = − ln (x)
xy + xyH (xy) dx x
This step allows us to determine f (x). Finally the exact solution of (1.33) is
Z
dα
ψ= − ln (x) (1.42)
α + αH (α)
Note that we had to know the integrating factor in order to solve the problem. Sometimes
what appears to be a trick will work. For example you can also get to the solution (1.42)
by simply defining a new variable α = xy. This allows (1.33) to be solved by separation of
variables. All such tricks are equivalent to finding the integrating factor.
In the general case the required integrating factor is not known. For example the equa-
tion
dy xy + y + y 2
= (1.43)
dx 2x2 − xy
comes up in the study of viscous boundary layer flow along a flat plate. We know an inte-
grating factor exists but it has never been found. Fortunately when it comes to streamline
patterns in two dimensions, conservation of mass can be used to determine the integrating
factor.
1.3.4 Incompressible flow in two dimensions
The flow of an incompressible fluid in 2-D is constrained by the continuity equation (1.6)
that in two dimensions reduces to
∂U ∂V
+ =0 (1.44)
∂x ∂y
This is exactly the condition (1.28) and so in this case the stream function and the velocity
field are related by
∂Ψ ∂Ψ
U= V =− . (1.45)
∂y ∂x
So the 1-form, −V dx+U dy is guaranteed to be a perfect differential; The integrating factor

is one. Therefore one can always write for 2-D incompressible flow
dψ = −V dx + U dy. (1.46)
If U and V are known functions, the stream function is determined analytically by integra-
tion of (1.46). In practice the easiest way to assign values to the streamlines is to integrate
(1.46) beginning at some reference state well upstream of the airfoil where the velocity is
uniform (U, V ) = (U∞ , 0) .
Incompressible irrotational flow in two dimensions
If the flow is incompressible and irrotational then the velocity field is described by both
a stream function (1.45) and a potential function (1.15). The two functions are related
by
∂Ψ ∂Φ
=
∂y ∂x
. (1.47)
∂Ψ ∂Φ
− =
∂x ∂y
These are the well known Cauchy-Riemann equations from the theory of complex variables.
Solutions of Laplace’s equation (1.16) which is the equation of motion for this class of flows
can be determined using the powerful methods of complex analysis. Interestingly, the flow
can be solved just from the continuity equation (1.13) without the use of the equation for
conservation of momentum. The irrotationality condition (1.12) essentially supplants the
need for the momentum equation.
1.3.5 Compressible flow in two dimensions
The continuity equation for the steady flow of a compressible fluid in two dimensions
is
∂ (ρU ) ∂ (ρV )
+ = 0. (1.48)
∂x ∂y
In this case, the required integrating factor for (1.26) is the density ρ (x, y) and we can
write
dψ = −ρV dx + ρU dy. (1.49)
The stream function in a compressible flow is proportional to the mass flux with units
mass/(area − sec) and the convergence and divergence of lines in the flow over the flap
shown in Figure 1.2 is a reflection of variations in mass flux over different parts of the flow
field.
1.4 Particle paths in three dimensions
Figure 1.6 shows the trajectory in space traced out by a particle under the action of a
general three-dimensional, unsteady flow, Ū (x̄, t).
Figure 1.6: Particle trajectory in three dimensions.
Using index notation, the equations governing the motion of the particle are:
dxi (t)
= Ui (x1 (t) , x2 (t) , x3 (t) , t) i = 1, 2, 3. (1.50)
dt
Formally, these equations are solved by integrating the velocity field.
Z t
xi (t) = xi0 + Ui (x1 (t) , x2 (t) , x3 (t) , t) dt i = 1, 2, 3 (1.51)
0
We shall return to the discussion of flow patterns in both 2 and 3 dimensions when we
discuss the kinematics of flow fields in Chapter 4.
1.5 The substantial derivative
The acceleration of a particle is
d2 xi (t) d ∂Ui ∂Ui dxk

2
= Ui (x1 (t) , x2 (t) , x3 (t) , t) = + . (1.52)
dt dt ∂t ∂xk dt
Remember, according to the Einstein convention, the repeated index denotes a sum over
k = 1, 2, 3. Use (1.50) to replace dxk /dt by Uk in (1.52). The result is called the substantial
or material derivative and is usually denoted by D () /Dt.
D () ∂ ()
= + Ū · ∇ () (1.53)
Dt ∂t
The substantial derivative of the velocity is
DUi ∂Ui ∂Ui

= + Uk . (1.54)
Dt ∂t ∂xk
The time derivative of any flow variable evaluated on a fluid element is given by a similar
formula. For example the time rate of change of the density ρ (x (t) , y (t) , z (t) , t) of a
given fluid particle is:
Dρ ∂ρ ∂ρ
= + Uk . (1.55)
Dt ∂t ∂xk
The substantial derivative is the time derivative of some property of a fluid element referred
to a fixed frame of reference within which the fluid element moves as shown in Figure
1.6.
1.5.1 Frames of reference
Occasionally it is necessary to transform variables between a fixed and moving set of

coordinates as shown in Figure 1.7. The transformation of position and velocity is
x0 = x − X (t)
y 0 = y − Y (t)
z 0 = z − Z (t)
(1.56)
U 0 = U − Ẋ (t)
V 0 = V − Ẏ (t)
W 0 = W − Ż (t) .
where X̄ = (X (t) , Y (t) , Z

(t)) is the displacement
of the moving frame in three coordinate
directions and dX̄/dt = Ẋ (t) , Ẏ (t) , Ż (t) is the velocity of the frame. Note that in
general, the frame may be accelerating.
Figure 1.7: Fixed and moving frames of reference.
The substantial derivative of some property of a fluid element, (1.53), is sometimes referred
to as the derivative moving with the fluid element. Do not interpret this as a transformation
to a coordinate system that is attached to the fluid element. If for some reason it is actually
desirable to make such a transformation the velocity of the particle in that frame is zero and
the substantial derivative of the velocity is also zero. The substantial derivative of a variable
such as the density of the particle at the origin of the moving coordinates is Dρ/Dt = ∂ρ/∂t.
Both the momentum and kinetic energy of a fluid particle of mass m depend on the frame
of reference. The momentum transforms linearly like the velocity.
mŪ 0 = mŪ − mdX̄/dt (1.57)
The transformation of kinetic energy is a little more complex. In the two coordinate
systems the definition of kinetic energy is the same,
1
kinetic energy in moving coordinates = m U 02 + V 02 + W 02

(1.58)
2
and
1
kinetic energy in f ixed coordinates = m U 2 + V 2 + W 2 .

(1.59)
2
The transformation of kinetic energy between frames is

2
1 02 02 02
1 2 2
m U +V +W = m U − Ẋ + V − Ẏ + W − Ż (1.60)
2 2
which, when expanded to show the kinetic energies explicitly, becomes
1 1
m U 02 + V 02 + W 02 = m U 2 + V 2 + W 2 +

2 2 (1.61)
1 1 1
mẊ Ẋ − 2U + mẎ Ẏ − 2V + mŻ Ż − 2U
2 2 2
or
1 1 1
k 0 = k + mẊ Ẋ − 2U + mẎ Ẏ − 2V + mŻ Ż − 2W . (1.62)
2 2 2
The transformation of kinetic energy depends nonlinearly on the velocity of the moving
coordinate system. In contrast thermodynamic properties such as density, temperature
and pressure are intrinsic properties of a given fluid element and so do not depend on the
frame of reference.
1.6 Momentum transport due to convection
The law of conservation of momentum is stated as
 
 rate of momentum 
accumulation inside =
the control volume
 
      . (1.63)

 rate of 
 
 rate of 
 
 sum of 

momentum momentum f orces acting
     
− +

 f low into 
  f low out of   on the 
  
  
the control volume the control volume control volume
   
As a fluid moves it carries its momentum with it. This is called convective momentum
transport. To study this kind of momentum transfer we use the same stationary control
volume element ∆x∆y∆z that we used to develop the continuity equation. As before,
the fluid velocity vector has components (U, V, W ) in the (x, y, z) directions and the fluid
density is ρ.
Figure 1.8: Fluxes of x-momentum through a fixed control volume. Arrows denote the
velocity component carrying momentum into or out of the control volume.
Figure 1.8 shows the contribution to the x-momentum inside the control volume from
the movement of fluid through all six faces of the control volume. Consider a pair of faces
perpendicular to the x-axis. The flux of x-momentum is ρU U with units momentum/(area-
time). The rate at which the x-component of momentum enters the face at x is ρU U |x ∆y∆z
and the rate at which it leaves through the face at x + ∆x is ρU U |x+∆x ∆y∆z.
The rate at which the x-component of momentum enters the face at y is ρU V |y ∆x∆z and
the rate at which it leaves through the face at y + ∆y is ρU V |y+∆y ∆x∆z. To understand
how fluid motion in the y-direction contributes to the rate of x-momentum transfer into
or out of the control volume one can make the following interpretation of the momentum
transfer.

x − momentum
ρU =
per unit volume
 

 volume of f luid per second 

passing into the control volume through
 
V |y ∆x∆z =
 the f ace normal to the V − velocity
(1.64)
 

component at position y
 
 

 volume of f luid per second 

passing out of the control volume through
 
V |y+∆y ∆x∆z = .
 the f ace normal to the V − velocity
 

component at position y + ∆y
 
Based on similar considerations, the rate at which the x-component of momentum enters
the face at z is ρU W |z ∆x∆y and the rate at which it leaves through the face at z + ∆z is
ρU W |z+∆z ∆x∆y. The rate of x-component of momentum accumulation inside the control
volume is ∆x∆y∆z (∂ρU/∂t). For the x-component of momentum, the balance (1.63) is
expressed mathematically as

∂ρU
∆x∆y∆z = ∆y∆zρ U U |x − ∆y∆zρ U U |x+∆x +
∂t
(1.65)
∆x∆zρU V |y − ∆x∆zρU V |y+∆y + ∆x∆yρ U W |z − ∆x∆yρU W |z+∆z +
{the sum of x − component f orces acting on the system}
which is rearranged to read

∂ρU
∆x∆y∆z + ∆y∆z ρ U U |x+∆x − ρ U U |x +
∂t
(1.66)

∆x∆z ρU V |y+∆y − ρU V |y + ∆x∆y ρ U W |z+∆z − ρU W |z =
{the sum of x − component f orces acting on the system} .
Divide (1.66) through by the volume of the control element ∆x∆y∆z.

∂ρU ρ U U |x+∆x − ρ U U |x ρU V |y+∆y − ρU V |y ρ U W |z+∆z − ρU W |z

+ + + =
∂t ∆x ∆y ∆z
{the sum of x − component f orces acting on the system per unit volume}
(1.67)
Let (∆x → 0, ∆y → 0, ∆z → 0). In this limit (1.67) becomes
 

 the sum of 

∂ρU ∂ρU U ∂ρU V ∂ρU W x − component f orces
 
+ + + = . (1.68)
∂t ∂x ∂y ∂z 
 per unit volume acting 

on the control volume
 
Similar equations describe conservation of momentum in the y and z directions.
 

 the sum of 

∂ρV ∂ρV U ∂ρV V ∂ρV W y − component f orces
 
+ + + =
∂t ∂x ∂y ∂z 

 
  (1.69)

 the sum of 

∂ρW ∂ρW U ∂ρW V ∂ρW W z − component f orces
 
+ + + =
∂t ∂x ∂y ∂z  per unit volume acting
 

 
Equations (1.68) and (1.69) are the components of the vector partial differential equation
that describes conservation of momentum for any moving continuum. In index notation,
equations (1.68) and (1.69) are concisely written in the form
 
 the sum of
 

∂ρUi ∂ (ρUi Uj )  ith − component f orces 
+ = i = 1, 2, 3 (1.70)
∂t ∂xj 

 
where the subscript refers to the ith vector component and the sum is over the repeated
index j. Since the repeated index is always summed it makes no difference what symbol
is used and so it is called a dummy index. Rearrange (1.70) by carrying out the indicated
differentiation.

∂Ui ∂ (Ui ) ∂ρ ∂ (ρUj )
ρ + ρUj + Ui + =
∂t ∂xj ∂t ∂xj
 

 the sum of 
 (1.71)
ith − component f orces
 
i = 1, 2, 3


 
The term in parentheses on the left hand side of (1.71) is zero by continuity and so the
momentum equation becomes
 
 the sum of
 

DUi  ith − component f orces 
ρ = i = 1, 2, 3. (1.72)
Dt 

 
Equation (1.72) can be stated in words as follows.
 
 the vector sum of
the rate of momentum change

= f orces acting (1.73)
of a f luid element
on the f luid element
 
1.7 Momentum transport due to molecular motion at mi-

croscopic scales
Appendix A describes the molecular origin of two of the most basic forces that act on
a fluid element; pressure and viscous friction. Although the derivations are restricted to
gases, both types of forces exist in all fluid motion and both arise from momentum transfer
due to the motion and collisions of the molecules that make up the fluid. Molecules in
the fluid move in random and chaotic ways and across any selected section of a surface
within the fluid, there are molecules moving in both directions through the surface. The
pressure and viscosity are defined through an average of the momentum transfer due to the
collective motion of a very large set of molecules over a time scale that is short compared
to the unsteady macroscopic motion of the fluid.
1.7.1 Pressure
Pressure forces arise from collisions between the molecules within the fluid and with what-
ever form of containment is used to hold or bound the fluid. In a fluid at rest, in the
absence of gravitational forces, the pressure is uniform everywhere. When the fluid is set
into motion the pressure becomes a scalar function of space and time. For example, vari-
ations in the pressure over a wing give rise to the lift of the wing. Pressure variations
in the atmosphere are the primary driving forces behind wind patterns. The drag of a
motor vehicle is primarily due to variations in the pressure field over the surface of the
vehicle.
1.7.2 Viscous friction - plane Couette flow
All simple fluids exhibit resistance to flow due to the physical property of viscosity. The
effects of viscosity are readily apparent in many everyday activities such as the spreading
of honey on a piece of bread or the slow dripping of paint from a brush. Anyone who has
purchased oil for their car knows that different grades of oil are characterized by different
viscosities and that the viscosity depends on the oil temperature. The viscosity of air
and other gases is less apparent but nonetheless just as real. The most important effect
of viscosity on the motion of a fluid is that it causes the fluid to stick to the surface of
a moving solid body, imposing the so-called no-slip condition on the velocity field at the
surface. Even though the viscosity of air may be very small, viscous forces profoundly
affect the flow and are critical to the generation of lift and drag by moving bodies.
To illustrate viscous friction, consider a fluid contained between two large parallel plates
with area A as shown in Figure 1.9. The fluid is initially at rest. At t = 0 the upper plate
is set into motion and, due to viscous forces, the adjacent fluid layers are dragged along
with it. Eventually the velocity field reaches a steady state. If the speed of the upper plate
is small compared to the speed of sound in the fluid, the final velocity profile is a straight
line as shown. This is called plane Couette flow after Maurice Frederic Alfred Couette
a Professor of Physics at the French University of Angers at the end of the nineteenth
century. Couette actually studied the flow between inner and outer concentric rotating
cylinders with the goal of using his device to measure viscosity.
Once the flow reaches steady state a constant force F is required to maintain the motion
of the upper plate. Provided the flow is laminar and the plate speed is not too large the
force may be expressed as
F U∞
=µ . (1.74)
A d
The viscous stress on the plate f orce/area is proportional to the velocity gradient and the
constant of proportionality is the viscosity of the fluid. Later we will generalize the notion
of viscous stress to flow in three dimensions and so it is useful to write (1.74) in a somewhat
Figure 1.9: Build-up to a steady laminar velocity profile for a viscous fluid contained between
two parallel plates. At t=0 the upper plate is set into motion at a constant speed U∞ .
more general form. The shear stress acting in the x direction exerted on a layer of fluid
with an outward unit normal vector n in the positive y direction (as shown in Figure 1.9)
is designated τxy . In terms of the velocity derivative (1.74) is written,
dU
τxy = µ (1.75)
dy
Equation (1.75) is called Newton’s law of viscosity after Sir Isaac Newton who first proposed
it. In Chapter 8 we will consider the compressible case where the plate speed is not small
and the steady-state velocity profile deviates from a straight line.
1.7.3 A question of signs
There is a subtle sign issue that comes up when one defines the viscous stress. In the
discussion just above we have adopted the convention normally used in aeronautics that
the stress is the force on a surface defined by its outward normal vector. In Figure 1.9
the outward normal on the lower surface is positive and the force is in the positive x-
direction; on the upper surface the force is in the negative x-direction on a surface defined
by an outward normal that points in the negative y-direction. Either way τxy is positive
as indicated in (1.75).
But in certain fields such as chemical engineering a different interpretation of the stress is
used. Here the stress is viewed as the transport of momentum in a certain direction. In this
view the flow in Figure 1.9 would be interpreted as the transport of positive x-momentum
in the negative y-direction and the expression in (1.75) would have a negative sign. If the
general momentum conservation equation is developed with this interpretation of stress
then another negative sign appears in front of the stress term. When the constitutive
relation between stress and rate-of- strain is introduced both camps wind up solving the
same system of equations, as they must.
1.7.4 Newtonian fluids
A fluid for which the viscous stress is linearly proportional to the velocity gradient is called
Newtonian. Virtually all gases and most liquids are Newtonian. The few exceptions include
polymers, pastes and slurries that exhibit non-Newtonian behavior where the stress depends
nonlinearly on the rate of strain and may exhibit dependence on the total strain as in a solid.
Some fluids are said to have memory where the stress can depend on the past history of the
rate-of-strain. We will introduce the Newtonian stress-rate-of-strain relation in Chapter
5. For the present the stress is left in general terms. A truly exceptional case is liquid
Helium that, when cooled below 2.3 K undergoes a transition to a strange macroscopic
quantum state called Helium II. A sample of Helium below this temperature consists of
two fluids that appear to co-exist. The normal fluid component exhibits viscosity whereas
the superfluid component appears to flow without any viscous friction. In fact Helium
II is an extraordinarily difficult fluid to contain since the superfluid component can pass
through ultra-fine leaks in the dewar used to hold it. Exceptions to the no-slip condition
occur in the flow of highly rarefied gases such as in the upper reaches of the atmosphere
during re-entry of a spacecraft. When the density of the gas is extremely low the fluid can
no longer be treated as a continuous medium. In this case the flow must be treated by
modeling the motion and collisions of individual molecules. Near a wall molecular collisions
may not be perfectly specular and may on average exhibit a small difference between the
velocity of the surface and the mean velocity of fluid particles near the surface. This is
called the slip velocity and can become quite large as the density of the gas decreases to
extremely low values.
1.7.5 Forces acting on a fluid element
At macroscopic scales the transport of momentum due to the net molecular motion of
a very large number of molecules is equivalent to continuous pressure and viscous forces
acting within the fluid. Figure 1.10 shows the contribution to the x-momentum inside our
small control volume from the pressure and viscous stresses acting on the six faces of the
control volume.
Figure 1.10: Pressure and viscous stresses acting in the x-direction.
Consider a pair of faces perpendicular to the x-axis in Figure 1.10. The force on the face at
x with outward normal in the negative x-direction is − (−P + τxx )|x ∆y∆z and the force on
the face at x+∆x with outward normal in the positive x-direction is (−P + τxx )|x+∆x ∆y∆z.
The x-component of force on the face at y due to viscous stress is − τxy |y ∆x∆z and the
x-component of force on the face at y + ∆y is τxy |y+∆y ∆x∆z. Finally, the x-component of
force on the face at z is −τxz |z ∆y∆z and the x-component of force on the face at z + ∆z
is τxz |z+∆z ∆y∆z. Note that the pressure is a scalar that acts normal to any surface in the
fluid and therefore there is no component of the pressure acting in the x-direction on the
faces perpendicular to the y and z directions. The pressure- viscous stress forces acting in
all three directions on the control volume are

Fx = ∆y∆z (−P + τxx )|x+∆x − (−P + τxx )|x + ∆x∆z τxy |y+∆y − τxy |y +

∆x∆y τxz |z+∆z − τxz |z

Fy = ∆y∆z τxy |x+∆x − τxy |x + ∆x∆z (−P + τyy )|y+∆y − (−P + τyy )|y +
. (1.76)
∆x∆y τyz |z+∆z − τyz |z

Fz = ∆y∆z τxz |x+∆x − τxz |x + ∆x∆z τyz |y+∆y − τyz |y +

∆x∆y (−P + τzz )|z+∆z − (−P + τzz )|z
With convection, pressure and viscous forces included, the x-component momentum bal-
ance over the control volume is, from (1.65)

∂ρU
∆x∆y∆z = ∆y∆z ρ U U |x − ρ U U |x+∆x +
∂t

∆x∆z ρU V |y − ρU V |y+∆y + ∆x∆y ρ U W |z − ρU W |z+∆z +
. (1.77)

∆y∆z (−P + τxx )|x+∆x − (−P + τxx )|x +

∆x∆z τxy |y+∆y − τxy |y + ∆x∆y τxz |z+∆z − τxz |z
Divide (1.77) through by ∆x∆y∆z and move the convective, pressure and viscous stress
terms to the left hand side. The result is
∂ρU ρ U U |x+∆x − ρ U U |x + (P − τxx )|x+∆x − (P − τxx )|x

+ +
∂t ∆x
.
ρU V |y+∆y − ρU V |y − τxy |y+∆y − τxy |y

ρ U W |z+∆z − ρU W |z − τxz |z+∆z − τxz |z
+ =0
∆y ∆z
(1.78)
Let ∆x → 0, ∆y → 0, ∆z → 0. In this limit (1.78) becomes
∂ρU ∂ (ρU U + P − τxx ) ∂ (ρU V − τxy ) ∂ (ρU W − τxz )

+ + + = 0. (1.79)
∂t ∂x ∂y ∂z
The equations for conservation of momentum in the y and z directions are derived in a
similar way using the expressions in (1.69) and (1.76).
∂ρV ∂ (ρV U − τxy ) ∂ (ρV V + P − τyy ) ∂ (ρV W − τyz )

+ + + =0
∂t ∂x ∂y ∂z
(1.80)
∂ρW ∂ (ρW U − τxz ) ∂ (ρW V − τyz ) ∂ (ρW W + P − τzz )
+ + + =0
∂t ∂x ∂y ∂z
In index form, using the Einstein convention, the momentum conservation equation is
∂ρUi ∂ (ρUi Uj + P δij − τij )

+ =0 i = 1, 2, 3. (1.81)
∂t ∂xj
The coordinate independent form of (1.81) is
∂ρŪ
+ ∇ · ρŪ Ū + P I¯ − τ̄¯ = 0. (1.82)
∂t
The tensor product of vectors that appears in (1.81), Ū Ū = Ui Uj , also called a dyadic
product, is sometimes written Ū ⊗ Ū . Fully written out, this product is
 
UU UV UW
Ū Ū =  V U VV V W . (1.83)
WU WV WW
Note that the matrix (1.83) is symmetric and as we shall see in Chapter 5, so is the stress
tensor τij .
1.8 Conservation of energy
The law of conservation of energy is stated as

   

 rate of energy 
 
 rate of energyf low 

accumulation into the
   
= −

 inside the control 
  control volume 
  
volume by convection
  
   

 rate of energy f low 
 
 work done on the 

out of the control volume . (1.84)
   
+ +

 control volume 
  by pressure and 
  
by convection viscous f orces
  
 
 energy generation
rate of energy addition

+ due to sources inside
due to heat conduction
the control volume
 
The energy per unit mass of a moving continuum is e + k where e is the internal energy
per unit mass and
k = (1/2) U 2 + V 2 + W 2

(1.85)
is the kinetic energy per unit mass. The convection of energy into and out of the control
volume is depicted in Figure 2.11.
Figure 1.11: Convection of energy into and out of a control volume.

1.8.1 Pressure and viscous stress work
The forces that act on the surface of a control volume do work on the fluid in the control
volume and therefore contribute to the energy balance. The power (energy per second)
put into the fluid within the control volume is given by the classical relation from mechan-
ics
P ower input to the control volume = F̄ · Ū (1.86)
where Ū is the flow velocity and F̄ is the net vector force on the control volume by pressure
and viscous forces with components (1.76). Fully written out (1.86) is
P ower input to the control volume = F̄ · Ū =

(−P + τxx )|x+∆x− (−P + τxx )|x U +
∆y∆z +
τxy |x+∆x − τxy |x V + τxz |x+∆x − τxz |x W
 
 τxy | − τ xy | U + (−P + τyy )| − (−P + τyy )| V +  (1.87)
y+∆y y y+∆y y
∆x∆z +
 τyz | − τyz | W 
y+∆y y

τxz |z+∆z − τxz |z U + τyz |z+∆z − τyz |z V +
∆x∆y .
(−P + τzz )|z+∆z − (−P + τzz )|z W
The right-hand-side of (1.87) can be parsed into a series of energy fluxes into and out of
the various faces of the control volume.
P ower input to the control volume = F̄ · Ū =

(−P U + τxx U + τxy V + τxz W )|x+∆x −
∆y∆z +
(−P U + τxx U + τxy V + τxz W )|x
( ) (1.88)
(−P V + τxy U + τyy V + τyz W )|y+∆y −
∆x∆z +
(−P V + τxy U + τyy V + τyz W )|y

(−P W + τzx U + τzy V + τzz W )|z+∆z −
∆x∆y
(−P W + τzx U + τzy V + τzz W )|z
The contribution of these fluxes to the energy balance over the control volume is illustrated
in Figure 2.12.
Figure 1.12: Energy fluxes due to the work done on the control volume by pressure and
viscous forces.
In addition to the energy input due to work done by pressure and viscous forces one
must also include the flux of heat Q into and out of the control volume due to thermal
conduction. The rate of internal and kinetic energy accumulation inside the control volume
is ∆x∆y∆z (∂ρ (e + k) /∂t). With all the various sources of energy taken into account, the
energy balance (1.84) is expressed mathematically as
∂ρ (e + k)
∆x∆y∆z =
∂t

∆y∆z ρ (e + k) U |x − ρ (e + k) U |x+∆x +

∆x∆z ρ (e + k) V |y − ρ (e + k) V |y+∆y +

∆x∆y ρ (e + k) W |z − ρ (e + k) W |z+∆z +

(−P + τxx )|x+∆x− (−P + τxx )|x U +
∆y∆z +
τxy |x+∆x − τxy |x V + τxz |x+∆x − τxz |x W
  (1.89)
 τxy | − τ |
xy y U + (−P + τ )|
yy y+∆y − (−P + τ )|
yy y V + 
y+∆y
∆x∆z +
 τyz |
y+∆y − τyz |y W


τxz |z+∆z − τxz |z U + τyz |z+∆z − τyz |z V +
∆x∆y +
(−P + τzz )|z+∆z − (−P + τzz )|z W

∆y∆z Qx |x − Qx |x+∆x + ∆x∆z Qy |y − Qy |y+∆y +

∆x∆y Qz |z − Qz |z+∆z +
{power generation due to energy sources inside the control volume} .
Divide (1.89) through by the infinitesimal volume ∆x∆y∆z and take the limit ∆x →
0, ∆y → 0, ∆z → 0. The conservation equation for the energy becomes
∂ρ (e + k)
+
∂t
∂ (ρ (e + k) U + P U − τxx U − τxy V − τxz W + Qx )
+
∂x
∂ (ρ (e + k) V + P V − τyx U − τyy V − τyz W + Qy ) . (1.90)
+
∂y
∂ (ρ (e + k) W + P W − τzx U − τzy V − τzz W + Qz )

=
∂z
{power generation due to energy sources inside the control volume}
In index form the energy conservation equation is
∂ρ (e + k) ∂ (ρ (e + k) Uj + P Uj − Ui τij + Qj )
+ =
∂t ∂xj
(1.91)
{power generation due to energy sources inside the control volume}
and the coordinate independent form of (1.91) is
∂ρ (e + k)
+ ∇ · ρ (e + k) Ū + P Ū − τ̄¯ · Ū + Q̄ =
∂t (1.92)
{power generation due to energy sources inside the control volume} .
Power sources within the control volume can take on a wide variety of forms including
chemical reactions, phase change, light absorption from an external source, nuclear reac-
tions, etc.
1.9 Summary - the equations of motion
The combined conservation equations in Cartesian coordinates are

Conservation of mass

+ + + =0
∂t ∂x ∂y ∂y
Conservation of momentum

+ + + =0
∂t ∂x ∂y ∂z

+ + + =0
∂t ∂x ∂y ∂z

+ + + =0
∂t ∂x ∂y ∂z
(1.93)
Conservation of energy
∂ρ (e + k)
+
∂t
∂x
∂ (ρ (e + k) V + P V − τyx U − τyy V − τyz W + Qy )
∂y

=
∂z
{P ower generation due to energy sources inside the control volume} .
See Appendix B for the equations of motion in cylindrical and spherical polar coordinates.
In their current form (1.93) the equations cannot be solved until there is an additional
equation that relates the viscous stresses to the velocity field. In Chapter 5 the general
form of the Newtonian constitutive law (1.75) will be defined.
Also required is a way of relating the density and pressure that appear in the momentum
equation to the internal energy that appears in the energy equation. This will come from
thermodynamics and the definition of an equation of state. The heat flux in the energy
equation will be related to the fluid temperature using Fick’s law for linear heat conduction.
Inclusion of these relations leads to a closed system of governing equations for viscous
compressible flow where the number of equations is equal to the number of unknowns.
Under certain flow assumptions that are of considerable aeronautical interest it will be
possible to directly relate the density and pressure without having to know the temperature.
In this case the momentum and mass conservation equations become a closed system that
is sufficient to fully describe the flow. The two most common cases where this occurs
are incompressible flow where the density is constant and isentropic compressible flow
where the pressure is related to the density by a power law. Some of the tools of vector
calculus will be established in Chapter 3. We will then return to the equations of motion
in Chapter 5 and re-derive them using a much more general approach based on methods
for differentiation of integrals over an arbitrarily shaped control volume.
1.10 Problems
Problem 1 - Show that the continuity equation can be expressed as
1 Dρ ∂Uj
+ =0 (1.94)
ρ Dt ∂xj
Problem 2 - Use direct measurements from the streamlines in Figure 2.13 to estimate the
percent change from the free stream velocity at points A , B , C and D.
Figure 1.13: Streamlines about a wing in potential flow.
Problem 3 - The general, first order, linear ODE

dy
= −g (x) y + f (x) (1.95)
dx
can be written as the differential form
(g (x) y − f (x)) dx + dy = 0 (1.96)
Show that (1.96) can be converted to a perfect differential by multiplying by the integrating
factor.
R
g(x)dx
M =e (1.97)
Work out the solution of (1.95) in terms of integrals. What is the solution for the case
g = sin (x), f = cos (x)? Sketch the corresponding streamline pattern.
Problem 4 - Solve
∂Ψ x ∂Ψ
y − = 0. (1.98)
∂x 3 ∂y
Sketch the resulting streamline pattern.

Problem 5 - Show that the following expression is a perfect differential.
− sin (x) sin (y) dx + cos (x) cos (y) dy = 0 (1.99)
Integrate (1.99) to determine the stream function and sketch the corresponding flow pat-
tern. Work out the substantial derivatives of the velocity components and sketch the
acceleration vector field.
Problem 6 - Determine the acceleration of a particle in the 1-D velocity field
x
Ū = k , 0, 0 (1.100)
t
where k is constant.
Problem 7 - In a fixed frame of reference a fluid element has the velocity components
(U, V, W ) = (100, 60, 175) meters/ sec . (1.101)

Suppose the same fluid element is observed in a frame of reference moving at
X̄˙ = (25, 110, 90) meters/ sec (1.102)
with respect to the fixed frame. Determine the velocity components measured by the ob-
server in the moving frame. Determine the kinetic energy per unit mass in each frame.
Problem 8 - The stream function of a steady, 2-D compressible flow in a corner is shown
in Figure 2.14.
Figure 1.14: Streamlines for potential flow in a corner.
Determine plausible expressions for the velocity components and density field. Does a
pressure field exist for this flow if it is assumed to be inviscid?
Problem 9 - The expansion into vacuum of a spherical cloud of a monatomic gas such as
helium has a well-known exact solution of the equations for compressible isentropic flow.
The velocity field is
xt yt zt
U= V = W = . (1.103)
t20 + t2 t20 + t2 t20 + t2
The density and pressure are
3/2
t30 t20 x2 + y 2 + z 2

ρ
= 1− 2 t20 + t2
ρ0 3/2
t20 + t2 Rinitial
(1.104)
5/3
P ρ
=
P0 ρ0
where Rinitial is the initial radius of the cloud. This problem has served as a model of the
expanding gas nebula from an exploding star.
1) Determine the particle paths (x(t), y(t), z(t)).
2) Work out the substantial derivative of the density Dρ/Dt.

Problem 10 - A moving fluid contains a passive non-diffusing scalar contaminant. Smoke
in a wind tunnel would be a pretty good example of such a contaminant. Let the concen-
tration of the contaminant be C (x, y, z, t). The units of C are
mass of contaminant/unit mass of f luid. (1.105)
Derive a conservation equation for C.

Problem 11 - Include the effects of gravity in the equations of motion (1.93). You can
check your answer with the equations derived in Chapter 5.
Chapter 2
Thermodynamics of dilute gases
2.1 Introduction
The motion of a compressible fluid is directly affected by its thermodynamic state which
is itself a consequence of the motion. For this reason, the powerful principles of thermody-
namics embodied in the first and second law are a central part of the theory of compressible
flow. Thermodynamics derives its power from the fact that the change in the state of a
fluid is independent of the actual physical process by which the change is achieved. This
enables the first and second laws to be combined to produce the famous Gibbs equation
which is stated exclusively in terms of perfect differentials of the type that we studied
in Chapter1. The importance of this point cannot be overstated and the reduction of
problems to integrable perfect differential form will be a recurring theme throughout the
course.
2.2 Thermodynamics
Thermodynamics is the science that deals with the laws that govern the relationship be-
tween temperature and energy, the conversion of energy from one form to another especially
heat, the direction of heat flow, and the degree to which the energy of a system is available
to do useful work.
2.2.1 Temperature and the zeroth law
In his classic textbook The Theory of Heat (SpringerVerlag 1967)Richard Becker begins
2-1
CHAPTER 2. THERMODYNAMICS OF DILUTE GASES 2-2
with the following description of temperature.

”The concept of temperature is basic in thermodynamics. It originates from our sensations,
warm and cold. The most salient physical property of temperature is its tendency to equal-
ize. Two bodies in contact (thermal contact!) will eventually have the same temperature,
independent of their physical properties and the special kind of contact. Just this property
is used to bring a substance to a given temperature, namely, by surrounding it with a
heat bath. Then, by definition, substance and heat bath have the same temperature. To
measure the temperature one can employ any physical property which changes continu-
ously and reproducibly with temperature such as volume, pressure, electrical resistivity,
and many others. The temperature scale is fixed by convention.”
In this description, Becker postulates an incomplete law of equilibrium whereby two systems
placed in thermal contact will spontaneously change until the temperature of each system
is the same. This is sometimes called the zeroth law of thermodynamics. James Clerk
Maxwell (1831-1879), the famous British physicist, who published his own text entitled
Theory of Heat in 1870, expressed the zeroth law as follows: ”When each of two systems
is equal in temperature to a third, the first two systems are equal in temperature to each
other.” A key concept implicit in the zeroth law is that the temperature characterizes the
state of the system at any moment in time and is independent of the path used to bring
the system to that state. Such a property of the system is called a variable of state.
2.2.2 The first law
During the latter part of the nineteenth century heat was finally recognized to be a form
of energy. The first law of thermodynamics is essentially a statement of this equivalence.
The first law is based on the observation that the internal energy E of an isolated system
is conserved. An isolated system is one with no interaction with its surroundings. The
internal energy is comprised of the total kinetic, rotational and vibrational energy of the
atoms and molecules contained in the system. Chemical bond and nuclear binding energies
must also be included if the system is undergoing a chemical or nuclear reaction.
The value of the internal energy can only be changed if the system ceases to be isolated.
In this case E can change by the transfer of mass to or from the system, by the transfer
of heat, and by work done on or by the system. For an adiabatic (δQ = 0), constant mass
system, dE = −δW .
By convention, δQ is positive if heat is added to the system and negative if heat is removed.
The work, δW is taken to be positive if work is done by the system on the surroundings
and negative if work is done on the system by the surroundings.
Because energy cannot be created or destroyed the amount of heat transferred into a system
must equal the increase in internal energy of the system plus the work done by the system.
For a nonadiabatic system of constant mass
δQ = dE + δW. (2.1)
This statement, which is equivalent to a law of conservation of energy, is known as the

first law of thermodynamics. In Equation (2.1) it is extremely important to distinguish
between a small change in internal energy which is a state variable and therefore is expressed
as a perfect differential d, and small amounts of heat added or work done that do not
characterize the system per se but rather a particular interaction of the system with its
surroundings. To avoid confusion, the latter small changes are denoted by a δ.
The internal energy of a system is determined by its temperature and volume. Any change
in the internal energy of the system is equal to the difference between its initial and final
values regardless of the path followed by the system between the two states. Consider the
piston cylinder combination shown in Figure 2.1.
Figure 2.1: Exchange of heat and work for a system comprising a cylinder with a movable
frictionless piston.
The cylinder contains some homogeneous material with a fixed chemical composition. An
infinitesimal amount of heat, δQ is added to the system causing an infinitesimal change in
internal energy and an infinitesimal amount of work δW to be done. The differential work
done by the system on the surroundings is the conventional mechanical work done by a
force acting over a distance and can be expressed in terms of the state variables pressure
and volume. Thus
δW = F dx = (F/A) d (Ax) = P dV. (2.2)
where A is the cross sectional area of the piston and F is the force by the material inside
the cylinder on the piston. The first law of thermodynamics now takes the following
form.
δQ = dE + P dV (2.3)
We will be dealing with open flows where the system is an infinitesimal fluid element. In
this context it is convenient to work in terms of intensive variables by dividing through by
the mass contained in the cylinder. The first law is then
δq = de + P dv. (2.4)
where δq is the heat exchanged per unit mass, e is the internal energy per unit mass and
v = 1/ρ is the volume per unit mass. As noted above, the δ symbol is used to denote that
the differential 1-form on the right-hand-side of Equation (2.4) is not a perfect differential
and in this form the first law is not particularly useful.
The first law, Equation (2.4), is only useful if we can determine an equation of state for
the substance contained in the cylinder. The equation of state is a functional relationship
between the internal energy per unit mass, specific volume and pressure, P (e, v). For a
general substance an accurate equation of state is not a particularly easy relationship to
come by and so most applications tend to focus on approximations based on some sort of
idealization. One of the simplest and most important cases is the equation of state for an
ideal gas which is an excellent approximation for real gases over a wide range of conditions.
We will study ideal gases shortly, but first let’s see how the existence of an equation of
state gives us a complete theory for the equilibrium states of the material contained in the
cylinder shown in Figure 2.1.
2.2.3 The second law
Assuming an equation of state can be defined, the first law becomes,
δq = de + P (e, v) dv. (2.5)
According to Pfaff’s theorem, discussed in Chapter 1, the differential form on the right-
hand-side of (2.5) must have an integrating factor which we write as 1/T (e, v). Multiplying
the first law by the integrating factor turns it into a perfect differential.
δq de P (e, v)
= + dv = ds (e, v) (2.6)
T (e, v) T (e, v) T (e, v)
In effect, once one accepts the first law (2.5) and the existence of an equation of state,
P (e, v), then the existence of two new variables of state is implied. By Pfaff’s theorem
there exists an integrating factor which we take to be the inverse of the temperature
postulated in the zeroth law, and there is an associated integral called the entropy (per
unit mass) s(e, v) . In essence, the second law implies the existence of stable states of
equilibrium of a thermodynamic system. The final result is the famous Gibbs equation,
usually written
T ds = de + P dv. (2.7)
This fundamental equation is the starting point for virtually all applications of thermody-
namics. Gibbs equation describes states that are in local thermodynamic equilibrium i.e.,
states that can be reached through a sequence of reversible steps. Since (2.7) is a perfect
differential we can conclude that the partial derivatives of the entropy are

∂s 1 ∂s P (e, v)
= = . (2.8)
∂e v=constant T (e, v)
∂v e=constant
T (e, v)
The cross derivatives of the entropy are equal and so one can state that for any homogeneous
material
∂2s

∂ 1 ∂ P (e, v)
= = . (2.9)
∂e∂v ∂v T (e, v) ∂e T (e, v)
Note that the integrating factor (the inverse of the temperature) is not uniquely defined.
In particular, there can be an arbitrary, constant scale factor since a constant times ds
is still a perfect differential. This enables a temperature scale to be defined so that the
integrating factor can be identified with the measured temperature of the system.
2.3 The Carnot cycle
Using the Second Law one can show that heat and work are not equivalent though each
is a form of energy. All work can be converted to heat but not all heat can be converted
to work. The Carnot cycle involving heat interaction at constant temperature is the most
efficient thermodynamic cycle and can be used to illustrate this point. Consider the piston
cylinder combination containing a fixed mass of a working fluid shown in Figure 2.2 and
the sequence of piston strokes representing the four basic states in the Carnot cycle. In
the ideal Carnot cycle the adiabatic compression and expansion strokes are carried out
isentropically and the piston is frictionless.
A concrete example in the P − V plane and T − S plane is shown in Figure 2.3 and Figure
2.4. The working fluid is nitrogen cycling between the temperatures of 300 and 500 Kelvin
with the compression stroke moving between one and six atmospheres. The entropy per
unit mass of the compression leg comes from tabulated data for nitrogen computed from
Figure 2.2: The Carnot cycle heat engine.
Figure 2.3: P-V diagram of the Carnot cycle.

Figure 2.4: T-S diagram of the Carnot cycle.
Equation (2.120) at the end of this chapter. The entropy per unit mass of the expansion
leg is specified to be 7300 J/(kg − K) . The thermodynamic efficiency of the cycle is
work output by the system during the cycle W

η= = . (2.10)
heat added to the system during the cycle Q2
According to the first law of thermodynamics
δQ = dE + δW. (2.11)
Over the cycle the change in internal energy (which is a state variable) is zero and the
work done is
W = Q2 + Q1 . (2.12)
where Q1 is negative and so the efficiency is
Q1
η =1+ . (2.13)
Q2
The change in entropy per unit mass over the cycle is also zero and so from the Second
Law
I I
δQ
ds = = 0. (2.14)
T
Since the temperature is constant during the heat interaction we can use this result to
write
Q1 Q2
=− . (2.15)
T1 T2
Thus the efficiency of the Carnot cycle is
T1
ηC = 1 − < 1. (2.16)
T2
For the example shown ηC = 0.4. At most only 40% of the heat added to the system can
be converted to work. The maximum work that can be generated by a heat engine working
between two finite temperatures is limited by the temperature ratio of the system and is
always less than the heat put into the system.
2.3.1 The absolute scale of temperature
For any Carnot cycle, regardless of the working fluid
Q1 T1
=− . (2.17)
Q2 T2
Equation (2.17) enables an absolute scale of temperature to be defined that only depends
on the general properties of a Carnot cycle and is independent of the properties of any
particular substance. As noted earlier the temperature, which is the integrating factor in
the Gibbs equation (2.7), is only defined up to an arbitrary constant of proportionality.
Similarly, any scale factor would divide out of (2.17) and so it has to be chosen by conven-
tion. The convention once used to define the Kelvin scale was to require that there be 100
degrees between the melting point of ice and the boiling point of water. Relatively recently
there was an international agreement to define the ice point as exactly 273.15 K above
absolute zero and allow the boiling point to be no longer fixed. As a result the boiling
point of water at standard pressure is 273.15 + 99.61 = 372.76 K rather than 373.15 K.
Note that the standard pressure is not one atmosphere at sea level, which is 101.325 kP a,
but since 1982 has been defined by the International Union of Pure and Applied Chemistry
(IUPAC) as 100 kP a.
Absolute temperature is generally measured in degrees Rankine or degrees Kelvin and the
scale factor between the two is

9
TRankine = TKelvin . (2.18)
5
The usual Farenheit and Centigrade scales are related to the absolute scales by
TRankine = TF arenheit + 459.67

(2.19)
TKelvin = TCentigrade + 273.15.
2.4 Enthalpy
It is often useful to rearrange Gibbs’ equation so as to exchange dependent and independent

variables. This can be accomplished using the so-called Legendre transformation. In this
approach, a new variable of state is defined called the enthalpy per unit mass.
h = e + Pv (2.20)
In terms of this new variable of state, the Gibbs equation becomes
dh v
ds = − dP. (2.21)
T T
Using this simple change of variables, the pressure has been converted from a dependent
variable to an independent variable.
dh v (h, P )
ds (h, P ) = − dP (2.22)
T (h, P ) T (h, P )
Note that 1/T is still the integrating factor. With enthalpy and pressure as the independent
variables the partial derivatives of the entropy are

∂s 1 ∂s v (h, P )
= =− . (2.23)
∂h P =cons tan t T (h, P ) ∂P h=Cons tan t
T (h, P )
and for any homogeneous material we can write
∂2s

∂ 1 ∂ v (h, P )
= =− . (2.24)
∂h∂P ∂P T (h, P ) ∂h T (h, P )
It is relatively easy to re-express Gibbs’ equation with any two variables selected to be
independent by defining additional variables of state, the free energy f = e − T s and the
free enthalpy g = h − T s (also called the Gibbs free energy). The Gibbs free energy is
very useful in the analysis of systems of reacting gases. Using the Gibbs equation and an
equation of state, any variable of state can be determined as a function of any two others.
For example,
e = ϕ (T, P ) s = ς (T, v)
g = ξ (e, P ) h = φ (T, P ) (2.25)
s = θ (h, P ) s = β (e, v) .
and so forth.
2.4.1 Gibbs equation on a fluid element
One of the interesting and highly useful consequences of (2.25) is that any differentiation
operator acting on the entropy takes on the form of Gibbs equation. Let
s = s (h (x, y, z, t) , P (x, y, z, t)) . (2.26)
Take the derivative of (2.26) with respect to time.
∂s ∂s ∂h ∂s ∂P
= + (2.27)
∂t ∂h ∂t ∂P ∂t
Use (2.23) to replace ∂s/∂h and ∂s/∂P in equation (2.27).
∂s 1 ∂h v (h, P ) ∂P
= − (2.28)
∂t T (h, P ) ∂t T (h, P ) ∂t
This is essentially identical to Gibbs equation with the replacements

ds → ∂s/∂t
dh → ∂h/∂t (2.29)
dP → ∂P/∂t.
Obviously, we could do this with any spatial derivative as well. For example
∂s 1 ∂h v (h, P ) ∂P
= −
∂x T (h, P ) ∂x T (h, P ) ∂x
∂s 1 ∂h v (h, P ) ∂P
= − (2.30)
∂y T (h, P ) ∂y T (h, P ) ∂y
∂s 1 ∂h v (h, P ) ∂P
= − .
∂z T (h, P ) ∂z T (h, P ) ∂z
The three equations in (2.30) can be combined to form the gradient vector
∇h v
∇s = − ∇P (2.31)
T T
which is valid in steady or unsteady flow. All these results come from the functional form
(2.26) in which the entropy depends on space and time implicitly through the functions
h(x, y, z, t) and P (x, y, z, t). The entropy does not depend explicitly on x, y, z, or t. Take
the substantial derivative, D/Dt, of the entropy. The result is

∂s 1 ∂h v ∂P
+ Ū · ∇s = + Ū · ∇h − + Ū · ∇P . (2.32)
∂t T ∂t T ∂t
The result (2.32) shows how Gibbs equation enables a direct connection to be made between
the thermodynamic state of a particular fluid element and the velocity field. One simply
replaces the differentials in the Gibbs equation with the substantial derivative.
Ds 1 Dh v DP
= −
Dt T Dt T Dt
or (2.33)
Ds 1 De P Dρ
= − 2
Dt T Dt ρ T Dt
Gibbs equation is the key to understanding the thermodynamic behavior of compressible

fluid flow. Its usefulness arises from the fact that the equation is expressed in terms
of perfect differentials and therefore correctly describes the evolution of thermodynamic
variables over a selected fluid element without having to know the flow velocity explicitly.
This point will be clarified as we work through the many applications to compressible flow
described in the remainder of the text.
2.5 Heat capacities
Consider the fixed volume shown in Figure 2.5. An infinitesimal amount of heat per unit
mass is added causing an infinitesimal rise in the temperature and internal energy of the
material contained in the volume.
Figure 2.5: Heat addition at constant volume.
The heat capacity at constant volume is defined as

δq de + P dv ∂e
Cv = = = . (2.34)
dT v=const dT
v=const ∂T v=const
Now consider the piston cylinder combination shown below.
Figure 2.6: Heat addition at constant pressure.

An infinitesimal amount of heat is added to the system causing an infinitesimal rise in

temperature. There is an infinitesimal change in volume while the piston is withdrawn
keeping the pressure constant. In this case the system does work on the outside world.
The heat capacity at constant pressure is

δq dh − vdP ∂h
Cp = = = . (2.35)
dT P =const dT
P =const ∂T P =const
For a process at constant pressure the heat added is used to increase the temperature of
the gas and do work on the surroundings. As a result more heat is required for a given
change in the gas temperature and thus Cp > Cv . The enthalpy of a general substance can
be expressed as
Z
h (T, P ) = Cp (T, P ) dT + f (P ) (2.36)
where the pressure dependence needs to be determined by laboratory measurement. Heat

capacities and enthalpies of various substances are generally tabulated purely as a functions
of temperature by choosing a reference pressure of Pref = P ◦ = 105 N/m2 for the integra-
tion. This leads to the concept of a standard enthalpy, h◦ (T ). The standard enthalpy can
be expressed as
Z Tf usion
◦
h (T ) = CP◦ dT + ∆Hf usion +
0
(2.37)
Z Tvaporization Z T
CP◦ dT + ∆Hvaporization + CP◦ dT
Tf usion Tvaporization
where the superscript ◦ implies evaluation at the standard pressure. The heat capacity
of almost all substances goes to zero rapidly as the temperature goes to zero and so the
integration in (2.37) beginning at absolute zero generally does not present a problem. We
shall return to the question of evaluating the enthalpy shortly after we have had a chance
to introduce the concept of an ideal gas.
2.6 Ideal gases
For an ideal gas, the equation of state is very simple.

nRu T
P = (2.38)
V
where n is the number of moles of gas in the system with volume, V . The universal gas
constant is
Ru = 8314.472 Joules/ (kgmole − K) . (2.39)
It is actually more convenient for our purposes to use the gas law expressed in terms of the
density
P = ρRT (2.40)
where
R = Ru /Mw (2.41)
and Mw is the mean molecular weight of the gas. The physical model of the gas that
underlies (2.38) assumes that the gas molecules have a negligible volume and that the
potential energy associated with intermolecular forces is also negligible. This is called the
dilute gas approximation and is an excellent model over the range of gas conditions covered
in this text.
For gas mixtures the mean molecular weight is determined from a mass weighted average
of the various constituents. For air
Mw |air = 28.9644 kilograms/(kg − mole)

. (2.42)
R = 287.06 m2 / sec2 −K .

The perfect gas equation of state actually implies that the internal energy per unit mass
of a perfect gas can only depend on temperature, e = e (T ). Similarly the enthalpy of a
perfect gas only depends on temperature
h (T ) = e + P/ρ = e (T ) + RT. (2.43)
Since the internal energy and enthalpy only depend on temperature, the heat capacities
also depend only on temperature, and we can express changes in the internal energy and
enthalpy as
de = Cv (T ) dT dh = Cp (T ) dT. (2.44)
In this course we will deal entirely with ideal gases and so there is no need to distinguish
between the standard enthalpy and the enthalpy and so there is no need to use the dis-
tinguishing character ◦ . We will use it in AA283 when we develop the theory of reacting
gases.
Differentiate RT = h (T ) − e (T ).
RdT = dh − de = (Cp − Cv ) dT (2.45)
The gas constant is equal to the difference between the heat capacities.
R = Cp − Cv (2.46)
The heat capacities themselves are slowly increasing functions of temperature. But the gas
constant is constant, as long as the molecular weight of the system doesn’t change (there
is no dissociation and no chemical reaction). Therefore the ratio of specific heats
Cp
γ= (2.47)
Cv
tends to decrease as the temperature of a gas increases. All gases can be liquefied and the
highest temperature at which this can be accomplished is called the critical temperature
Tc . The pressure and density at the point of liquefaction are called the critical pressure Pc
and critical density ρ. The critical temperature and pressure are physical properties that
depend on the details of the intermolecular forces for a particular gas. An equation of state
that takes the volume of the gas molecules and intermolecular forces into account must
depend on two additional parameters besides the gas constant R. The simplest extension
of the ideal gas law that achieves this is the famous van der Waals equation of state

1 aρ
P = ρRT − (2.48)
1 − bρ RT
where
a 27 a
= RTc = 27Pc . (2.49)
b 8 b2
The van der Waals equation provides a somewhat useful approximation for gases near the
critical point where the dilute gas approximation loses validity.
2.7 Constant specific heat
The heat capacities of monatomic gases are constant over a wide range of temperatures.
For diatomic gases the heat capacities vary only a few percent between the temperatures of
200 K and 1200 K. For enthalpy changes in this range one often uses the approximation of
a calorically perfect gas for which the heat capacities are assumed to be constant and
e2 − e1 = Cv (T2 − T1 ) h2 − h1 = Cp (T2 − T1 ) . (2.50)
For constant specific heat the Gibbs equation becomes
ds dT dρ
= − (γ − 1) (2.51)
Cv T ρ
which can be easily integrated.

Figure 2.7 shows a small parcel of gas moving along some complicated path between two
points in a flow. The thermodynamic state of the gas particle at the two endpoints is
determined by the Gibbs equation.
Figure 2.7: Conceptual path of a fluid element moving between two states.
Integrating (2.51) between 1 and 2 gives an expression for the entropy of an ideal gas with
constant specific heats.
−(γ−1)
s2 − s1 T2 ρ2
exp = (2.52)
Cv T1 ρ1
We can express Gibbs equation in terms of the enthalpy instead of internal energy.
dh dP dT dP
ds = − = Cp −R (2.53)
T ρT T P
Integrate between states 1 and 2.

− γ−1
s2 − s1 T2 P2 γ
exp = . (2.54)
Cp T1 P1
If we eliminate the temperature in (2.54) using the ideal gas law the result is
−γ
s2 − s1 P2 ρ2
exp = . (2.55)
Cv P1 ρ1
In a process where the entropy is constant these relations become
γ γ 1
P2 T2 γ−1 P2 ρ2 ρ2 T2 γ−1
= = = . (2.56)
P1 T1 P1 ρ1 ρ1 T1
Lines of constant entropy in P − T space are shown in Figure 2.8.
Figure 2.8: Lines of constant entropy for a calorically perfect gas.
The relations in (2.56) are sometimes called the isentropic chain.

Note that when we expressed the internal energy and enthalpy for a calorically perfect gas
in (2.50) we were careful to express only changes over a certain temperature range. There
is a temptation to simply express the energy and enthalpy as e = Cv T and h = Cp T . This
is incorrect! The correct values require the full integration from absolute zero shown in
(2.37). As it happens, the gases we deal with in this course, air, oxygen, nitrogen, hydrogen,
etc, all condense at very low temperatures and so the contributions to the enthalpy from
the condensed phase and phase change terms in (2.37) tend to be relatively small. This is
also true for the monatomic gas Helium in spite of the fact that, unlike virtually all other
substances, its heat capacity becomes very large in a narrow range of temperatures near
absolute zero.
2.8 The entropy of mixing
The second law of thermodynamics goes beyond the description of changes that relate
solely to equilibrium states of a system and quantifies the distinction between reversible
and irreversible processes that a system may undergo. For any change of a system
δq ≤ T ds (2.57)
where ds is the change in entropy per unit mass. Substitute the first law into (2.57). For
any change of a system
T ds ≥ de + P dv. (2.58)
Equation (2.58) is the basis of a complete theory for the equilibrium of a thermodynamic
system. The incomplete notion of thermal equilibrium expressed by the zeroth law is only
one facet of the vast range of phenomena covered by the second law (2.58).
2.8.1 Sample problem - thermal mixing
Equal volumes of an ideal gas are separated by an insulating partition inside an adiabatic
container. The gases are at the same pressure but two different temperatures. Assume
there are no body forces acting on the system (no gravitational effects).
Figure 2.9: Thermal mixing of an ideal gas at two temperatures.

The partition is removed and the temperature of the system is allowed to come to equilib-
rium.
Part 1 - What is the final temperature of the system?
Energy is conserved and all the gas energy is in the form of internal energy.
E − Eref = ma Cv (Ta − Tref ) + mb Cv (Tb − Tref ) = (ma + mb ) Cv (Tf inal − Tref ) (2.59)
Canceling the reference energies on both sides we can write
ma Cv Ta + mb Cv Tb = (ma + mb ) Cv Tf inal . (2.60)
The ideal gas law is
P V = ma RTa = mb RTb . (2.61)
Rearrange (2.61) to read
mb Ta
= . (2.62)
ma Tb
Solve (2.59) for Tf inal .

mb
ma Ta + mb Tb Ta + ma Tb 2Ta Tb
Tf inal = = = = 400 K (2.63)
(ma + mb ) mb
1+ m Ta + Tb
a
Part 2 - What is the change in entropy per unit mass of the system? Express the result in
dimensionless form.
The process takes place at constant pressure. In this case the entropy change per unit mass
of the two gases is

sf inal − sa Tf inal
= ln
Cp Ta (2.64)
sf inal − sb Tf inal
= ln .
Cp Tb
The entropy change per unit mass of the whole system is


sf inal −sa sf inal −sb
sf inal − sinitial ma CP + mb Cp
= . (2.65)
Cp ma + mb
This can be expressed in terms of the initial and final temperatures as

Tf inal Ta Tf inal
ln + ln 400 400

sf inal − sinitial Ta Tb Tb ln 600 = 2 ln 300
= = =
Cp 1+ Ta 1 + (2)
Tb
(2.66)
sf inal − sinitial −0.405465 + 2 (0.28768)
= = 0.0566.
Cp 3
The entropy of the system increases as the temperatures equalize.
2.8.2 Entropy change due to mixing of distinct gases
The second law states that the entropy change of an isolated system undergoing a change in
state must be greater than or equal to zero. Generally, non-equilibrium processes involve
some form of mixing such as in the thermal mixing problem just discussed. Consider
two ideal gases at equal temperatures and pressures separated by a partition that is then
removed as shown below.
Figure 2.10: Mixing of two ideal gases at constant pressure and temperature.
For an ideal gas the Gibbs equation is
dT dP
ds = Cp −R . (2.67)
T P
The entropy per unit mass is determined by integrating the Gibbs equation.
Z
dT
s= Cp − R ln (P ) + α (2.68)
T
where α is a constant of integration. A fundamental question revolves around the evaluation

of the entropy constant α for a given substance. This is addressed by the third law of
thermodynamics discussed later. For the two gases shown in the figure
Z
dT Ru
sa = CPa − ln (P ) + αa
T Mwa
Z (2.69)
dT Ru
sb = CPb − ln (P ) + αb .
T Mwb
where Mwa,b refers to the molecular weights of the two distinctly different gases. The
entropy of the whole system is
S = ma sa + mb sb . (2.70)
Define the mass fractions
ma mb
χa = χb = . (2.71)
ma + mb ma + mb
The overall entropy per unit mass before mixing is
Sbef ore
sbef ore = = χa s a + χb s b . (2.72)
ma + mb
Substitute (2.69).
Z Z
dT Ru dT Ru
sbef ore = χa CPa − ln (P ) + αa +χb CPb − ln (P ) + αb (2.73)
T Mwa T Mwb
After mixing each gas fills the whole volume V with the partial pressures given by
ma Ru mb Ru
Pa = T Pb = T. (2.74)
V Mwa V Mwb
where
P = Pa + Pb . (2.75)
The entropy of the mixed system is
Z Z
dT Ru dT Ru
saf ter = χa CPa − ln (Pa ) + αa + χb CPb − ln (Pb ) + αb .
T Mwa T Mwb
(2.76)
Therefore the entropy change of the system is

Ru P Ru P
saf ter − sbef ore = χa ln + χb ln > 0. (2.77)
Mwa Pa Mwb Pa
The initially separated volumes were each in a state of local thermodynamic equilibrium.
When the partition is removed the gases mix and until the mixing is complete the system
is out of equilibrium. As expected the entropy increases. The nice feature of this example
is that at every instant of the non-equilibrium process the pressure and temperature of the
system are well defined. By the way it should be noted that as long as the gases in Figure
2.10 are dilute and the enthalpy and internal energy depend only on temperature then the
mixing process depicted in Figure 2.10 occurs adiabatically without any change in enthalpy
or internal energy. If the gases are very dense so that intermolecular forces contribute to
the internal energy then the enthalpy and internal energy depend on the pressure and the
mixing process may release heat. In this case heat must be removed through the wall to
keep the gas at constant temperature. This is called the heat of mixing. Throughout this
course we will only deal with dilute gases for which the heat of mixing is negligible.
2.8.3 Gibbs paradox
If the gases in Figure 2.10 are identical, then there is no diffusion and no entropy change
occurs when the partition is removed. But the full amount of entropy change is produced
as long as the gases are different in any way no matter how slight. If we imagine a limiting
process where the gas properties are made to approach each other continuously the same
finite amount of entropy is produced at each stage until the limit of identical gases when it
suddenly vanishes. This unexpected result is called Gibbs paradox after J.W. Gibbs who
first noticed it.
However the atomistic nature of matter precludes the sort of continuous limiting process
envisioned. As long as the two gases are different by any sort of experimentally measure-
able property whatsoever, the full entropy change (2.77) is produced. This is true even if
the two gases are chemically similar isotopes of the same element. For example, the inter-
diffusion of ortho and para forms of hydrogen which differ only by the relative orientation
of their nuclear and electronic spins would produce the same entropy increase. The entropy
disappears only if the two molecules are identical.
A full understanding of this statement requires a combination of statistical thermodynam-
ics and quantum mechanics. The founder of statistical mechanics is generally regarded
to be Ludwig Boltzmann (1844-1906) an Austrian physicist who in 1877 established the
relationship between entropy and the statistical model of molecular motion. Boltzmann
is buried in the Central Cemetery in Vienna and on his grave marker is inscribed the
equation
S = k ln (W ) (2.78)
that is his most famous discovery.

Boltzmann showed that the entropy is equal to a fundamental constant k times the loga-
rithm of W which is equal to the number of possible states of the thermodynamic system
with energy, E. A state of the system is a particular set of values for the coordinates and
velocities of each and every molecule in the system. Boltzmann’s constant is essentially
the universal gas constant per molecule
k = Ru /N (2.79)
where N is Avagadro’s number. For a monatomic ideal gas, statistical mechanics gives
3Np
W ≈ V Np E 2 (2.80)
where V is the volume and Np is the number of atoms in the system. When (2.80) is
substituted into (2.78) the result is

S ≈ Ru ln V (Cv T )3/2 (2.81)
which is essentially the same expression that would be generated from Gibbs equation. See
Appendix A for more detail.
When a volume of gas molecules is analyzed using quantum mechanics the energy of the
system is recognized to be quantized and the statistical count of the number of possible
states of the system is quite different depending on whether the individual molecules are
the same or not. If the molecules are different the number of possible states for a given
energy is much larger and this is the basis for the explanation of the Gibbs paradox.
2.9 Isentropic expansion
2.9.1 Blowdown of a pressure vessel
Shown in Figure 2.11 is the blowdown through a small hole of a calorically perfect gas
from a large adiabatic pressure vessel at initial pressure Pi and temperature Ti to the
surroundings at pressure Pa and temperature Ta . Determine the final temperature of the
gas in the sphere.
Figure 2.11: A spherical, thermally insulated pressure vessel exhausts to the surroundings
through a small hole.
To determine the temperature imagine a parcel of gas that remains inside the sphere during
the expansion process as shown in Figure 2.12. As long as the gas is not near the wall
where viscosity might play a role, the expansion of the gas parcel is nearly isentropic. The
final temperature is
γ−1
Tf Pa γ
= (2.82)
Ti Pi
Determine the entropy change per unit mass during the process for the gas ejected to the
surroundings. The ejected gas mixes with the infinite surroundings and comes to a final
temperature and pressure, Ta and Pa . The entropy change is

sf − si Ta γ−1 Pa
= ln − ln (2.83)
Cp Ti γ Pi
Since the second term in (2.83) is clearly positive, the entropy change for the parcel is likely
to be positive unless the ambient temperature is much lower than the initial temperature
in the vessel.
Figure 2.12: Expansion of a parcel of gas as the pressure vessel exhausts to the surroundings.
2.9.2 Work done by an expanding gas
The gun tunnel is a system for studying the flow over a projectile at high speed in rarefied
conditions typical of very high altitude flight. High pressure gas is used to accelerate the
projectile down a gun barrel. The projectile exits into a large chamber at near vacuum
pressure. The figure below depicts the situation.
Figure 2.13: Projectile energized by an expanding gas.
We wish to determine the kinetic energy of the projectile when it exits into the vacuum
chamber. The work done by the gas on the projectile is equal to the kinetic energy of the
projectile
Z L2
1
W = P dV = mU22 (2.84)
L1 2
where the friction between the projectile and the gun barrel has been neglected as well as
any work done against the small pressure in the vacuum chamber.
In order to solve this problem it is necessary to postulate a relationship between the gas
pressure in the gun barrel and the volume. The simplest approach is to assume that the
gas expands isentropically. In this case the pressure and density of the gas behind the
projectile are related by
4mgas
γ ! γ
P ρ πd2 L L1
= = 4mgas = (2.85)
P1 ρ1 L
πd2 L1
where mgas is the mass of the gas expanding behind the projectile. The work integral
(2.84) becomes
L2 γ
πd2
Z
1 L1
mU22 = P1 dL. (2.86)
2 4 L1 L
Carry out the integration
πd2 P1 Lγ1 1−γ

1
mU22 = L2 − L11−γ (2.87)
2 4 1−γ
or
γ−1 !
πd2 L1

1 P1 L1
mU22 = 1− . (2.88)
2 4 γ−1 L2
The first term in brackets on the right side of (2.88) is the initial volume of gas
πd2 L1
V1 = . (2.89)
4
Using (2.89) Equation (2.88) takes the form
γ−1 !
1 P1 V1 L1
mU22 = 1− . (2.90)
2 γ−1 L2
Replace P1 V1 with mgas RT1 and recall that R = Cp − Cv and γ = Cp /Cv . The kinetic
energy of the projectile when it leaves the barrel is
γ−1 !
1 L1
mU22 = mgas Cv T1 1− . (2.91)
2 L2
Note that in the limit where the barrel of the gun is extremely long so that L1 /L2 << 1
all of the gas thermal energy is converted to kinetic energy of the projectile.
2.9.3 Example - Helium gas gun
Suppose the tunnel is designed to use Helium as the working gas. The gas is introduced into
the gun barrel and an electric arc discharge is used to heat the Helium to very high pressure
and temperature. Let the initial gas pressure and temperature be P1 = 4 × 108 N/m2 and
T1 = 2000 K. The initial length is L1 = 0.1 m , the final length is L2 = 2.0 m and the barrel
diameter is d = 0.04 m. The projectile mass is 0.1 kg. Determine the exit velocity of the
projectile. Compare the exit velocity with the speed of sound in the gas at the beginning
and end of the expansion.
Solution: Helium is a monatomic gas with an atomic mass of 4.0026 . The mass of Helium
used to drive the projectile is determined from the ideal gas law

Ru T1
P1 V1 = mgas (2.92)
Mw
or
!
4 × 108 π (0.04)2 (0.1)

P1 V 1 Mw 4.0026
mgas = = = 0.01208 kg. (2.93)
T1 Ru 2000 4 8314.472
The velocity of the projectile at the exit of the barrel is

2/3 !!1/2
0.01208 3 8314.472 0.1
U2 = 2 2000 1 − = 1140.7 m/sec (2.94)
0.1 2 4.0026 2
where the relation Cv = (3/2)R has been used. Note that the final projectile speed is fairly
small compared to the initial speed of sound in the gas
p
a1 = γRT1 = 2631 m/sec. (2.95)
The temperature of the gas at the end of the expansion is determined using the isentropic
relations (2.56).
γ−1 2/3
L1 0.1
T2 = T1 = 2000 = 271 K (2.96)
L2 2
The corresponding speed of sound is
p
a2 = γRT2 = 968 m/ sec . (2.97)
The main assumption used to solve this problem is embodied in the use of (2.85) to
determine the pressure behind the projectile. This equation effectively neglects the motion
of the gas and assumes that the pressure, temperature and density are uniform over the
volume behind the projectile during the expansion. This is an excellent assumption if the
expansion is slow but not so good if the expansion is fast. But fast compared to what? This
is where the speed of sound calculation comes in. If the projectile speed is small compared
to the sound speed, a, then the gas speed over the length of the barrel must also be small
compared to a. In that limit the pressure variation is also small and the uniform property
assumption works quite well. But notice that by the end of the expansion the speed of the
projectile exceeds the speed of sound. A more accurate treatment of this problem requires
a full analysis of the unsteady gas-dynamics of the flow.
2.9.4 Entropy increase due to viscous friction
In the gas gun example viscous friction is an important generator of entropy. The gas
behind the projectile, away from the barrel wall is moving at very high speed but the gas
near the wall is subject to the no slip condition. In the small gap between the projectile
and the barrel the flow is similar to Couette flow considered in Chapter 8. Everywhere near
the wall the flow is subject to very high shear rates leading to high viscous stresses.
So far we have seen how gradients in temperature and gas concentration lead to an increase
in the entropy. In Chapter 7 we will show that flow kinetic energy dissipation due to viscous
friction always leads to an increase in the entropy per unit mass. An accurate calculation
of viscous effects in the gun tunnel problem requires a numerical analysis of the full viscous
equations of motion and remains today a difficult research problem.
One of the most difficult challenges in the development of new power and propulsion systems
is the accurate prediction of entropy changes in the system. Literally billions of dollars
are spent by manufacturers in the pursuit of small reductions in the entropy generated in
compressors and turbines and small excesses can spell the difference between success and
failure of a new design.
2.10 Some results from statistical mechanics
A discussion of this topic can be found in Appendix A. Here we shall state the main results
that will be used in our investigations of compressible flow. Classical statistical mechanics
leads to a simple expression for Cp and Cv in terms of β, the number of degrees of freedom
of the appropriate molecular model.

β+2 β β+2
Cp = R Cv = R γ= (2.98)
2 2 β
For a mass point, m, with three translational degrees of freedom and no internal structure,
β = 3. The law of equipartition of energy says that any term in the expression for the
energy of the mass point that is quadratic in either the position or velocity contributes
(1/2) kT to the thermal energy of a large collection of such mass points. Thus the thermal
energy (internal energy) per molecule of a gas composed of a large collection of mass points
is
ẽ = (3/2) kT (2.99)
where k is Boltzmann’s constant,
k = 1.3806505 × 10−23 Joules/K. (2.100)
Over one mole of gas,
N ẽ = (3/2) Ru T (2.101)
where Ru = N k is the universal gas constant and N is Avogadro’s number,
N = 6.0221415 × 1026 molecules/kgmole. (2.102)
On a per unit mass of gas basis the internal energy is,
e = (3/2) RT. (2.103)
This is a good model of monatomic gases such as Helium, Neon,Argon, etc. Over a very
wide range of temperatures,
5 3
Cp = R Cv = R (2.104)
2 2
from near condensation to ionization.
2.10.1 Diatomic gases
At room temperature, diatomic molecules exhibit two additional rotational degrees of free-
dom and
7 5
Cp = R Cv = R. (2.105)
2 2
At very low temperatures, Cp can decrease below (7/2) R because rotational degrees of
freedom can freeze out; A phenomenon that can only be understood using quantum statis-
tical mechanics. However, for N2 and O2 the theoretical transition temperature is in the
neighborhood of 3 degrees Kelvin, well below the temperature at which both gases liquefy.
For H2 the transition temperature is about 90 K.
At high temperatures, Cp can increase above (7/2) R because the atoms are not rigidly
bound but can vibrate. This brings into play two additional vibrational degrees of freedom.
At high temperatures the heat capacities approach
9 7
Cp = R Cv = R. (2.106)
2 2
2.10.2 Characteristic vibrational temperature
The determination of the temperature at which the specific heat changes from (7/2) R to
(9/2) R is also beyond classical statistical mechanics but can be determined using quantum
statistical mechanics. The specific heat of a diatomic gas from room temperature up to
high combustion temperatures is accurately predicted from theory to be
( )2
θv
Cp 7 2T
= + θv
. (2.107)
R 2 sinh 2T
The characteristic vibrational temperatures for common diatomic gases are
θv |O2 = 2238 K θv |N2 = 3354 K θv |H2 = 6297 K. (2.108)

The increasing values of θv with decreasing molecular weight reflect the increasing bond
strength as the interatomic distance decreases.
2.11 Enthalpy - diatomic gases
The enthalpy of a diatomic gas is

 ( )2 
Z T Z T θv
h (T ) − h (Tref ) = Cp dT = R 7 + 2T dT. (2.109)
θv

Tref Tref 2 sinh 2T
This integrates to

h (T ) − h (Tref ) 7 T − Tref 1 1
= + − . (2.110)
Rθv 2 θv e(θv /T ) − 1 e(θv /Tref ) − 1
The enthalpy is plotted below for Tref = 0.
Figure 2.14: Enthalpy versus temperature for a diatomic gas.
2.12 Speed of sound
The speed of sound in a homogeneous medium is,

2 ∂P
a = . (2.111)
∂ρ s=constant
For an ideal gas,
γP
a2 = = γRT. (2.112)
ρ
For a flow at velocity U the Mach number is
U
M=√ . (2.113)
γRT
2.13 Atmospheric models
In a stable atmosphere where the fluid velocity is zero, the pressure force on a fluid element
is balanced by the force of gravity
∇P = −ρ∇Ψ (2.114)
where Ψ is the gravitational potential. Near the Earth’s surface the variation of gravita-
tional acceleration with height is relatively small and so we can write
dP
= −ρg (2.115)
dz
where g = 9.80665 m/ sec2 and z is the height above the ground.

In an atmosphere where the entropy is constant the pressure and density at a given height
are related by
γ
P ρ
= (2.116)
P0 ρ0
where P0 and ρ0 are the pressure and density at ground level. Use (2.116) to replace the
pressure in (2.115). The resulting equation can be integrated from the ground to a height
z and the result is
1
ρ gz γ−1
= 1 − (γ − 1) 2 . (2.117)
ρ0 a0
In this model the atmospheric density decreases algebraically with height and goes to zero
(vacuum) when gz/a20 = 1/ (γ − 1).
An alternative model that is more accurate in the upper atmosphere is to assume that the
atmosphere is isothermal. In this case the pressure and density are related by the ideal gas
law P = ρRT where the temperature is constant at the ground value T = T0 . In this case
(2.115) integrates to

γgz
ρ −
a2
=e 0 (2.118)
ρ0
The length scale
a20 RT0
H= = (2.119)
γg g
is called the scale height of the atmosphere. Roughly speaking it is the height where the
gravitational potential energy of a fluid element is comparable to its thermal energy. For
Air at 288.15 K the scale height is 8435 m .
How accurate are these models? A comparison is shown in Figure 2.15. Note that the two
models tend to bracket the actual behavior of the atmosphere. Below a scale height of one,
the atmosphere is approximately isentropic and the temperature falls off almost linearly.
Above a scale height of about 1.5 the temperature is almost constant.
2.14 The third law of thermodynamics
The first law is a statement of conservation of energy and shows that heat and work can be
exchanged with one another. The second law restricts the possible occurrence of certain
processes and can be utilized to predict the direction of a process. Moreover, the second
law shows that no engine can be built that converts all the input heat energy to an equal
amount of useful work. The first and second laws are well supported by a long history of
agreement with experimental measurements in a vast variety of applications. The third
law of thermodynamics is still a source of leading edge research in the thermodynamics
of materials. It was first formulated by the German physicist Walther Hermann Nernst
in 1906 whose work won him the Nobel prize for Chemistry in 1920. One statement of
the Nernst theorem is that the entropy of a system at zero absolute temperature is a well-
defined constant. Another statement is that the entropy of a pure perfect crystalline solid
at absolute zero is zero.
Figure 2.15: Density and temperature of the standard atmosphere.

The standard entropy of a gas at the standard pressure (105 N/m2 ) is determined by inte-
grating the Gibbs equation,
Z Tf usion
◦ dT ∆Hf usion
s (T ) = CP◦ + +
0 T Tf usion
(2.120)
Z Tvaporization Z T
dT ∆Hvaporization dT
CP◦ + + CP◦ .
Tf usion T Tvaporization Tvaporization T
The heats of fusion and vaporization in (2.120) are at standard pressure. The third law of
thermodynamics requires that the heat capacity Cp → 0 as T → 0 at a rate that is fast
enough to insure convergence of the first integral in (2.120). Calorimetric measurements
show that Cp ≈ T 3 for nonmetals at very low temperatures. For metals Cp is proportional
to T 3 at low temperatures but becomes proportional to T at extremely low temperatures.
At very low temperature the atomic motion in a metallic crystal freezes out and the residual
heat capacity comes from the motion of the conduction electrons in the metal.
Experimentally it does appear that the entropy at absolute zero approaches the same
constant value for all pure substances. The third law codifies this result and sets S (0) = 0
for all pure elements and compounds in their most stable, perfect crystalline state at
absolute zero. So for example the entropy of water vapor at 125 C and one atmosphere
would be calculated from heat capacity data as follows.
Z 273.15
◦ dT ∆Hf usion
s (398.15) = CP◦ + +
0 T 273.15
(2.121)
Z 373.15 Z 398.15
dT ∆Hvaporization dT
CP◦ + + CP◦
273.15 T 373.15 373.15 T
This calculation assumes that the ice crystal is perfect. From a quantum mechanical view-
point a perfect crystal at zero temperature would exist in only one (nonzero) energy state.
The Boltzmann relation (2.78) would reduce to S (0) = k ln (1) = 0 which is consistent
with the Nernst theorem.
In a pure, perfect crystal there is no entropy of mixing but in a mixed crystal containing
atomic or molecular species A and B there are different ways A and B can be arranged
and so the entropy at absolute zero is not zero.
The entropy at absolute zero is called the residual entropy. There can be a significant
residual entropy in a variety of common substances where imperfections can occur due to
variations in the molecular orientation. Glasses (that are amorphous materials) and alloys
(that are mixtures of metals) also have residual entropy. The determination of the residual
entropy can be quite complex and is responsible for deviations between calculated and
experimental values of the entropy.
Even in the case of a pure solid if the crystal is not perfect the entropy is nonzero. For
example, the CO molecule has a small dipole moment and so there is a finite probability
that, as it freezes, the molecules can align in the crystal as CO − OC − CO instead of
CO − CO − CO. A given crystal will contain a mixture of both types of alignment and so
it is not perfect and the entropy at 0 K is not zero.
All these matters are the subject of ongoing research. In virtually all applications of
thermodynamics to compressible flow the interesting result is based on changes in entropy
associated with some process such as the mixing problem described earlier. Such changes
are unaffected by possible errors in the entropy constant.
2.15 Problems
Problem 1 - Use the Gibbs equation to determine each of the following for an ideal
gas.

∂s (T, ρ) ∂s (T, ρ)
∂T ρ=constant ∂ρ T =constant
(2.122)
∂s (T, P ) ∂s (T, P )
∂T
P =constant ∂P
T =constant
Problem 2 - In Section 2 it was stated that the internal energy and enthalpy of an ideal
gas depend only on temperature. Show that this is true. First show that for an ideal gas
the Gibbs equation can be written in the form
1 R v
ds (T, P ) = de (T, P ) + dT − dP. (2.123)
T T T
Work out the partial derivatives of the entropy, and show by the cross-derivative test that
∂e (T, P ) /∂P = 0.
Problem 3 - Use the Gibbs equation to show that for a general substance.

∂h ∂v (T, P )
= −T + v (T, P ) (2.124)
∂P T ∂T
where v (T, P ) is the volume per unit mass.

Problem 4 - The temperature, entropy and pressure in a calorically perfect ideal gas
moving in an unsteady, three-dimensional flow are related by the function

− γ−1
s − sref T P γ
exp = . (2.125)
Cp Tref Pref
Take the gradient of (2.125) and show directly that the flow satisfies
∇P
T ∇s = ∇h − . (2.126)
ρ
Problem 5 - Show that the internal energy of a van der Waals gas is of the form e (T, v) =
f (T ) − a/v.
Problem 6 - A heavy piston is dropped from the top of a long, insulated, vertical shaft
containing air. The shaft above the piston is open to the atmosphere. Determine the
equilibrium height of the piston when it comes to rest. Feel free to introduce whatever
data or assumptions you feel are required to solve the problem. Suppose you actually
carried out this experiment. How do you think the measured height of the piston would
compare with your model?
Problem 7 - In problem 6 what would be the equilibrium height if the gas in the shaft is
helium.
Problem 8 - Consider the nearly isentropic flow of an ideal gas across a low pressure fan
such as an aircraft propeller. Assume that the pressure change ∆P is small. Show that
the corresponding density change is
∆ρ 1 ∆P
≈ (2.127)
ρ0 γ P0
where ρ0 and P0 are the undisturbed values ahead of the fan.

Problem 9 - Mars has an atmosphere that is about 96% Carbon Dioxide at a temperature
of about 200 K. Determine the scale height of the atmosphere and compare it with Earth.
The pressure at the surface of Mars is only about 1000 P ascals. Entry, descent and landing
of spacecraft on Mars is considered to be in some ways more difficult than on Earth. Why
do you think this is?
Problem 10 - Suppose you are driving and a child in the back seat is holding a helium
filled balloon. You brake for a stoplight. In surprise, the child releases the balloon. The
x-momentum equation governing the motion of the air in the car can be simplified to
∂U ∂P
ρ =− . (2.128)
∂t ∂x
Use this result to show in which direction the balloon moves. What assumptions are needed
to reduce the momentum equation to (2.128)? Compare this problem to the material
developed in Section 2.13
Chapter 3
Control volumes, vector calculus
3.1 Control volume definition
The idea of the control volume is an extremely general concept used widely in fluid me-
chanics. In Chapter 1 we derived the equations for conservation of mass and momentum
on a small cubic control volume fixed in space. In this chapter we will provide a general
definition of the control volume and review some of the powerful mathematical tools of
vector calculus used in conjunction with control volumes. This will enable us to re-derive
in Chapter 5 the equations of motion on a general, arbitrarily moving, control volume
suitable for a wide variety of applications.
The control volume V (t) is a closed, simply connected region in space which may be
finite or infinitesimal in size. The size and geometry of the volume is selected to provide
a convenient imaginary vessel for systematically accounting for the fluxes of various flow
quantities related to the conservation of mass, momentum and energy. Part or all of the
surface of the control volume may be moving and the motion can be arbitrary as long as
the control volume is not torn apart.
The surface A (t) is the instantaneous surface of the control volume and dA is an infinites-
imal surface element. The vector n (t) is a unit vector normal to the surface pointing
outward from the interior of the control volume.
Let F (x1 , x2 , x3 ) be any field variable. It could be a scalar or a component of a vector or
tensor.
3-1
CHAPTER 3. CONTROL VOLUMES, VECTOR CALCULUS 3-2
Figure 3.1: Simply connected control volume.
3.2 Vector calculus
The gradient operator is

∂ ∂ ∂ ∂
∇≡ , , = (3.1)
∂x ∂y ∂z ∂xi
where the subscript i on the right refers to the index of the vector component. As noted
in Chapter 1 we will use this index notation to denote vectors and tensors throughout the
text.
The gradient of a scalar function of space is written

∂F ∂F ∂F ∂F
∇F ≡ , , = (3.2)
∂x ∂y ∂z ∂xi
and the gradient of a vector function of space is

∂F1 ∂F1 ∂F1

 

 ∂x1 ∂x2 ∂x3 

 
 ∂F2 ∂F2 ∂F2  ∂F
i
∇F̄ =  = . (3.3)
 
 ∂x1 ∂x2 ∂x3  ∂xj
 
 
 ∂F3 ∂F3 ∂F3 
∂x1 ∂x2 ∂x3
The divergence of a vector is
3 X
3
X ∂Fi ∂Fi
∇ · F̄ = trace ∇F̄ = δij = δij =
∂xj ∂xj
i=1 j=1
(3.4)
∂F1 ∂F2 ∂F3 ∂Fi
+ + =
∂x1 ∂x2 ∂x3 ∂xi
where the Kronecker unit tensor is defined as
 
1 0 0
δij =  0 1 0  . (3.5)
0 0 1
Equation (3.4) is the first time we have encountered a vector calculus relation involving
two sets of repeated indices and the summation symbols have been included for clarity.
For each index i there is a sum over j and the i’s are also summed. From here on we
will use the Einstein convention and the summation symbols will be dropped. Note that
multiplying by the Kronecker unit tensor and summing over both indices is equivalent to
simply equating the two indices and summing as indicated in (3.4).
The dot (or inner) product of a vector and a tensor is
∂F1 ∂F1 ∂F1 ∂F1 ∂F1 ∂F1

   
 ∂x1 ∂x2 ∂x3   F   F1 ∂x1 + F2 ∂x2 + F3 ∂x3 
  1  
    
∂Fi  ∂F2 ∂F2 ∂F2    ∂F2 ∂F2 ∂F2 
F̄ · ∇F̄ = Fj =   F2  =  F1 + F2 + F3 .
   
∂xj  ∂x1 ∂x2 ∂x3  
  ∂x1 ∂x2 ∂x3 
   
 F3
   
 ∂F3 ∂F3 ∂F3  ∂F3 ∂F3 ∂F3 
F1 + F2 + F3
∂x1 ∂x2 ∂x3 ∂x1 ∂x2 ∂x3
(3.6)
The curl of a vector is defined by the skew-symmetric operation

ē1 ē2 ē3

∂ ∂ ∂
∇ × F̄ = =
∂x1 ∂x2 ∂x3

(3.7)
F1 F2 F3

∂F3 ∂F2 ∂F1 ∂F3 ∂F2 ∂F1
− ē1 + − ē2 + − ē3
∂x2 ∂x3 ∂x3 ∂x1 ∂x1 ∂x2
or we can just write

∂F3 ∂F2 ∂F1 ∂F3 ∂F2 ∂F1
∇ × F̄ = − , − , − . (3.8)
∂x2 ∂x3 ∂x3 ∂x1 ∂x1 ∂x2
The quantities (ē1 , ē2 , ē3 ) are unit vectors in the three orthogonal coordinate directions.
The curl can also be expressed using index notation
∂Fk
∇ × F̄ i
= εijk (3.9)
∂xj
where the alternating unit tensor (also called the Levi-Civita tensor or the permutation
tensor) is defined as follows.

 0, if any two indices are alike
εijk = 1, ijk an even permutation {(1, 2, 3) , (2, 3, 1) , (3, 1, 2)} (3.10)
−1 , ijk an odd permutation {(1, 3, 2) , (3, 2, 1) , (2, 1, 3)}

The alternating unit tensor satisfies the following identities.
δij εijk = 0
εipq εjpq = 2δij

(3.11)
εijk εjpq = 2δij
εijk εijk = 6
where δij is the Kronecker unit tensor defined in (3.5)

3.2.1 Useful vector identities
Some useful vector identities involving first derivatives are as follows.
∇ (ψφ) = ψ∇φ + φ∇ψ

∇ · φŪ = φ∇ · Ū + Ū · ∇φ

∇ × φŪ = φ∇ × Ū + ∇φ × Ū
(3.12)
∇ Ū · V̄ = Ū · ∇ V̄ + V̄ · ∇ Ū + Ū × ∇ × V̄ + V̄ × ∇ × Ū

∇ · Ū × V̄ = V̄ · ∇ × Ū − Ū · ∇ × V̄

∇ × Ū × V̄ = Ū (∇ · V ) + V̄ · ∇ Ū − V̄ ∇ · Ū − Ū · ∇ V̄
Note that dot products involving the gradient operator are not always commutative. For
example consider the difference between
∂Vi
Ū ∇ · V̄ = Uj =
∂xi
(3.13)
∂V1 ∂V2 ∂V3 ∂V1 ∂V2 ∂V3 ∂V1 ∂V2 ∂V3
U1 + + , U2 + + , U3 + +
∂x1 ∂x2 ∂x3 ∂x1 ∂x2 ∂x3 ∂x1 ∂x2 ∂x3
and
∂Uj
V̄ · ∇ Ū = Vi =
∂xi

∂U1 ∂U1 ∂U1 ∂U2 ∂U2 ∂U2 ∂U3 ∂U3 ∂U3
V1 + V2 + V3 , V1 + V2 + V3 , V1 + V2 + V3
∂x1 ∂x2 ∂x3 ∂x1 ∂x2 ∂x3 ∂x1 ∂x2 ∂x3
(3.14)
These two cases illustrate why the parentheses are needed and how useful the index nota-
tion can be in avoiding the kind of visual ambiguities that can arise in the use of vector
notation.
Here are some more vector identities involving second derivatives
∇ · (∇F ) = ∇2 F
∇ · ∇F̄ = ∇2 F̄

(3.15)
∇ · ∇ × F̄ = 0
∇ × ∇ × F̄ = ∇ ∇ · F̄ − ∇2 F̄

where the Laplacian is
∂2 ∂2 ∂2 ∂2
∇2 = + + = . (3.16)
∂x1 2 ∂x2 2 ∂x3 2 ∂xi ∂xi
The Laplacian is a scalar operator. The Laplacian of a scalar is a scalar.
∂ 2 Fj
∇2 F̄ = (3.17)
∂xi ∂xi
The Laplacian of a vector is a vector.
∂ 2 Fj
∇2 F̄ = (3.18)
∂xi ∂xi
Note that the index that is summed is a dummy index and any repeated symbol will do,
for example
∂ 2 Fj
(3.19)
∂xk ∂xk
has the same meaning as the right hand side of (3.18).
3.3 The Gauss theorem
Gauss’s theorem can be used to convert a volume integral involving the gradient operator
to a surface integral involving the outward unit normal. For example
Z Z
∇F dV = F n̄dA (3.20)
V A
where F is a scalar. Here is another example.

Z Z

∇ · F̄ dV = F̄ · n̄dA (3.21)
V A
where F is a vector, and another

Z Z
∂Fij
dV = Fij nj dA (3.22)
V ∂xj A
where Fij is a tensor. A volume integral involving the curl can be converted to a surface
integral.
Z Z

∇ × F̄ dV = n̄ × F̄ dA (3.23)
V A
3.4 Stokes’ theorem
Figure 3.2: Unclosed control volume used to define Stokes’ theorem.
Stokes’ theorem can be used to convert a surface integral involving the curl of a vector
to a line integral involving the unit tangent over the boundary of an unclosed control
volume.
Although we defined the control volume to be closed there are important instances when
an unclosed control volume such as that depicted above (looking kind of like a wind sock!)
can be useful.
The differential length dC is an infinitesimal line segment along the bounding curve C (t)
with unit tangent c (t). Stokes theorem tells us that
Z I

∇ × F̄ · n̄dA = F̄ · c̄dC
A C
Z I (3.24)
∂Fk
εijk ni dA = Fk ck dC
A ∂xj C
and
Z I
(n̄ × ∇F ) dA = F c̄dC. (3.25)
A C
Gauss’s theorem is used repeatedly in the derivation of the equations of motion. Stokes’
theorem comes up in the theory of rotational flow particularly in the development of wing
theory.
3.5 Problems
Problem 1 - Working in Cartesian coordinates and using index notation, prove each of
the following the vector identities.

∇ · ρF̄ = F̄ · ∇ρ + ρ∇ · F̄ (3.26)
∇ × ∇ × F̄ = ∇ ∇ · F̄ − ∇2 F̄

(3.27)

F̄ · F̄
F̄ · ∇F̄ = ∇ × F̄ × F̄ + ∇ (3.28)
2
Problem 2 - Let ēi , ēj and ēk be the unit vectors in a right hand orthogonal coordinate
system. Show that
εijk = ēi · (ēj × ēk ) . (3.29)

Problem 3 - Demonstrate Stokes’ theorem by integration of the curl of some smooth

vector field variable over a square boundary.
Problem 4 - Find a unit vector normal to each of the following surfaces.
i) x + y + z = 2
ii) ax2 + by 2 + cz 2 = 1 (3.30)
iii) xyz = 1
Problem 5 - Show that the unit vector normal to the plane
ax + by + cz = d (3.31)
has the components

a b c
n̄ = √ ,√ ,√ . (3.32)
2 2
a +b +c2 2 2
a +b +c2 a + b2 + c2
2
Why doesn’t n̄ depend on d ?

Problem 6 - Verify Gauss’s theorem
Z Z

∇ · F̄ dV = F̄ · n̄dA (3.33)
V A
in each of the following cases,

i) F̄ = (x, y, z) and V is a cube of side b aligned with the x, y, z axes,
ii) F̄ = n̄r r2 where nr is a unit vector in the radial direction, V is a sphere of radius b
surrounding the origin and r2 = x2 + y 2 + z 2 .
Problem 7 - Verify Stokes’ theorem
Z I

∇ × F̄ · n̄dV = F̄ · c̄dC (3.34)
A C
where F̄ = (x, y, z) and A is the surface of a cube of side b aligned with the x, y, z axes.
The open face of the cube has an outward normal aligned with the positive x-axis.
Chapter 4
Kinematics of fluid motion
4.1 Elementary flow patterns
Recall the discussion of flow patterns in Chapter 1. The equations for particle paths in a
three-dimensional, steady fluid flow are
dx dy dz
= U (x̄) = V (x̄) = W (x̄) . (4.1)
dt dt dt
Although the position of a particle depends on time as it moves with the flow, the flow
pattern itself does not depend on time and the system (4.1) is said to be autonomous.
Autonomous systems of differential equations arise in a vast variety of applications in
mechanics, from the motions of the planets to the dynamics of pendulums to velocity
vector fields in steady fluid flow. A great deal about the flow can be learned by plotting
the velocity vector field Ui (x̄). When the flow pattern is plotted one notices that among
the most prominent features are stagnation points also known as critical points that occur
where
Ui (x̄c ) = 0. (4.2)
Quite often the qualitative features of the flow can be almost completely described once
the critical points of the flow field have been identified and classified.
4.1.1 Linear flows
If the Ui (x̄) are analytic functions of x; The velocity field can be expanded in a Taylor
4-1
CHAPTER 4. KINEMATICS OF FLUID MOTION 4-2
series about the critical point and the result can be used to gain valuable information about
the geometry of the flow field. Retaining just the lowest order term in the expansion of
Ui (x̄) the result is a linear system of equations.
dxi
= Aik (xk − xck ) + O (xk − xck )2 + · · · (4.3)
dt
The matrix Aik is the gradient tensor of the velocity field evaluated at the critical point
and x̄c is the position vector of the critical point.

∂Ui
Aik = (4.4)
∂xk x̄=x̄c
The linear, local solution is expressed in terms of exponential functions and only a relatively
small number of solution patterns are possible. These are determined by the invariants
of Aik . The invariants arise naturally as traces of various powers of Aik . They are all
derived as follows. Transform Anm
Bik = Min Anm M̄mk (4.5)
where M is a non-singular matrix and M̄ is its inverse. Take the trace of (4.5). The dot
(or inner) product of a vector and a tensor is
Bii = Min Anm M̄mi = M̄mi Min Anm = δmn Anm = Amm . (4.6)
The trace is invariant under the affine transformation Mik . One can think of the vector
field Ui as if it is imbedded in an n - dimensional block of rubber. An affine transformation
is one which stretches or distorts the rubber block without ripping it apart or reflecting it
through itself. For traces of higher powers the proof of invariance is similar.
tr (B α ) =
Mjn1 An1 m1 M̄m1 j1 Mj1 n2 An2 m2 M̄m2 j2 · · · Mjα−1 nα Anα mα M̄mα j = (4.7)
= tr (Aα )
The traces of all powers of the gradient tensor remain invariant under an affine transfor-
mation. Likewise any combination of the traces is invariant.
4.1.2 Linear flows in two dimensions
In two dimensions the eigenvalues of Aik satisfy the quadratic
λ2 + P λ + Q = 0 (4.8)
where P and Q are the invariants
P = −Aii
(4.9)
Q = Det (Aik ) .
The eigenvalues of Aik are
P 1p 2
λ=− ± P − 4Q (4.10)
2 2
and the character of the local flow is determined by the quadratic discriminant
P2
D =Q− . (4.11)
4
Figure 4.1: Classification of linear flows in two dimensions
The various possible flow patterns can be summarized on a cross-plot of the invariants as
shown in Figure 4.1. If D > 0 the eigenvalues are complex and a spiraling motion can be
expected in the neighborhood of the critical point. Depending on the sign of P the spiral
may be stable or unstable (spiraling in or spiraling out). If D < 0 the eigenvalues are
real and a predominantly straining flow can be expected. In this case the directionality of
the local flow is defined by the two eigenvectors of Aik . The case P = 0 corresponds to
incompressible flow for which there are only two possible kinds of critical points, centers
with Q > 0 and saddles with Q < 0 . The line Q = 0 in Figure 4.1 corresponds to a
degenerate case where (4.8) reduces to λ (λ + P ) = 0. In this instance the critical point
becomes a line with trajectories converging from either side of the line.
4.1.3 Linear flows in three dimensions
In three dimensions the eigenvalues of Aik satisfy the cubic
λ3 + P λ2 + Qλ + R = 0 (4.12)
where the invariants are
P = −tr [Aij ] = −Aii
1 1
P 2 − tr [Aik Akj ] = P 2 − Aik Aki

Q= (4.13)
2 2
1
−P 3 + 3P Q − Aik Akm Ami .

R=
3
Any cubic polynomial can be simplified as follows. Let
P
λ=α− . (4.14)
3
Then α satisfies
α3 + Q̂α + R̂ = 0 (4.15)
where
1
Q̂ = Q − P 2
3
(4.16)
1 2
R̂ = R − P Q + P 3 .
3 27
Let
1/2 !1/3
R̂ 1 27
a1 = − + √ Q̂3 + R̂2
2 3 3 4
(4.17)
1/2 !1/3
R̂ 1 27
a2 = − − √ Q̂3 + R̂2 .
2 3 3 4
The real solution of (4.15) is expressed as
α1 = a1 + a2 (4.18)
and the complex (or remaining real) solutions are
√
1 i 3
α2 = − (a1 + a2 ) + (a1 − a2 )
2 2
√ (4.19)
1 i 3
α3 = − (a1 + a2 ) − (a1 − a2 ) .
2 2
When (4.12) is solved for the eigenvalues one is led to the cubic discriminant

27 2 3 9 2 1 2
D = R + P − PQ R + Q Q − P . (4.20)
4 2 4
The surface D = 0, is depicted in Figure 4.2.

To help visualize the surface in Figure 4.2 it is split down the middle on the plane P = 0
and the two parts are rotated apart to provide a better view. Note that (4.20) can be
regarded as a quadratic in R and so the surface D = 0 is really composed of two roots for
R that meet in a cusp. If D > 0 the point (P, Q, R) lies above the surface and there is
one real eigenvalue and two complex conjugate eigenvalues. If D < 0 all three eigenvalues
are real. The invariants can be expressed in terms of the eigenvalues as follows. If the
eigenvalues are real
P = − (λ1 + λ2 + λ3 )
Q = λ1 λ2 + λ1 λ3 + λ2 λ3 (4.21)
R = −λ1 λ2 λ3 .
Figure 4.2: The surface dividing real and complex eigenvalues in three dimensions.
If the eigenvalues are complex
P = − (2σ + b)
Q = σ 2 + ω 2 + 2σb (4.22)
R = −b σ 2 + ω 2 .

The variable b is the real eigenvalue and σ and ω are the real and imaginary parts of the
complex conjugate eigenvalues.
4.1.4 Incompressible flow
Flow patterns in incompressible flow are characterized by
∂Ui
∇ · Ū = = Aii = 0. (4.23)
∂xi
This corresponds to P = 0 . In this case the discriminant is

27 2
D = Q3 + R (4.24)
4
and the invariants simplify to
1
Q = − Aik Aki
2
(4.25)
1
R = − Aik Akm Ami .
3
The various possible elementary flow patterns for this case can be categorized on a plot of
Q versus R shown in Figure 4.3
Figure 4.3: Three-dimensional flow patterns in the plane P = 0.
Figure 4.1 and Figure 4.3 are cuts through the surface (4.20) at R = 0 and P = 0 respec-
tively.
We introduced the transformation of coordinates between a fixed and moving frame in

Chapter 1. Here we briefly revisit the subject again in the context of critical points. For a
general smooth flow, the particle path equations (4.1) can be expanded as a Taylor series
about any point x̄0 as follows
dxi
= Ui |x̄=x̄0 + Aik |x̄=x̄0 (xk − x0k ) + O (xk − x0k )2 + · · · (4.26)
dt
If a coordinate system is attached to and moves with the particle at x̄0 with the velocity
Ui |x̄=x̄0 so that
x̄0 = x̄ − x̄0
(4.27)
Ū 0 = Ū − Ū x̄=x̄0

then in that frame of reference the origin of coordinates in effect becomes a critical point
(since the velocity is zero there) and the flow pattern that an observer in this coordinate
system would see is determined by the second and higher order terms in (4.26).
dx0 i
2

= Aik x0k + O x0 k + · · · (4.28)
dt
The elementary flow patterns described above are what would be seen locally at an instant
by an observer moving with a fluid element. Notice that the velocity gradient tensor referred
to either frame is the same. In this way the velocity gradient tensor can be used to infer
the geometry of the local flow pattern at any point in an unambiguous, frame-invariant
manner.
Categorizing flow patterns using the invariants of the velocity gradient tensor has a long
history of applications in fluid mechanics particularly in the kinematic description of flow
separation and reattachment near a solid surface. More recently these methods have been
used to describe light propagation near complex apertures and to describe changes in the
electron charge density field in molecules during the making and breaking of chemical
bonds.
4.2 Rate-of-strain and rate-of-rotation tensors
The velocity gradient tensor

Aij = ∂Ui /∂xj (4.29)
can be split into a symmetric and antisymmetric part

∂Ui 1 ∂Ui ∂Uj 1 ∂Ui ∂Uj
Aij = = + + − . (4.30)
∂xj 2 ∂xj ∂xi 2 ∂xj ∂xi
The symmetric part is the rate-of-strain tensor

1 ∂Ui ∂Uj
Sij = + . (4.31)
2 ∂xj ∂xi
The anti-symmetric part is the rate-of-rotation tensor or spin tensor

1 ∂Ui ∂Uj
Wij = − . (4.32)
2 ∂xj ∂xi
The vorticity vector Ω̄ = ∇ × Ū is related to the velocity gradients by

∂Uk
Ωi = εijk (4.33)
∂xj
and the spin tensor is related to the vorticity by
1
Wik = εijk Ωj . (4.34)
2
All local flow patterns can be regarded as a linear sum of a purely rotational motion and a
purely straining motion. The balance between these two components determines which of
the local flow fields shown in Figure 4.1 or Figure 4.3 will exist at the point. As we move
into our studies of compressible flow we shall see that a natural division exists between
flows that are irrotational, where the effects of viscosity can often be neglected, and flows
that are strain-rate dominated where viscosity plays an important and sometimes dominant
role.
4.3 Problems
Problem 1 - The simplest 2-D flows imaginable are given by the linear system
dx
= ax + by
dt
(4.35)
dy
= cx + dy.
dt
Sketch the corresponding flow pattern for the following cases.

i) (a, b, c, d) = (1, −1, −1, −1)
ii) (a, b, c, d) = (1, −3, 1, −1)
iii) (a, b, c, d) = (−1, 0, 0, −1)
Work out the invariants of the velocity gradient tensor as well as the various components
of the rate-of-rotation and rate-of-strain tensors and the vorticity vector. Which flows are
incompressible?
Problem 2 - An unforced, damped, pendulum is governed by the second order ODE
d2 θ dθ g
2
+β + Sin (θ) = 0. (4.36)
dt dt L
Let x = θ(t) and y = dθ(t)/dt. Use these variables to convert the equation to the canonical
form.
dx
= U (x, y)
dt
(4.37)
dy
= V (x, y) .
dt
Sketch the ”streamlines” defined by (4.37). Locate and categorize any critical points using
the methods developed in this chapter. Identify which points are dominated by rotation
and which are dominated by the rate-of-strain. You can do this graphically by drawing line
segments of the appropriate slope in (x, y) coordinates. The picture of the flow that results
is called the phase portrait of the flow in reference to the fact that, for the pendulum, a
point in the phase portrait represents the instantaneous relation between the position and
velocity of the pendulum. For what value of β can the ”flow” defined by the phase portrait
be used as a model of an incompressible fluid flow?
Problem 3 - Use (4.14) to reduce (4.12) to (4.15).

Problem 4 - Sketch the flow pattern generated by the 3-D linear system
dx
= −y
dt
dy
=x (4.38)
dt
dz
= z.
dt
Work out the invariants of the velocity gradient tensor as well as the components of the
rate-of-rotation and rate-of-strain tensors and vorticity vector. The vector field plotted in
three dimensions is called the phase space of the system of ODEs. In fluid mechanics the
phase portrait or phase space is the physical space of the flow.
Problem 5 - Show that
Sij Aji = Sij Sji (4.39)
and is therefore greater than or equal to zero.

Problem 6 - Work out the formulas for the components of the vorticity vector and show
that the spin tensor is related to the vorticity vector by
1
Wik = εijk Ωj . (4.40)
2
Problem 7 - The velocity field given below has been used in the fluid mechanics literature
to model a two dimensional separation bubble.
U (x, y) = −y + 3y 2 + x2 y − (2/3) y 3
(4.41)
2
V (x, y) = −3xy .
Draw the phase portrait and identify critical points.

Chapter 5
The conservation equations
5.1 Leibniz’ rule for differentiation of integrals
5.1.1 Differentiation under the integral sign
According to the fundamental theorem of calculus if f is a smooth function and the integral
of f is
Z x
f x0 dx0

I (x) = (5.1)
cons tan t
then the derivative of I(x) is
dI
= f (x) . (5.2)
dx
Similarly if
Z cons tan t
f x0 dx0

I (x) = (5.3)
x
then
dI
= −f (x) . (5.4)
dx
Suppose the function f depends on two variables and the integral is a definite integral
5-1
CHAPTER 5. THE CONSERVATION EQUATIONS 5-2
Z b
f x0 , t dx0

I (t) = (5.5)
a
where a and b are constant. The derivative with respect to t is
Z b
dI (t) ∂
f x0 , t dx0 .

= (5.6)
dt a ∂t
The order of the operations of integration and differentiation can be exchanged and so it
is permissible to bring the derivative under the integral sign.
We are interested in applications to compressible flow and so from here on we will interpret
the variable t as time. Now suppose that both the kernel of the integral and the limits of
integration depend on time.
Z b(t)
f x0 , t dx0

I (t, a (t) , b (t)) = (5.7)
a(t)
This situation is shown schematically below with movement of the boundaries indicated.
Figure 5.1: Integration with a moving boundary. The function f (x, t) is shown at one
instant of time.
Using the chain rule, the substantial derivative of (5.7) is
DI ∂I ∂I da ∂I db
= + + (5.8)
Dt ∂t ∂a dt ∂b dt
Now make use of the results in (5.2), (5.4) and (5.6). Equation (5.8) becomes,
b(t)
∂f (x0 , t) 0
Z
DI db da
= dx + f (b (t) , t) − f (a (t) , t) . (5.9)
Dt a(t) ∂t dt dt
The various terms in (5.9) can be interpreted as follows. The first term is the time rate of
change of I due to the integrated time rate of change of f (x, t) within the domain [a, b].
The second and third terms are the contributions to the time rate of change of I due
to the movement of the boundaries enclosing more or less f at a given instant in time.
The relation (5.9) is called Leibniz’ rule for the differentiation of integrals after Gottfried
Wilhelm Leibniz (1646-1716) who, along with Isaac Newton, is credited with independently
inventing differential and integral calculus.
5.1.2 Extension to three dimensions
Let F (x1 , x2 , x3 , t) be some field variable defined as a function of space and time and V (t)
be a time-dependent control volume that encloses some finite region in space at each instant
of time. The time dependent surface of the control volume is A(t).
Figure 5.2: Control volume definition.
Leibniz’ rule extended to three dimensions describes the time rate of change of the amount
of F contained inside V .
Z Z Z
D ∂F
F dV = dV + F ŪA · n̄dA (5.10)
Dt V (t) V (t) ∂t A(t)
When (5.9) is generalized to three dimensions the boundary term in (5.9) becomes a surface
integral. Equation (5.10) can be expressed in words as follows.
   
Rate of    Rate of change due to
  Rate of change  

 
 

change of the movement of the
  
= due to changes +
total amount of F  surf ace A enclosing
of F within V

     

inside V more or less F within V
   
(5.11)
The Leibniz relationship (5.10) is fundamental to the development of the transport theory
of continuous media. The velocity vector ŪA is that of the control volume surface itself.
If the medium is a moving fluid the surface velocity ŪA is specified independently of the
fluid velocity Ū . Consider a fluid with velocity vector Ū which is a function of space and
time.
Figure 5.3: Control volume defined in a flow field subject to surface stresses, body forces
and heat conduction.
Now let the velocity of each surface element of the control volume be the same as the
velocity of the flow, Ū = ŪA . In effect, we assume that the surface is attached to the fluid
and therefore the control volume always contains the same set of fluid elements. This is
called a Lagrangian control volume. In this case Leibnitz’ rule becomes
Z Z Z
D ∂F
F dV = dV + F Ū · n̄dA (5.12)
Use the Gauss theorem to convert the surface integral to a volume integral. The result is
the Reynolds transport theorem.
Z Z
D ∂F
F dV = + ∇ · F Ū dV (5.13)
Dt V (t) V (t) ∂t
5.2 Conservation of mass
Let F = ρ where ρ is the density of the fluid. The Reynolds transport theorem gives
Z Z
D ∂ρ
ρdV = + ∇ · ρŪ dV . (5.14)
Dt V (t) V (t) ∂t
The left-hand-side is the rate of change of the total mass inside the control volume. If there
are no sources of mass within the control volume, the left-hand-side must be zero. Since
the choice of control volume is arbitrary, the kernel of the right- hand-side must therefore
be zero at every point in the flow.
Thus the continuity equation in the absence of mass sources is
∂ρ
+ ∇ · ρŪ = 0 (5.15)
∂t
This equation, expressed in coordinate independent vector notation, is the same one that
we derived in Chapter 1 using an infinitesimal, cubic, Eulerian control volume. Expand
(5.15)
∂ρ
+ U · ∇ρ + ρ∇ · Ū = 0. (5.16)
∂t
In terms of the substantial derivative the continuity equation is
Dρ
+ ρ∇ · Ū = 0. (5.17)
Dt
If the medium is incompressible then ρ = constant and ∇ · Ū = 0.
5.3 Conservation of momentum
In this case the generic variable in Leibniz’ rule is the vector momentum per unit volume,
F = ρŪ . Momentum is convected about by the motion of the fluid itself and spatial vari-
ations of pressure and viscous stresses act as sources of momentum. Restricting ourselves
to the motion of a continuous, viscous fluid (liquid or gas), the stress in a fluid is com-
posed of two parts; a locally isotropic part proportional to the scalar pressure field and a
non-isotropic part due to viscous friction. The stress tensor is
σij = −P δij + τij (5.18)
where P is the thermodynamic pressure, δij is the Kronecker unit tensor defined in Chapter
3,
 
1 0 0 
   δij = 1; i = j
¯  
 0
I= 1 0 
= (5.19)
  6 j
δij = 0; i =
0 0 1
and τij is the viscous stress tensor. The net force acting on the control volume is the
integral of the stress tensor, σij , over the surface plus the integral of any body force
vectors per unit mass, G (gravitational acceleration, electromagnetic acceleration, etc.),
over the volume.
The isotropy of the pressure implies that it acts normal to any surface element in the fluid
regardless of how it is oriented. The viscous part of the stress can take on many different
forms. In Aeronautics and Astronautics we deal almost exclusively with Newtonian fluids
discussed in Chapter 1 such as air or water where the viscous stress is linearly propor-
tional to the rate-of-strain tensor of the flow. The general form of the stress-rate-of-strain
constitutive relation in Cartesian coordinates for a compressible Newtonian fluid is

2
τij = 2µSij − µ − µv δij Skk (5.20)
3
where

1 ∂Ui ∂Uj
Sij = + . (5.21)
2 ∂xj ∂xi
Recall that Skk = ∇ · Ū . The stress components in cylindrical and spherical polar coordi-
nates are given in Appendix B.
Interestingly, there are actually two viscosity coefficients that are required to account for
all possible stress fields that depend linearly on the rate-of-strain tensor. The so-called
shear viscosity µ arises from momentum exchange due to molecular motion. A simple
model of µ is described in Appendix A. The bulk viscosity µv is a little more mysterious.
It contributes only to the viscous normal force and seems to arise from the exchange of
momentum that can occur between colliding molecules and the internal degrees of freedom
of the molecular system. Some typical values of the bulk viscosity are shown in Figure
5.4.
Figure 5.4: Physical properties of some common fluids at one atmosphere and 298.15K.
For monatomic gases that lack such internal degrees of freedom, µv = 0. For some poly-
atomic gases such as CO2 the bulk viscosity is much larger than the shear viscosity. Recall
the discussion of elementary flow patterns from Chapter 4. Any fluid flow can be de-
composed into a rotational part and a straining part. According to the Newtonian model
(5.20), only the straining part contributes to the viscous stress. Although µ is called the
shear viscosity, it is clear from the diagonal terms in (5.20) that there are viscous normal
force components proportional to µ . However they make no net contribution to the mean
normal stress defined as
σmean = (1/3) σii = −P + µv Skk . (5.22)
This is not to say that viscous normal stresses are unimportant. They play a key role in
many compressible flow phenomena we will study later, especially shock waves.
A common assumption called Stokes’ hypothesis is to assume that the bulk viscosity µv is
equal to zero. Then only the so-called shear viscosity µ appears in the constitutive relation
for the stress. While this assumption is strictly valid only for monatomic gases, it is applied
very widely and works quite well, mainly because ∇ · Ū tends to be relatively small in most
situations outside of shock waves and high Mach number flow.
The rate of change of the total momentum inside the control volume is
Z Z Z
D
ρŪ dV = −P I + τ · n̄dA + ρḠdV . (5.23)
Dt V (t) A(t) V (t)
Use the Reynolds transport theorem to replace the left-hand-side of (5.23) and the Gauss
theorem to convert the surface integral to a volume integral.
Z
∂ρŪ
+ ∇ · ρŪ Ū + P I − τ − ρḠ dV = 0. (5.24)
V (t) ∂t
Since the equality must hold over an arbitrary control volume, the kernel must be zero at
every point in the flow and we have the differential equation for conservation of momen-
tum.
∂ρŪ
+ ∇ · ρŪ Ū + P I − τ − ρḠ (5.25)
∂t
This is the same momentum equation we derived in Chapter 1 except for the inclusion of
the body force term.
5.4 Conservation of energy
The energy per unit mass of a moving fluid element is e + k where e is the internal energy
per unit mass of the medium and
1 1
U1 2 + U2 2 + U3 2

k = Ui Ui = (5.26)
2 2
is the kinetic energy per unit mass. In this case we use F = ρ(e+k) in the Leibniz rule
The stress tensor acting over the surface does work on the control volume as do the body
force vectors. In addition, there may be conductive heat flux Q through the surface. There
could also be sources of heat within the flow due to chemical reactions, radiative heating,
etc. For a general fluid the internal energy per unit mass is a function of temperature and
pressure e = f (T, P ) .
The rate of change of the energy inside the control volume in Figure 5.3 is
Z Z Z
D
ρ (e + k)dV = −P I + τ · Ū − Q̄ · n̄dA + ρḠ · Ū dV. (5.27)
Dt V (t) A(t) V (t)
The Reynolds transport theorem and Gauss’s theorem lead to
Z
∂ρ (e + k) P
+ ∇ · ρŪ e + + k − τ · Ū + Q̄ − ρḠ · Ū dV = 0. (5.28)
V (t) ∂t ρ
Since the equality holds over an arbitrary volume, the kernel must be zero and we have
the differential equation for conservation of energy.

∂ρ (e + k) P
+ ∇ · ρŪ e + + k − τ · Ū + Q̄ − ρḠ · Ū = 0 (5.29)
∂t ρ
The sum of enthalpy and kinetic energy
P 1
ht = e + + k = h + Ui Ui (5.30)
ρ 2
is called the stagnation or total enthalpy and plays a key role in the transport of energy
in compressible flow systems. Take care to keep in mind that the flow energy is purely the
sum of internal and kinetic energy, e + k . Some typical gas transport properties at 300 K
and one atmosphere are shown in Figure 5.4. According to Fick’s law, the heat flux vector
in a linear heat conducting medium is
Qi = −κ (∂T /∂xi ) (5.31)
where κ is the thermal conductivity.

The rightmost column in Figure 5.4 is the Prandtl number.
µCp
Pr = (5.32)
κ
The Prandtl number can be thought of as comparing the rate at which momentum is trans-
ported by viscous diffusion to the rate at which temperature diffuses through conductivity.
For most gases the Prandtl number is around 0.7. This number is close to one due to the
fact that heat and momentum transport are accomplished by the same basic mechanism of
molecular collision with lots of space between molecules. Liquids are often characterized
by large values of the Prandtl number and the underlying mechanisms of heat and mo-
mentum transport in a condensed fluid are more complex, since they depend much more
on collisions that are heavily influenced by intermolecular forces.
5.5 Summary - differential form of the equations of mo-

tion
The coordinate-independent form of the equations of motion is
∂ρ
+ ∇ · ρŪ = 0
∂t
∂ρŪ
+ ∇ · ρŪ Ū + P I − τ − ρḠ = 0 (5.33)
∂t

∂ρ (e + k) P
+ ∇ · ρŪ e + + k − τ · Ū + Q̄ − ρḠ · Ū = 0.
∂t ρ
Using index notation the same equations in Cartesian coordinates are
∂ρ ∂
+ (ρUj ) = 0
∂t ∂xj
∂ρUi ∂
+ (ρUi Uj + P δij − τij ) − ρGi = 0 (5.34)
∂t ∂xj

∂ρ (e + k) ∂ P
+ ρUi e + + k − τij Uj + Qi − ρGi Ui = 0.
∂t ∂xi ρ
The equations of motion in cylindrical and spherical polar coordinates are given in Ap-
pendix B.
5.6 Integral form of the equations of motion
In deriving the differential form of the conservation equations (5.33) we used a general
Lagrangian control volume where the surface velocity is equal to the local fluid velocity
and the surface always encloses the same set of fluid elements. In Chapter 1 we derived
the same equations on a rectangular Eulerian control volume. It is important to recognize
that these partial differential equations are valid at every point and at every instant in the
flow of a compressible continuum and are completely independent of the particular control
volume approach (Eulerian or Lagrangian) used to derive them.
With the equations of motion in hand we will now reverse the process and work out the
integral form of these equations on control volumes that are adapted to solving useful
problems. In this endeavor it is useful to consider other kinds of control volumes where
the control surface may be stationary or where part of the control surface is stationary and
part is moving but not necessarily attached to the fluid. Recall the general form of the
Leibniz rule.
Z Z Z
D ∂F
F dV = dV + F ŪA · n̄dA (5.35)
5.6.1 Integral equations on an Eulerian control volume
The simplest case to consider is the Eulerian control volume used in Chapter 1 where
ŪA = 0. This is a stationary volume fixed in space through which the fluid moves. In this
case the Leibniz rule reduces to
Z Z
d ∂F
F dV = dV . (5.36)
dt V V ∂t
Note that, since the Eulerian volume is fixed in space and not time dependent, the lower
case form of the derivative d/dt is used in (5.36). See for comparison (5.6) and (5.9).
Let F = ρ in Equation (5.36)
Z Z
d ∂ρ
ρdV = dV (5.37)
dt V V ∂t
Use (5.15) to replace ∂ρ/∂t in (5.37).

Z Z
d
ρdV = − ∇ · ρŪ dV (5.38)
dt V V
Now use the Gauss theorem to convert the volume integral on the right-hand-side of (5.38)
to a surface integral. The integral form of the mass conservation equation valid on a finite
Eulerian control volume of arbitrary shape is
Z Z
d
ρdV + ρŪ · n̄dA = 0. (5.39)
dt V A
The integral equations for conservation of momentum and energy are derived in a similar
way using (5.25), (5.29), and (5.36). The result is
Z Z
d
ρdV + ρŪ · n̄dA = 0
dt V A
Z Z Z
d
ρŪ dV + ρŪ Ū + P I − τ · n̄dA − ρḠ = 0
dt V A V
Z Z Z
d P
ρ (e + k) dV + ρŪ e + + k − τ · Ū + Q̄ · n̄dA − ρḠ · Ū dV = 0.
dt V A ρ V
(5.40)
5.6.2 Mixed Eulerian-Lagrangian control volumes
More general control volumes where part of the surface may be at rest and other parts
may be attached to the fluid are of great interest especially in the analysis of propulsion
systems. Now use the general form of the Leibniz rule with F = ρ.
Z Z Z
D ∂ρ
ρdV = dV + ρŪA · n̄dA = 0 (5.41)
Use (5.15) to replace ∂ρ/∂t in (5.41) and use the Gauss theorem to convert the volume
integral to a surface integral. The result is the integral form of mass conservation on an
arbitrary moving control volume.
Z Z Z
D
ρdV = − ρŪ · n̄dA + ρŪA · n̄dA (5.42)
Dt V (t) A(t) A(t)
The integral equations for conservation of momentum and energy on a general, moving,
finite control volume are derived in a similar way using (5.25), (5.29), and (5.35). Finally,
the most general integrated form of the conservation equations is
Z Z
D
ρdV + ρ Ū − UA · n̄dA = 0
Dt V (t) A(t)
Z Z Z
D
ρŪ dV + ρŪ Ū − UA + P I − τ · n̄dA − ρḠ = 0
Dt V (t) A(t) V (t)
(5.43)
Z Z
D
ρ (e + k) dV + ρ (e + k) Ū − UA + P I · Ū − τ · Ū + Q̄ · n̄dA−
Dt V (t) A(t)
Z

ρḠ · Ū dV = 0.
V (t)
Remember, ŪA is the velocity of the control volume surface and can be selected at the
convenience of the user.
5.7 Applications of control volume analysis
5.7.1 Example 1 - Solid body at rest in a steady flow
This is a simply connected Eulerian control volume where the segment A2 surrounding (and
attached to) the body is connected by a cut to the surrounding boundary A1 . All fluxes
on the cut A3 cancel and therefore make no contribution to the integrated conservation
laws. There is no mass injection through the surface of the body thus
Z

ρŪ · n̄dA = 0. (5.44)
A1
Momentum fluxes integrated on A1 are directly related to the lift and drag forces exerted
on the body. The integrated momentum equation gives
Z Z
ρŪ Ū + P I − τ · n̄dA + P I − τ · n̄dA = 0. (5.45)
A1 A2
Figure 5.5: Eulerian control volume about a solid body at rest in a steady flow
The drag and lift forces by the flow on the body are
Z

Drag = P I − τ · n̄dA
A2 x1
Z (5.46)

Lif t = P I − τ · n̄dA .
A2 x2
The integral momentum balance in the streamwise direction is
Z

ρŪ Ū + P I − τ · n̄dA + Drag = 0 (5.47)
A1 x1
and in the vertical direction is
Z

ρŪ Ū + P I − τ · n̄dA + Lif t = 0. (5.48)
A1 x2
5.7.2 Example 2 - Channel flow with heat addition
Heat addition to the compressible flow shown above occurs through heat transfer through
the channel wall, Q . There is no net mass addition to the control volume.
Z Z

ρŪ · n̄dA + ρŪ · n̄dA = 0 (5.49)
A1 A2
Figure 5.6: Steady flow in a channel with heat addition.
The flow is steady with no body forces. In this case, the energy balance is
Z

ρŪ (e + k) + P Ū − τ · Ū + Q̄ · n̄dA = 0. (5.50)
A
The contribution of the viscous

stresses to the energy balance is zero along the wall because
of the no-slip condition, Ū Aw = 0 and, as long as the streamwise velocity gradients are
not large, the term −τ̄¯ · Ū is very small on the upstream and downstream faces, A1 and
A2 . In this approximation, The energy balance becomes
Z Z
P
ρ e + + k Ū · n̄dA = − Q̄ · n̄dA. (5.51)
A ρ A
The effects of heat transfer through the wall and conduction through the upstream and
downstream faces, A1 and A2 are accounted for by the change in the flux of stagnation
enthalpy. Heat transfer through the upstream and downstream faces is usually small and
so most of the conductive heat transfer into the flow is through the wall.
Z Z
− Q̄ · n̄dA ∼
=− Q̄ · n̄dA = δQ (5.52)
A Aw
The energy balance for this case reduces to

Z Z
ρht Ū · n̄dA + ρht Ū · n̄dA = δQ. (5.53)
A2 A1
When the vector multiplication is carried out (5.53) becomes

Z Z
ρ2 U2 ht2 dA − ρ1 U1 ht1 dA = δQ. (5.54)
A2 A1
The heat added to the flow is directly and simply related to the change of the integrated
flow rate of stagnation enthalpy along the channel.
5.7.3 Example 3 - Stationary flow about a rotating fan
Figure 5.7: Mixed Eulerian-Lagrangian control volume about a rotating fan in a steady
upstream flow.
This is the prototypical example for propellers, compressors and turbines. In contrast to
a steady flow, a stationary flow is one where time periodic motions, such as the rotation
of the fan illustrated above, do occur, but the properties of the flow averaged over one fan
rotation period or one blade passage period are constant. This will be the case if the fan
rotation speed is held constant.
Here we make use of a mixed Eulerian-Lagrangian control volume. The Lagrangian part
is attached to and moves with the fan blade surfaces and fan axle. Remember the fluid
is viscous and subject to a no slip condition at the solid surface. The Eulerian surface
elements are the upstream and downstream faces of the control volume as well as the
cylindrical surrounding surface aligned with the axis of the fan. We will assume that
the fan is adiabatic, Q̄ = 0 and there is no mass injection through the fan surface. The
integrated mass fluxes are zero.
Z

ρŪ · n̄dA = 0 (5.55)
Ae
Momentum fluxes integrated on Ae are equal to the surface forces exerted by the flow on
the fan
Z Z

ρŪ Ū + P I − τ · n̄dA + ρŪ Ū − ŪA + P I − τ · n̄dA = 0 (5.56)
Ae Af
or
Z
ρŪ Ū + P I − τ · n̄dA + F̄ = 0 (5.57)
Ae
where the vector force by the flow on the fan is

Z
F̄ = P I − τ · n̄dA. (5.58)
Af
Note that the flow and fan velocity on Af are the same due to the no-slip condition

Ū − ŪA A = 0. (5.59)
f
For stationary flow, the integrated energy fluxes on Ae are equal to the work/sec done by
the flow on the fan.
Z Z

ρŪ (e + k) + P Ū − τ · Ū + Q̄ · n̄dA + P Ū − τ · Ū · n̄dA = 0 (5.60)
Ae Af
Z

W ork = P Ū − τ · Ū · n̄dA = δW (5.61)
Af
If the flow is adiabatic, and viscous normal stresses are neglected on Ae and Af the energy
equation becomes
Z
P
ρ e + + k Ū · n̄dA + δW = 0. (5.62)
Ae ρ
5.7.4 Example 4 - Combined heat transfer and work
In a general situation where there is heat transfer and work done
Figure 5.8: A general flow situation with heat transfer and work in a steady upstream flow.
the energy equation has the concise form

Z Z
ρ2 U2 ht2 dA − ρ1 U1 ht1 dA = δQ − δW. (5.63)
A2 A1
The effects of heat addition and work done are both accounted for by changes in the
stagnation enthalpy of the flow. In general, changes due to thermal exchange are irreversible
whereas changes due to the work done can be very nearly reversible except for entropy
changes due to viscous friction.
5.8 Stagnation enthalpy, temperature and pressure
5.8.1 Stagnation enthalpy of a fluid element
It is instructive to develop an equation for the stagnation enthalpy change of a fluid element
in a general unsteady flow. Using the energy equation

∂ρ (e + k) P
+ ∇ · ρŪ e + + k − τ · Ū + Q̄ − ρḠ · Ū = 0 (5.64)
∂t ρ
along with the continuity equation
Dρ
= −ρ∇ · Ū (5.65)
Dt
and the identity

D P 1 DP P Dρ
= − 2 (5.66)
Dt ρ ρ Dt ρ Dt
one can show that
Dht ∂P
ρ = ∇ · τ · Ū − Q̄ + ρḠ · Ū + . (5.67)
Dt ∂t
Equation (5.67) shows that changes in the stagnation enthalpy of a fluid element may be
due to heat conduction as well as the work done by body forces and work by viscous forces.
In addition, nonsteady changes in the pressure, generally associated with exchange of work
by the system, can also change ht .
In most aerodynamic problems body forces are negligible. Exceptions can occur at low
speeds where density changes due to thermal gradients lead to gravity driven flows.
The work done by viscous forces is usually small (the work is zero at a wall where Ū = 0)
and energy transport by heat conduction, Q̄, is often small. In this case the stagnation
enthalpy of a fluid element is preserved if the flow is steady, Dht /Dt = 0.
The last term in (5.67) can be quite large in some nonsteady processes. For example, large
temperature changes can occur in unsteady vortex formation in the wake of a bluff body
in high speed flow due to this term.
If the flow is steady, inviscid and non-heat-conducting then (5.67) reduces to
Ū · ∇ (ht + Ψ) = 0 (5.68)
where the gravitational potential is related to the gravitational force by Ḡ = ∇Ψ.
5.8.2 Blowdown from a pressure vessel revisited
In Chapter 2 we looked at the blowdown of gas from an adiabatic pressure vessel through
a small nozzle. The situation is shown below.
Figure 5.9: An adiabatic pressure vessel exhausting to the surroundings.
The stagnation enthalpy of the flow inside the pressure vessel is governed by (5.67). If
we drop the body force term and assume the divergence term involving heat transfer and
viscous work are small then (5.67) simplifies to
Dht ∂P
ρ = . (5.69)
Dt ∂t
If we further assume that the vessel is very large and the hole is very small then the
contribution of the flow velocity to the stagnation enthalpy can be neglected since most
of the gas is almost at rest. Under these conditions the temperature and pressure can be
regarded as approximately uniform over the interior of the vessel. With these assumptions
(5.69) simplifies to
dh dP
ρ = . (5.70)
dt dt
Equation (5.70) is the Gibbs equation for an isentropic process. If the gas is calorically
perfect the final temperature is given by
γ−1
Tf Pa γ
= . (5.71)
Ti Pi
This is the result we were led to in Chapter 2 when we assumed the process in the vessel
was isentropic. The difference is that by beginning with (5.67) the assumptions needed to
reach (5.71) are clarified. This example nicely illustrates the role of the unsteady pressure
term in (5.67).
5.8.3 Stagnation enthalpy and temperature in steady flow
The figure below shows the path of a fluid element in steady flow stagnating at the leading
edge of an airfoil.
Figure 5.10: Schematic of a stagnation process in steady flow.
The dashed lines outline a small streamtube surrounding the stagnation streamline. Lets
interpret this problem in two ways, first as an extension of section 5.7 example 2 and then
in the context of equation (5.67).
The result in example 2 applies to the flow in the streamtube.
Z Z
ρ2 U2 ht2 dA − ρ1 U1 ht1 dA = δQ. (5.72)
A2 A1
If we assume the airfoil is adiabatic and that there is no net heat loss or gain through the
stream tube then
Z Z
ρ2 U2 ht2 dA = ρ1 U1 ht1 dA. (5.73)
A2 A1
Since the mass flow at any point in the stream tube is the same then one can conclude
that the enthalpy per unit mass is also the same at any point along the stream tube (at
least in an average sense across the tube which we can make arbitrarily narrow).
According to (5.67) the stagnation enthalpy of the fluid element can also change due to
viscous work which we neglected in example 3. However as the element decelerates on
its approach to the airfoil leading edge where viscous forces might be expected to play a
role, the velocity becomes small and goes to zero at the stagnation point so one can argue
that the viscous work terms are also small. An exception where this assumption needs
to be re-examined is at very low Reynolds number where the viscous region can extend a
considerable distance from the airfoil.
By either argument, we can conclude that to a good approximation in a steady flow
P1 1 1
ht1 = e1 + + k1 = h1 + U12 = h2 + U22 = h2 . (5.74)
ρ1 2 2
With viscous work neglected, the stagnation enthalpy is conserved along an adiabatic path.
When a fluid element is brought to rest adiabatically the enthalpy at the rest state is the
stagnation enthalpy at the initial state.
The stagnation temperature is defined by the enthalpy relation
Z Tt
1
ht − h = Cp dT = Ui Ui . (5.75)
T 2
The stagnation temperature is the temperature reached by an element of gas brought to

rest adiabatically. For a calorically perfect ideal gas with constant specific heat in the
range of temperatures between T and Tt (5.75) can be written
1
Cp Tt = Cp T + Ui Ui . (5.76)
2
In Figure 5.10 we would expect T2 = Tt1 to the extent that the heat capacity is constant.
Divide through by Cp T and introduce the speed of sound a = γRT . Equation (5.76)
becomes

Tt γ−1
=1+ M2 (5.77)
T 2
where M is the Mach number of the fluid element.
In Chapter 1 we discussed how the momentum and kinetic energy of a fluid element change
when the frame of reference is transformed from fixed to moving coordinates. Clearly since
the stagnation temperature (5.76) depends on the kinetic energy then it too depends on the
frame of reference. The stagnation temperatures in fixed and moving coordinates are
1
Cp Tt = Cp T + Ui Ui
2
(5.78)
1
Cp Tt0 = Cp T + Ui0 Ui0 .
2
The transformation of the kinetic energy is
1 0 0 1 1 1 1
Ui Ui = Ui Ui + Ẋ Ẋ − 2U + Ẏ Ẏ − 2V + Ż Ż − 2W (5.79)
2 2 2 2 2
where (Ẋ, Ẏ , Ż) are the velocity components of the moving origin of coordinates. The
thermodynamic variables density, temperature and pressure do not change and so the
transformation of the stagnation temperature is
1 1 1
Cp Tt0 = Cp Tt + Ẋ Ẋ − 2U + Ẏ Ẏ − 2V + Ż Ż − 2W . (5.80)
2 2 2
5.8.5 Stagnation pressure
The stagnation pressure is defined using the Gibbs equation for an ideal gas.
dT dP
ds = Cp −R (5.81)
T P
Note that while Cp and Cv depend on temperature, the gas constant Cp −Cv is independent
of temperature as long as there are no chemical reactions that might change the molecular
weight of the gas. Integrate (5.81) along an isentropic path ds = 0. Between state 1 and
state 2
Z P2 Z T2
dP dT
R = Cp (T ) (5.82)
P1 P T1 T
which becomes
Z T2
P2 1 dT
= Exp Cp (T ) . (5.83)
P1 R T1 T
If an element of flowing gas at pressure P and temperature T is brought to rest adiabatically

and isentropically to a temperature Tt , the stagnation pressure Pt is defined by
Z Tt
Pt 1 dT
= Exp Cp (T ) . (5.84)
P R T T
If the heat capacities are constant (5.84) reduces to the isentropic relation
γ γ
Pt Tt γ−1 γ−1 2
γ−1
= = 1+ M . (5.85)
P T 2
If the stagnation path in Figure 5.10 is isentropic as well as adiabatic and the gas is
calorically perfect then one would expect
γ
γ−1 γ−1
P2 = Pt1 = P1 1 + M12 . (5.86)
2
The stagnation state is just that, a thermodynamic state, and changes in the stagnation
state of a material are described by the Gibbs equation
Ds Dht 1 DPt
Tt = − (5.87)
Dt Dt ρt Dt
where for an ideal gas Pt = ρt RTt . Note that there is no distinction between the entropy
and stagnation entropy. They are one and the same and the identity
1 Dht 1 DPt 1 Dh 1 DP
− = − (5.88)
Tt Dt ρt Tt Dt T Dt ρT Dt
is occasionally useful.
Let’s return to Figure 5.10 for a moment and ask, what pressure and temperature would
one measure at the stagnation point of the airfoil in a real experiment. In practice it
is impossible to make the airfoil truly adiabatic and so some heat loss would be expected
through conduction. In addition, as the gas approaches the airfoil, gradients in temperature
arise that tend to conduct heat away from the stagnation point. In addition, gradients
in temperature and velocity near the stagnation point tend to produce an increase in
entropy. As a result, the measured temperature and pressure both tend to be lower than
predicted. The difference may vary by anywhere from a fraction of a percent to several
percent depending on the Reynolds number and Mach number of the flow.
5.8.6 Transforming the stagnation pressure between fixed and moving

frames
The stagnation pressure and stagnation temperature in the fixed and moving frames
are
γ
Pt Tt γ−1
=
P T
(5.89)
γ
Pt0 Tt0

γ−1
= .
P T
Divide out the static pressure and temperature in (5.89).
γ
Pt0 Tt0

γ−1
= (5.90)
Pt Tt
Substitute (5.80) into (5.90). The transformation of stagnation pressure is
  γ
γ−1
Ẋ Ẋ − 2U + Ẏ Ẏ − 2V + Ż Ż − 2W
0
Pt = Pt 1 +
  . (5.91)
2Cp Tt
5.9 Problems
Problem 1 - Work out the time derivative of the following integral.
Z Sin(t)
I (t) = ext dx (5.92)
t2
Obtain dI/dt in two ways: (1) by directly integrating, then differentiating the result and
(2) by applying Leibniz’ rule (5.9) then carrying out the integration.
Problem 2 - In Chapter 2, Problem 2 we worked out a hypothetical incompressible steady
flow with the velocity components
(U, V ) = (Cos (x) Cos (y) , Sin (x) Sin (y)) . (5.93)
This 2-D flow clearly satisfies the continuity equation (conservation of mass), could it
possibly satisfy conservation of momentum for an inviscid fluid? To find out work out the
substantial derivatives of the velocity components and equate the results to the partial
derivatives of the pressure that appear in the momentum equation. The differential of the
pressure is
∂P ∂P
dP = dx + dy. (5.94)
∂x ∂y
Show by the cross derivative test whether a pressure field exists that could enable (5.93)
to satisfy momentum conservation. If such a pressure field exists work it out.
Problem 3 - Consider steady flow in one dimension where Ū = (U (x), 0, 0) and all velocity
gradients are zero except
∂U
A11 = . (5.95)
∂x
Work out the components of the Newtonian viscous stress tensor τij . Note the role of the
bulk viscosity.
Problem 4 - A cold gas thruster on a spacecraft uses Helium (atomic weight 4) at a
chamber temperature of 300 K and a chamber pressure of one atmosphere. The gas ex-
hausts adiabatically through a large area ratio nozzle to the vacuum of space. Estimate
the maximum speed of the exhaust gas.
Problem 5 - Work out equation (5.67).
Problem 6 - Steady flow through the empty test section of a wind tunnel with parallel
walls and a rectangular cross-section is shown below. Use a control volume balance to
relate the integrated velocity and pressure profiles at stations 1 and 2 to an integral of the
wall shear stress.
Figure 5.11: Steady flow in an empty wind tunnel
State any assumptions used.

Problem 7 - Use a control volume balance to show that the drag of a circular cylinder at
low Mach number can be related to an integral of the velocity and stress profile in the wake
downstream of the cylinder. Be sure to use the continuity equation to help account for the
x-momentum convected out of the control volume through the upper and lower surfaces.
State any assumptions used.
Problem 8 - Use a control volume balance to evaluate the lift of a three dimensional wing
in an infinite steady stream. Assume the Mach number is low enough so that there are no
shock waves formed.
Figure 5.12: 3-D wing in an infinite stream
1) Select an appropriate control volume.

2) Write down the integral form of the mass conservation equation.
3) Write down the integral form of the momentum conservation equation.
4) Evaluate the various terms on the control volume boundary so as to express the lift of
the wing in terms of an integral over the downstream wake.
5) Why did I stipulate that there are no shock waves? Briefly state any other assumptions
that went in to your solution.
Problem 9 - Suppose a model 3-D wing is contained in a finite sized wind tunnel test
section with horizontal and vertical walls as shown below.
Figure 5.13: 3-D wing in a wind tunnel
What would a test engineer have to measure to determine lift and drag in the absence of
sensors on the model or a mechanical balance for directly measuring forces? Consider a
control volume that coincides with the wind tunnel walls.
Chapter 6
Several forms of the equations of

motion
6.1 The Navier-Stokes equations
Under the assumption of a Newtonian stress-rate-of-strain constitutive equation and a

linear, thermally conductive medium, the equations of motion for compressible flow become
the famous Navier-Stokes equations. In Cartesian coordinates,
∂ρ ∂
+ (ρUi ) = 0
∂t ∂xi

∂ρUi ∂ 2
+ ρUi Uj + P δij − 2µSij + µ − µv δij Skk − ρGi = 0
∂t ∂xj 3

∂ρ (e + k) ∂ 2
+ ρUi ht − κ (∂T /∂xi ) − 2µUj Sij + µ − µv δij Uj Skk − ρUi Gi = 0
∂t ∂xi 3
(6.1)
The Navier-Stokes equations are the foundation of the science of fluid mechanics. With the
inclusion of an equation of state, virtually all flow solving revolves around finding solutions
of the Navier-Stokes equations. Most exceptions involve fluids where the relation between
stress and rate-of-strain is nonlinear such as polymers, or where the equation of state is
not very well understood (for example supersonic flow in water) or rarefied flows where
the Boltzmann equation must be used to explicitly account for particle collisions. The
equations can take on many forms depending on what approximations or assumptions may
6-1
CHAPTER 6. SEVERAL FORMS OF THE EQUATIONS OF MOTION 6-2
be appropriate to a given flow. In addition, transforming the equations to different forms

may enable one to gain insight into the nature of the solutions. It is essential to learn the
many different forms of the equations and to become practiced in the manipulations used
to transform them.
6.1.1 Incompressible Navier-Stokes equations
If there are no body forces and the flow is incompressible, ∇ · Ū = 0, the Navier-Stokes
equations reduce to what is probably their most familiar form.
∂Ui
=0
∂xi

∂Ui ∂ P µ
+ Ui Uj + δij − 2 Sij = 0 (6.2)
∂t ∂xj ρ ρ

∂T ∂ κ ∂T µ
+ Ui T − −2 Sij Sij = 0
∂t ∂xi ρC ∂xi ρC
where the internal energy is assumed to be e = CT and C is the heat capacity of the mate-
rial. The equation of state in this case is just ρ = constant. Notice that for incompressible
flow the continuity and momentum equations are completely decoupled from the energy
equation and can be solved separately. Once the velocity is known the energy equation
can be solved for the temperature. The last term in the energy equation is always positive
and represents a source of internal energy due to dissipation of kinetic energy by viscous
friction. This will be discussed in much more detail in Chapter 7.
6.2 The momentum equation expressed in terms of vortic-

ity
If the transport coefficients, µ and µv are assumed to be constant (a reasonable assump-

tion if the Mach number is not too large), the compressible momentum equation can be
written,

∂ρŪ 1
+ ∇ · ρŪ Ū + ∇P − µ∇2 Ū −

µ + µv ∇ ∇ · Ū − ρḠ = 0 (6.3)
∂t 3
where the body force is the gradient of a potential function Ḡ = −∇Ψ The vorticity is
defined as the curl of the velocity.
Ω̄ = ∇ × Ū (6.4)
If we use the vector identities

Ū · Ū
Ū · ∇Ū = ∇ × Ū × Ū + ∇
2
(6.5)
∇ × ∇ × Ū = ∇ ∇ · Ū − ∇2 Ū

together with the continuity equation, the momentum equation can be written in the
form,

∂ Ū Ū · Ū 4
ρ + ρ Ω̄ × Ū + ρ∇ + ∇P − µ + µv ∇ ∇ · Ū + µ∇ × Ω̄ + ρ∇Ψ = 0 (6.6)
∂t 2 3
If the flow is irrotational Ω̄ = 0 equation (6.6) reduces to

∂ Ū Ū · Ū 4
ρ + ρ∇ + ∇P − µ + µv ∇ ∇ · Ū + ρ∇Ψ = 0 (6.7)
∂t 2 3
According to (6.7) viscous forces play very little role in momentum transport when the
flow is irrotational. If the flow is compressible the effect of viscosity is only through a
term that depends on ∇ · Ū which is usually relatively small in flows at moderate Mach
number. Remarkably, if the flow is incompressible the viscous term disappears altogether
and viscosity plays no role whatsoever in momentum transport. Viscous stresses do act in
the fluid but they generate no net momentum transfer in an irrotational flow. An example
would be the flow over a wing outside the boundary layer near the surface of the wing and
outside the viscous wake.
6.3 The momentum equation expressed in terms of the en-

tropy and vorticity
In Chapter 2 we noted that when an equation of state is combined with Gibbs equation
any thermodynamic variable can be expressed in terms of any two others. In an unsteady,
three-dimensional flow of a continuous medium
s = s (h (x, y, z, t) , P (x, y, z, t)) (6.8)
Taking the gradient of (6.8) leads to
∇P
T ∇s = ∇h − (6.9)
ρ
which we derived in Chapter 2.

Equation (6.9) can be used to replace the gradient of the pressure in (6.6). Thus the
momentum equation in terms of the entropy is

∂ Ū Ū · Ū 1 4 µ
+ Ω̄ × Ū + ∇ h + + Ψ − T ∇s − µ + µv ∇ ∇ · Ū + ∇ × Ω̄ = 0
∂t 2 ρ 3 ρ
(6.10)
6.3.1 Crocco’s theorem
The inviscid form of the above equation with µ = 0, µv = 0 is

∂ Ū Ū · Ū
+ Ω̄ × Ū + ∇ h + + Ψ − T ∇s = 0 (6.11)
∂t 2
Equation (6.11) is called Croccos theorem and demonstrates the close relationship between
vorticity and the gradient of the entropy in a compressible flow.
The relation
Ω̄ × Ū = 0 (6.12)
is satisfied if Ū = 0 (a trivial situation), if the vorticity and velocity are parallel (this is
called a Beltrami flow) or if the flow is irrotational,∇ × Ū = 0.
6.3.2 The energy equation for inviscid, non-heat conducting flow
With the viscosities and thermal conductivity, κ , set to zero, the energy equation in (6.1)
becomes
∂ρ (e + k) ∂
+ (ρUi ht ) − ρGi Ui = 0 (6.13)
∂t ∂xi
If we are going to assume that the fluid is inviscid then it is consistent to also assume
that the heat conductivity is zero since the underlying molecular collision mechanisms for
both fluid properties are fundamentally the same in a gas. In Chapter 7 we will come
to recognize that assuming the gas is inviscid and non-heat-conducting is equivalent to
assuming that the entropy of a fluid element cannot change.
In terms of the stagnation enthalpy ht = h + k (6.13) can be written as
∂ρht ∂ ∂P
+ (ρUi ht ) − ρGi Ui = (6.14)
∂t ∂xi ∂t
Use the continuity equation in (6.14) and Gi = −∂Ψ/∂xi to produce
∂ht ∂ 1 ∂P
+ Ui (ht + Ψ) = (6.15)
∂t ∂xi ρ ∂t
We developed the viscous, heat-conducting version of (6.15) previously in Chapter 5.
6.3.3 Steady flow
If the flow is steady, the inviscid energy equation (6.15) reduces to
U2 U2

∂
Ui h+ +Ψ = Ū · ∇ h + +Ψ =0 (6.16)
∂xi 2 2
The quantity
U2
h+ +Ψ (6.17)
2
is called the steady Bernoulli integral or Bernoulli function. In a general, inviscid, non-
heat-conducting, steady flow the energy equation reduces to the statement (6.16) that the
velocity field is normal to the gradient of the Bernoulli function. In the absence of body
forces, Dht /Dt = 0 in such a flow. The flow is necessarily adiabatic and the stagnation
enthalpy of a fluid element is preserved. If the flow is fed from a reservoir where the
enthalpy is everywhere uniform then, since ht is preserved for each fluid element, the
enthalpy remains uniform and ∇ht = 0 everywhere. In this case the momentum equation
(Croccos theorem) reduces to
Ω̄ × Ū = T ∇s (6.18)
If the flow is steady, inviscid and irrotational then both the stagnation enthalpy and the en-
tropy are constant everywhere. A flow where ∇s = 0 is called homentropic. In this instance
both the momentum equation and energy equations reduce to the same result
U2
h+ + Ψ = Const (6.19)
2
6.4 Inviscid, irrotational, homentropic flow
It is instructive to examine this case further. Again, assume there are no body forces
Ψ = 0, set µ = µv = 0, κ = 0 in the energy equation and ∇ × Ū = 0 in the momentum
equation (6.6). The governing equations become
∂ρ ∂
+ (ρUk ) = 0
∂t ∂xk

∂Ui ∂ Uk Uk ∂P
ρ +ρ + =0 (6.20)
∂t ∂xi 2 ∂xi
γ
P ρ
=
P0 ρ0
In this approximation the energy equation reduces to the isentropic relation between pres-
sure and density. These equations are the starting point for the development of inviscid
flow theory as well as the theory for the propagation of sound.
6.4.1 Steady, inviscid, irrotational flow
If the flow is steady the equations (6.20) reduce to

∂
(ρUk ) = 0
∂xk

∂ Uk Uk 1 ∂P
+ =0 (6.21)
∂xi 2 ρ ∂xi
γ
P ρ
=
P0 ρ0
Take the gradient of the isentropic relation. The result is
∇P = a2 ∇ρ (6.22)
For an ideal gas

P ∇P P γ−1 ∇P
∇ = − 2 ∇ρ = (6.23)
ρ ρ ρ γ ρ
where (6.22) is used. Using (6.23) the momentum equation now becomes

γ P Ū · Ū
∇ + =0 (6.24)
γ−1 ρ 2
The term in parentheses in (6.24) is the stagnation enthalpy. For this class of flows the
momentum equation reduces to ∇ht = 0 and the entropy is the same everywhere as just
noted in the previous section. Substitute (6.22) into the continuity equation.

∇P
Ū · + a2 ∇ · Ū = 0 (6.25)
ρ
Using (6.23) the continuity equation becomes

The energy per unit mass of a moving fluid element is e + k where e is the internal energy
per unit mass of the medium and
Ū · ∇ a2 + (γ − 1) a2 ∇ · Ū = 0

(6.26)
Using (6.24) we can write

The rate of change of the energy inside the control volume in Figure is
a2

Ū · Ū
= ht − (6.27)
γ−1 2
The continuity equation now becomes

Ū · Ū Ū · Ū
(γ − 1) ht − ∇ · Ū − Ū · ∇ =0 (6.28)
2 2
The equations governing steady, inviscid, irrotational motion reduce to a single equation
for the velocity vector Ū .
6.5 The velocity potential
The condition ∇ × Ū = 0 implies that the velocity can be expressed in terms of a scalar
potential.
Ū = ∇Φ (6.29)
Substitute (6.29) into (6.28). The result is the full potential equation for steady, irrotational
flow.

∇Φ · ∇Φ 2 ∇Φ · ∇Φ
(γ − 1) ht − ∇ Φ − ∇Φ · ∇ =0 (6.30)
2 2
The range of flows where (6.30) applies includes subsonic flow over bodies at Mach numbers
below the critical Mach number at which shocks begin to form and supersonic flows that in-
volve smooth expansion and compression such as the flow in a nozzle without shocks.
6.5.1 Unsteady potential flow, the unsteady Bernoulli integral
The irrotational, unsteady momentum equation is, from (6.10),
U2

∂ Ū 1 4
+∇ h+ + Ψ − T ∇s − µ + µv ∇ ∇ · Ū = 0 (6.31)
∂t 2 ρ 3
Insert (6.29) into (6.31). The equation becomes

U2

∂Φ 1 4
µ + µv ∇ ∇ 2 Φ = 0

∇ +h+ + Ψ − T ∇s − (6.32)
∂t 2 ρ 3
If the flow is inviscid,
U2

∂Φ
∇ +h+ + Ψ − T ∇s = 0 (6.33)
∂t 2
If the flow is inviscid and homentropic (homogeneously isentropic) with ∇s = 0, the un-
steady momentum equation reduces to
U2

∂Φ
∇ +h+ +Ψ =0 (6.34)
∂t 2
The quantity
∂Φ U2
+h+ + Ψ = F (t) (6.35)
∂t 2
which can be at most a function of time is called the unsteady Bernoulli integral and is
widely applied in the analysis of unsteady, inviscid, compressible flows.
If the flow is a calorically perfect gas with h = Cp T , then (6.35) can be written
U2

∂Φ γ P
+ + + Ψ = F (t) (6.36)
∂t γ−1 ρ 2
Generally F (t) can be taken to be a constant independent of time. The sort of unusual
situation where it could be time dependent might occur in a closed wind tunnel where the
pressure throughout the system was forced to change due to some overall volume change
of the tunnel. Such changes can occur but they are usually negligible unless there is some
intention to change the volume for a particular purpose, perhaps to exert some form of
flow control.
In summary, the equations of inviscid, homentropic flow in terms of the velocity potential
are
1 ∂ρ ∇Φ · ∇ρ
+ + ∇2 Φ = 0
ρ ∂t ρ

∂Φ γ P ∇Φ · ∇Φ
∇ + + +Ψ =0 (6.37)
∂t γ−1 ρ 2
γ
P ρ
=
Pref ρref
Note that the isentropic relation provides the needed equation to close the system in lieu
of the energy equation. Equation (6.30) is the steady version of (6.37) all boiled down to
one equation for the potential. We can reduce (6.37) to a single equation for the velocity
potential as follows.

∂Φ γ P ∇Φ · ∇Φ
+ + + Ψ = F (t) (6.38)
∂t γ−1 ρ 2
Replace the pressure in (6.38) with the density using the isentropic relation.

1
ρref γ

γ−1 ∂Φ ∇Φ · ∇Φ γ−1
ρ= F (t) − − −Ψ
Pref γ ∂t 2
−1
1 ∂ρ 1 dF (t) ∂Φ ∇Φ · ∇Φ
= − Φtt − ∇Φt · ∇Φ F (t) − − −Ψ
ρ ∂t γ−1 dt ∂t 2
−1
1 1 ∇Φ · ∇Φ ∂Φ ∇Φ · ∇Φ
∇ρ = −∇Φt − ∇ − ∇Ψ F (t) − − −Ψ
ρ γ−1 2 ∂t 2
(6.39)
where the body force potential is assumed to be only a function of space. Substitute (6.39)
into the continuity equation. The full unsteady potential equation becomes

dF (t) ∇Φ · ∇Φ
− Φtt − ∇Φt · ∇Φ − ∇Φ · ∇Φt + ∇ + ∇Ψ +
dt 2
(6.40)
∂Φ ∇Φ · ∇Φ
(γ − 1) F (t) − − − Ψ ∇2 Φ = 0
∂t 2
6.5.2 Incompressible, irrotational flow
If the flow is incompressible, ∇ · Ū = 0 , and irrotational, Ω̄ = 0, the momentum equation

in the absence of gravity ,Ψ = 0, reduces to,

∂ Ū Ū · Ū P
+∇ +∇ =0 (6.41)
∂t 2 ρ
Note again that viscous forces, though present, have no effect on the momentum in an
incompressible, irrotational flow. Since ∇ × Ū = 0 we can write Ū = ∇Φ and in terms
of the velocity potential, the momentum equation for incompressible, irrotational flow
becomes
∂Φ U 2 P

∇ + + =0 (6.42)
∂t 2 ρ
The quantity in parentheses is the incompressible form of the Bernoulli integral and, as in
the compressible case, is at most a function of time.
∂Φ U 2 P
+ + = F (t) (6.43)
∂t 2 ρ
An incompressible, irrotational flow (steady or unsteady) is governed by Laplace’s equa-

tion
∇ · Ū = ∇2 Φ = 0 (6.44)
which can be solved using a wide variety of well established techniques. Once the veloc-
ity potential Φ is known, the velocity field is generated by differentiation (6.29) and the
pressure field is determined from the Bernoulli integral (6.43).
6.6 The vorticity equation
Earlier we cast the momentum equation in terms of the vorticity. Now let’s derive a
conservation equation for the vorticity itself. If we take the curl of the momentum equation
the result is
 
∂ Ū Ū · Ū
 ∂t + Ω̄ × Ū + ∇ h + 2 + Ψ − T ∇s− 
∇×
 1 4
=0 (6.45)
µ 
µ + µv ∇ ∇ · Ū + ∇ × Ω̄
ρ 3 ρ
or
∂ Ω̄
+ ∇ × Ω̄ × Ū − ∇ × (T ∇s) −
∂t ! (6.46)
∇ ∇ · Ū

4 ∇ × Ω̄
µ + µv ∇ × + µ∇ × =0
3 ρ ρ
Using vector identities, (6.46) can be rearranged to read
∂ Ω̄
+ Ū · ∇Ω̄ − Ω̄ · ∇Ū + Ω̄∇ · Ū − ∇T × ∇s−
∂t ! (6.47)
∇ ∇ · Ū

4 ∇ × Ω̄
µ + µv ∇ × + µ∇ × =0
3 ρ ρ
For inviscid, homentropic flow, ∇s = 0, µ = 0, µv = 0 , the vorticity equation reduces

to
DΩ̄ ∂ Ω̄
= + Ū · ∇Ω̄ = Ω̄ · ∇Ū − Ω̄∇ · Ū (6.48)
Dt ∂t
Equation (6.48) is interpreted to mean that, in an inviscid fluid, if the flow is initially
irrotational it will always remain irrotational. Conversely, if the flow has vorticity to begin
with, then that vorticity can be convected or amplified through stretching or volume change
but cannot disappear.
If the flow is incompressible and viscous, the vorticity equation is

DΩ̄ ∂ Ω̄ µ
∇2 Ω̄

= + Ū · ∇Ω̄ = Ω̄ · ∇Ū + (6.49)
Dt ∂t ρ

For planar, two-dimensional flow, Ω̄ · ∇ Ū = 0 and there is only one non-zero component
of the vorticity, Ωz , which satisfies the diffusion equation
2
∂ 2 Ωz

DΩz µ ∂ Ωz
= + (6.50)
Dt ρ ∂x2 ∂y 2
This equation is of the same form as the incompressible equation for the transport of
the temperature. See the energy equation in (6.2) with the viscous dissipation term ne-
glected.

DT κ
= ∇2 T (6.51)
Dt ρ
In two dimensions vorticity diffuses like a scalar such as temperature or concentration.

But note that vorticity can be positive or negative and where mixing between opposite
signs occurs, vorticity can disappear through diffusion. In three- dimensional flow, the
vorticity can be amplified or reduced due to vortex stretching or compression arising from
the nonlinear term Ω̄ · ∇ Ū .
6.7 Fluid flow in three dimensions, the dual stream func-

tion
The equations for particle paths in a three-dimensional, steady fluid flow are
dx
= U (x, y, z)
dt
dy
= V (x, y, z) (6.52)
dt
dz
= W (x, y, z)
dt
The particle paths are determined by integrating (6.52)
x = f (x̃, ỹ, z̃, t)
y = g (x̃, ỹ, z̃, t) (6.53)
z = h (x̃, ỹ, z̃, t)
where (x̃, ỹ, z̃) is the vector coordinate of a particle at t = 0. Elimination of t among these
three relations leads to two infinite families of integral surfaces; the dual stream-function
surfaces
ψ1 = Ψ1 (x, y, z)
(6.54)
ψ2 = Ψ2 (x, y, z)
The total differential of the stream function is
∂Ψ ∂Ψ ∂Ψ
dψ = dx + dy + dz (6.55)
∂x ∂y ∂z
Use (6.52) to replace the differentials in (6.55) and note that on a stream surface dψ = 0,
the stream-functions (6.54) are each integrals of the first order PDE
∂Ψj ∂Ψj ∂Ψj

Ū · ∇Ψj = U +V +W j = 1, 2 (6.56)
∂x ∂y ∂z
A given initial point, (x̃, ỹ, z̃), defines two stream-surfaces. The velocity vector through the
point lies along the intersection of the surfaces as implied by (6.56) and shown in figure
6.1.
Figure 6.1: Dual stream-function surfaces.
Given the dual stream-functions, the velocity field can be reconstructed from
Ū = ∇Ψ1 × ∇Ψ2 (6.57)

6.8 The vector potential
The velocity field of an incompressible flow can be represented as the curl of a vector
potential, Ā also called the vector stream function.
Ū = ∇ × Ā (6.58)
Use of the vector potential in (6.58) guarantees that the incompressibility condition ∇· Ū =
0 is satisfied. In two-dimensional flow only the out-of-plane component of Ā is nonzero
and this corresponds to the stream function discussed in Chapter 1. Take the curl of
(6.58).

∇ × Ū = ∇ × ∇ × Ā (6.59)
Now use the vector identity
∇ ∇ · Ā − ∇2 Ā = ∇ × ∇ × Ā

(6.60)
Equation (6.59) becomes
Ω̄ = ∇ ∇ · Ā − ∇2 Ā

(6.61)
6.8.1 Selection of a Coulomb gauge
There is a certain arbitrariness to the vector potential and we can take advantage of this to
simplify (6.61). Suppose we add the gradient of a scalar function to the vector potential.
Let
Ā0 = Ā + ∇f (6.62)
¯ × ∇f ≡ 0 the velocity field generated by A0 or A is identically the same.

Since ∇
∇ × Ā0 = ∇ × Ā (6.63)
Imagine that we are given A0 . We can always find a function f such that
∇ · Ā0 = ∇2 f (6.64)
which implies that
∇ · Ā = 0 (6.65)
This is called choosing a Coloumb gauge for Ā. In effect we are free to impose the condi-
tion (6.65) on Ā without affecting the velocity field generated from Ā when we take the
curl.
Using (6.65) in (6.61) generates the Poisson equation for the vector potential.
∇2 Ā = −Ω̄ (6.66)
parIn order for (6.66) to be useful one has to be given the vorticity field Ω. The vector
potential is then determined using standard techniques for solving equations of Poisson
type. The velocity field is then generated by taking the curl.
The vector potential is directly related to the dual stream-functions discussed in the pre-
vious section.
Ā = Ψ1 ∇Ψ2 = −Ψ2 ∇Ψ1 (6.67)
6.9 Incompressible flow with mass and vorticity sources
Consider an incompressible flow constructed from a distribution of mass sources, Q(x, t)

and distributed sources of vorticity, Ω̄(x, t). The velocity field is constructed from a linear
superposition of two fields.
Ū = Ūsources + Ūvortices (6.68)
The velocity field generated by the mass sources is irrotational and that generated by the
vorticity sources is divergence free. The continuity equation for such a flow has a source
term
∇ · Ū = ∇ · Ūsources = Q (x̄, t) (6.69)
and the curl of the velocity is
∇ × Ū = ∇ × Ūvortices = Ω̄ (x̄, t) (6.70)

In terms of the potentials the velocity field is
Ū = ∇Φ + ∇ × Ā (6.71)
The potentials satisfy the system of Poisson equations
∇2 Φ = Q (x̄, t)
(6.72)
2
∇ Ā = −Ω̄ (x̄, t)
This approach allows one to construct fairly complex flow fields that can be rotational while
retaining the simplicity of working in terms of potentials governed by linear equations (6.72)
and the associated law of superposition (6.68). Notice that the theory of potential flow
is exactly analogous to the theory of potentials in electricity and magnetism. The mass
sources coincide with the distribution of electric charges and the vorticity coincides with
the electric currents.
6.10 Turbulent flow
Turbulent flows are characterized by unsteady eddying motions with a wide range of scales
superimposed on a slowly changing or steady mean flow. The eddies mix fluid rapidly and
are responsible for much higher rates of mass, momentum and heat transport than would
occur if the flow was laminar. The primary difficulty with analyzing or numerically com-
puting turbulent flows stems from the fact that, at high Reynolds number, the the range
of scales involved can be very large with velocity gradients at the finest scale, where most
kinetic energy dissipation occurs, of order Re larger than the gradients at the largest scale
that drive the flow. Computations designed to fully simulate the flow must be resolved
on such a fine grid that often only a few simulations can be carried out even at moderate
Reynolds number. Very high Reynolds numbers are still beyond todays largest comput-
ers. Experiments designed to measure a turbulent flow are faced with a similar challenge
requiring extremely small measurement volumes as well as the ability to measure over a
large field comparable to the overall size of the flow.
Recall the equations of motion, repeated here for convenience
∂ρ ∂
+ (ρUi ) = 0
∂t ∂xi

∂ρUi ∂ 2
+ ρUi Uj + P δij − 2µSij + µ − µv δij Skk − ρGi = 0
∂t ∂xj 3

∂ρ (e + k) ∂ 2
+ ρUi ht − κ (∂T /∂xi ) − 2µUj Sij + µ − µv δij Uj Skk − ρUi Gi = 0
∂t ∂xi 3
(6.73)
In this section we will discuss an alternative to fully resolving the unsteady flow introduced
by Osborne Reynolds in 1895 in which each flow variable is decomposed into a mean and
fluctuating part. When the equations of motion are suitably averaged, the result is a
new set of equations that relate the mean flow to the correlated part of the fluctuating
motion.
Imagine an ensemble of realizations of the time dependent flow that all satisfy the same set
of initial and boundary conditions, perhaps in the form of a series of laboratory experiments
repeated over and over. Decompose each flow variable into an ensemble mean and a
fluctuation
F (x, y, z, t) = F̂ (x, y, z, t) + F 0 (x, y, z, t) (6.74)
where F is a generic variable. The mean is defined by
N
1 X
F̂ = Fn (6.75)
N
n=1
where the number of realizations of the flow is N and the index refers to the nth member
of the ensemble. All members of the ensemble are aligned to have the same time origin and
evolve with the same clock which is reset each time a new realization is initiated.
Apply the decomposition and averaging to a linear term in the equations of motion such
as
∂\ ∂ \ 0 ∂ ∂ \0 ∂
(P δij ) = P̂ + P δij = P̂ δij + (P δij ) = P̂ δij (6.76)
∂xj ∂xj ∂xj ∂xj ∂xj
Any term that is linear in the fluctuation is zero. Nonlinear terms are more complex. The
decomposition of the density, pressure and velocity is
ρ = ρ̂ + ρ0
P = P̂ + P 0
U = Û + U 0 (6.77)
V = V̂ + V 0
W = Ŵ + W 0
Consider averaging the momentum flux term ρ U V .
\
ρU
[ V = (ρ̂ + ρ0 ) Û + U 0 V̂ + V 0 =
(6.78)
ρ̂[
Û V̂ + ρ\ \
0 Û V̂ + ρ̂U 0 V̂ + ρ\0 U 0 V̂ + ρ̂\Û V 0 + ρ\
0 Û V 0 + ρ̂U
\ 0 V 0 + ρ\
0U 0V 0 =
0 U 0 V̂ + ρd
ρ̂Û V̂ + ρd 0 V 0 Û + ρ̂U[ 0 V 0 + ρ\0U 0V 0
The transformation of the kinetic energy is

[ 0
d 0 d0 0 [0 0 0
\ 0 0
ρU V = ρ̂Û V̂ + ρ U V̂ + ρ V Û + ρ̂U V + ρ U V = ρ̂Û V̂ + τxy |turbulent (6.79)
The terms in parentheses can be thought of as effective stresses due to the background
fluctuations. Similar terms arise when generating the turbulent normal stresses in com-
pressible flow

ρU
[ 0 U 0 Û + ρ̂U
U = ρ̂Û Û + 2ρd [ 0 U 0 + ρ\
0 U 0 U 0 = ρ̂Û Û + τ |
xx turbulent
(6.80)
ρV
[ 0 V 0 V̂ + ρ̂V
V = ρ̂V̂ V̂ + 2ρd [ 0 V 0 + ρ\
0 V 0 V 0 = ρ̂V̂ V̂ + τ |
yy turbulent
Time dependent terms are ensemble averaged in the same way.
N
∂ρ0

∂ρ
c 1 X ∂ρ ∂ ρ̂ d ∂ ρ̂
= = + =
∂t N ∂t n ∂t ∂t ∂t
n=1
(6.81)
N \ \0 0U 0 0U 0
∂ ρ̂Ubi ∂ρ0 Ubi ∂
\
∂ρU
di 1 X ∂ρUi ρ̂Ui ∂ρ i ∂ ρ̂Ubi ∂ ρ[ i
= = + + + = +
∂t N ∂t n ∂t ∂t ∂t ∂t ∂t ∂t
n=1
Since fluid properties such as µ, muv and κ are temperature dependent they must also be
included in the decomposition and averaging process.
6.10.1 Turbulent, incompressible, isothermal flow
If the flow is incompressible, the density is constant. If the temperature is also constant
the equations of motion simplify to
∂Uj
=0
∂xj
2 (6.82)
∂Ui ∂Ui Uj 1 ∂P µ ∂ Ui
+ + − =0
∂t ∂xj ρ ∂xi ρ ∂xj ∂xj
Substitute the Reynolds decomposition (6.77) with density constant into (6.82) and en-
semble average the equations.
∂ Ûj
=0
∂xj
(6.83)
\ 0 0 2
∂ Ûi ∂ Ûi Ûj ∂Ui Uj 1 ∂ P̂ µ ∂ Ûi
+ + + − =0
∂t ∂xj ∂xj ρ ∂xi ρ ∂xj ∂xj
In incompressible flow the only nonlinear term is the convective term and the only term
involving the fluctuations is the last term on the right which can be viewed as an effective
turbulent stress term, the Reynolds stress.
\ 0 0
τij |turbulent = −ρUi Uj (6.84)
The advantage of this approach is that the entire unsteady motion does not need to be
resolved, only the correlations (6.84) are needed. The fundamental difficulty is that the
Reynolds stresses constitute an additional set of unknowns and the equations (6.83) are
unclosed. The field of turbulence modeling is basically all about searching for model
equations that can be used to relate the Reynolds stresses to the mean flow so the system
(6.83) can be closed.
In free shear flows the laminar stress term in (6.83), µ∂ 2 Ui /∂xj ∂xj is generally much smaller
than the Reynolds stress term and some simplification can be achieved by neglecting the
viscous term in the momentum equation altogether. In wall bounded flows this cannot be
done. The Reynolds stresses tend to dominate the flow away from the wall but near the
wall the velocity fluctuations are damped to zero and the viscous stress term is responsible
for the friction on the wall. We will have more to say about the flow near a wall in Chapter
8.
6.11 Problems
Problem 1 - Derive equation (6.6) beginning with the Navier-Stokes equations. Do the
same for equation (6.47).
Problem 2 - Show that for homentropic flow of an ideal gas ∇P = a2 ∇ρ where a is the
local speed of sound.
Chapter 7
Entropy generation and

transport
7.1 Convective form of the Gibbs equation
In this chapter we will address two questions.

1) How is Gibbs equation related to the energy conservation equation?
2) How is the entropy of a fluid affected by its motion?
We have touched on the latter questions several times in this text but without a rigorous
analysis. In this chapter the precise form of the flow properties that represent the sources
of entropy will be derived from first principles. In this endeavor we will use the Gibbs
equation on a fluid element.
Ds De Dv
T = +P (7.1)
Dt Dt Dt
The Navier-Stokes equations are the foundation of the science of fluid mechanics. With the
inclusion of an equation of state, virtually all flow solving revolves around finding solutions
of the Navier-Stokes equations. Most exceptions involve fluids where the relation between
stress and rate-of-strain is nonlinear such as polymers, or where the equation of state is
not very well understood (for example supersonic flow in water) or rarefied flows where
the Boltzmann equation must be used to explicitly account for particle collisions. The
equations can take on many forms depending on what approximations or assumptions may
be appropriate to a given flow. In addition, transforming the equations to different forms
may enable one to gain insight into the nature of the solutions. It is essential to learn the
7-1
CHAPTER 7. ENTROPY GENERATION AND TRANSPORT 7-2
many different forms of the equations and to become practiced in the manipulations used
to transform them.
7.2 The kinetic energy equation
To answer these questions we first need to form a conservation equation for the kinetic en-
ergy. This can be accomplished by manipulating the momentum and continuity equations.
Multiply the momentum equation by Ui (sum over i).
∂ρUi ∂
Ui + Ui (ρUi Uj + P δij − τij ) − ρUi Gi = 0 (7.2)
∂t ∂xj
First let’s look at the temporal and convective terms in this equation.
∂ (ρUi ) ∂ (ρUi Uj )
Ui + Ui =
∂t ∂xj
∂ρ ∂Ui ∂ρ ∂Uj ∂Ui (7.3)
Ui Ui + Ui ρ + Ui Ui Uj + ρUi Ui + ρUi Uj
∂t ∂t ∂xj ∂xj ∂xj
The sum of the underlined terms is zero by continuity. Thus
∂ (ρUi ) ∂ (ρUi Uj ) ∂k ∂k
Ui + Ui =ρ + ρUj (7.4)
where, k = Ui Ui /2. Use continuity again in the form,
∂ρ ∂ρ ∂Uj
k + kUj + kρ =0 (7.5)
∂t ∂xj ∂xj
Add the above two equations to get the temporal-convective part of the kinetic energy
equation.
∂ (ρUi ) ∂ (ρUi Uj ) ∂ (ρk) ∂ (ρkUj )

Ui + Ui = + (7.6)
Now let’s rearrange the pressure term.
∂ (P δij ) ∂ (P Uj ) ∂Uj
Ui = −P (7.7)
∂xj ∂xj ∂xj
and the viscous stress term
∂τij ∂ (Ui τij ) ∂Ui

Ui = − τij (7.8)
∂xj ∂xj ∂xj
which we derived in Chapter 2.

Finally, the kinetic energy equation is
∂ (ρk) ∂ ∂Uj ∂Ui

+ (ρUj k + P Uj − Ui τij ) = ρUj Gj + P − τij (7.9)
∂t ∂xj ∂xj ∂xj
7.3 Internal energy
When the kinetic energy equation (7.9) is subtracted from the energy equation (7.2) the
result is the conservation equation for internal energy.
∂ (ρe) ∂ ∂Uj ∂Ui

+ (ρUj e + Qj ) = −P + τij (7.10)
∂t ∂xj ∂xj ∂xj
Note that both of these equations are in the general conservation form
∂( )
+ ∇ · ( ) = sources (7.11)
∂t
The series of steps used to generate these equations are pretty standard and illustrate
that the equations of motion can be (and often are) manipulated to create a conservation
equation for practically any variable we wish.
The source term
∂Uj
P = P ∇ · Ū (7.12)
∂xj
appears in both the kinetic and internal energy equations with opposite sign. It can
be either positive or negative depending on the whether the fluid is expanding or being
compressed. Recall the continuity equation in the form
1 Dρ
= −∇ · Ū (7.13)
ρ Dt
Multiply by the pressure

P Dρ Dv
P ∇ · Ū = − =ρ P (7.14)
ρ Dt Dt
The term in parentheses we recognize as the work term in the Gibbs equation (the whole
term is actually the work per second per unit volume). This term represents a reversible
exchange between internal and kinetic energy due to the work of compression or expansion
of a fluid element.
7.3.1 Viscous dissipation of kinetic energy
Now let’s look at the term
∂Ui
Φ = τij (7.15)
∂xj
This needs to be worked a little further. Substitute the expression for the Newtonian
stress tensor and decompose the velocity gradient into a symmetric and antisymmetric
part.

2
Φ = 2µSij − µ − µv δij Skk (Sij + Wij ) (7.16)
3
The sum over the antisymmetric part is zero. Thus
2
Φ = 2µ (Sij Sij ) − µ (Sii Skk ) + µv (Sii Skk ) (7.17)
3
We can write the first two terms as a square.

1 1
Φ = 2µ Sij − δij Skk Sij − δij Skk + µv (Sii Skk ) (7.18)
3 3
The scalar quantity, Φ , is the viscous dissipation of kinetic energy and it is clear from the
fact that (7.18) is a sum of squares that it is always positive. The dissipation term appears
as a sink in the kinetic energy equation and as a source in the internal energy equation. It is
the irreversible conversion of kinetic energy to internal energy due to viscous friction.
We can attach two physical interpretations to Stokes’ hypothesis; the assumption that
µv = 0. Recall that the mean normal stress is
σmean = (1/3) σii = −P + µv Skk (7.19)
In terms of flow forces, Stokes’ hypothesis implies that, in a viscous Newtonian fluid, the
mean normal stress is everywhere equal to the pressure.
We can give a second interpretation to Stokes’ hypothesis in terms of flow energy by noting
that
trace (Sij − (1/3) δij Skk ) = 0 (7.20)
Under Stokes’ hypothesis the second term in the dissipation expression (7.18) is zero and
the only contributor to kinetic energy dissipation is the anisotropic (off diagonal) part of
the rate of strain tensor.
7.4 Entropy
We are now in a position to say something about the entropy of the system. Recall the
Gibbs equation.

De P Dρ Ds
ρ − = ρT (7.21)
Dt ρ Dt Dt
Write the internal energy equation,
∂ρe ∂ ∂Uj
+ (ρUj e + Qj ) = −P +Φ (7.22)
∂t ∂xj ∂xj
in terms of the substantial derivative.
De ∂Uj ∂Qj
ρ +P =− +Φ (7.23)
Dt ∂xj ∂xj
We have already written down the continuity equation in the form required to replace the
second term on the left hand side.

P Dρ ∂Uj
= −P (7.24)
ρ Dt ∂xj
The energy equation now becomes

De P Dρ ∂Qj
ρ − =− +Φ (7.25)
Dt ρ Dt ∂xj
Comparing this form of the energy equation with the Gibbs equation, we see that the
left-hand-side corresponds to the entropy term
Ds ∂Qj
ρT =− +Φ (7.26)
Dt ∂xj
Let’s put this in the form of a conservation equation for the entropy. The approach is to
use the continuity equation yet again!
∂s ∂s 1 ∂Qj Φ
ρ + ρUj =− +
∂t ∂xj T ∂xj T
(7.27)
∂ρ ∂ρ ∂Uj
s + sUj + sρ =0
∂t ∂xj ∂xj
Add the two equations in (7.28) to get
∂ρs ∂ 1 ∂Qj Φ
+ (ρUj s) = − + (7.28)
The heat flux term can be rearranged into a divergence term and a squared term.

∂ Qj 1 ∂Qj Qj ∂T
= − 2 (7.29)
∂xj T T ∂xj T ∂xj
For a linear conducting material, the heat flux is
∂T
Qj = −k (7.30)
∂xj
Let

k ∂T ∂T
Υ= (7.31)
T ∂xj ∂xj
Finally, the conservation equation for the entropy becomes

∂ρs ∂ k ∂T Υ+Φ
+ ρUj s − = (7.32)
The integral form of the entropy transport equation on a general control volume is
Z Z Z
D k ∂T Υ+Φ
ρsdV + ρ Uj − UAj s − nj dA = dV (7.33)
Dt V (t) A(t) T ∂xj V (t) T
The right hand side of (7.32) is the entropy source term and is always positive. This re-
markable result provides an explicit expression (in terms of squared temperature gradients
and squared velocity gradients) for the irreversible changes in entropy that occur in a com-
pressible flow. Notice that the body force term contributes nothing to the entropy.
Consider a closed adiabatic box containing a viscous, heat conducting fluid subject to the
no slip condition on the walls. The fluid is stirred by a fan that is then turned off and the
flow is allowed to settle.
Figure 7.1: Fluid stirred in an adiabatic box.
The entropy of the fluid in the box behaves according to
Z Z Z
D k ∂T Υ+Φ
ρsdV + ρUj s − nj dA = dV (7.34)
Dt V A T ∂xj V T
The surface integral is zero by the no-slip condition and the assumption of adiabaticity. The
volume integral of the density times the intensive entropy is the extensive entropy.
Z
S= ρsdV (7.35)
V
Thus as the closed, adiabatic, system comes to rest, the entropy continues to increase until
all the gradients in velocity and temperature have vanished.
Z
dS Υ+Φ
= dV > 0 (7.36)
dt V T
7.5 Problems
Problem 1 - Show that (7.17) can be expressed in the form (7.18).

Problem 2 - Imagine a spherically symmetric flow in the absence of shear.
Figure 7.2: Radially directed viscous flow.
Let the fluid velocity in the radial direction be f (r). Work out the expression for the
kinetic energy dissipation in terms of f and the two viscosities µ and µv . Suppose the
fluid is a monatomic gas for which µv = 0 . Does there exist a function f for which the
dissipation is zero? Refer to Appendix B for the dissipation function in spherical polar
coordinates.
Problem 3 - Suppose the fluid in the box shown in Section 7.5 is Air initially at 300 K
and one atmosphere. Let the tip velocity of the propeller be 50 m/sec and the typical
boundary layer thickness over the surface of the propeller be 1 millimeter. The surface
area of the propeller is 1 m2 and the volume of the box is 1 m3 . Roughly estimate the rate
at which the temperature of the air in the box increases due to viscous dissipation.
Problem 4 - A simple pendulum of mass m = 1.0 kilogram is placed inside a rigid
adiabatic container filled with 0.1 kilograms of Helium gas initially at rest (state 1). The
gas pressure is one atmosphere and the temperature is 300 K. The acceleration due to
gravity is g = 9.8 meters/sec2 .
Figure 7.3: Pendulum released from rest in a viscous fluid in an adiabatic box.
At t = 0 the pendulum is released from an initial displacement height h = 0.1 meters

and begins to oscillate and stir the gas. Eventually the pendulum and gas all come to rest
(state 2).
1) Assume all the potential energy of the pendulum is converted to internal energy of the
gas. What is the change in temperature of the Helium in going from state 1 to state
2?
2) What is the change in entropy per unit mass of the Helium in going from state 1 to
state 2?
Chapter 8
Viscous flow along a wall
8.1 The no-slip condition
All liquids and gases are viscous and, as a consequence, a fluid near a solid boundary
sticks to the boundary. The tendency for a liquid or gas to stick to a wall arises from mo-
mentum exchanged during molecular collisions with the wall. Viscous friction profoundly
changes the fluid flow over a body compared to the ideal inviscid approximation that is
widely applied in aerodynamic theory. Figure 8.1 below illustrates the effect of the no-slip
condition on the flow over a curved boundary. Later we will see how the velocity Ue at
the wall derived from irrotational flow theory is used as the outer boundary condition for
the equations that govern the thin region of rotational flow near the wall imposed by the
no-slip condition. This thin layer is called the boundary layer.
Figure 8.1: Slip versus no-slip flow near a solid surface.
The presence of low speed flow near the surface of a wing can lead to flow separation or
stall. Separation can produce large changes in the pressure field surrounding the body
leading to a decrease in lift and an increase in pressure related drag. The pressure drag
due to stall may be much larger than the drag due to skin friction. In a supersonic flow
the no-slip condition insures that there is always a subsonic flow near the wall enabling
pressure disturbances to propagate upstream through the boundary layer. This may cause
8-1
CHAPTER 8. VISCOUS FLOW ALONG A WALL 8-2
flow separation and shock formation leading to increased wave drag. In Appendix A we
worked out the mean free path between collisions in a gas
1
λ= √ (8.1)
2πnσ 2
where n is the number of molecules per unit volume and σ is the collision diameter of
the molecule. On an atomic scale all solid materials are rough and adsorb or retard gas
molecules near the surface. When a gas molecule collides with a solid surface a certain
fraction of its momentum parallel to the surface is lost due to Van der Walls forces and
interpenetration of electron clouds with the molecules at the solid surface. Collisions of
a rebounding gas molecule with other gas molecules within a few mean free paths of the
surface will slow those molecules and cause the original molecule to collide multiple times
with the surface. Within a few collision times further loss of tangential momentum will
drop the average molecular velocity to zero. At equilibrium, the velocity of the gas within
a small fraction of a mean free path of the surface and the velocity of the surface must
match.
The exception to this occurs in the flow of a rarified gas where the distance a molecule
travels after a collision with the surface is so large that a return to the surface may never
occur and equilibrium is never established. In this case there is a slip velocity that can be
modeled as
∂U
vslip = Cλ (8.2)
∂y
where C is a constant of order one. The slip velocity is negligible for typical values of the
mean free path in gases of ordinary density. In situations where the velocity gradient is
extremely large as in some gas lubrication flows there can be significant slip even at ordinary
densities. For similar reasons there can be a discontinuity in temperature between the wall
and gas in rarified flows or very high shear rate flows.
In this Chapter we will discuss two fundamental problems involving viscous flow along
a wall at ordinary density and shear rate. The first is plane Couette flow between two
parallel plates where the flow is perfectly parallel. This problem can be solved exactly.
The second is the plane boundary layer along a wall where the flow is almost but not quite
perfectly parallel. In this case an approximate solution to the equations of motion can be
determined.
8.2 Equations of motion
Recall the equations of motion in Cartesian coordinates from Chapter 1; In the absence of
body forces and internal sources of energy
Conservation of mass

+ + + =0
∂t ∂x ∂y ∂y
Conservation of momentum

+ + + =0
∂t ∂x ∂y ∂z

+ + + =0
∂t ∂x ∂y ∂z

+ + + =0
∂t ∂x ∂y ∂z
(8.3)
Conservation of energy
∂ρ (e + k)
+
∂t
∂x
∂ (ρ (e + k) V + P V − τyx U − τyy V − τyz W + Qy )
∂y

=
∂z
{P ower generation due to energy sources inside the control volume} .
Throughout this chapter we will be looking at the steady flow over a flat wall and so the
first simplification is to reduce the problem to the equations governing steady flow in two
space dimensions.
∂ρU ∂ρV
+ =0
∂x ∂y
∂ (ρU U + P − τxx ) ∂ (ρU V − τxy )

+ =0
∂x ∂y
∂ (ρV U − τxy ) ∂ (ρV V + P − τyy )

+ =0 (8.4)
∂x ∂y

∂ (ρhU + Qx ) ∂ (ρhV + Qy ) ∂P ∂P
+ − U +V
∂x ∂y ∂x ∂y

∂U ∂U ∂V ∂V
− τxx + τxy − τxy + τyy =0
∂x ∂y ∂x ∂y
8.3 Plane, compressible Couette flow
First let’s study the two-dimensional, compressible flow produced between two parallel
plates in relative motion as shown in Figure 8.2. This is the simplest possible compressible
flow where viscous forces and heat conduction dominate the motion. We studied the
incompressible version of this flow in Chapter 1. There the gas temperature is constant
and the velocity profile is a straight line. In the compressible problem the temperature
varies between the plates, as does the gas viscosity, leading to a more complex and more
interesting problem.
Figure 8.2: Flow produced between two parallel plates in relative motion.
The upper wall moves at a velocity U∞ doing work on the fluid while the lower wall is
at rest. The temperature of the upper wall is T∞ . The flow is assumed to be steady
with no variation in the z direction. The plates extend to plus and minus infinity in the x
direction.
All gradients in the x direction are zero and the velocity in the y direction is zero.
V =W =0
(8.5)
∂ ∂
() = () = 0
∂x ∂z
With these simplifications the equations of motion reduce to
∂τxy
=0
∂y
∂P
=0 (8.6)
∂y
∂ (Qy − τxy U )
= 0.
∂y
The U velocity component and temperature only depend on y. The y momentum equation
implies that the pressure is uniform throughout the flow (the pressure does not depend
on x or y). The implication of the x momentum balance is that the shear stress τxy also
must be uniform throughout the flow just like the pressure. Assume the fluid is Newtonian.
With V = 0 , the shear stress is
dU
τxy = µ = τw = constant (8.7)
dy
where τw is the shear stress at the lower wall. For gases the viscosity depends only on
temperature.
µ = µ (T ) (8.8)
Since the pressure is constant the density also depends only on the temperature. According
to the perfect gas law
P
ρ (y) = . (8.9)
RT (y)
Using these equations, the solution for the velocity profile can be written as an inte-
gral.
Z y
dy
U (y) = τw (8.10)
0 µ (T )
To determine the velocity profile we need to know how the viscosity depends on temperature
and the temperature distribution. The temperature distribution across the channel can be
determined from the energy equation. Fourier’s law for the heat flux is
dT
Qy = −κ . (8.11)
dy
The coefficient of heat conductivity, like the viscosity is also only a function of tempera-
ture.
κ = κ (T ) (8.12)
The relative rates of diffusion of momentum and heat are characterized by the Prandtl
number.
Cp µ
Pr = (8.13)
κ
In gases, heat and momentum are transported by the same mechanism of molecular colli-
sions. For this reason the temperature dependencies of the viscosity and heat conductivity
approximately cancel in (8.13). The heat capacity is only weakly dependent on temper-
ature and so to a reasonable approximation, the Prandtl number for gases tends to be a
constant close to one. For Air P r = 0.71.
8.3.1 The energy integral in plane Couette flow
The energy equation expresses the balance between heat flux and work done on the
gas.
d
(−Qy + τw U ) = 0 (8.14)
dy
Integrate (8.14).
−Qy + τw U = −Qw (8.15)
The constant of integration, Qw = κ(∂T /∂y)y=0 , is the heat flux on the lower wall. Now
insert the expressions for the shear stress and heat flux into (8.15).

dT dU d 1 1
k + µU =µ Cp T + U 2 = −Qw (8.16)
dy dy dy Pr 2
Integrate (8.16) from the lower wall.

Z y
1 dy
Cp (T − Tw ) + Pr U 2 = −Qw Pr (8.17)
2 0 µ (T )
The quantity Tw is the temperature of the lower wall. The integral on the right of (8.17) can
be replaced by the velocity using (8.10). The result is the so-called energy integral.
1 Qw
Cp (T − Tw ) + Pr U 2 = − Pr U (8.18)
2 τw
At the upper wall, the temperature is T∞ and this can now be used to evaluate the lower
wall temperature.
U∞ 2 Qw

Cp Tw = Cp T∞ + Pr + U∞ (8.19)
2 τw
8.3.2 The adiabatic wall recovery temperature
Suppose the lower wall is insulated so that Qw = 0 . What temperature does the lower
wall reach? This is called the adiabatic wall recovery temperature Twa .
Pr
Twa = T∞ + U∞ 2 (8.20)
2Cp
Introduce the Mach number M∞ = U∞ /a∞ . Now

Twa γ−1
= 1 + Pr M∞ 2 . (8.21)
T∞ 2
Equation (8.21) indicates that the recovery temperature equals the stagnation temperature
at the upper wall only for a Prandtl number of one. The stagnation temperature at the
upper wall is

Tt∞ γ−1
=1+ M∞ 2 . (8.22)
T∞ 2
The recovery factor is defined as
Twa − T∞
= r. (8.23)
Tt∞ − T∞
In Couette flow for a perfect gas with constant Cp equation (8.21) tells us that the recovery
factor is the Prandtl number.
Twa − T∞
= Pr (8.24)
Tt∞ − T∞
Equations (8.19) and (8.20) can be used to show that the heat transfer and shear stress
are related by
Qw Cp (Tw − Twa )
= . (8.25)
τw U∞ P r U∞ 2
Equation (8.25) can be rearranged to read

τw Qw
1 2 = 2Pr ρ∞ U∞ Cp (Tw − Twa )
. (8.26)
2 ρ∞ U∞
The left side of (8.26) is the friction coefficient.
τw
Cf = 1 2 (8.27)
2 ρ∞ U∞
On the right side of (8.26) there appears a dimensionless heat transfer coefficient called
the Stanton number.
Qw
St = (8.28)
ρ∞ U∞ Cp (Tw − Twa )
In order to transfer heat into the fluid, the lower wall temperature must exceed the recovery
temperature. Equation (8.26) now becomes
Cf = 2Pr St . (8.29)
The coupling between heat transfer and viscous friction indicated in (8.29) is a general
property of all compressible flows near a wall. The numerical factor 2 may change and
the dependence on Prandtl number may change depending on the flow geometry and on
whether the flow is laminar or turbulent, but the general property that heat transfer and
viscous friction are coupled is universal.
8.3.3 Velocity distribution in Couette flow
Now that the relation between temperature and velocity is known we can integrate the
momentum relation for the stress. We use the energy integral written in terms of T∞
Qw 1
(U∞ − U ) + Pr U∞ 2 − U 2

Cp (T − T∞ ) = Pr (8.30)
τw 2
or
U2

T Qw 2 U γ−1 2
= 1 + Pr (γ − 1) M∞ 1 − + Pr M∞ 1 − . (8.31)
T∞ U∞ τw U∞ 2 U∞ 2
Recall the momentum equation.
dU
µ (T ) = τw (8.32)
dy
The temperature is a monotonic function of the velocity and so the viscosity can be regarded
as a function of U . This enables the momentum equation to be integrated.
Z U
µ (U ) dU = τw y (8.33)
0
For gases, the dependence of viscosity on temperature is reasonably well approximated by

Sutherland’s law
3
µ T 2 T∞ + TS
= (8.34)
µ∞ T∞ T + TS
where TS is the Sutherland reference temperature, which is 110.4K for Air. An approxi-
mation that is often used is the power law
ω
µ T
= 0.5 < ω < 1.0. (8.35)
µ∞ T∞
Using (8.31) and (8.35), the momentum equation becomes
U ω
U2

γ−1
Z
Qw 2 U 2 τw
1 + Pr (γ − 1) M∞ 1 − + Pr M∞ 1 − 2 dU = y.
0 U∞ τw U∞ 2 U∞ µ∞
(8.36)
For Air the exponent ω is approximately 0.76. The simplest case, and a reasonable ap-
proximation, corresponds to ω = 1. In this case the integral can be carried out explicitly.
For an adiabatic wall Qw = 0, the wall stress and velocity are related by
!
1 U 3

τw U γ−1 U
y= + Pr M∞ 2 − . (8.37)
µ∞ U∞ U∞ 2 U∞ 3 U∞
The shear stress is determined by evaluating (8.37) at the upper wall.

τw γ−1
d = 1 + Pr M∞ 2 (8.38)
µ∞ U∞ 3
The velocity profile is expressed implicitly as
3
U γ−1 2 U 1 U
U∞ + Pr 2 M∞ U∞ − 3 U∞
y
= . (8.39)
d 1 + Pr γ−1
M∞ 2
3
At M∞ → 0 the profile reduces to the incompressible limit U/U∞ = y/d. At high Mach
number the profile is
!
1 U 3
y
3 U
lim = − . (8.40)
M∞ →∞ d 2 U∞ 3 U∞
At high Mach number the profile is independent of the Prandtl number and Mach number.
The two limiting cases are shown in Figure 8.3.
Using (8.38) the wall friction coefficient for an adiabatic lower wall is expressed in terms
of the Prandtl, Reynolds, and Mach numbers.
 
γ−1
1 + Pr 3 M∞ 2
Cf = 2   (8.41)
Re
Figure 8.3: Velocity distribution in plane Couette flow for an adiabatic lower wall and
ω = 1.
The Reynolds number is
ρ∞ U∞ d
Re = (8.42)
µ∞
and can be expressed as
1
ρ∞ U∞ d ρ∞ U∞ 2 dynamic pressure at the upper plate
Re = = 21 U
= . (8.43)
µ∞ 2 µ∞ d
∞ characteristic shear stress
The last result nicely illustrates the interpretation of the Reynolds number as a ratio of
convective to viscous forces.
Notice that if the upper and lower walls are both adiabatic the work done by the upper wall
would lead to a continuous accumulation of energy between the two plates and a continuous
rise in temperature. In this case a steady state solution to the problem would not exist.
The upper wall has to be able to conduct heat into or out of the flow for a steady state
solution to be possible.
8.4 The viscous boundary layer on a wall
The Couette problem provides a useful insight into the nature of compressible flow near
a solid boundary. From a practical standpoint the more important problem is that of a
compressible boundary layer where the flow originates at the leading edge of a solid body
such as an airfoil. A key reference is the classic text Boundary Layer T heory by Hermann
Schlichting. Recent editions are authored by Schlichting and Gersten.
To introduce the boundary layer concept we begin by considering viscous, compressible flow
past a flat plate of length L as shown in Figure 8.4. The question of whether a boundary
layer is present or not depends on the overall Reynolds number of the flow.
ρ∞ U∞ L
ReL = (8.44)
µ∞
Figure 8.4 depicts the case where the thickness of the viscous region, delineated, schemati-
cally by the parabolic boundary, is a significant fraction of the length of the plate and the
Reynolds number based on the plate length is quite low. In this case there is no boundary
layer and the full equations of motion must be solved to determine the flow.
Figure 8.4: Low Reynolds number flow about a thin flat plate of length L. ReL is less than
a hundred or so. The parabolic envelope which extends upstream of the leading edge roughly
delineates the region of rotational flow produced as a consequence of the no slip condition
on the plate.
If the Reynolds number based on plate length is large then the flow looks more like that
depicted in Figure 8.5.
At high Reynolds number the thickness of the viscous region is much less than the length
of the plate.
δ
<< 1 (8.45)
L
Moreover, there is a region of the flow away from the leading and trailing edges of the plate
where the guiding effect of the plate produces a flow that is nearly parallel. In this region
the transverse velocity component is much less than the streamwise component.
V
<< 1 (8.46)
U
Figure 8.5: High Reynolds number flow developing from the leading edge of a flat plate of
length L. ReL is several hundred or more .
The variation of the flow in the streamwise direction is much smaller than the variation in
the cross-stream direction. Thus
∂ () ∂ ()
<<
∂x ∂y
(8.47)
∂ () ∂ ()
U ∼V .
∂x ∂y
Moreover continuity tells us that
∂U ∂V
∼− . (8.48)
∂x ∂y
The convective and viscous terms in the stream wise momentum equation are of the same
order, suggesting the following estimate for δ/L.
ρ∞ U∞ 2 U∞ δ 1
≈µ 2 ⇒ ≈ (8.49)
L δ L (ReL )1/2
With this estimate in mind let’s examine the y momentum equation.
∂ (ρV U − τxy ) ∂ (ρV V + P − τyy )

+ =0 (8.50)
∂x ∂y
Utilizing (8.46) and (8.47), equation (8.50) reduces to

∂ (P − τyy )
= 0. (8.51)
∂y
Now integrate this equation at a fixed position x and evaluate the constant of integration
in the free stream.
P (x, y) = τyy (x, y) + Pe (x) (8.52)
The pressure (8.52) is inserted into the x momentum equation in (8.4). The result is the
boundary layer approximation to the x momentum equation.
∂U ∂U dPe ∂ ∂τxy
ρU + ρV =− + (τxx − τyy ) + (8.53)
∂x ∂y dx ∂x ∂y
Now consider the energy equation.

∂ (ρhU + Qx ) ∂ (ρhV + Qy ) ∂P ∂P ∂U ∂U ∂V ∂V
+ − U +V − τxx + τxy − τxy + τyy =0
∂x ∂y ∂x ∂y ∂x ∂y ∂x ∂y
(8.54)
Using (8.46), (8.47) and (8.48) the energy equation simplifies to
∂h ∂h ∂Qy ∂Pe ∂ ∂ (V τyy ) ∂ (U τxx ) ∂U

ρU +ρV + −U +U (τxx − τyy )− − −τxy = 0. (8.55)
∂x ∂y ∂y ∂x ∂x ∂y ∂x ∂y
In laminar flow the normal stresses τxx and τyy are very small and the normal stress terms
that appear in (8.53) and (8.55) can be neglected. In turbulent flow the normal stresses
are not particularly small, with fluctuations of velocity near the wall that can be 10 to 20
percent of the velocity at the edge of the boundary layer. Nevertheless a couple of features
of the turbulent boundary layer (compressible or incompressible) allow these normal stress
terms in the boundary layer equations to be neglected.
1) τxx and τyy tend to be comparable in magnitude so that τxx − τyy is small and the
streamwise derivative ∂ (τxx − τyy ) /∂x is generally quite small.
2) τyy has its maximum value in the lower part of the boundary layer where V is very small
so the product V τyy is small.
3) τxx also has its maximum value relatively near the wall where U is small and the
streamwise derivative of U τxx is small.
Using these assumptions to remove the normal stress terms, the compressible boundary
layer equations become
∂ρU ∂ρV
+ =0
∂x ∂y
∂U ∂U dPe ∂τxy
ρU + ρV + − =0 (8.56)
∂x ∂y dx ∂y
∂h ∂h ∂Qy dPe ∂U
ρU + ρV + −U − τxy = 0.
∂x ∂y ∂y dx ∂y
The only stress component that plays a role in this approximation is the shearing stress
τxy . In a turbulent boundary layer both the laminar and turbulent shearing stresses are
important.
∂U
τxy = τxy |laminar + τxy |turbulent = µ + τxy |turbulent (8.57)
∂y
In the outer part of the layer where the velocity gradient is relatively small the turbulent
stresses dominate. But near the wall where the velocity fluctuations are damped and the
velocity gradient is large, the viscous stress dominates.
The flow picture appropriate to the boundary layer approximation is shown in Figure
8.6.
Figure 8.6: High Reynolds number flow developing from the leading edge of a semi-infinite
flat plate.
The boundary layer is assumed to originate from a virtual origin near the plate leading edge
and the wake is infinitely far off to the right. The velocity, temperature and pressure at
the boundary layer edge are assumed to be known functions of the stream-wise coordinate.
The appropriate measure of the Reynolds number in this flow is based on the distance from
the leading edge.
ρ∞ U∞ x
Rex = (8.58)
µ∞
One would expect a thin flat plate to produce very little disturbance to the flow. So it is a
bit hard at this point to see the origin of the variation in free stream velocity indicated in
Figure 8.6. We will return to this question at the end of the chapter. For now we simply
accept that the free stream pressure and velocity can vary with x even along a thin flat
plate. By the way, an experimental method for generating a pressure gradient is to put the
plate into a wind tunnel with variable walls that can be set at a geometry to accelerate or
decelerate the flow.
For Newtonian laminar flow, the stress is

2
τij = 2µSij − µ − µv δij Skk
3
(8.59)
∂U ∂V ∼ ∂U
τxy = µ + =µ .
∂y ∂x ∂y
The diffusion of heat is governed by Fourier’s law introduced earlier.
∂T
Qy = −k (8.60)
∂y
For laminar flow, the compressible boundary layer equations are
∂ρU ∂ρV
+ =0
∂x ∂y

∂U ∂U dPe ∂ ∂U
ρU + ρV + − µ =0 (8.61)
∂x ∂y dx ∂y ∂y
∂U 2

∂T ∂T dPe ∂ ∂T
ρU Cp + ρV Cp −U − κ −µ = 0.
∂x ∂y dx ∂y ∂y ∂y
8.4.1 Measures of boundary layer thicknessl
The thickness of the boundary layer depicted in Figure 8.6 is denoted δ. There are several
ways to define the thickness. The simplest is to identify the point where the velocity is
some percentage of the free stream value, say δ0.95 or δ0.99 . This is a somewhat arbitrary
measure and it is generally better to use a thickness defined in terms of an integral over
the boundary layer that converges at the boundary layer edge. The two most common
measures are the displacement thickness and the momentum thickness.
Displacement thickness
Z δ
ρU
δ∗ = 1− dy (8.62)
0 ρe Ue
This is a measure of the distance by which streamlines are shifted away from the plate by
the blocking effect of the boundary layer. This outward displacement of the flow comes
from the reduced mass flux in the boundary layer compared to the mass flux that would
occur if the flow were inviscid. Generally δ ∗ is a fraction of δ0.99 . The integral (8.62) is
terminated at the edge of the boundary layer where the velocity equals the free stream
value Ue . This is important in a situation involving a pressure gradient where the velocity
profile might look something like that depicted in Figure 8.1. The free stream velocity
used as the outer boundary condition for the boundary layer calculation comes from the
potential flow solution for the irrotational flow about the body evaluated at the wall. If
the integral (8.62) is taken beyond this point it will begin to diverge.
M omentum thickness
Z δ
ρU U
θ= 1− dy (8.63)
0 ρe Ue Ue
This is a measure of the deficit in momentum flux within the boundary layer compared to
the free stream value and is smaller than the displacement thickness. The evolution of the
momentum thickness along the wall is directly related to the skin friction coefficient.
8.5 The Von Karman integral momentum equation
Often the detailed structure of the boundary layer velocity profile is not the primary
object of interest. The most important properties of the boundary layer are the skin
friction and the displacement effect. Boundary layer models usually focus primarily on
these variables.
Figure 8.7: Boundary layer velocity and density profiles.
The steady compressible boundary layer equations, excluding the energy equation, are
∂ρU ∂ρV
+ =0
∂x ∂y
(8.64)
∂ρU 2 ∂ρU V dPe ∂τxy
+ + − = 0.
∂x ∂y dx ∂y
Integrate the continuity and momentum equations over the thickness of the boundary
layer.
Z δ(x) Z δ(x)
∂ρU ∂ρV
dy + dy = 0
0 ∂x 0 ∂y
δ(x) Z δ(x) Z δ(x) Z δ(x)

∂ρU 2
Z
∂ρU V dPe ∂τxy
dy + dy + dy − dy = 0
0 ∂x 0 ∂y 0 dx 0 ∂y
(8.65)
Upon integration, the momentum equation becomes
δ(x)
∂ρU 2
Z
dPe
dy + ρe Ue Ve + δ (x) + τxy |y=0 = 0. (8.66)
0 ∂x dx
The partial derivatives with respect to x in (8.65) can be taken outside the integral using
Liebniz’ rule.
Z δ(x) Z δ(x)
d ∂ρU dδ
ρU dy = dy + ρe Ue
dx 0 0 ∂x dx
(8.67)
δ(x) δ(x)
∂ρU 2
Z Z
d dδ
ρU 2 dy = dy + ρe Ue 2
dx 0 0 ∂x dx
Use the second equation in (8.67) in (8.66)
Z δ(x)
d dδ dPe
ρU 2 dy − ρe Ue 2 + ρe Ue Ve + δ (x) = −τw (8.68)
dx 0 dx dx
where τw (x) = τxy (x, 0) and τw is positive. We will make use of the following rela-
tion.
Z δ(x) Z δ(x)
d ∂ρU Ue dδ
(ρU Ue )dy = dy + ρe Ue 2 =
dx 0 0 ∂x dx
Z δ(x)
∂ρU dUe dδ
Ue + ρU dy + ρe Ue 2 = (8.69)
0 ∂x dx dx
δ(x)
dUe δ(x)
Z Z
∂ρU dδ
Ue dy + ρU dy + ρe Ue 2
0 ∂x dx 0 dx
Multiply the continuity equation in (8.65) by Ue and integrate with respect to y.
Z δ(x)
∂ρU
Ue dy + ρe Ue Ve = 0 (8.70)
0 ∂x
Insert (8.70) into the last relation in (8.69).
Z δ(x) Z δ(x)
d dUe dδ
(ρU Ue )dy + ρe Ue Ve − ρU dy − ρe Ue 2 =0 (8.71)
dx 0 dx 0 dx
Subtract (8.71) from (8.68)
Z δ(x) Z δ(x)
d dUe dPe
ρU 2 − ρU Ue dy +

ρU dy + δ (x) = −τw (8.72)
dx 0 dx 0 dx
and subtract the identity
Z δ(x)
dUe dUe
ρe Ue dy − ρe Ue δ (x) = 0 (8.73)
dx 0 dx
from (8.72). Now the integral momentum equation takes the form
Z δ(x) Z δ(x)
d dUe dPe dUe
ρU 2 − ρU Ue dy +

(ρU − ρe Ue ) dy + + ρe Ue δ (x) = −τw .
dx 0 dx 0 dx dx
(8.74)
Recall the definitions of displacement thickness (8.62) and momentum thickness (8.63).
Z δ(x)
∗ ρU
δ (x) = 1− dy
0 ρe Ue
(8.75)
Z δ(x)
ρU U
θ (x) = 1− dy
0 ρe Ue Ue
Substitute (8.75) into (8.74). The result is

d dUe dPe dUe
ρe Ue 2 θ − ρe Ue δ ∗

− + + ρe Ue δ (x) = −τw . (8.76)
dx dx dx dx
At the edge of the boundary layer both ∂U/∂y and τxy go to zero and the x-boundary
layer momentum equation reduces to the Euler equation.
dPe + ρe Ue dUe = 0 (8.77)
Using (8.77), the integral equation (8.76) reduces to
d dUe
ρe Ue 2 θ + ρe Ue δ ∗

= τw . (8.78)
dx dx
The significance of this last step is that (8.78) does not depend explicitly on the perceived
boundary layer thickness but only on the more precisely defined momentum and displace-
ment thicknesses. It is customary to write (8.78) in a slightly different form. Introduce the
wall friction coefficient and carry out the differentiation of the first term in (8.78).
τw
Cf = 1 2 (8.79)
2 ρe Ue
dθ 1 dUe θ dρe Cf
+ (2θ + δ ∗ ) + = (8.80)
dx Ue dx ρe dx 2
If the free stream flow is adiabatic (not necessarily isentropic), then

dTe dUe
+ (γ − 1) Me 2 =0
Te Ue
dPe dUe
+ γMe 2 =0 (8.81)
Pe Ue
dρe dPe dTe dUe

= − = −Me 2
ρe Pe Te Ue
Using the last relation in (8.81) in (8.78), the compressible Von Karman integral momentum
equation is
dθ 1 dUe Cf
2 − Me 2 θ + δ ∗

+ = . (8.82)
dx Ue dx 2
Another common form of (8.82) is generated by introducing the shape factor
δ∗
H= . (8.83)
θ
dθ θ dUe Cf
+ 2 − Me 2 + H = . (8.84)
dx Ue dx 2
8.6 The laminar boundary layer in the limit M 2 → 0.
At very low Mach number the density is constant, temperature variations throughout the
flow are very small, and the boundary layer equations (8.61) reduce to their incompressible
form
∂U ∂V
+ =0
∂x ∂y
(8.85)
∂2U

∂U ∂U 1 dPe
U +V =− +ν
∂x ∂y ρ dx ∂y 2
where the kinematic viscosity ν = µ/ρ has been introduced. The boundary conditions
are
U (0) = V (0) = 0
(8.86)
U (δ) = Ue .
The pressure at the edge of the boundary layer at y = δ is determined using the Bernoulli
relation.
1 1 dPe dUe
Pt = Pe (x) + ρUe (x)2 ⇒ = −Ue (8.87)
2 ρ dx dx
The stagnation pressure is constant in the irrotational flow outside the boundary layer. The
pressure at the boundary layer edge, Pe (x), is assumed to be a given function determined
from a potential flow solution for the flow outside the boundary layer.
The continuity equation is satisfied identically by introducing a stream function.
∂ψ
U=
∂y
(8.88)
∂ψ
V =− .
∂x
In terms of the stream function, the governing momentum equation becomes a third-order
partial differential equation.
dUe
ψy ψxy − ψx ψyy = Ue + νψyyy (8.89)
dx
8.6.1 The zero pressure gradient, incompressible boundary layer
For dUe /dx = 0 the governing equation reduces to
ψy ψxy − ψx ψyy = νψyyy (8.90)
with boundary conditions
ψ (x, 0) = ψy (x, 0) = 0
(8.91)
ψy (x, ∞) = Ue .
We can solve this problem using a symmetry argument. Transform (8.90) using the fol-
lowing three parameter dilation Lie group.
x̃ = ea x
ỹ = eb y (8.92)
ψ̃ = ec ψ
Equation (8.90) transforms as follows.
ψ̃ỹ ψ̃x̃ỹ − ψ̃x̃ ψ̃ỹỹ − ν ψ̃ỹỹỹ = e2c−a−2b (ψy ψxy − ψx ψyy ) − ec−3b (νψyyy ) (8.93)
Equation (8.90) is invariant under the group (8.92) if and only if c = a − b. The boundaries
of the problem at the wall and at infinity are clearly invariant under (8.92).
ỹ = eb y = 0 ⇒ y = 0
ψ̃ (x̃, 0) = ec ψ (ea x, 0) = 0 ⇒ ψ (x, 0) = 0 (8.94)
ψ̃ỹ (x̃, 0) = ec−b ψy (ea x, 0) = 0 ⇒ ψy (x, 0) = 0
The free stream boundary condition requires some care.
ψ̃ỹ (x̃, ∞) = ec−b ψy (ea x, ∞) = Ue (8.95)
Invariance of the free stream boundary condition only holds if c = b. So the problem as a
whole, equation and boundary conditions, is invariant under the one-parameter group
x̃ = e2b x
ỹ = eb y (8.96)
ψ̃ = eb ψ.
This process of showing that the problem is invariant under a Lie group is essentially a
proof of the existence of a similarity solution to the problem. We can expect that the
solution of the problem will also be invariant under the same group (8.96). That is we can
expect a solution of the form

ψ y
√ =F √ . (8.97)
x x
The problem can be further simplified by using the parameters of the problem to non-
dimensionalize the similarity variables. Introduce
ψ = (2νU∞ x)1/2 F (α)

1/2 (8.98)
U∞
α=y .
2νx
In terms of these variables the velocities are
U
= Fα
U∞
1/2 (8.99)
V ν
= (αFα − F ) .
U∞ 2U∞ x
The Reynolds number in this flow is based on the distance from the leading edge. Recall
Equation (8.58).
U∞ x
Rex = (8.100)
ν
As the distance from the leading edge increases, the Reynolds number increases, V /U
decreases, and the boundary layer approximation becomes more and more accurate. The
vorticity in the boundary layer is
U∞ 1/2

∂V ∂U ∼
ω= − = −U∞ Fαα . (8.101)
∂x ∂y 2νx
The remaining derivatives that appear in (8.90) are

U∞
ψxy = − αFαα
2x
1/2
U∞
ψyy = U∞ Fαα (8.102)
2νx
U∞ 2
ψyyy = Fααα .
2νx
Substitute (8.99) and (8.102) into (8.90).
1/2 ! 1/2
U∞ 2

U∞ ν U∞
U∞ Fα − αFαα − U∞ (αFα − F ) U∞ Fαα = ν Fααα
2x 2U∞ x 2νx 2νx
−Fα (αFαα ) + (αFα − F ) Fαα = Fααα
−αFα Fαα + αFα Fαα − F Fαα − Fααα = 0

(8.103)
Canceling terms in (8.103) leads to the Blasius equation
Fααα + F Fαα = 0 (8.104)
subject to the boundary conditions
F (0) = 0
Fα (0) = 0 (8.105)
Fα (∞) = 1.
The numerical solution of the Blasius equation is shown below.

The friction coefficient derived from evaluating the velocity gradient at the wall is
τw 0.664
Cf = 2 =
√ . (8.106)
(1/2) ρU∞ Rex
The transverse velocity component at the edge of the layer is

Figure 8.8: Solution of the Blasius equation (8.104) for the stream function velocity and
stress (or vorticity) profile in a zero pressure gradient laminar boundary layer.
Ve 0.8604
= √ . (8.107)
U∞ Rex
Notice that at a fixed value of x this velocity does not diminish with vertical distance
from the plate which may seem a little surprising given our notion that the region of flow
disturbed by a body should be finite and the disturbance should die away. But remember
that in the boundary layer approximation, the body is semi-infinite. In the real flow over
a finite length plate where the boundary layer solution only applies over a limited region,
the disturbance produced by the plate would die off at infinity.
The various thickness measures of the Blasius boundary layer are
δ0.99 4.906
=√
x Rex
δ∗ 1.7208
= √ (8.108)
x Rex
θ 0.664
=√ .
x Rex
In terms
√ of the similarity variable, the edge of the boundary layer at δ0.99 is at αe =
4.906/ 2 = 3.469.
We can use (8.104) to get some insight into the legitimacy of the boundary layer idea
whereby the flow is separated into a viscous region where the vorticity is nonzero and an
inviscid region where the vorticity is zero as depicted in Figure 8.8. The dimensionless
vorticity (or shear stress) is given in (8.101). Let τ = Fαα . The Blasius equation (8.104)
can be expressed as follows
dτ
= −F dα. (8.109)
τ
Integrate (8.109).
τ Rα
= e− 0 F dα (8.110)
τw
The dimensionless stream function, the left panel in Figure 8.8, can be represented by
F (α) = α − G (α) . (8.111)
The limiting behavior of G is lim G (α) = C1 where C1 is a positive constant related to

α→∞
the displacement thickness of the boundary layer. If we substitute (8.111) into (8.110) and
integrate beyond the edge of the boundary layer the result is

τ Rα α2
= e− 0 (α−G(α))dα = C2 eC1 α− 2 . (8.112)
τw α>αe

Equation (8.112) indicates that the shearing stress and vorticity decay exponentially at the
edge of the layer. This rapid drop-off is the key point because it supports the fundamental
boundary layer assumption noted above that the flow can be separated into two distinct
zones.
8.7 The Falkner-Skan boundary layers
Finally, we address the question of free stream velocity distributions that lead to other
similarity solutions beside the Blasius solution. We again analyze the stream function
equation.
dUe
ψy ψxy − ψx ψyy − Ue − νψyyy = 0 (8.113)
dx
Let
Ue = M x β (8.114)
where M has units
[M ] = L1−β /T. (8.115)

Similarity solutions of (8.113) exist for the class of power law free-stream velocity distri-
butions given by (8.114). This is the well-known Falkner-Skan family of boundary layers
and the exponent β is the Falkner-Skan pressure gradient parameter.
The form of the similarity solution can be determined using a symmetry arguement similar
to that used to solve the zero pressure gradient case. Insert (8.115) into (8.113). The
governing equation becomes
ψy ψxy − ψx ψyy − βM 2 x(2β−1) − νψyyy = 0. (8.116)
Now transform (8.116) using a three parameter dilation Lie group.
x̃ = ea x
ỹ = eb y (8.117)
ψ̃ = ec ψ
Equation (8.116) transforms as
ψ̃ỹ ψ̃x̃ỹ − ψ̃x̃ ψ̃ỹỹ − βM 2 x̃(2β−1) − ν ψ̃ỹỹỹ =

(8.118)
e2c−a−2b (ψy ψxy − ψx ψyy ) − e(2β−1)a βM 2 x(2β−1) − ec−3b (νψyyy ) = 0.
Equation (8.116) is invariant under the group Equation (8.117) if and only if
2c − a − 2b = (2β − 1) a = c − 3b. (8.119)
The boundaries of the problem at the wall and at infinity are invariant under (8.117).
ỹ = eb y = 0 ⇒ y = 0
ψ̃ (x̃, 0) = ec ψ (ea x, 0) = 0 ⇒ ψ (x, 0) = 0 (8.120)
ψ̃ỹ (x̃, 0) = ec−b ψy (ea x, 0) = 0 ⇒ ψy (x, 0) = 0
As in the case of the Blasius problem, the free stream boundary condition requires some
care.
ψ̃ỹ (x̃, ∞) = ec−b ψy (ea x, ∞) = eβa M xβ (8.121)
The boundary condition at the outer edge of the boundary layer is invariant if and only
if
c − b = βa. (8.122)
Solving (8.119) and (8.122) for a and c in terms of b leads to the group
2
b
x̃ = e 1−β x
ỹ = eb y (8.123)
1+β
b
ψ̃ = e 1−β ψ.
We can expect that the solution of the problem will be invariant under the two-parameter
group (8.123). That is we can expect a solution of the form

ψ y
1+β =F 1−β . (8.124)
x( 2 ) x( 2 )
As in the Blasius problem, we use M and ν to nondimensionalize the problem. The

similarity variables are
1/2
M y
α= 1−β
2ν (x + x0 ) 2
(8.125)
1/2
1 ψ
F = 1+β .
2νM (x + x0 ) 2
Upon substitution of (8.125) and (8.114), the streamfunction equation (8.113) becomes
(x + x0 )2β−1 (Fα ((1 + β) F − (1 − β) αFα )α −

(8.126)
Fαα ((1 + β) F − (1 − β) αFα ) − 2β − Fααα ) = 0.
Cancelling terms produces the Falkner-Skan equation

Fααα + (1 + β) F Fαα − 2β(Fα )2 + 2β = 0 (8.127)
F (0) = 0
Fα (0) = 0 (8.128)
Fα (∞) = 1.
Note that β = 0 reduces (8.127) to the Blasius equation. It is fairly easy to reduce the
order of (8.127) by one. The new variables are
φ=F
(8.129)
G = Fα .
Differentiate G with respect to φ.
DG ∂G ∂G ∂G
Dα dG ∂α dα + ∂F dF + ∂Fα dFα Fαα
Dφ
= = ∂β ∂β
= (8.130)
dφ Fα
Dα ∂α dα + ∂F dF
Rearrange (8.130) to read
d2 G Fα Fααα − Fαα 2

1
= =
dφ2 Fα 2 Fα
(8.131)
Fα − (1 + β) F Fαα + 2β(Fα )2 − 2β − Fαα 2
Fα 3
where the Falkner-Skan equation (8.127) has been used to replace the third derivative.
Using (8.129) and (8.130), equation (8.131) can be expressed as the following second order
ODE governing G (φ)

2 1
GGφφ + (1 + β) φGφ + (Gφ ) + 2β −G =0 (8.132)
G

G (0) = 0
(8.133)
G (∞) = 1.
Several velocity profiles are shown in Figure 8.9.
Figure 8.9: Falkner-Skan velocity profiles.
The various measures of boundary layer thickness including shape factor (8.83) and wall
stress, are shown in Figure 8.10.
8.7.1 The case β = −1.
A particularly interesting case occurs when β = −1. The pressure gradient term is
M dUe M2
Ue = ⇒ Ue =− 3 (8.134)
x dx x
and the original variables become

2ν y
α=
|M | x + x0
(8.135)
ψ
F =± .
(2ν |M |)1/2
Figure 8.10: Falkner-Skan boundary layer parameters versus β.
The units of the governing parameter, M̂ = L2 /T , are the same as the kinematic viscosity.
The ratio |M | /ν is the Reynolds number for the β = −1 flow. Note that the Reynolds
number of this flow is a constant independent of space in contrast to flows governed by
other values of β. Equation (8.127) becomes

Fααα ± 2(Fα )2 − 2 = 0. (8.136)
The quantity M is an area flow rate and can change sign depending on whether the flow
is created by a source or a sink. The plus sign corresponds to a source while the minus
sign represents a sink. To avoid an imaginary root, the absolute value of M is used to
non-dimensionalize the stream function in (8.135). The once reduced equation is
G2 Gφφ + G(Gφ )2 ± 2 G2 − 1 = 0

(8.137)
where
φ=F
(8.138)
G = Fα .
Choose new variables.

γ=G
(8.139)
H = Gφ
Differentiate the new variables in (8.139) with respect to φ and divide.
dGφ Gφ 2
dH Hφ + HG dG
dφ + HGφ dφ Gφφ − G − G22 + 2
= = = (8.140)
dγ γφ + γG dG
dφ
Gφ Gφ
Equation (8.137) finally reduces to
−γH 2 ± 2 1 − γ 2

dH
= . (8.141)
dγ γ2H
8.7.2 Falkner-Skan sink flow.
At this point we will restrict ourselves to the case of a sink flow (choose the minus sign
in (8.137) and the plus sign in (8.141)). The flow we are considering is sketched in Figure
8.11.
Figure 8.11: Falkner-Skan sink flow for β = −1.
The negative sign in front of the F in (8.135) insures that the velocity derived from the
stream function is directed in the negative x direction. The first order ODE, (8.141) (with
the minus sign selected) can be broken into the autonomous pair
dH
= −γH 2 − 2 1 − γ 2

ds
(8.142)
dγ
= γ2H
ds
with critical points at (γ, H) = (0, ±1). The phase portrait of (8.142) is shown in Figure
8.12.
Figure 8.12: Phase portrait of the Falker-Skan case β = −1.
Equation (8.141) is rearranged to read
γH 2 + 2 − 2γ 2 dγ + γ 2 H dH = 0

(8.143)
which, by the cross derivative test can be shown to be a perfect differential with the
integral
2 1
C = 2γ − γ 3 + γ 2 H 2 . (8.144)
3 2
Recall that
γ = G = Fα
(8.145)
Fαα
H = Gφ = .
Fα
At the edge of the boundary layer

)
lim Fα = 1
α→∞ ⇒ H (1) = 0. (8.146)
lim Fαα = 0
α→∞
This allows us to evaluate C in (8.144). The result is
4
C= . (8.147)
3
Solve (8.144) for H.
1/2
4γ 4 8
H= − + 2
3 γ 3γ (8.148)
(0 < γ < 1)
where the positive root is recognized to be the physical solution. The solution (8.148) is
shown as the thicker weight trajectory in Figure 8.12. Equation (8.148) can be written
as
r
4 1/2
γH = (γ − 1)2 (γ + 2) . (8.149)
3
In terms of the original variables
r
4 1/2
Fαα = (Fα − 1)2 (Fα + 2) (8.150)
3
and
r ! r !
−1 Fα + 2 −1 2
α = T anh − T anh . (8.151)
3 3
The latter result can be solved for the negative of the velocity, Fα .
r !
2 −1 2
Fα = 3T anh α + T anh −2 (8.152)
3
This is shown plotted in Figure 8.13.

Figure 8.13: Falkner-Skan sink flow velocity profile β = −1.
The Falkner-Skan sink flow represents one of the few known exact solutions of the boundary
layer equations. However the fact that an exact solution exists for the case β = −1 is no
accident. Neither is the fact that this case corresponds to an independent variable of the
form α ≈ y/x where both coordinate directions are in some sense equivalent. Remember
the essence of the boundary layer approximation is that the streamwise direction is in a
sense ”convective” while the transverse direction is regarded as ”diffusive” producing a
flow that is progressively more slender in the y direction as x increases. In the case of the
Falkner-Skan sink flow the aspect ratio of the flow is constant.
8.8 Thwaites’ method for approximate calculation of lami-

nar boundary layer characteristics
At the wall, the momentum equation reduces to
∂ 2 U

Ue dUe
2
=− . (8.153)
∂y y=0 ν dx
Rewrite (8.84) as
Ue dUe Ue 2 dθ

∂U
= (2 + H) θ + . (8.154)
∂y y=0 ν dx ν dx
Choose θ and Ue as length and velocity scales to non-dimensionalize the left sides of (8.153)
and (8.154).
θ2 ∂ 2 U θ2 dUe

2
=−
Ue ∂y y=0 ν dx
(8.155)
θ2 dUe Ue dθ2

θ ∂U
= (2 + H) +
Ue ∂y y=0 ν dx 2ν dx
In a landmark paper in 1948 Bryan Thwaites argued that the normalized derivatives on the
left of (8.155) should depend only on the shape of the velocity profile and not explicitly on
the free stream velocity or thickness. Moreover he argued that there should be a universal
function relating the two. He defined
θ2 ∂ 2 U

m=
Ue ∂y 2 y=0
(8.156)
θ ∂U
l (m) = .
Ue ∂y y=0
In terms of m the Von Karman equation is
Ue dθ2
= 2 ((2 + H) m + l (m)) = L (m) (8.157)
ν dx
where (8.153) has been used. Thwaites proceeded to examine a variety of known exact and
approximate solutions of the boundary layer equations with a pressure gradient. The main
results are shown in Figure 8.14.
The correlation of the data was remarkably good, especially the near straight line behavior
of L (m). Thwaites proposed the linear approximation
L (m) = 0.45 + 6m. (8.158)
One of the classes of solutions included in Thwaites’ data is the Falkner-Skan boundary
layers discussed earlier. For these solutions the Thwaites functions can be calculated ex-
plicitly.
Figure 8.14: Data on skin friction collected by Thwaites: l (m) , shape factor H (m) and
L (m) for a variety of boundary layer solutions.
Z α 2 Z α 2
m = Fααα (0) Fα (1 − Fα ) dα = −2β Fα (1 − Fα ) dα
0 0
Z α
l (m) = Fαα (0) Fα (1 − Fα ) dα (8.159)
0
Rα
(1 − Fα ) dα
H (m) = R α0
0 Fα (1 − Fα ) dα
The various measures of the Falkner-Skan solutions shown in Figure 8.10 can be related to
m instead of β using the first equation in (8.159). The relation between m and β is shown
in Figure 8.15.
Figure 8.15: The variable m defined in (8.156) versus the free stream velocity exponent β
for Falkner-Skan boundary layers.
The functions l (m) and H (m) for the Falkner-Skan boundary layer solutions are shown in
Figure 8.16.
Thwaites’ correlations were re-examined by N. Curle who came up with a very similar set
of functions but with slightly improved prediction of boundary layer evolution in adverse
pressure gradients. Figure 8.17 provides a comparison between the two methods.
Curle’s tabulation of his functions for Thwaites’ method is included in Figure 8.18.
A reasonable linear approximation to the data for L (m) in Figure 8.18 is
Figure 8.16: Thwaites functions for the Falkner-Skan solutions (8.159).
Figure 8.17: Comparison between Curle’s functions and Thwaite’s functions.

Figure 8.18: Curle’s functions for Thwaites’ method.
L (m) = 0.441 + 6m. (8.160)
Equation (8.160) is not the best linear approximation to Curle’s data in Figure 8.18 but
is consistent with√the value of the friction coefficient for the zero-pressure gradient Blasius
boundary layer( 0.441 = 0.664). Insert (8.160) into (8.157) and use the definition of m
in (8.155) and (8.156).
θ2 θ2

d dUe
Ue = 0.441 − 6 (8.161)
dx ν ν dx
Equation (8.161) can be integrated exactly.

Z x
0.441ν 5
θ2 = Ue x0 dx0 (8.162)
Ue (x)6 0
Equation (8.60) provides the momentum thickness of the boundary layer directly from
the distribution of velocity outside the boundary layer Ue (x). Although the von Karman
integral equation (8.157) was used to generate the data for Thwaites method, it is no longer
needed once (8.60) is known.
The procedure for applying Thwaites’ method is as follows.
1) Given Ue (x) use (8.162) to determine θ2 (x).
At a given x:
2) The parameter m is determined from (8.156) and (8.153).
θ2 dUe
m=− (8.163)
ν dx
3) The functions l (m) and H (m) are determined from the data in Figure 8.18.
4) The friction coefficient is determined from
2ν
Cf = l (m) . (8.164)
Ue θ
5) The displacement thickness δ ∗ (m) is determined from H (m).

The process is repeated while progressing along the wall to increasing values of x. Sep-
aration of the boundary layer is assumed to have occurred if a point is reached where
l (m) = 0.
The key references used in this section are
1) Thwaites, B. 1948 Approximate calculations of the laminar boundary layer, VII Inter-
national Congress of Applied Mechanics, London. Also Aeronautical Quarterly Vol. 1,
page 245, 1949.
2) Curle, N. 1962 T he Laminar Boundary Layer Equations, Clarendon Press.
8.8.1 Example - free stream velocity from the potential flow over a cir-
cular cylinder.
To illustrate the application of Thwaites’ method let’s see what it predicts for the flow over a
circular cylinder where we take as the free stream velocity the potential flow solution.
Ue x
= 2Sin (8.165)
U∞ R
According to (8.162)
2 Z φ
θ 0.441
Sin5 φ0 dφ0

Re = 6
(8.166)
R Sin (φ) 0
where
Figure 8.19: Example for Thwaites’ method.
U∞ (2R)
Re = . (8.167)
ν
Near the forward stagnation point
2
0.441 φ 05 0 0.441
Z
θ
lim Re = φ dφ = . (8.168)
φ→0 R φ6 0 6
Interestingly the method gives a finite momentum thickness at the stagnation point. This
is useful to know when we apply the method to an airfoil where the radius of the leading
edge at the forward stagnation point will define the initial thickness for the boundary layer
calculation. Next the relationship between m and φ is determined using (8.163).
θ2 dUe 1 θ 2 0.441Cos (φ) φ

Z
d Ue
Sin5 φ0 dφ0

m=− =− Re =− 6
(8.169)
ν dx 2 R dφ U∞ Sin (φ) 0
Once m (φ) is known, l (m (φ)) is determined from the data in Figure 8.18. The friction
coefficient determined using (8.164) is shown in Figure 8.21.
According to this method the boundary layer on the cylinder would separate. The boundary
layer parameters, momentum thickness, displacement thickness, shape factor and friction
coefficient are shown in figures 8.21 and 8.22. But the predicted separation point is well
beyond the separation point measured in experiments. Part of the reason for this is that
the pressure distribution is forced to be the potential solution. In practice the boundary
Figure 8.20: Thwaites’ functions for the freestream distribution (8.165).
Figure 8.21: Friction coefficient for the freestream distribution (8.165).

layer and the free stream interact with each other leading to a forward surface pressure
distribution that is quite different from the potential flow solution. If the experimentally
measured pressure is used, the predicted separation point is much closer to the measured
position of about φseparation = 83◦ .
Figure 8.22: Boundary layer thicknesses and shape factor for the freestream distribution
(8.165).
Notice the role of the Reynolds number. The momentum thicknesspand displacement
thickness as well as the friction coefficient are all proportional to 1/ Re . We saw this
same behavior in the Blasius solution, (8.106) and (8.108).
8.9 Compressible laminar boundary layers
The equations of motion for laminar flow are
∂ρU ∂ρV
+ =0
∂x ∂y

∂U ∂U dPe ∂ ∂U
ρU + ρV + − µ =0 (8.170)
2
∂T ∂T dPe ∂ ∂T ∂U
ρU Cp + ρV Cp −U − κ −µ = 0.
∂x ∂y dx ∂y ∂y ∂y
Using the same approach as in the Couette flow problem, let the temperature in the
boundary layer layer be expressed as a function of the local velocity, T = T (U ). Substitute
this functional form into the energy equation.
∂U 2

∂U ∂U dT dPe ∂ dT ∂U
ρU + ρV Cp −U − κ −µ =0 (8.171)
∂x ∂y dU dx ∂y dU ∂y ∂y
Use the momentum equation to replace the factor in parentheses on the left hand side of
(8.171)
d2 T ∂U 2 ∂U 2

dPe ∂ ∂U dT dPe dT ∂ ∂U
− + µ Cp −U − κ −κ 2 −µ =0
dx ∂y ∂y dU dx dU ∂y ∂y dU ∂y ∂y
(8.172)
which we can write as
∂U 2
2
dPe dT dT ∂ κ ∂U d T
− Cp + U + Cp µ− + k 2 +µ = 0. (8.173)
dx dU dU ∂y Cp ∂y dU ∂y
Introduce the Prandtl number (8.13) which can be assumed to be constant independent of
position in the boundary layer. The energy equation becomes
∂U 2
2
dPe dT dT Pr − 1 ∂ ∂U d T
− Cp + U + Cp µ + k 2 +µ = 0. (8.174)
dx dU dU Pr ∂y ∂y dU ∂y
There are several important cases to consider.
8.9.1 The energy integral for a compressible boundary layer with an

adiabatic wall and Pr = 1.
In this case (8.174) reduces to
∂U 2
2
dPe dT d T
− Cp +U + k 2 +µ = 0. (8.175)
dx dU dU ∂y
The flow at the wall satisfies
U |y=0 = 0
(8.176)
dT dT ∂U
= = 0.
dy y=0 dU ∂y y=0
The velocity gradient at the wall is finite as is the wall shear stress (∂U/∂y)y=0 6= 0 so the
second condition in (8.176) implies that (∂T /∂y)y=0 = 0. The pressure gradient along the
wall is not necessarily zero so flow conditions at the edge of the boundary layer can vary
with the streamwise coordinate. The temperature must satisfy
d2 T µ
=− (8.177)
dU 2 κ
and
dT U
=− . (8.178)
dU Cp
Both (8.177) and (8.178) are consistent with the definition of the Prandtl number and the
assumption Pr = 1 and integrate to
U2
Twa − T = (8.179)
2Cp
where Twa is the adiabatic wall temperature defined in (8.122). This temperature can be
expressed in terms of the temperature and velocity at the edge of the boundary layer.
Ue 2
Twa = Te + (8.180)
2Cp
Introduce the Mach number at the edge of the boundary layer, Me = Ue /ae . Now

Twa γ−1 Tte
=1+ Me 2 = . (8.181)
Te 2 Te
For a Prandtl number of one, the stagnation temperature is constant through the boundary
layer at the value at the boundary layer edge.
8.9.2 Non-adiabatic wall with dP/dx = 0 and Pr = 1
Again, although for different reasons than in the previous section, the temperature is
governed by
d2 T µ
=− . (8.182)
dU 2 κ
In this case, the temperature profile across the boundary layer is
U2

dT
Cp (T − Tw ) + = Cp U (8.183)
2 dU y=0
where Tw is the temperature at the no-slip wall. The heat transfer at the wall is

∂T dT ∂U κ ∂U dT
Qw = −κ = −κ =− µ . (8.184)
∂y y=0 dU ∂y y=0 µ ∂y y=0 dU y=0
Thus

dT Cp µ Qw
Cp =− . (8.185)
dU y=0 κ τw
The temperature profile with heat transfer is
U2 Qw
Cp (T − Tw ) + =− U (8.186)
2 τw
which is identical to the temperature profile (8.18) derived for the Couette flow case for Pr =
1. Evaluate (8.186) at the edge of the boundary layer where the velocity and temperature
Ue and Te are the same as the free stream values U∞ and T∞ . The result is
U∞ 2 Qw U∞
Tw = T∞ + + . (8.187)
2Cp τw Cp
The Stanton number was defined in the previous section on Couette flow.
Qw
St = (8.188)
ρ∞ U∞ Cp (Tw − Twa )
The adiabatic wall temperature is the free stream stagnation temperature. Using (8.187)
in (8.188) gives the friction coefficient in terms of the Stanton number
Cf = 2St (8.189)
which can be compared with the Couette flow result (8.29).

Using (8.186) and (8.187) the temperature profile can be expressed in terms of the free
stream and wall temperatures as follows.
U∞ 2

T − Tw Tw U U U
= 1− + 1− (8.190)
T∞ T∞ U∞ 2Cp T∞ U∞ U∞
8.10 Mapping a compressible to an incompressible boundary

layer
In the late 1940’s L. Howarth (P roc. R. Soc. London A 194, 16-42, 1948) and K. Stewartson
(P roc. R. Soc. London A 200, 84-100, 1949) introduced a remarkable transformation that
can be used to map the compressible boundary layer equations to the incompressible form
including the effects of free stream velocity variation. The basic idea is to define a stream
function for a virtual incompressible flow that carries the same mass flow, integrated to
the wall, as the real compressible flow.
Figure 8.23 illustrates the idea. To satisfy the mass balance requirement, the constant
density of the virtual flow is taken to be the stagnation density of the real flow at the edge
of the boundary layer.
1/(γ−1)
γ−1 2
ρt = ρe 1 + Me (8.191)
2
The flow at the edge of the boundary layer is assumed to be isentropic.
γ/(γ−1) 2γ/(γ−1)
Pt Tt at
= = (8.192)
Pe Te ae
The pressure through the boundary layer is constant, P = Pe and P̃ = P̃e .

Sutherland’s law is referenced to the stagnation temperature Tt of the compressible flow
at the edge of the boundary layer.
3/2
µ T Tt + TS
= (8.193)
µt Tt T + TS
The viscosity of the virtual flow, µt , is the viscosity of the gas evaluated at Tt . The key
assumption needed to make the mapping work is that the viscosity of the gas is linearly
proportional to temperature.
Figure 8.23: Mapping of a compressible to an incompressible flow.
µ T
=σ (8.194)
µt Tt
The constant σ is chosen to provide the best approximation to (8.193).
1/2
Tw Tt + TS
σ= (8.195)
Tt Tw + TS
If Pr = 1 then σ = 1 . The continuity and momentum equations governing the compressible

flow are
∂ρU ∂ρV
+ =0
∂x ∂y
(8.196)
∂U ∂U 1 dPe 1 ∂ ∂U 1 ∂τxy
U +V + − µ − =0
∂x ∂y ρ dx ρ ∂y ∂y ρ ∂y
where it is understood that τxy refers to the turbulent shearing stress. The coordinates of
the virtual flow are defined as
Z x
Pe ae
x̃ = σ dx0 = f (x)
0 Pt at
(8.197)
y
ρ (x, y 0 )
Z
ae
ỹ = dy 0 = g (x, y) .
at 0 ρt
Note that ỹ is a function of x and y since the density depends on both spatial variables.
The variables x0 and y 0 are dummy variables of integration. From the fundamental theorem
of calculus the derivatives of the coordinates are

∂ x̃ Pe ae
= fx = σ
∂x Pt at
∂ x̃
= fy = 0
∂y
(8.198)
∂ ỹ
= gx
∂x

∂ ỹ ae ρ
= gy = .
∂y at ρt
Each of the derivatives is known explicitly except ∂ ỹ/∂x. However, as we will see in the
analysis to follow, terms that involve this derivative will cancel. The continuity equation is
satisfied identically through the introduction of a compressible stream function. Let
∂ψ
ρU = ρt
∂y
(8.199)
∂ψ
ρV = −ρt .
∂x
The stream function of the virtual incompressible flow is a function of x̃ (x) and ỹ (x, y)
and has the same value as its counterpart in the real compressible flow since there is the
same mass flow between the streamline and the wall in the two flows.
ψ (x, y) = ψ̃ (x̃ (x) , ỹ (x, y)) (8.200)
The mapping we are about to carry out is a bit complicated. In the discussion below I
have tried to retain as much detail as possible to help the reader get through the derivation
without requiring a lot of side work to see how one relation follows from another. Maybe
there is more detail than needed but I decided to err on the side of more rather than less
to try to give maximum help to the reader. The partial derivatives of the stream function
are
∂ψ ∂ ψ̃ ∂ x̃ ∂ ψ̃ ∂ ỹ ∂ ψ̃ dx̃ ∂ ψ̃ ∂ ỹ
= + = +
∂x ∂ x̃ ∂x ∂ ỹ ∂x ∂ x̃ dx ∂ ỹ ∂x
(8.201)
∂ψ ∂ ψ̃ ∂ x̃ ∂ ψ̃ ∂ ỹ ∂ ψ̃ ∂ ỹ
= + = .
∂y ∂ x̃ ∂y ∂ ỹ ∂y ∂ ỹ ∂y
The velocities can be expressed in terms of the new coordinates as

ρt ∂ψ ae ∂ ψ̃ ae
U= = = Ũ
ρ ∂y at ∂ ỹ at
(8.202)

ρt ∂ψ ρt Pe ae ∂ ψ̃ ρt ∂ ψ̃ ∂ ỹ
V =− = −σ − .
ρ ∂x ρ Pt at ∂ x̃ ρ ∂ ỹ ∂x
The streamwise velocity in the virtual flow is Ũ = ∂ ψ̃/∂ ỹ. The mass flow between the
streamline and the wall in the two flows is depicted in Figure 8.23.
Z y Z ỹ
ρU dy = ρt Ũ dỹ (8.203)
0 0
Differentiate (8.203) at a fixed x and use (8.202). The result is

dỹ ρU ae ρ
= = (8.204)
dy ρt Ũ at ρt
which is consistent with the partial derivative ∂ ỹ/∂y in (8.198). Equation (8.204) confirms
that the definition of ỹ defined in (8.197) insures that the streamline values in the two flows
are the same, that (8.200) holds. Use the chain rule to determine the first derivatives of U
that appear in (8.196).
! !
∂U ∂ ae ∂ ψ̃ dx̃ ∂ ae ∂ ψ̃ ∂ ỹ
= + =
∂x ∂ x̃ at ∂ ỹ dx ∂ ỹ at ∂ ỹ ∂x
2 ! !
∂ 2 ψ̃ ∂ 2 ψ̃ ∂ ỹ

Pe ae 1 ∂ae ∂ ψ̃ ae
σ + + (8.205)
Pt at ae ∂ x̃ ∂ ỹ ∂ x̃∂ ỹ at ∂ ỹ 2 ∂x
! ! 2 !
∂ 2 ψ̃

∂U ∂ ae ∂ ψ̃ dx̃ ∂ ae ∂ ψ̃ ∂ ỹ ae ρ
= + =
∂y ∂ x̃ at ∂ ỹ dy ∂ ỹ at ∂ ỹ ∂y at ρt ∂ ỹ 2
Now we can form the convective terms in (8.196).
∂U ∂U
U +V =
∂x ∂y
 !2 !
! !
Pe ae 3  1 ∂ae ∂ ψ̃ ∂ 2 ψ̃ ∂ ψ̃  ae 2 ∂ ψ̃ ∂ 2 ψ̃ ∂ ỹ
σ + + −
Pt at ae ∂ x̃ ∂ ỹ ∂ x̃∂ ỹ ∂ ỹ at ∂ ỹ ∂ ỹ 2 ∂x
2 ! 2 !!
∂ 2 ψ̃ ∂ 2 ψ̃ ∂ ψ̃ ∂ ỹ

Pe ae ae ∂ ψ̃ ae
σ +
Pt at at ∂ ỹ 2 ∂ x̃ at ∂ ỹ 2 ∂ ỹ ∂x
(8.206)
Note the terms involving ∂ ỹ/∂x cancel in (8.206). Thus
∂U ∂U
U +V =
∂x ∂y

3 ! !2 !  (8.207)
Pe ae  1 ∂ae ∂ ψ̃ ∂ 2 ψ̃∂ ψ̃ ∂ 2 ψ̃ ∂ ψ̃ 
σ + − .
Pt at ae ∂ x̃ ∂ ỹ ∂ x̃∂ ỹ ∂ ỹ ∂ ỹ 2 ∂ x̃
Now consider the pressure gradient term. Assume the flow at the edge of the boundary
layer is isentropic.
2γ
Pe ae γ−1
= (8.208)
Pt at
The pressure gradient term can be expressed in terms of the speed of sound at the boundary
layer edge.
γ+1 γ+1
dPe 2γPt ae γ−1 1 dae dx̃ Pe ae 2γPt ae γ−1 1 dae
= =σ
dx γ − 1 at at dx̃ dx Pt at γ − 1 at at dx̃
(8.209)
2 ! γ+1
dPe Pe ae 2γPt ae γ−1 1 dae
=σ
dx Pt at γ − 1 at ae dx̃
Now
3 ! ! !
∂ 2 ψ̃ ∂ ψ̃ ∂ 2 ψ̃

∂U ∂U 1 dPe Pe ae ∂ ψ̃
U +V + =σ − +
∂x ∂y ρ dx Pt at ∂ x̃∂ ỹ ∂ ỹ ∂ ỹ 2 ∂ x̃
 !2  (8.210)
3 ! 2
Pe ae  ∂ ψ̃ 1 2γPt ae γ−1
 1 dae .
σ +
Pt at ∂ ỹ ρ γ − 1 at ae dx̃
Note that, from (8.202).
!2 2
∂ ψ̃ 2 at
=U . (8.211)
∂ ỹ ae
So far
3 ! ! !
∂ 2 ψ̃ ∂ ψ̃ ∂ 2 ψ̃

U +V + =σ − +
∂x ∂y ρ dx Pt at ∂ x̃∂ ỹ ∂ ỹ ∂ ỹ 2 ∂ x̃
(8.212)
3 ! 2 2 !
Pe ae at 1 2γPt ae γ−1 1 dae
σ U2 + .
Pt at ae ρ γ − 1 at ae dx̃
The flow at the edge of the boundary layer is adiabatic

2 2 γ−1
at = ae + Ue 2 (8.213)
2
and

1 dae γ−1 dUe
=− Ue . (8.214)
ae dx̃ 2ae 2 dx̃
Now
3 ! ! !
∂ 2 ψ̃ ∂ ψ̃ ∂ 2 ψ̃

U +V + =σ − −
∂x ∂y ρ dx Pt at ∂ x̃∂ ỹ ∂ ỹ ∂ ỹ 2 ∂ x̃
(8.215)
3 ! 2 2 !
Pe ae γ−1 at 1 γPt ae γ−1 dUe
σ U2 + Ue .
Pt at 2ae 2 ae ρ ae 2 at dx̃
Note that
2 2
2 γ−1 at 1 γPt ae γ−1
U + =
2ae 2 ae ρ ae 2 at
2 2γ 2
2 γ−1 at ρe γPe at γ−1 ae γ−1
U + = (8.216)
2ae 2 ae ρ ρe ae 2 ae at
 
γ−1

at
4 a2 + 2 U2
 
ae at 2
where we have used the isentropic relation Pt /Pe = (at /ae )2γ/(γ−1) and γPe /ρ = γP/ρ = a2 .
Now work out the viscous term using ρT = P/R = Pe /R.
! !
∂U T ∂ ae ∂ ψ̃ T ae ∂ ∂ ψ̃
τxy |laminar =µ = σµt = σµt =
∂y Tt ∂y at ∂ ỹ Tt at ∂y ∂ ỹ
2 ! 2 ! 2
ae T ρ ∂ ∂ ψ̃ ae ρT ∂ Ũ ae Pe
σµt =σ µt =σ τ̃x̃ỹ |laminar
at Tt ρt ∂ ỹ ∂ ỹ at ρt Tt ∂ ỹ at Pt
(8.217)
Recall Ũ = ∂ ψ̃/∂ ỹ and τ̃ x̃ỹ|laminar = µt ∂ Ũ /∂ ỹ. The viscous stress term in the boundary
layer equation is
3 !
∂ 3 ψ̃

1 ∂ ∂U µt P e ae
µ =σ . (8.218)
ρ ∂y ∂y ρt Pt at ∂ ỹ 3
The turbulent stress term transforms as
2
ae Pe
τxy |turbulent =σ τ̃x̃ỹ |turbulent (8.219)
at Pt
which can be expressed as
3
1 ∂ ae Pe ∂
τxy |turbulent = σ τ̃x̃ỹ |turbulent . (8.220)
ρ ∂y at ρt Pt ∂ ỹ
Now the boundary layer momentum equation (8.196)becomes

∂U ∂U 1 dPe 1 ∂ ∂U 1 ∂τxy
U +V + − µ − =
∂x ∂y ρ dx ρ ∂y ∂y ρ ∂y
3 ! ! !
∂ 2 ψ̃ ∂ ψ̃ ∂ 2 ψ̃

Pe ae ∂ ψ̃
σ − −
Pt at ∂ x̃∂ ỹ ∂ ỹ ∂ ỹ 2 ∂ x̃
3 !
 
4 a2 + γ−1 U 2
 (8.221)
Pe ae  at 2
 Ue dUe −
σ 
2
Pt at ae at dx̃
3 ! ! 3 !
∂ 3 ψ̃

Pe ae µt Pe ae 1 ∂
σ −σ τ̃x̃ỹ |turbulent = 0.
Pt at ρt ∂ ỹ 3 Pt at ρt ∂ ỹ
Drop the common factor multiplying each term on there right of (8.221). The transformed
equation is nearly in incompressible form.
 
γ−1
a2 + U2
! ! 4
∂ 2 ψ̃ ∂ 2 ψ̃

∂ ψ̃ ∂ ψ̃ at 2
 Ue dUe −
− − 
∂ x̃∂ ỹ ∂ ỹ ∂ ỹ 2 ∂ x̃ ae at 2 dx̃
(8.222)
!
µt ∂ 3 ψ̃ 1 ∂
− τ̃x̃ỹ |turbulent = 0
ρt ∂ ỹ 3 ρt ∂ ỹ
The real and virtual free stream velocities are related by
at
Ũe = Ue (8.223)
ae
which comes from the expression for U in (8.202). Differentiate (8.223)
2 4
dŨe at 1 2 γ−1 2 dUe at dUe
Ũe = ae + Ue Ue = Ue (8.224)
dx̃ ae ae 2 2 dx̃ ae dx̃
and substitute (8.224) into (8.222).

 
γ−1
a2 + U2
! !
∂ 2 ψ̃ ∂ ψ̃ ∂ 2 ψ̃ ∂ ψ̃ 2
 Ũe dŨe −
− −
∂ x̃∂ ỹ ∂ ỹ ∂ ỹ 2 ∂ x̃ at 2 dx̃
(8.225)
!
µt ∂ 3 ψ̃ 1 ∂
− τ̃x̃ỹ |turbulent = 0
ρt ∂ ỹ 3 ρt ∂ ỹ
The velocities in the virtual flow are
∂ ψ̃
Ũ =
∂ ỹ
(8.226)
∂ ψ̃
Ṽ = − .
∂ x̃
The transformed boundary layer momentum equation finally becomes
!!  2 γ−1 2  !
∂ Ũ ∂ Ũ a + 2 U ∂ 2 Ũ
 Ũe dŨe −νt 1 ∂
Ũ + Ṽ − − τ̃x̃ỹ |turbulent = 0.
∂ x̃ ∂ ỹ a 2+ γ−1
U 2 dx̃ ∂ ỹ 2 ρt ∂ ỹ
e 2 e
(8.227)
For an adiabatic wall and Pr = 1 the factor in brackets is equal to one. In this case the
momentum equation maps exactly to the incompressible form
!! !
∂ Ũ ∂ Ũ dŨe ∂ 2 Ũ 1 ∂
Ũ + Ṽ − Ũe − νt − τ̃x̃ỹ |turbulent = 0 (8.228)
∂ x̃ ∂ ỹ dx̃ ∂ ỹ 2 ρt ∂ ỹ
Ũ (0) = 0
Ṽ (0) = 0 (8.229)

Ũ δ̃ = Ũe .
The implication of (8.228) and (8.229) is that the effects of compressibility on the boundary
layer can be almost completely accounted for by the scaling of coordinates presented in
(8.197) which is driven in the y direction by the decrease in density near the wall due to
heating and in the x direction by the isentropic changes in free stream temperature and
boundary layer pressure due to flow acceleration or deceleration imposed by the surrounding
potential flow.
Let’s use this transformation to relate the skin friction in the compressible case to the
incompressible skin friction. The definition of the friction coefficient is

2
1 at Pt
σ ae τw
Pe

τ̃w 1 ρe Pt τw 1 Tt
C̃f = = 2 = 2 = Cf .
(1/2) ρt Ũe2 1
ρt at 2 σ ρt Pe (1/2) ρt Ue σ Te
2 ρe ρe ae Ue
(8.230)
Recall that for Pr = 1 the constant σ = 1. In terms of the Mach number, the ratio of
friction coefficients in the real compressible flow and the virtual incompressible flow is
Cf 1
= . (8.231)
C̃f γ−1
1+ 2 Me 2
The physical velocity profiles in the compressible flow cannot be determined without solving
for the temperature in the boundary layer. Here we will restrict our attention to the lam-
inar, zero pressure gradient, Blasius case, dUe /dx = 0, τ xy|turbulent = 0. The temperature
equation was integrated earlier to give
1 2
T = Tt − U . (8.232)
2Cp
Equation (8.232) can be expressed as
2 !
T γ−1 U
=1+ Me 2 1− . (8.233)
Te 2 Ue
Using the first relation in (9.18) U/Ue = Ũ /Ũe and the equality ρT = ρe Te (8.233) be-
comes
 !2 

T γ−1 Ũ  = ρe .
=1+ Me 2  1 − (8.234)
Te 2 Ũe ρ
Our goal is to relate the wall normal coordinate in compressible and incompressible flows.
From (8.198)
  !2 

at ρt ρe at ρt γ−1 Ũ
dy = dỹ = 1 + Me 2 1 −  dỹ.
ae ρe ρ ae ρe 2 Ũe
(8.235)
The spatial similarity variable in the virtual flow is
!1/2
Ũe
α̃ = ỹ . (8.236)
2νt x̃
Now (8.235) becomes
  !2    !1/2 
1/2
at ρt 2νt x̃ γ−1 Ũ  d ỹ Ũe
dy = 1 + Me 2 1 − 
ae ρe Ũe 2 Ũe 2νt x̃
(8.237)
and we can write
1/2 !
Ue
d y =
2νe x
  !2    !1/2 
1/2
at ρt 2νt x̃ Ue γ−1 Ũ  d ỹ Ũe
1 + Me 2 1 − .
ae ρe Ũe 2νe x 2 Ũe 2νt x̃
(8.238)
Rearrange the coefficient on the right hand side of (8.238) using (8.197) and (8.202).
1/2 1/2
at ρt 2νt x̃ Ue at ρt µt ρe Ue x̃
= =
ae ρe Ũe 2νe x ae ρe µe ρt Ũe x
(8.239)
1/2 1/2
at ρt Tt ρe ae Pe ae ρt Tt ρe ρe Te
= =1
ae ρe Te ρt at Pt at ρe Te ρt ρt Tt
Finally
  !2 

γ−1 Ũ
dα = 1 + Me 2  1 −  dα̃. (8.240)
2 Ũe
Integrate (8.240). The similarity variables in the real and virtual flows are related by
 !2 
α̃
γ−1
Z
Ũ  dα̃0 .
α (α̃) = α̃ + Me 2 1 − (8.241)
2 0 Ũe
The velocity ratio, Ũ /Ũe is a known function F̃α̃ (α̃) from the√Blasius solution. The outer
edge of the incompressible boundary layer is at α̃e = 4.906/ 2 = 3.469. The integral of
the velocity term in (8.241) is
 !2 
Z α̃e
1 − Ũ  dα̃0 = 1.67912. (8.242)
0 Ũe
This allows us to determine how the thickness of the compressible layer depends on Mach
number. The outer edge of the compressible boundary layer is at
 !2 
α̃e
γ−1 γ−1
Z
2 Ũ  dα̃0 = 3.469+1.67912
αe = α̃e + Me 1 − Me 2 . (8.243)
2 0 Ũe 2
The thickness of the compressible layer grows rapidly with Mach number. Knowing α (α̃)
enables the temperature and density of the compressible layer to be determined.
 !2 

T (α (α̃)) ρe γ−1 Ũ (α̃)
= =1+ Me 2  1 −  (8.244)
Te ρ (α (α̃)) 2 Ũe
Numerically determined solutions for the velocity profile at several Mach numbers are
shown in Figure 8.24.
The compressible layer is considerably thicker than the incompressible layer due to the high
temperature and low density near the wall caused by the deceleration of the boundary layer
flow imposed by the no-slip condition.
Figure 8.24: Compressible boundary layer profiles on an adiabatic plate for Pr = 1, viscosity
exponent ω = 1 and γ = 1.4.
8.11 Turbulent boundary layers
As the Reynolds number increases along the plate a point is reached where the laminar
boundary layer begins to be unstable to small disturbances. A complex series of events
occurs by which the flow becomes turbulent characterized by very rapid mixing of mo-
mentum in the transverse direction. As a result the velocity profile becomes much fuller
and the velocity gradient near the wall becomes much steeper compared to the velocity
gradient that would have occurred if the layer had remained laminar. The friction at the
wall is also correspondingly much larger and the growth rate of the boundary layer is much
faster as suggested schematically in Figure 8.25.
Figure 8.25: Sketch of boundary layer growth in the laminar and turbulent regions.
There is no ab-initio theory for the Reynolds shearing stress in a turbulent boundary layer
and so there is no fundamental theory for the velocity profile, boundary layer thickness or
skin friction. A reasonable and commonly used impirical formula for the thickness of an
incompressible turbulent boundary layer is
δ 0.37
= . (8.245)
x Rex 1/5
There is no theoretical reason to assume that the thickness of a turbulent boundary layer
grows according to a power law as there is in the laminar case. A relation that applies over
a wider range of Reynolds number than (8.245) is Hansen’s formula
δ 0.14
= G (ln (Rex )) . (8.246)
x ln (Rex )
Where G is a very slowly changing function of the Reynolds number with an asymptotic
value of one at Ln (Rex ) → ∞ . In the Reynolds number range 105 < Rex < 106 , G =
1.5.
The critical Reynolds number for transition is generally taken to be approximately 5 × 105
although in reality transition is affected by a wide variety of flow phenomena including
such things as plate roughness, free stream turbulence, compressibility and the presence of
acoustic noise in the free stream. Figure 8.26 is intended to illustrate this idea with the
theoretical laminar and impirical turbulent skin friction lines connected by several curves
reflecting the disturbance environment of a particular flow. In the transition zone the
physical wall shear stress can actually increase in the x direction due to the rapid filling
out of the velocity profile as the turbulence develops.
Figure 8.26: Sketch of boundary layer growth in the laminar and turbulent regions.
8.11.1 The incompressible turbulent boundary layer velocity profile
A commonly used impirical form of the velocity profile for a turbulent boundary layer is
the so-called 1/7th power law.
U y 1/7
= (8.247)
Ue δ
This profile can be useful in a limited range of Reynolds numbers but it fails to capture one
of the most important features of the turbulent boundary layer which is that the actual
shape of the velocity profile depends on the Reynolds number. In contrast, the Blasius
profile shape is completely independent of Reynolds number. Only the thickness changes
with Reynolds number.
While there is no good theory for the velocity profile in a turbulent boundary, there is a
very good correlation that accurately reflects certain fundamental flow properties and can
be used to analyze and compare flows at different Reynolds numbers. The first thing to
know is that the flow near the wall, when properly normalized, has a universal shape that
is independent of the parameters that govern the outer flow. The idea is to normalize the
velocity near the wall by the so-called friction velocity.
r
∗ τw
u =
ρ
(8.248)
∂U
τw = µ
∂y y=0
This leads to the definition of dimensionless wall variables.
yu∗
y+ =
ν
(8.249)
U
U+ = ∗
u
The thickness of the boundary layer in wall units is
δu∗
δ+ = . (8.250)
ν
If we use the empirical formulas for thickness and skin friction described above then
1/2 1/2 1/2

u∗

τw Cf 0.0592 0.172
= = = = (8.251)
Ue ρUe 2 2 2Rex 1/5
Rex 1/10
and
δ u∗ Ue x

+ 0.37 0.172
δ = = 1/5 1/10
Rex = 0.0636Rex 7/10 . (8.252)
x Ue ν Rex Rex
In other words, once Rex is known most of the important properties of the boundary layer
are known. A typical velocity profile is shown in Figure 8.27. I have chosen a relatively
low Reynolds number so that the flow near the wall can be seen at a reasonable scale. The
boundary layer is generally thought of as comprising several layers and these are indicated
in Figure 8.27. Usually they are delineated in terms of wall variables.
Viscous sublayer - Wall to A, 0 ≤ y + < 7. In this region closest to the wall the velocity
profile is linear.
U + = y+ (8.253)
Buffer layer - A to B, 7 ≤ y + < 30 . Several alternative formulations are used to approx-

imate the velocity in this region. An implicit relation that works reasonably well all the
way to the wall is

+ 1 2 1 3 1 4
y + = U + + e−κC eκU − 1 − κU + − κU + − κU + − κU + . (8.254)
2 6 24
Logarithmic and outer layer - B to C to D, 30 ≤ y + < δ + . An impirical formula that

works well and includes the effect of pressure gradient is
π y+

+ 1 Π (x)
= ln y + + C + 2 Sin2

U . (8.255)
κ κ 2 δ+
The slope of the profile in the logarithmic region is inversely proportional to the Karman
constant κ which is generally taken to be κ = 0.4 and is viewed as a universal constant
of turbulent flows. However modelers of meteorological flows such as the atmospheric
boundary layer have deduced values of κ as low as 0.35 while experiments in very high
Reynolds number pipe flow at Princeton have suggested values as high as 0.436.
Figure 8.27: Turbulent boundary layer velocity profile in linear and log-linear coordinates.
The Reynolds number is Rex = 106 .
The constant C depends on the wall roughness. If ks is a measure of the height of roughness
elements at the wall (so-called sand grain roughness) one can define Res = u∗ ks /ν as a
roughness Reynolds number. The wall is considered hydraulically smooth if Res < 3 and
hydraulically rough if Res > 100. For a hydraulically smooth wall C = 5.1 in (8.255) and
increases with increasing Res . The function Π (x) is determined by the pressure gradient
and layer Reynolds number. For dPe /dx = 0, Π (x) = 0.62. The logarithmic term in
(8.255) is remarkably robust in the presence of pressure gradients and roughness. The
constant κ hardly changes at all with roughness.
The Sin2 term in (8.255) is called the wake function because of the vaguely wake-like shape
of the outer velocity profile. Something to notice is that the velocity profile (8.255) is not
valid beyond y + = δ + and the velocity profile has a small but finite slope at y + = δ +
because of the logarithmic term.
Figure 8.28 compares turbulent boundary layer profiles at various Reynolds numbers. The
effect of Reynolds number on the flow expressed in wall variables is felt through the value
of δ + in the wake function term in (8.255). In Figure 8.28 there is a slight mismatch
between (8.254) and (8.255) at the outer edge of the buffer layer at a Reynolds number of
105 . This is because a turbulent boundary layer can hardly exist at such a low Reynolds
number and the functions (8.254) and (8.255) which have been developed from boundary
layer data were not designed to fit this case.
The most remarkable feature of the turbulent boundary layer is the logarithmic region B
to C governed by the so-called universal law of the wall. In this region the wake function is
negligible and the velocity profile has the same shape regardless of the Reynolds number.
The shape remains the same even in the presence of a pressure gradient which is remarkable
in view of the sensitivity of the shape of the laminar velocity profile to pressure gradient. If
the plate is rough the main effect is to increase the value of the constant C while preserving
the logarithmic shape and the slope1/κ.
Figure 8.28: Incompressible turbulent boundary layer profiles at several Reynolds numbers
compared to the Blasius solution for a laminar boundary layer.
According to (8.252) δ + increases fairly rapidly with Reynolds number and this leads to an
increasingly full velocity profile. At high Reynolds numbers the viscous sublayer becomes
extremely thin and it is very difficult to make direct measurements of the linear part of the
velocity profile to determine skin friction. Fortunately the skin friction can be determined
using measurments in the much more accessible logarithmic region by utilizing the law of
the wall.
yU ∗

U 1
∗
= ln +C (8.256)
U κ ν
Measurements of U versus y in region B-C can be used to determine U ∗ and hence Cf .

The friction coefficient in the compressible case can be estimated using (8.231).
The turbulent fluctuations about the mean velocity tend to peak near the wall in the
general vicinity of the log layer. Figure 8.29 shows some classical data from Klebanoff
(NACA report 1247, 1955). Generally the stream-wise velocity fluctuations are a factor of
two larger than the wall normal or span-wise fluctuations. The peak turbulent shearing
stress is generally quite close to the value at the wall indicating that there is a region of
nearly constant shearing stress that extends from the wall out to the log layer. This can
be used to develop a heuristic argument for the existence of a logarithmic segment of the
mean velocity profile.
There is a great temptation to argue that the structure of a turbulent boundary layer near
the wall is of universal character independent of the Reynolds number when the turbulent
velocity fluctuations are normalized by the wall friction velocity. When turbulence data
Figure 8.29: Profiles of turbulent velocity fluctuations and turbulent shearing stress at
Rex = 4.2 × 106 normalized by the free stream velocity.
near the wall is normalized this way u0 /U ∗ seems to have a value of about 2.5 independent
of the Reynolds number. The great problem with this idea is that according to (8.251)
the friction velocity decreases as a fraction of the free stream velocity as the Reynolds
number increases. If the turbulent velocity fluctuations also decreased in the same way
then in the limit of infinite Reynolds number the turbulent fluctuations would become a
negligible fraction of the free stream speed. On the other hand if the turbulent velocity
fluctuations scale with the free stream speed then the correlation between streamwise and
wall normal fluctuations would have to decrease to be consistent with the decrease of the
friction coefficient with increasing Reynolds number. If an ab-initio theory of the turbulent
boundary layer is ever developed this is the sort of issue that it should resolve.
8.12 Transformation between flat plate and curved wall bound-

ary layers
One of the consequences of the boundary layer approximation is that the solution for the
flow on a flat plate with a variable free stream velocity can be transformed directly to the
solution on a curved wall with the same free stream velocity function. Once again the
boundary layer equations are
∂ρU ∂ρV
+ =0
∂x ∂y
∂U ∂U dPe ∂τxy
ρU + ρV + − =0 (8.257)
∂x ∂y dx ∂y
∂T ∂T dPe ∂Qy ∂U
ρU Cp + ρV Cp −U + − τxy = 0.
where τxy includes both laminar and turbulent components of shearing stress and Qy
includes laminar and turbulent components of heat flux. The transformation of variables
between the flat plate and curved wall is
x̃ = x
ỹ = y + g (x)
Ũ (x̃, ỹ) = U (x, y)
dg (x)
Ṽ (x̃, ỹ) = V (x, y) + U (x, y)
dx (8.258)
ρ̃ (x̃, ỹ) = ρ (x, y)
τ̃x̃ỹ (x̃, ỹ) = τxy (x, y)
Q̃ỹ (x̃, ỹ) = Qy (x, y)
P̃e (x̃) = Pe (x) .
The transformation (8.258) generates the transformation of derivatives using the chain
rule.
∂ Ũ ∂U dg ∂U
= −
∂ x̃ ∂x dx ∂y
∂ Ũ ∂U
=
∂ ỹ ∂y
∂ 2 Ũ ∂2U
=
∂ ỹ 2 ∂y 2
∂ Ṽ ∂V dg ∂U
= +
∂ ỹ ∂y dx ∂y
∂ ρ̃ ∂ρ dg ∂ρ
= −
∂ x̃ ∂x dx ∂y
(8.259)
∂ ρ̃ ∂ρ
=
∂ ỹ ∂y
∂ T̃ ∂T dg ∂T
= −
∂ x̃ ∂x dx ∂y
∂ T̃ ∂T
=
∂ ỹ ∂y
∂ Q̃ỹ ∂Qy
=
∂ ỹ ∂y
∂ τ̃x̃ỹ (x̃, ỹ) ∂τxy (x, y)

= .
∂ ỹ ∂y
The transformation (8.258) and its extension to derivatives (8.259) maps (8.257) to it-
self.
∂ ρ̃Ũ ∂ ρ̃Ṽ ∂ρU ∂ρV

+ = + =0
∂ x̃ ∂ ỹ ∂x ∂y
∂ Ũ ∂ Ũ dP̃e ∂ τ̃x̃ỹ ∂U ∂U dPe ∂τxy

ρ̃Ũ + ρ̃Ṽ + − = ρU + ρV + − =0
∂ x̃ ∂ ỹ dx̃ ∂ ỹ ∂x ∂y dx ∂y
(8.260)
∂ T̃ ∂ T̃ dP̃e ∂ Q̃ỹ ∂ Ũ
ρ̃Ũ Cp + ρ̃Ṽ Cp − Ũ + − τ̃x̃ỹ =
∂ x̃ ∂ ỹ dx̃ ∂ ỹ ∂ ỹ
∂T ∂T dPe ∂Qy ∂U
ρU Cp + ρV Cp −U + − τxy =0
In principle the function g (x) added to the y coordinate is arbitrary, but in practice g (x)
is restricted to be smooth and slowly varying so as not to violate the boundary layer
approximation.
It is the absence of the derivative ∂V /∂x from (8.257) that enables the transformation (8.258)
and (8.259) to reproduce the same equations in the new coordinates. Figure 8.30 illustrates
the idea.
Figure 8.30: Mapping of the boundary layer developing over an airfoil to the boundary layer
on a flat plate with the same pressure gradient.
The function g (x̃) defines the surface of the airfoil in the Cartesian coordinates (x̃, ỹ). The
flow at a given x̃ distance from the leading edge and a distance ỹ − g (x̃) above the surface
of the airfoil is mapped to the same position in the boundary layer on a flat plate with the
same free stream flow velocity Ũe (x̃) = Ue (x). The main restriction on g (x̃) is that it vary
slowly enough so that the boundary layer does not approach separation and so that the
approximation x̃ = x remains valid. The exponential decay of the vorticity at the edge of
the boundary layer described earlier, together with this simple invariance of the boundary
layer equations enables boundary layer theory to be applied to a wide variety of slender
body shapes, creating one of the most powerful tools in fluid mechanics.
An iterative algorithm can be used to determine the viscous flow over a complex shape
such as the airfoil shown in Figure 8.29. The procedure is the following.
1) Solve for the potential flow over the airfoil.
2) Use the potential flow velocity at the airfoil surface as the Ue (x) for a boundary layer
calculation beginning at the leading edge.
3) Determine the displacement thickness of the boundary layer and use the data to define
a new airfoil shape. Repeat the potential flow calculation using the new airfoil shape to
determine a new Ue (x).
4) Using the new Ue (x) repeat the boundary layer calculation.
A few iterations of this viscous-inviscid interaction procedure will converge to an accurate
solution for the viscous, compressible flow over the airfoil.
8.13 Head’s method for approximate calculation of turbu-

lent boundary layer characteristics
In section 8.8 we discussed Thwaites’ method for calculating the integral characteristics of
laminar boundary layers in the presence of free stream velocity variation. Similar schemes
have been developed for calculating the integral characteristics of turbulent boundary lay-
ers. One of the simplest to apply is the method described in the paper Entrainment in the
turbulent boundary layer, Aero. Res. Council Research and Memoranda 3152, 1960 by M.
R. Head.
Figure 8.31: Incompressible turbulent boundary over a flat plate with pressure gradient.
In the boundary layer shown in figure 8.31 the area flow at any position x is
Z δ
Q= U dy. (8.261)
0
Equation (8.261) can be rearranged to express the area flow in terms of the boundary layer
displacement thickness.
Z δ Z δ Z δ
U
Q= U dy = Ue dy − Ue 1 − dy = Ue (δ − δ ∗ ) (8.262)
0 0 0 Ue
The entrainment velocity of the boundary layer is defined as the rate of change of Q with
x.
dQ d
Ve = = (Ue (δ − δ ∗ )) (8.263)
dx dx
Head defined a modified boundary layer shape factor as
(δ − δ ∗ )
H1 = . (8.264)
θ
He then analyzed a variety of boundary layer data and came up with the correlations shown
in figure 8.32.
Figure 8.32: Correlated data from Head’s paper.

The following quote from Head’s paper is interesting in the insight that it provides into his
thinking about his own model. He is brutally honest in describing the weaknesses of the
assumptions of the model.
Figure 8.33: Head’s frank comments about his method.
Based on the results in figure 8.32 Head assumed the existence of two universal functions
for turbulent boundary layers, F and G where
Ve 1 d 1 d
= (Ue (δ − δ ∗ )) = (Ue θH1 ) = F (H1 ) (8.265)
Ue Ue dx Ue dx
and
H1 = G (H) (8.266)
where H is the conventional shape factor
δ∗
H= . (8.267)
θ
In addition he assumed that the skin friction coefficient followed the well-established em-
pirical formula developed by Ludweig and Tillman. See Investigations of the wall-shearing
stress in turbulent boundary layers, NACA TM 1285, 1950. The Ludweig-Tillman formula
is
0.246
Cf = 0.268 (8.268)
100.678H UU∞e Rθ
where the reference velocity for Rθ is the free stream velocity.
U∞ θ
Rθ = (8.269)
ν
In the paper Calculation of separation points in incompressible turbulent flows, in J. Air-

craft vol 9, No. 9, 1972 by T. Cebeci, G. J. Mosinskis and A. M. O. Smith of Douglas
Aircraft, the authors suggest
0.0299
F (H1 ) = (8.270)
(H1 − 3.0)0.6169
and
0.8234


 3.3 + H ≤ 1.6
(H − 1.1)1.287



G (H) = (8.271)

 1.5501
 3.3 + H > 1.6.


(H − 0.6778)3.064
In the book Boundary Layer Theory, H. Schlichting suggests the same functions but uses
the constant 0.0306 instead of 0.0299 in (8.270). The function (8.271) and its derivative
are shown in figure 8.34.
Figure 8.34: Empirical function G suggested in the references described above.

Unfortunately there is a discontinuity in the function and its derivative at H = 1.6. I

prefer to use the smooth function
0.8702
G (H) = 3.0445 + (8.272)
(H − 1.1)1.2721
shown in figure 8.35.
Figure 8.35: My preferrred function G.
Figure 8.36 shows a comparison between (8.271) and (8.272).
Figure 8.36: Comparison between G0 s.
Expand the derivative in (8.265) to generate an equation for the shape factor.
− G(H)
 
dθ 1 dUe 0.0299
dH θ dx − G (H) Ue dx + θ(G(H)−3.0)0.6169
= dG
 (8.273)
dx dH
Recall the von Karman integral momentum equation.

dθ θ dUe Cf
+ (2 + H) = (8.274)
dx Ue dx 2
Substitute (8.274) into (8.273).
G(H) Cf
(1 + H) G(H)
 
dUe 0.0299
dH Ue dx − θ 2 + θ(G(H)−3.0)0.6169
= dG
 (8.275)
dx dH
Equations (8.274) and (8.275) constitute two equations for θ and H with known boundary
layer edge velocity distribution Ue (x). It is useful to put the equations into a normalized
form that will make it easier to identify the inherent dependence of boundary layer param-
eters on Reynolds number. Refer the Reynolds number based on distance from the leading
edge to the free stream velocity.
U∞ x
Rex = (8.276)
ν
Using the Bernoulli equation
1 1
P∞ + ρU∞ 2 = Pe + ρUe 2 (8.277)
2 2
the pressure coefficient can be expressed in terms of the velocity at the edge of the boundary
layer.
2
Pe − P∞ Ue
Cp = 1 2 =1− U∞
(8.278)
2 ρU∞
Using (8.276), (8.278), and (8.268) express the von Karman equation as

dRθ 1 2+H dCp 0.246
= Rθ + 0.678H (8.279)
dRex 2 1 − Cp dRex 10 (1 − Cp )0.268 Rθ 0.268
and the shape factor equation, (8.275) as

G(H) 1+H dCp 0.246×G(H) 0.0299
dH 2 1−Cp dRex + 100.678H (1−Cp )0.268 Rθ 1.268
− (G(H)−3.0)0.6169 Rθ
= 0.8702×1.2721 . (8.280)
dRex
(H−1.1)0.2721
For given initial conditions on Rθ , H and known free stream pressure distribution Cp (Rex ),
(8.279), (8.280) and (8.272) are solved for Rθ (Rex ) and H (Rex ).
8.13.1 Head’s method applied to the zero pressure gradient flat plate
In this case the governing equations (8.279) and (8.280) reduce to
dRθ 0.246
= 0.678H 0.268 (8.281)
dRex 10 Rθ
and
 
1.2721
dH  0.193381 1 + 3.49862(H − 1.1) 0.0310393(H − 1.1)1.5442
= − 0.6169  .

dRex 100.678H (H − 1.1) Rθ 1.268 1 + 0.0511377(H − 1.1)1.2721 Rθ
(8.282)
Figure 8.37 shows the results of a calculation of the momentum thickness and shape factor
for this case. The calculation is initiated at a streamwise Reynolds number of 10, 000 and
the initial momentum thickness and shape factor are taken to be the Blasius values at that
Reynolds number. Note the rapid decrease of the shape factor at low Reynolds number as
the method adjusts to the turbulent value which is less than 1.5.
Figure 8.37: Turbulent boundary layer parameters calculated using Head’s method initiated
with laminar values.
Figure 8.38 shows the Ludweig-Tillman value of the friction coefficient calculated using
Head’s method and compared to the Blasius friction coefficient as well as the power law
formula used in figure 8.26. The agreement is reasonable in a narrow range of Reynolds
numbers around 10, 000, 000.
Figure 8.39 shows data from Head’s paper for a variety of experiments for the shape factor of
flat-plate zero-pressure-gradient turbulent boundary layers compared with the prediction of
Figure 8.38: Data for the shape factor compared with Head’s method.
Head’s method over almost three orders of magnitude in Rex . The method works reasonably
well although the prediction seems a little high at high Reynolds numbers.
Figure 8.39: Data for the shape factor compared with Head’s method.
8.13.2 Head’s method used to study the effect of Reynolds number on

flow separation on a circular cylinder
Since Head’s method can be used to determine skin friction for an evolving boundary
layer one might expect the method to be able to predict the point where the skin friction
approaches zero and flow separation occurs, say on a wing at high angle of attack. With
a healthy dose of skepticism and accepting a fair amount of uncertainty in the actual
predicted separation point, the method can reasonably be used to predict trends in the
separation point with, say, changing angle of attack.
Perhaps the most challenging problem for predicting separation is flow past a bluff body
such as a circular cylinder shown in Figure 8.40. In section 8.8 we used Thwaites’ method
and the potential flow solution for Ue /U∞ to predict the separation point of a laminar
boundary layer developing from the forward stagnation point. See figure 8.20. The pre-
dicted separation point was somewhat aft of φ = 90◦ mainly because the pressure distri-
bution was forced to be the potential flow solution.
Figure 8.40: Flow past a circular cylinder.
Figures 8.41 and 8.42 show results for a relatively low cylinder Reynolds number that would
not be expected to have a turbulent boundary layer on the forward surface unless it was
artificially tripped. Note that separation is pretty clearly indicated by a rapidly increasing
shape factor and decreasing skin friction with increasing Rex . The calculation becomes
singular when the friction goes to zero and there is no data beyond separation.
Figure 8.41: Assumed turbulent boundary layer on a circular cylinder at low cylinder
Reynolds number computed using Head’s method. The flow is initiated with laminar values
and the pressure distribution is taken to be the potential flow solution.
Figures 8.43 and 8.44 show results for a high cylinder Reynolds number that would be
expected to have a turbulent boundary layer on the forward surface. The separation point
moves aft compared to the low Reynolds number case but is also probably too far aft
because of the use of the potential flow solution for the pressure.
In both cases the effect of a turbulent boundary layer is to move the separation point
aft with the high Reynolds number case furthest back. But in both cases the separation
point is probably too far back compared the prediction that would come from using the
Figure 8.42: Skin friction for an assumed turbulent boundary layer on a low Reynolds
number circular cylinder computed using Head’s method. Separation is predicted to occur
at the back of the cylinder.
Figure 8.43: Assumed turbulent boundary layer on a circular cylinder at high cylinder
Reynolds number computed using Head’s method. The flow is initiated with laminar values
and the pressure distribution is taken to be the potential flow solution.
Figure 8.44: Skin friction for an assumed turbulent boundary layer on a high Reynolds
number circular cylinder computed using Head’s method. Separation is predicted to occur
at the back of the cylinder somewhat further aft than in the low Reynolds number case.
experimental pressure distribution. See the paper by Cebici et al mentioned earlier, where
the actual pressure distribution is used.
8.14 Problems
Problem 1 - Figure 8.45 depicts Couette flow of an ideal gas between two infinite parallel
plates. The lower wall is adiabatic. Determine the entropy difference between the lower
and upper walls.
Figure 8.45: Couette flow, adiabatic lower wall.
Problem 2 - Figure 8.46 depicts Couette flow of helium gas between two infinite parallel
walls spaced 1 cm apart. The lower wall is adiabatic and the speed of the upper wall is
400 meters/sec. The temperature of the upper wall is 300K.
Figure 8.46: Couette flow of helium, adiabatic lower wall.
Assume the viscosity depends linearly on temperature.
µ/µ∞ = T /T∞ (8.283)
Set up and solve the compressible flow equations for this simple flow. Note that the flow is
assumed to be steady and all flow variables depend only on the coordinate normal to the
wall.
1) Determine the speed of sound at the upper wall.
2) Determine the temperature of the lower wall.
3) Determine the shear stress.
4) Is there work done on the flow? How much?

5) Determine the heat flux through the upper wall.
6) Sketch the distribution of stagnation temperature across the channel.
7) Sketch the distribution of entropy across the channel.
Problem 3 - Figure 8.47 depicts Couette flow of a gas between two infinite parallel walls
spaced a distance
p d apart. The lower wall is adiabatic. The reference Mach number
is M∞ = U∞ / γRT∞ . The viscosity is assumed to depend linearly on temperature
µ/µ∞ = T /T∞ and the reference Reynolds number is Re = ρ∞ U∞ d/µ∞ .
Figure 8.47: Couette flow of an ideal gas, adiabatic lower wall.
Sketch how the friction coefficient Cf depends on U∞ . At what Mach number is the friction
coefficient an extremum? Is it a maximum or a minimum? Express your answer in terms
of γ and the Prandtl number. What are the values for helium and air?
Problem 4 - Figure 8.48 shows the unsteady flow produced by a flat plate set into motion
impulsively at velocity U∞ .
Figure 8.48: Impulsively started flat plate.
The plate extends to infinity in both directions. Simplify the compressible flow equations.
Note that the fluid near the wall is heated when the plate is started. Solve for the velocity
and vorticity in the incompressible case.
Problem 5 - In the discussion of boundary layers we considered several definitions of the
thickness. How would you define a thickness based on the vorticity distribution? What
might be the advantage of such a definition?
Problem 6 - Use the Howarth-Stewartson transformation to generate the velocity and

temperature profiles in a laminar, compressible, zero pressure-gradient boundary layer at
a free stream Mach number M∞ = 8.
Problem 7 - The 2-D stream function for potential flow over an elliptically shaped body
at zero angle of attack is produced by the superposition of a uniform flow plus a source
and a sink of equal strength. The stream function and flow pattern are

Q y Q y
Ψ = U∞ y + ArcT an − ArcT an (8.284)
2π x+a 2π x−a
Figure 8.49: Potential flow over a 2-D elliptical body.
Choose two aspect ratios for the ellipse (length/width = 2, length/width = 20).
1) Use the potential flow solution to determine the pressure coefficient on the body.
2) Use Thwaites’ method to calculate the properties of the laminar boundary layer up to
separation. How does the separation point depend on the aspect ratio of the ellipse? Use
the radius of curvature at the forward stagnation point to initiate the calculation.
3) Use Head’s method to do the same for a turbulent boundary layer. For a given aspect
ratio how does the separation point depend on the Reynolds number based on the length
of the body?
Chapter 9
Quasi-one-dimensional flow
9.1 Control volume and integral conservation equations
In this chapter we will treat the very general stationary flow of a compressible fluid in a
channel without body forces (Gi = 0) shown in figure 9.1 .
Figure 9.1: Control volume enclosing a general stationary channel flow.
Viscous friction imposes a no-slip condition at all solid surfaces. There may be mass
addition as well as heat conduction through the wall. The added mass has total enthalpy,
htm and stream wise velocity component, Uxm . In addition there may be work done on
the flow by a fan or by the flow on a turbine. The work exchanged with the control volume
is symbolized by a fan assumed to be rotating at a constant angular frequency.
9-1
CHAPTER 9. QUASI-ONE-DIMENSIONAL FLOW 9-2
The appropriate Eulerian-Lagrangian control volume is shown in figure 9.1 . The Eulerian
inlet and outlet surfaces of the control volume, A1 , and A2 are fixed in space as is Aw
which is attached to the non-moving wall of the channel. The Lagrangian surface Af is
attached to, and moves with, the surface of the rotating fan. The area where the control
volume is constantly being stretched and twisted at the corner where A2 and Af meet is
negligibly small and therefore involves no net contribution to the mass momentum and
energy entering or leaving the control volume.
In chapter 5 we briefly discussed the distinction between steady flow and stationary flow.
The flow in the neighborhood of the rotating fan with moving blades is surely unsteady
but as long as the rotation speed of the fan is constant there will be no net gain or loss
of mass, momentum or energy inside the control volume. This justifies dropping the time
derivative terms in the conservation equations. Another way to look at this is to imagine
that the fan blades are smeared out onto a continuous disk called an actuator disk. This
would make the flow truly steady but begs the question as to how such a disk could be
constructed. A third way to justify the use of the steady flow equations is to view any flow
variable as a time average over precisely one blade passage period, or one full fan rotation
period, or for an averaging time that is so long compared to the blade passage period that
a small additional contribution from a portion of a period is negligible.
There are other aspects of the flow that might have to be treated as stationary. For
example the flow in figure 9.1 could be turbulent in which case unsteady fluctuations in
the flow properties, particularly near the wall, would be characterized by a wide range
of time scales lacking any obvious periodicity. In this case the flow is stationary as long
as the time average of any flow property such as velocity, pressure or density is constant.
This brings in some subtle questions as to the length of the time average and specific times
when the averaging begins and ends. For our purposes it is sufficient to require that the
averaging is long compared to both the rotation period of the fan and any low frequency
variations in the turbulent fluctuations.
As a practical matter we are not attempting to treat the case where the flow is being turned
on or off or where the fan is changing speed. These cases would require a full unsteady
analysis. This is quite do-able but beyond the scope of this course.
9.1.1 Conservation of mass
The integral form of the mass conservation equation in steady flow is evaluated on each
surface of the control volume.
Z

ρ Ū − ŪA · n̄dA =
A(t)
Z Z Z Z (9.1)

ρ2 U2 dA − ρ1 U1 dA + ρŪ · n̄dA + ρ Ū − ŪA · n̄dA = 0
A1 A1 Aw Af (t)
The integral of the mass flux over the wall is simply the total mass added.
Z
ρŪ · n̄dA = −δ ṁ (9.2)
Aw
There is no mass addition through the fan surface (the last term in equation (9.1)) and so
the integrated form of the mass equation becomes
Z Z
ρ2 U2 dA − ρ1 U1 dA = δ ṁ. (9.3)
A2 A1
9.1.2 Conservation of x -momentum
Now evaluate the integral form of the momentum equation over the control volume. Note
that U − UA = 0 on the fan surface by the no-slip condition.
Z

ρŪ Ū − ŪA + P¯δ −¯τ · n̄dA =

A(t)
x
Z Z
(ρ2 U2 U2 + P2 − τxx2 )dA − (ρ1 U1 U1 + P1 − τxx1 )dA+ (9.4)
A2 A1
Z Z

ρŪ Ū + P¯δ −¯τ · n̄dA +
ρŪ Ū − ŪA + P¯δ −¯τ · n̄dA = 0
Aw x Af (t)
Over most of the wall the velocity is zero due to the no slip condition. However over the
duct through which the added mass enters the control volume there is a contribution to
the x -momentum in the amount
Uxm δ ṁ. (9.5)
Any small contributions from the pressure and stress forces acting over the control volume
in the neighborhood of the mass injector as well as any non-uniformities in the injector
velocity field have been included in the effective value that might be assigned to Uxm in
(9.5). The integrated x -momentum equation now becomes
Z

ρŪ Ū − ŪA + P¯δ −¯τ · n̄dA =

A(t)
x
Z Z
(ρ2 U2 U2 + P2 − τxx2 )dA − (ρ1 U1 U1 + P1 − τxx1 )dA+ (9.6)
A2 A1
Z

(P¯δ −¯τ ) · n̄dA − Uxm δ ṁ + δFx = 0.
Aw x
On the fan U = UA and so only the surface pressure and stress contribute an amount Fx
to the momentum flux. The force by the flow on the fan is
Z

δFx = (P ¯δ −¯τ ) · n̄dA . (9.7)

Af (t)
x
If the Reynolds number of the flow is relatively high the major contribution to the total
force in the x -direction, (9.7), comes from pressure forces integrated over the fan surface.
The way to evaluate this force is to sum the stream-wise lift components of all the fan
blades. If the sum is negative the fan is adding momentum to the flow. If the sum is
positive, the fan is actually a windmill or turbine taking momentum out of the flow.
9.1.3 Conservation of energy
Next we evaluate the integral form of the energy equation over the control volume.
Z

ρ (e + k) Ū − ŪA + P Ū −¯τ · Ū + Q̄ · n̄dA =
A(t)
Z Z
(ρ2 ht2 U2 − τxx2 U2 + Q2 )dA − (ρ1 ht1 U1 − τxx1 U1 + Q1 )dA+
A2 A1
Z (9.8)

ρht Ū −¯τ · Ū + Q̄ · n̄dA+
Aw
Z

ρ (e + k) Ū − ŪA + P Ū −¯τ · Ū + Q̄ · n̄dA = 0
Af (t)
The integral over the wall includes the total enthalpy of the mass injected through the
wall.
Z

ρht Ū −¯τ · Ū + Q̄ · n̄dA = −δQw − htm δ ṁ (9.9)
Aw
The last integral in (9.8) is the work done by the flow on the fan.
Z

δW = P Ū −¯τ · Ū · n̄dA (9.10)
Af (t)
Using (9.9) and (9.10) the integrated energy conservation equation is
Z Z
(ρ2 ht2 U2 − τxx2 U2 + Q2 )dA − (ρ1 ht1 U1 − τxx1 U1 + Q1 )dA = δQw + htm δ ṁ − δW.
A2 A1
(9.11)
9.2 Area averaged flow
Every flow variable is a three-dimensional function of space. By averaging the flow across
the channel we can produce approximate flow variables that depend only on the stream-wise
coordinate. For example
Z
1
ρ̂ (x) = ρ (x, y, z)dydz. (9.12)
A (x)
Similarly we can define the following area averages.

Z
1
T̂ (x) = T (x, y, z)dydz
A (x)
Z
1
P̂ (x) = P (x, y, z)dydz
A (x)
Z
1
ŝ (x) = s (x, y, z)dydz
A (x)
Z (9.13)
1
Û (x) = U (x, y, z)dydz
A (x)
Z
1
τ̂xx (x) = τxx (x, y, z)dydz
A (x)
Z
1
Q̂x (x) = Qx (x, y, z)dydz
A (x)
Thus every variable in the flow can be written as a mean plus a deviation.
ρ (x, y, z) = ρ̂ (x) + ρ0 (x, y, z)
T (x, y, z) = T̂ (x) + T 0 (x, y, z)
P (x, y, z) = P̂ (x) + P 0 (x, y, z)
s (x, y, z) = ŝ (x) + s0 (x, y, z) (9.14)
U (x, y, z) = Û (x) + U 0 (x, y, z)
τxx (x, y, z) = τ̂xx (x) + τ 0 xx (x, y, z)
Qx (x, y, z) = Q̂x (x) + Q0 x (x, y, z)
Consider the mass flux integral.

Z Z Z Z
0
0

ρU dA = ρ̂ + ρ Û + U dA = ρ̂Û dA + ρ0 U 0 dA =
A A A A
(9.15)
0 U 0 A (x)
ρ̂ (x) Û (x) A (x) + ρ\
The terms that are linear in the deviations are zero by the definition of the average. As
0 U 0 are small and can be neglected,
long as spatial correlations of flow variables such as ρd
the area average is a useful approximation to the flow. In terms of area averaged variables,
the integral equations of motion become
ρ̂2 Û2 A2 − ρ̂1 Û1 A1 = δ ṁ

ρ̂2 Û2 Û2 + P̂2 − τ̂xx2 A2 − ρ̂1 Û1 Û1 + P̂1 − τ̂xx1 A1 +
Z

(P¯δ −¯τ ) · n̄dA − Uxm δ ṁ + δFx = 0 (9.16)
Aw x

ρ̂2 ĥt2 Û2 − τ̂xx2 Û2 + Q̂x2 A2 − ρ̂1 ĥt1 Û1 − τ̂xx1 Û1 + Q̂x1 A1 =
δQw + htm δ ṁ − δW.
9.2.1 The traction vector
The pressure-stress integral on the wall in (9.16) (see also (9.7)) needs to be examined a
little further.
Figure 9.2: Traction force on the wall due to pressure and viscous forces.
The unit outward normal to the wall is n̄ = (nx , ny , nz ) and the components of the so-called
traction vector are
 
P nx − τxx nx − τxy ny − τxz nz
T̄ = (P¯δ −¯τ ) · n̄ =  −τxy nx + P ny − τyy ny − τyz nz  . (9.17)
−τzx nx − τzy ny + P nz − τzz nz
The x -component of the traction vector is indicated in figure 9.2.

If we imagine that the length of the control volume is made very small (∆x → 0) then the
pressure and viscous stress variation along the wall of the control volume is small and so
the pressure-viscous stress integral over the surface in (9.16) can be approximated using
average values as follows.
Z Z
(P nx − τxx nx − τxy ny − τxz nz )dA ∼

(P¯δ −¯τ ) · n̄dA =
=
Aw x Aw
(9.18)
P1 + P2 τxx1 + τxx2 D1 + D2
(A1 − A2 ) − (A1 − A2 ) + τw π ∆x
2 2 2
where we have used

Z
nx dA = (A1 − A2 ) (9.19)
Aw
and
Z
D1 + D2
(τxy ny + τxz nz ) dA ≡ −τw π ∆x. (9.20)
Aw 2
This last relation is essentially a definition for the effective wall shear stress τw which is
always taken to be positive. The wetted surface of the channel is defined using the hydraulic
diameter of the channel defined as
1/2
4A
D= . (9.21)
π
The integrated equations of motion (for small ∆x) now take the form
ρ2 U2 A2 − ρ1 U1 A1 = δ ṁ
(ρ2 U2 U2 + P2 − τxx2 ) A2 − (ρ1 U1 U1 + P1 − τxx1 ) A1 −

P1 + P2 τxx1 + τxx2 D1 + D2
(A1 − A2 ) + (A1 − A2 ) − τw π ∆x = Uxm δ ṁ − δFx
2 2 2
(ρ2 H2 U2 − τxx2 U2 + Q2 ) A2 − (ρ1 H1 U1 − τxx1 U1 + Q1 ) A1 = δQw + htm δ ṁ − δW

(9.22)
where the averaging symbol has been dropped for convenience. Now take the limit ∆x → 0
of each of the equations in (9.22). Each of the terms in the integrated equations becomes
a differential.
ρ2 U2 A2 − ρ1 U1 A1 ⇒ d (ρU A)
ρ2 U2 U2 A2 − ρ1 U1 U1 A1 ⇒ d ρU 2 A

P2 A2 − P1 A1 ⇒ d (P A)
τxx2 A2 − τxx1 A1 ⇒ d (τxx A)

P1 + P2
(A1 − A2 ) ⇒ P dA
2
(9.23)
τxx1 + τxx2
(A1 − A2 ) ⇒ τxx dA
2

D1 + D2
τw π ∆x ⇒ τw πDdx
2
ρ2 ht2 U2 A2 − ρ1 ht1 U1 A1 ⇒ d (ρU ht A)
τxx2 U2 A2 − τxx1 U1 A1 ⇒ d (τxx U A)
Qx2 A2 − Qx1 A1 ⇒ d (Qx A)
Now the equations of motion (9.22) become

d (ρU A) = δ ṁ
d ρU 2 A + d (P A) − d (τxx A) − P dA + τxx dA = −τw πDdx + Uxm δ ṁ − δFx

(9.24)
d (ρU ht A) − d (τxx U A) + d (Qx A) = δQw + htm δ ṁ − δW.
Both the momentum and energy equations can be simplified using continuity.
U δ ṁ + ρU AdU + AdP − Adτxx = −τw πDdx + Uxm δ ṁ − δFx

τxx τxx Qx Qx (9.25)
ht δ ṁ + ρU Adht − δ ṁ − ρU Ad + δ ṁ + ρU Ad =
ρ ρ ρU ρU
δQw + htm δ ṁ − δW
Finally our quasi-one-dimensional equations of motion are
d (ρU A) = δ ṁ

πDdx (Uxm − U ) δ ṁ δFx
d (P − τxx ) + ρU dU = −τw + −
A A A (9.26)

τxx Qx δQw δW τxx Qx δ ṁ
d ht − + = − + htm − ht − + .
ρ ρU ρU A ρU A ρ ρU ρU A
It is convenient to introduce the friction coefficient defined as
τw
Cf = 1 2
. (9.27)
2 ρU
The heat added through the wall and work done per unit mass flow are
δQw
δqw =
ρU A
(9.28)
δW
δw = .
ρU A
With these definitions, the area-averaged equations of motion take on the fairly concise
form
δ ṁ dA
d (ρU ) = − ρU
A A

1 4Cf dx (Uxm − U ) δ ṁ δFx
d (P − τxx ) + ρU dU = − ρU 2 + − (9.29)
2 D A A

τxx Qx τxx Qx δ ṁ
d ht − + = δqw − δw + htm − ht − + .
ρ ρU ρ ρU ρU A
The heat added to the flow through the wall per unit mass is δqw and the work done by
the flow per unit mass is δw. The spirit of these equations is that the differential terms
on the right hand sides of (9.29) such as dA/A , δ ṁ , 4Cf dx/D , δqw and δw are assumed
to be known or given quantities. They can be thought of as inputs to the flow whereas
the resulting variations in ρ, U , P and temperature T (recall Qx = −k∂T /∂x for a linear
conducting medium) are the output changes in the flow.
Notice that up to this point we have not invoked an equation of state. Equations (9.29)
apply to the steady flow of any continuous medium in the absence of body forces. By
the same token (9.29) is not a closed system; We have four unknowns but only three
equations.
9.2.2 Example - steady, gravity-free, adiabatic flow of a compressible

fluid in a channel
Figure 9.3: Adiabatic channel flow
The integral form of the energy equation in this case is simply


τxx Qx
d ht − + = 0. (9.30)
ρ ρU
For typical cases such as the flow of gases at more than a few meters per second, the stress
and heat conduction terms on A1 and A2 are very small and can be neglected. Thus
ht2 = ht1 . (9.31)
This is, of course, a result we have seen before. It is extraordinarily useful because there are
so many flow situations where an adiabatic approximation can be assumed and where the
stagnation enthalpy of a fluid particle is approximately conserved. This applies generally,
except for flow near a nonadiabatic wall where significant heat conduction may take
place and within a shock wave where large stream-wise velocity and temperature gradients
occur.
9.3 Normal shock waves
Lets consider uni-directional flow in the absence of mass addition, wall friction, and heat
addition. The continuity equation in this case is
d (ρU ) = 0. (9.32)
If the fluid is incompressible (9.32) gives the trivial result dU = 0. But if the fluid is
compressible then a non-trivial flow can exist where U and ρ vary inversely. The prime
example of such a flow is the shock wave where the properties of the flow change almost
discontinuously.
The shaded region in figure 9.4 (the shock) contains gradients in temperature, pressure,
density and velocity while the upstream and downstream conditions are perfectly uniform.
The thickness of the shock is δ . The equations of motion in this case are
d (ρU ) = 0
d P − τxx + ρU 2 = 0

(9.33)

τxx Qx
d ht − + = 0.
ρ ρU
Figure 9.4: Shock wave schematic
Notice that the continuity equation has been used to make the momentum equation a
perfect differential. There are three conserved quantities in this flow.
ρU = constant1
P − τxx + ρU 2 = constant2 (9.34)

τxx Qx
ht − + = constant3
ρ ρU
These combinations of flow variables have the same value at every point in the flow including
within the shock. Integrate (9.34) between states 1 and 2.
ρ1 U1 = ρ2 U2
P1 − τxx1 + ρ1 U1 2 = P2 − τxx2 + ρ2 U2 2 (9.35)

τxx1 Qx1 τxx2 Qx2
ht1 − + = ht2 − +
ρ1 ρ1 U1 ρ2 ρ2 U2
Recall that for a Newtonian fluid

4 ∂U
τxx = µ + µν (9.36)
3 ∂x
where µv is the bulk viscosity, and

∂T
Qx = −k . (9.37)
∂x
The flow at 1 and 2 is uniform, i.e., the velocity and temperature gradients are zero and
so the equations of motion reduce to the classical shock jump conditions.
ρ1 U1 = ρ2 U2
P1 + ρ1 U1 2 = P2 + ρ2 U2 2 (9.38)
ht1 = ht2
It is important to recognize that in deriving (9.38) viscosity and heat conduction were
not neglected. They vanish from the jump conditions because of the uniformity of the
upstream and downstream flows. Nevertheless their effects are felt through the mechanism
of molecular collision by which changes in the state of the fluid are accomplished within
the shock. Notice also that we still have not invoked an equation of state. The equations
(9.38) govern the propagation of shock waves in any continuous medium, air, water, rock,
etc.
9.3.1 The Rankine-Hugoniot relations
A general set of jump relations in which the velocity does not appear can be derived from
(9.38). First, use mass and momentum to work out
P2 − P1
U1 U2 = . (9.39)
ρ2 − ρ1
An alternate form of (9.39) assuming an ideal gas law and constant specific heat is

P2 ρ1
− 1 = γM1 2 1 − . (9.40)
P1 ρ2
The energy equation is

γ P1 1 2 γ P2 1 2
+ U1 = + U2 . (9.41)
γ−1 ρ1 2 γ−1 ρ2 2
Use continuity to eliminate U1 in (9.41) and solve for U2 .


2 2γ P2 ρ1 − P1 ρ2
U2 = (9.42)
γ−1 (ρ2 − ρ1 ) (ρ2 + ρ1 )
Use (9.42) to replace U2 2 in the momentum jump condition. After some rearrangement
the result is

γ+1 ρ2
P2 γ−1 ρ1−1
= . (9.43)
P1 γ+1
− ρ2
γ−1 ρ1
Equation (9.43) is the famous Rankine-Hugoniot relation. Another form of (9.43) is
P2 ρ2
!
P1 −1 ρ1 −1
P2
=γ ρ2 . (9.44)
P1 +1 ρ1 +1

P2 − P1 P2 + P1
=γ . (9.45)
ρ2 − ρ1 ρ2 + ρ1
9.3.2 Shock property ratios in a calorically perfect ideal gas
For a calorically perfect gas the third jump condition ht1 = ht2 can be written
1 1
Cp T1 + U1 2 = Cp T2 + U2 2 . (9.46)
2 2
Consider a reference state where the flow speed equals the local speed of sound. The flow
variables at this state will be denoted with a superscript *; Thus U ∗ = a∗ , and ρ = ρ∗ ,
P = P ∗ , T = T ∗ . Each side of (9.46) can be equated with the reference state.
1 γ + 1 ∗2
Cp T1 + U1 2 = a
2 2 (γ − 1)
(9.47)
1 γ + 1 ∗2
Cp T2 + U2 2 = a
2 2 (γ − 1)
Use the ideal gas law to write (9.47) in the form

γ 1 γ+1
P1 + ρ2 U2 U1 = ρ1 a∗ 2
γ−1 2 2 (γ − 1)
(9.48)
γ 1 γ+1
P2 + ρ1 U2 U1 = ρ2 a∗ 2 .
γ−1 2 2 (γ − 1)
Subtract the relations in (9.48)
γ 1 γ + 1 ∗2
(P2 − P1 ) − (ρ2 − ρ1 ) U1 U2 = a (ρ2 − ρ1 ) . (9.49)
γ−1 2 2 (γ − 1)
Now use (9.39) to replace P2 − P1 in (9.49) and gather terms. The result is the famous
Prandtl relation
U1 U2 = a∗ 2 . (9.50)
One of the implications of (9.50) is that for a normal shock the flow must be supersonic
ahead of the shock and subsonic behind the shock. Note that a∗ is essentially defined by
(9.47) and can be determined entirely by values in either region 1 ahead of the shock or
region 2 behind the shock. Prandtl’s relation is the Rosetta stone for generating all of the
shock jump relations in terms of the shock Mach number. For example, substitute (9.50)
into the first relation in (9.47).
1 γ+1
Cp T1 + U1 2 = U1 U2 (9.51)
2 2 (γ − 1)
Divide (9.51) by U1 2 . The result is

U2 2 γRT1 γ−1
= + . (9.52)
U1 γ+1 U1 2 γ+1
The Mach number ahead of the shock is
U1
M1 = √ (9.53)
γRT1
and (9.52) becomes the basic relation for the velocity ratio (and density ratio) across the
shock in terms of the upstream Mach number.

γ−1
U2 1 + 2 M1 2 ρ1
= = (9.54)
U1 γ+1
M 2 ρ2
2 1
All of the important properties of a normal shock wave can be expressed in terms of the
upstream Mach number. Using (9.54) and the Rankine-Hugoniot relation (9.43) we can
work out the pressure ratio across the shock.

( γ+1 M 2
2 ) 1

γ+1
γ+1 ρ2
−1 −1
P2 γ−1 ρ1 γ−1 1+( γ−1 M 2
2 ) 1
= = γ+1 (9.55)
P1 γ+1
− ρ2 γ+1 ( 2 )M1 2
γ−1 ρ1 −
γ−1 1+( γ−1 M 2
2 ) 1
Simplify (9.55).
2γ 2
P2 γ−1 M1 −1
= γ+1 (9.56)
P1 γ−1
The pressure ratio (9.56) is called the shock strength. Note the rapid increase with Mach
number. The temperature ratio is generated using the ideal gas law.
   
γ−1 γ−1
T2

P2

ρ1
γM1 2 − 2 1+ 2 M1 2
= = γ+1
  (9.57)
T1 P1 ρ2 2
γ+1
M1 2
2
The downstream Mach number is
2 2
M2 U2 T1
= . (9.58)
M1 U1 T2
Substitute (9.54) and (9.57).

2
τij = 2µSij − µ − µv δij Skk
3
(9.59)
∂U ∂V ∼ ∂U
τxy = µ + =µ
∂y ∂x ∂y
Several factors in (9.59) cancel and the final result for the downstream Mach number
is
 
γ−1
1+ 2 M1 2
M2 2 =  . (9.60)
γ−1
γM1 2 − 2
The relation (9.60) is plotted below for γ = 1.4.
Figure 9.5: Downstream Mach number versus upstream Mach number for a normal shock.
Note that the downstream Mach number has a finite lower limit.
r
γ−1
lim M2 = (9.61)
M1 →∞ 2γ
9.3.3 Stagnation pressure ratio across a normal shock wave
The stagnation pressure is determined from the isentropic relations in regions 1 and 2.
γ
Pt1 γ−1 2
γ−1
= 1+ M1
P1 2
(9.62)
γ
Pt2 γ−1 2
γ−1
= 1+ M2
P2 2
Divide these two relations.
 γ
 γ−1
γ−1
Pt2 P2  1+ 2 M2 2
=  (9.63)
Pt1 P1 1 + γ−1 M 2
2 1
In (9.63) replace P2 /P1 with (9.56) and M 2 with (9.60).
γ
 1+ γ−1 M 2  γ−1
( 2 ) 1

γ−1
2γ 2
1 +
!
Pt2 γ−1 M1 −1 γM1 2 −( γ−1
2
2 )

= γ+1
  (9.64)
Pt1 γ−1
 γ−1
1 + 2 M1 2 
Rearrange (9.64).
 2  γ
γ−1
2γ 2
! γ+1 2
−1 M1
Pt2 γ−1 M1 2
= (9.65)
 
γ+1  γ−1
Pt1 2γ γ−1

2 2
γ−1 2 γ−1 M 1 − 1 1 + 2 M 1
Combine factors in (9.65). The stagnation pressure ratio across the shock wave is
γ
1 
 γ−1
γ+1
! γ−1 γ+1 2
Pt2 2 M1
γ−1
= 2γ 2
  . (9.66)
Pt1 γ−1 M1 −1 1 + γ−1 M 2
2 1
The stagnation pressure ratio across a normal shock wave is plotted in figure 9.6 for two
values of γ.
Two important features of figure 9.6 need to be noted.
1) At Mach numbers close to one the change in stagnation pressure across a normal shock
wave is very small.
2) At high Mach numbers the stagnation pressure loss is very large with
γ 1
Pt2 γ+1 γ−1 γ+1 γ−1
lim = . (9.67)
M1 →∞ Pt1 γ−1 2γM1 2
Recall that we can write the Gibbs equation in terms of the stagnation state of the
gas.
Figure 9.6: Stagnation pressure ratio across a normal shock as a function of shock Mach
number.
dht dPt
ds = −R (9.68)
Tt Pt
Integrate (9.68) from state 1 to state 2. Since dht = 0 the entropy change across the wave
is determined directly from the change in stagnation pressure.

s2 −s1
Pt2 −
=e R
(9.69)
Pt1
The connection (9.69) between the stagnation pressure and entropy in an adiabatic flow
is an extremely important one that we see often when we solve problems in compressible
channel flow.
The equations for the flow properties across the shock are the fundamental relations used
to generate the shock tables.
9.3.4 Example - stagnation at a leading edge in supersonic flow
The figure below shows a supersonic flow of Helium (atomic weight equals 4) over the
leading edge of a thick flat plate at a free stream Mach number M1 = 2.0.
The temperature of the free stream is 300 K and the pressure is one atmosphere.
Figure 9.7: Supersonic flow of helium over a leading edge.
1) Determine the energy per unit mass of a fluid element located at points 1 (free stream), 2
(just behind the shock) and 3 (at the stagnation point on the body). State the assumptions
used to solve the problem. Express your answer in Joules/kg.
Solution
The energy per unit mass of a flowing gas is the sum of internal energy and kinetic energy
per unit mass, e + k .
a) Assume the gas is calorically perfect - constant heat capacities.
b) Assume the flow is adiabatic from station 1 to station 3.
c) Assume the body is adiabatic.
For Helium the number of degrees of freedom equals 3 and at the conditions of the free
stream we have the following values.
8314.472
R= = 2078.62 m2 /sec2 − K
4
3
Cv = R = 3117.93 m2 /sec2 − K
2
3+2
Cp = R = 5196.55 m2 /sec2 − K
2
γ = 5/3
(9.70)
r
p 5
a = γRT = 2078.62 (300) = 1019.46 m2 /sec2
3
U1 = 2 (1019.46) = 2038.92 m/sec
1
e1 + k1 = Cv T1 + U1 2 = 3117.93 (300) + 0.5(2038.92)2 = 935379 + 2078597
2
e1 + k1 = 3013976 J/kg
The stagnation temperature of the free stream is determined from

Tt γ−1
=1+ M 2. (9.71)
T 2
Thus

4
Tt1 = 300 1 + = 700 K (9.72)
3
Across a normal shock at Mach 2 the temperature ratio is

γ−1
T2 1+ 2 M1 2 γM1 2 − γ−12
= 2 (9.73)
T1

γ+1
2 M1 2
which gives
4 5 1 7 19

(4) −

T2 1+ 3 3 3 3 3 7 19
= = = = 2.078. (9.74)
T1 4 2 4 16 4 16

3 (4) 3 3
Assume the flow is adiabatic from the free stream to the stagnation point.
1 1 1
h1 + U1 2 = h2 + U2 2 = h3 + U3 2 (9.75)
2 2 2
We can rewrite this equation as follows.
RT1 + (e1 + k1 ) = RT2 + (e2 + k2 ) = RT3 + (e3 + k3 ) (9.76)
The temperatures at stations 1, 2 and 3 are respectively
T1 = 300 K
T2 = 2.078 (300) = 623.44 K (9.77)
T3 = Tt1 = 700 K.
Now
e1 + k1 = 3.014 × 106 J/kg
(e2 + k2 ) = (e1 + k1 ) − R (T2 − T1 ) = 3.014 × 106 − 2078.62 (623.44 − 300)

= 3.014 × 106 − 0.6723 × 106 = 2.3417 × 106 J/kg (9.78)
(e3 + k3 ) = (e1 + k1 ) − R (T3 − T1 ) = 3.014 × 106 − 2078.62 (700 − 300)

= 3.014 × 106 − 0.8314 × 106 = 2.1826 × 106 J/kg.
The energy of a fluid element decreases considerably across the shock and then decreases
further to the stagnation point.
2) Describe the mechanism by which the energy of the fluid element changes as it moves
from station 1 to station 3.
The work done by the pressure and viscous normal force field on the fluid element is the
mechanism by which the energy decreases in moving from station 1 to station 3. The flow
energy decreases across the shock wave through a combination of pressure and viscous
normal stress forces of roughly equal magnitude that act to compress the fluid element
increasing its internal energy while decelerating it and reducing its kinetic energy. The loss
of kinetic energy dominates the increase in internal energy.
Between stations 2 and 3 the flow further decelerates as the pressure increases toward the
stagnation point. Viscous normal forces also act in region 2 to 3 but because the streamwise
velocity gradients are small (compared to the shock) viscous forces are generally much
smaller than the pressure forces.
9.4 Shock wave thickness
Let’s use the jump conditions and our knowledge of entropy generation to estimate the
thickness of a shock wave.
Figure 9.8: Linear model for temperature and pressure variation within a shock.
The integral form of the entropy transport equation, derived in Chapter 7, over an Eulerian
control volume is
Z Z Z
d k Υ+Φ
ρsdV + ρŪ s − ∇T · n̄dA = dV (9.79)
dt V A T V T
where for a Newtonian, linear, heat conducting medium

1 1
Φ = 2µ Sij − δij Skk Sij − δij Skk + µv (Sii Skk ) (9.80)
3 3
and

k ∂T ∂T
Υ= . (9.81)
T ∂xj ∂xj
The stress tensor is
τij = 2µSij − ((2/3) µ − µv ) δij Skk . (9.82)
For the unidirectional flow within the shock wave the stress tensor reduces to
 
4 dU
µ + µv 0 0

 3 dx


 2 dU 
τij =  0 − µ + µv 0 . (9.83)

 3 dx


 2 dU 
0 0 − µ + µv
3 dx
Note that the viscous normal stresses in the y and z directions are not zero since Skk =
∇ · Ū = dU/dx. The modified rate-of-strain tensor that appears in the dissipation is
2 dU
 
0 0
 3 dx 
1  1 dU 
Sij − δij Skk =  0 − 0  (9.84)
3  3 dx 
 1 dU 
0 0 −
3 dx
and therefore the kinetic energy dissipation within the shock is
2
4 dU
Φ= µ + µv . (9.85)
3 dx
The temperature dissipation is
2
k dT
Υ= . (9.86)
T dx
Now we integrate the entropy transport equation over the control volume indicated in the
figure 9.8. Since the flow is steady
Z
d
ρsdV = 0. (9.87)
dt V
The integrated entropy equation becomes

Z δ
k dT k dT Υ+Φ
ρU sA − A − ρU sA − A = dx A. (9.88)
T dx 2 T dx 1 0 T
The areas cancel on both sides of the equation and the temperature gradients at stations
1 and 2 are zero. The product ρU is constant through the wave.
! 2 2 !
δ 4
3µ + µv
Z
dU k dT
ρU (s2 − s1 ) = + 2 dx (9.89)
0 T dx T dx
In order to model the integral on the right-hand-side of (9.89) we make the following linear
approximations for the gradients inside the wave and for the mean temperature.
dU ∼ U2 − U1
=
dx δ
dT ∼ T2 − T1
= (9.90)
dx δ
T2 + T1
T ∼
=
2
We assume that the viscosity and thermal conductivity are evaluated at the mean temper-
ature. The control volume balance of entropy now becomes
!
(T2 − T1 )2

T2 + T1 4 2
ρU (s2 − s1 ) δ= µ + µv (U2 − U1 ) + 2k (9.91)
2 3 (T2 + T1 )
which can be expressed as
 2 
2 T2
U1 2 U2 −1
4 T1
2 3µ + µv T1 U1 − 1 + 4k
T2

 T1
+1 
ρU (s2 − s1 ) δ =  . (9.92)
 T2 
 T1 +1 
Use U2 /U1 = ρ1 /ρ2 and move the entropy term to the right-hand-side.
 2 
2 T2
−1
 2 (γ − 1) 4
3 + µµv M1 2 ρρ21 − 1 + 4 µC
k T1
p T2

ρU δ  T1
+1 
=  (9.93)
µ  1 T2 s2 −s1 
γ T1 + 1 Cv
 
Recall that the Prandtl number is defined as
µCp
Pr = . (9.94)
k
The entropy jump across the shock can be expressed in terms of the density and tempera-
ture ratio using Gibbs’ equation.
(γ−1) !
s2 − s1 T2 ρ1
= ln (9.95)
Cv T1 ρ2
Finally
 2 
2 T2
γ T1 −1

µ ρ
 2γ (γ − 1) 43 + µv M1 2 ρ21 − 1 + 4 Pr T2 +1 
ρU δ  T1 
= (γ−1) . (9.96)
µ 
T2 T2 ρ1

T1 + 1 ln T1 ρ2
 
The right hand side of (9.96) is expressed in terms of ratios across the shock and can
be written entirely in terms of the upstream Mach number using the shock jump rela-
tions.
ρ1 (γ − 1) M1 2 + 2
=
ρ2 (γ + 1) M1 2
(9.97)
2γM1 2 − (γ − 1) (γ − 1) M1 2 + 2

T2
=
T1 (γ + 1)2 M1 2
The left hand side of (9.96) can be interpreted as the shock Reynolds number where the
characteristic length is taken to be the shock thickness. In summary

ρU δ 2 µv
=F M1 , γ, Pr , . (9.98)
µ µ
The function F M1 2 , γ, Pr , µv /µ in (9.96) is plotted in figure 9.9 for two values of the

ratio of bulk to shear viscosity µv /µ.
Figure 9.9: Shock Reynolds number as a function of Mach number.
The results of our model indicate that the Reynolds number of a shock wave is small and
decreases with increasing Mach number, affirming the strongly viscous nature of the flow
through the wave. Note also that, at a given Mach number, a gas with a large bulk viscosity
will generate a much thicker shock wave.
Results from kinetic theory relating the viscosity to the mean free path are provided in
Appendix A. To a good approximation µ = ρaλ where λ is the mean free path and a is the
speed of sound. Equation (9.98) can be written

ρU δ ∼ 2 µv
=F M1 , γ, Pr , . (9.99)
ρaλ µ
In the spirit of the approximations used earlier let the Mach number appearing on the left
hand side of (9.99) be approximated as
 
ρ1
+1

U U1 + U2 2 ρ2
= = M1  q . (9.100)
a 2 a1 + a2 T2
+1
T1
Now the ratio of shock thickness to the mean free path in the gas (the inverse of the shock
Knudsen number) can be estimated.
q 
2 µv T2
δ ∼ F M 1 , γ, Pr , µ + 1
=  ρ T1  (9.101)
1
λ M1 ρ2 + 1
This function is plotted in figure 9.10.
Figure 9.10: Shock thickness number as a function of Mach number.
We can see from this final result that in fact a shock is extremely thin, on the order of a
few mean free paths thick. It is important that the model gives a shock thickness greater
than the mean free path so as to be consistent with the fact that the shock is viscous and
that a molecule experiences at least several collisions as it passes through the shock. It is
the increased randomization of the molecular motion due to these collisions that leads to
the entropy rise across the shock imposed by the shock jump conditions.
This result also serves as a warning that the model, based on the Navier-Stokes equations,
has its limitations! Notice that for Mach numbers above about 2, the shock is a very small
number of mean free paths thick. This is inconsistent with the assumption that the gas is
in local thermodynamic equilibrium. The equilibrium state of the gas, with well defined
values of temperature (to which the viscosity and thermal conductivity are related) and
pressure, can only be established through collisions and four or five collisions is barely
enough.
The state of the gas inside a very strong shock wave is far from thermodynamic equilibrium
and a proper theory for the flow requires a much more sophisticated treatment. One is
forced to a much more complex analysis based on the Boltzmann equation which describes
the space-time evolution of the velocity probability density function and explicitly accounts
for molecular collisions.
9.5 Problems
Problem 1 - Heat in the amount of 106 J/kg is added to a compressible flow of helium in
a diverging channel. The heat is distributed so that the area averaged velocities at stations
1 and 2 are the same.
Figure 9.11: Heat added to a diverging channel flow.
The temperature at station 1 is 1000 K and the area ratio is A2 /A1 = 2. Determine T2 /T1 ,
ρ2 /ρ1 , P2 /P1 , and (s2 − s1 ) /Cp .
Problem 2 - Recall Problem 5.3. Consider steady flow in one dimension where Ū =
(U (x) , 0, 0) and all velocity gradients are zero except
∂U
A11 = . (9.102)
∂x
Work out the components of the Newtonian viscous stress tensor τij . Note the role of
the bulk viscosity. Inside a normal shock wave the velocity gradient can be as high as
1010 sec−1 . Using values for Air at 300 K and one atmosphere, estimate the magnitude of
the viscous normal stress inside a shock wave. Express your answer in atmospheres.
Problem 3 - Consider a normal shock wave in helium with Mach number M1 = 3. The
temperature of the upstream gas is 300 K and the pressure is 105 N/m2 .
1) Determine the stagnation temperature in region 2 as measured by an observer at rest
with respect to the upstream gas. This is an observer that sees the shock wave propagating
Figure 9.12: Normal shock wave in helium
to the left at Mach 3.

2) Determine the stagnation pressure in region 2 as measured by an observer at rest with
respect to the upstream gas.
Problem 4 - The figure below shows supersonic flow of carbon dioxide,γ = 4/3, past a
cylindrical bullet at a free stream Mach number, M1 = 2.77.
Figure 9.13: Supersonic flow of carbon dioxide past a bullet. See Van Dyke page 163.
a) Determine the temperature, pressure and Mach number of the gas on the centerline just
downstream of the shock wave.
b) Estimate the temperature and pressure at the stagnation point on the upstream face of
the cylinder.
c) Determine the entropy increase across the shock wave.
d) Estimate the thickness of the shock wave.
e) Estimate the acceleration of the fluid element as it traverses the shock wave. Express
your answer in m/sec2 .
Problem 5 - Estimate the thickness of the shock wave in Helium discussed in section
9.3.4.
Problem 6 - The sketch below shows supersonic flow of air, γ = 1.4, past a sphere at a
free stream Mach number, M1 = 1.53. See Van Dyke page 164.
Figure 9.14: Supersonic flow past a sphere.
a) Compare each of the following properties of the gas; stagnation enthalpy, ht , stagnation
pressure, Pt and entropy per unit mass, s at locations 1, 2 and 3 identified in the figure.
State the assumptions needed to make your comparisons.
b) What can you say about the same properties of the gas at station 4? How certain is
your answer? Why?
c) Determine the Mach number at station 2.
d) Is the energy per unit mass (internal plus kinetic) of a gas particle at 1 and 2 the same?
Prove your answer.
Problem 7 - We often encounter practical situations involving weak shock waves where
the Mach number upstream of the wave is very close to one.
Figure 9.15: Weak normal shock
Let the Mach number ahead of the wave be M1 = 1 + ε where ε < 1. Derive the weak
shock jump relations M2 ∼
= 1 − ε and
U2 − U1 ∼ 4
=− ε
a1 γ+1
T2 − T1 ∼
=?
T1
(9.103)
P2 − P1 ∼
=?
P1
Pt2 − Pt1 ∼ 16 γ
=− ε3 .
Pt1 3 (γ + 1)2
The last result in (9.103) is extremely important in that it shows that the stagnation loss
across a weak shock is extremely small indeed. This fact is exploited in the design of
supersonic inlets. Note that first and second order terms in ε have cancelled. I suggest you
use symbol manipulation software such as Mathematica to derive this result.
Problem 8 - The photo below shows a rifle bullet moving in air at a Mach number of 1.1.
The air temperature is 300 K. On the centerline the flow from the left passes through a
normal shock wave and then stagnates on the nose of the bullet.
Figure 9.16: Supersonic rifle bullet.
a) Determine the temperature, pressure and density change across the wave.
b) Compare the temperature, pressure and density of the gas at the nose of the bullet to
values in the freestream.
c) Evaluate the entropy change.
d) State any assumptions used.
Problem 9 - Use the weak shock theory developed in problem 7 to estimate the thickness of
the shock wave depicted in Problem 8. Develop an expression for estimating the thickness
of a weak shock wave δ in terms of ε and γ.
Problem 10 - The figure below shows supersonic flow of air over a model of a re-entry
body at a free stream Mach number, M1 = 2.
Figure 9.17: Supersonic flow over a re-entry body.
1) Determine the stagnation temperature and pressure of a fluid element located at stations
1 (free stream), 2 (just behind the shock) and 3 (at the stagnation point on the body).
State the assumptions used to solve the problem. Express your answers in degrees Kelvin
and atmospheres.
2) What can you say about the state of the gas at point 4?
3) Refer the stagnation temperatures at 1, 2 and 3 to an observer at rest with respect to
the upstream gas. To such an observer the body is moving to the left at a Mach number
of 2.0.
Problem 11 - The photo below shows the flow of Helium gas past a sphere at a Mach
number of 1.05. The pressure is one atmosphere (1.01325×105 N/m2 ) and the temperature
is 300 K. The viscosity of Helium at that temperature is µ1 = 1.98 × 10−5 kg/m − sec.
Consider a fluid element that passes through the shock on the flow centerline.
Figure 9.18: Slightly supersonic flow past a sphere.
Estimate the acceleration of the fluid element as it traverses the shock wave. Express your
answer in m/sec2 .
Chapter 10
Gasdynamics of nozzle flow
A nozzle is an extremely efficient device for converting thermal energy to kinetic energy.
Nozzles come up in a vast range of applications. Obvious ones are the thrust nozzles
of rocket and jet engines. Converging-diverging ducts also come up in aircraft engine
inlets, wind tunnels and in all sorts of piping systems designed to control gas flow. The
flows associated with volcanic and geyser eruptions are influenced by converging-diverging
nozzle geometries that arise naturally in geological formations.
10.1 Area-Mach number function
In Chapter 8 we developed the area-averaged equations of motion.
δ ṁ dA
d (ρU ) = − ρU
A A

1 4Cf dx (Uxm − U ) δ ṁ δFx
d (P − τxx ) + ρU dU = − ρU 2 + − (10.1)
2 D A A

τxx Qx τxx Qx δ ṁ
d ht − + = δqw − δw + htm − ht − +
ρ ρU ρ ρU ρU A
Assume the only effect on the flow is streamwise area change so that
δ ṁ = Cf = δFx = δq = δw = 0. (10.2)
10-1
CHAPTER 10. GASDYNAMICS OF NOZZLE FLOW 10-2
Also assume that streamwise normal stresses and heat fluxes τxx , Qx are small enough to
be neglected. With these assumptions the governing equations (10.1) together with the
perfect gas law reduce to
d (ρU A) = 0
dP + ρU dU = 0
(10.3)
Cp dT + U dU = 0
P = ρRT.
Introduce the Mach number
U 2 = γRT M 2 . (10.4)
Each of the equations in (10.3) can be expressed in fractional differential form.
dρ dU 2 dA
+ + =0
ρ 2U 2 A
dP γM 2 dU 2
+ =0
P 2 U2
(10.5)
dT (γ − 1) M 2 dU 2
+ =0
T 2 U2
dP dρ dT
= +
P ρ T
Equation (10.4) can also be written in fractional differential form.
dU 2 dT dM 2
= + (10.6)
U2 T M2
Use the equations for mass, momentum and energy to replace the terms in the equation of
state.
γM 2 dU 2 dU 2 dA (γ − 1) M 2 dU 2
− = − − − (10.7)
2 U2 2U 2 A 2 U2
Solve for dU 2 /U 2 .
dU 2

2 dA
= (10.8)
U2 M2 − 1 A
Equation (10.8) shows the effect of streamwise area change on the speed of the flow. If the
Mach number is less than one then increasing area leads to a decrease in the velocity. But
if the Mach number is greater than one then increasing area leads to an increase in flow
speed. Use (10.8) to replace dU 2 /U 2 in each of the relations in (10.5).
M2

dρ dA
=−
ρ M2 − 1 A
γM 2

dP dA
=− (10.9)
P M2 − 1 A
(γ − 1) M 2

dT dA
=−
T M2 − 1 A
Equations (10.9) describe the effects of area change on the thermodynamic state of the
flow. Now use (10.8) and the temperature equation in (10.6).
(γ − 1) M 2 dA dM 2

2 dA
=− + (10.10)
M2 − 1 A M2 − 1 A M2
Rearrange (10.10). The effect of area change on the Mach number is
dA M2 − 1 dM 2
= . (10.11)
A 2 1 + γ−1 M 2 M2
2
Equation (10.11) is different from (10.8) and (10.9) in that it can be integrated from an
initial to a final state. Integrate (10.11) from an initial Mach number M to one.
1 A∗
M2 − 1 dM 2
Z Z
dA
= (10.12)
M2 2 1 + γ−1 M 2 M2 A A
2
The result is
( γ+1 !)1
A∗

γ−1 2(γ−1)
ln = −ln (M ) + ln 2 1 + M2 . (10.13)

A 2

M2
Evaluate (10.13) at the limits
γ+1 ! ( γ+1 !)
A∗

γ+1 2(γ−1) γ−1 2(γ−1)
ln = ln − −ln (M ) + ln 1+ M2
A 2 2
(10.14)
which becomes
 
γ+1
A∗

 γ+1 2(γ−1) M 
ln = ln  γ+1  .
 (10.15)
A  2
γ−1

2(γ−1)
1 + 2 M2
Exponentiate both sides of (10.15). The result is the all-important area-Mach number
equation.
γ+1
A∗

γ+1 2(γ−1) M
f (M ) = = γ+1 (10.16)
A 2
1 + γ−1
2 M 2 2(γ−1)
In (10.16) we referenced the integration process to M = 1. The area A∗ is a reference

area at some point in the channel where M = 1 although such a point need not actually
be present in a given problem. The area-Mach-number function is plotted below for three
values of γ.
Note that for smaller values of γ it takes an extremely large area ratio to generate high
Mach number flow. A value of γ = 1.2 would be typical of the very high temperature
mixture of gases in a rocket exhaust. Conversely, if we want to produce a high Mach
number flow in a reasonable size nozzle, say for a wind tunnel study, an effective method
is to select a monatomic gas such as Helium which has γ = 1.66. A particularly interesting
feature of (10.16) is the insensitivity of f (M ) to γ for subsonic flow.
10.1.1 Mass conservation
The result (10.16) can also be derived simply by equating mass flows at any two points in
the channel and using the mass flow relation.
Figure 10.1: Area-Mach number function.
ṁ = ρU A (10.17)
This can be expressed as
P
ṁ = ρU A = (γRT )1/2 M A. (10.18)
RT
Insert
Tt γ−1 2
=1+ M
T 2
γ (10.19)
Pt γ − 1 2 γ−1
= 1+ M
P 2
into (10.18) to produce

γ PA
ṁ = ρU A = γ+1 √ t f (M ) . (10.20)
γ+1

2(γ−1) γRTt
2
If we equate the mass flows at any two points in a channel (10.20) gives
ṁ1 = ṁ2
(10.21)
Pt1 A1 Pt2 A2
√ f (M1 ) = √ f (M2 ) .
Tt1 Tt2
In the case of an adiabatic (Tt = constant), isentropic (Pt = constant) flow in a channel
(10.16) provides a direct relation between the local area and Mach number.
A1 f (M1 ) = A2 f (M2 ) (10.22)
10.2 A simple convergent nozzle
Figure 10.2 shows a large adiabatic reservoir containing an ideal gas at pressure Pt . The gas
exhausts through a simple convergent nozzle with throat area Ae to the ambient atmosphere
at pressure Pambient . Gas is continuously supplied to the reservoir so that the reservoir
pressure is effectively constant. Assume the gas is calorically perfect, (P = ρRT , Cp and
Cv are constant) and assume that wall friction is negligible.
Let’s make this last statement a little more precise. Note that we do not assume that the
gas is inviscid since we want to accommodate the possibility of shock formation somewhere
in the flow. Rather, we make use of the fact that, if the nozzle is large enough, the boundary
layer thickness will be small compared to the diameter of the nozzle enabling most of the
flow to be treated as irrotational and isentropic.
Figure 10.2: Reservoir with a convergent nozzle.
The isentropic assumption works quite well for nozzles that are encountered in most appli-
cations. But if the plenum falls below a few centimeters in size with a nozzle diameter less
than a few millimeters then a fully viscous, non-isentropic treatment of the flow is required.
Accurate nozzle design, regardless of size, virtually always requires that the boundary layer
on the wall of the plenum and nozzle is taken into account.
If the ambient pressure equals the reservoir pressure there is, of course, no flow. If Pambient
is slightly below Pt then there is a low-speed, subsonic, approximately isentropic flow from
the plenum to the nozzle. If Pt /Pambient is less than a certain critical value then the
condition that determines the speed of the flow at the exit is that the exit static pressure
is very nearly equal to the ambient pressure.
Pe = Pambient (10.23)
The reason this condition applies is that large pressure differences cannot occur over small
distances in a subsonic flow. Any such difference that might arise, say between the nozzle
exit and a point slightly outside of and above the exit, will be immediately smoothed out by
a readjustment of the whole flow. Some sort of shock or expansion is required to maintain
a pressure discontinuity and this can only occur in supersonic flow. Slight differences in
pressure are present due to the mixing zone that exists outside the nozzle but in subsonic
flow these differences are very small compared to the ambient pressure. Since the flow up
to the exit is approximately isentropic the stagnation pressure Pt is approximately constant
from the reservoir to the nozzle exit and we can write
γ
Pt γ−1 2
γ−1
= 1+ Me . (10.24)
Pe 2
Using (10.23) and (10.24) we can solve for the Mach number at the nozzle exit in terms of
the applied pressure ratio.
1/2 γ−1 !1/2

2 Pt γ
Me = −1 (10.25)
γ−1 Pambient
Note that the nozzle area does not appear in this relationship.
10.2.1 The phenomenon of choking
The exit Mach number reaches one when
γ
Pt γ+1 γ−1
= . (10.26)
Pambient 2
For Air with γ = 1.4 this critical pressure ratio is Pt /Pambient = 1.893 and the condition ??
holds for 1 ≤ Pt /Pambient ≤ 1.893. At Pt /Pambient = 1.893 the area-Mach number function
f (M ) is at its maximum value of one. At this condition the mass flow through the nozzle
is as large as it can be for the given reservoir stagnation pressure and temperature and the
nozzle is said to be choked.
If Pt /Pambient is increased above the critical value the flow from the reservoir to the nozzle
throat will be unaffected; the Mach number will remain Me = 1 and Pt /Pambient = 1.893.
However condition (10.23) will no longer hold because now Pe > Pambient . The flow exiting
the nozzle will tend to expand supersonically eventually adjusting to the ambient pressure
through a system of expansions and shocks.
Figure 10.3: Plenum exhausting to very low pressure.
10.3 The converging-diverging nozzle
Now let’s generalize these ideas to the situation where the nozzle consists of a converging
section upstream of the throat and a diverging section downstream. Consider the nozzle
geometry shown below.
The goal is to completely determine the flow in the nozzle given the pressure ratio Pt /Pambient
and the area ratio Ae /Athroat . Before analyzing the flow we should first work out the critical
exit Mach numbers and pressures for the selected area ratio. Solving
γ+1
Athroat γ+1 2(γ−1) Me
= γ+1 (10.27)
Ae 2
γ−1

2 2(γ−1)
1 + 2 Me
gives two critical Mach numbers Mea < 1 and Meb > 1 for isentropic flow in the nozzle with
M = 1 at the throat. The corresponding critical exit pressures are determined from
Figure 10.4: Converging-diverging geometry
γ
Pt γ−1 2
γ−1
= 1+ Mea
Pea 2
(10.28)
γ
Pt γ−1 2
γ−1
= 1+ Meb .
Peb 2
There are several possible cases to consider.
10.3.1 Case 1 - Isentropic subsonic flow in the nozzle
If Pt /Pambient is not too large then the flow throughout the nozzle will be subsonic and
isentropic and the pressure at the exit will match the ambient pressure. In this instance
the exit Mach number is determined using (10.25).
If Pt /Pambient is increased there is a critical value that leads to choking at the throat. This
flow condition is sketched below.
Figure 10.5: Onset of choking.
The exit Mach number is Mea and the pressure ratio is Pt /Pambient = Pt /Pea . Note that
when a diverging section is present the pressure ratio that leads to choking is less than
that given by (10.26). The flow in the nozzle is all subsonic when the pressure ratio is in
the range
Pt Pt
1< < . (10.29)
Pambient Pea
10.3.2 Case 2 - Non-isentropic flow - shock in the nozzle
If the pressure ratio is increased above Pt /Pea a normal shock will form downstream of
the throat, the exit Mach number remains subsonic and the exit pressure will continue to
match the ambient pressure. This flow condition is shown below.
Figure 10.6: Shock in nozzle.
The entropy is constant up to the shock wave, increases across the wave and remains
constant to the exit. To work out the flow properties, first equate mass flows at the throat
and nozzle exit
ṁthroat = ṁexit (10.30)
or
Pt Athroat = Pte Ae f (Me ) . (10.31)
The key piece of information that enables us to solve for the flow is that the exit pressure
still matches the ambient pressure and so we can write
γ γ
γ−1 2
γ−1 γ−1 2
γ−1
Pte = Pe 1 + Me = Pambient 1 + Me . (10.32)
2 2
When (10.32) is incorporated into (10.31) the result is

γ+1 1
Pt Athroat γ+1 2(γ−1) γ−1 2
2
= Me 1 + Me . (10.33)
Pambient Ae 2 2
The items on the left side of (10.33) are known quantities and so one solves (10.33) implicitly
for Me < 1. With the exit Mach number known, (10.31) is used to determine the stagnation
pressure ratio across the nozzle.

Pte Athroat 1
= <1 (10.34)
Pt Ae f (Me )
Since the only mechanism for stagnation pressure loss is the normal shock, the value of
Pte /Pt determined from (10.34) can be used to infer the shock Mach number from
 γ 
 γ−1  1
γ+1

Pte

2M 2 shock γ+1 γ−1
2
=    . (10.35)
Pt 1 + γ−1 M 2 γM 2 − γ−1
2 shock shock 2
Thus all of the important properties of the flow in the nozzle are known given the plenum
to ambient pressure ratio and the nozzle area ratio.
As the nozzle pressure ratio is increased, the shock moves more and more downstream until
it is situated at the nozzle exit. This flow condition is shown below.
Figure 10.7: Normal shock at the nozzle exit.
In this case the Mach number just ahead of the shock is the supersonic critical value Meb
and the Mach number just behind is the corresponding subsonic value derived from normal
shock relations.

γ−1
1+ 2 M 2 eb
2
M e(behindshock) = (10.36)
γ−1
γM 2 eb − 2
The value of Pt /Pambient that produces this flow is
γ+1 1
Pt Ae γ+1 2(γ−1) γ−1 2
2
= Me(behindshock) 1 + Me(behindshock) .
Pambient exit⊥shock Athroat 2 2
(10.37)
The shock-in-the-nozzle case occurs over the range

Pt Pt Pt
< < . (10.38)
Pea Pambient Pambient exit⊥shock
10.3.3 Case 3 - Isentropic supersonic flow in the nozzle
If the nozzle pressure ratio exceeds the value given in (10.37) then no further changes occur
in the flow within the nozzle. Three different cases are distinguished.
i) Over expanded flow - This corresponds to the range

Pt Pt Pt
< < . (10.39)
Pambient exit⊥shock Pambient Peb
In this case the flow passes through an oblique shock as it exhausts.
Figure 10.8: Oblique shock at the nozzle exit.
ii) Fully expanded flow - This is the case where
Pt Pt
= . (10.40)
Pambient Peb
The exit pressure now matches the ambient pressure and the flow exhausts smoothly.
Figure 10.9: Fully expanded flow.
iii) Under expanded flow - This corresponds to the range
Pt Pt
> . (10.41)
Pambient Peb
In this case the exit pressure exceeds the ambient pressure and the flow expands outward
as it leaves the nozzle.
Figure 10.10: Expansion fan at the nozzle exit.
A good example of the occurrence of all three conditions is the Space Shuttle Main Engine
which leaves the pad in an over expanded state, becomes fully expanded at high altitude
and then extremely under expanded as the Shuttle approaches the vacuum of space.
10.4 Examples
10.4.1 Shock in a nozzle
A normal shock is stabilized in the diverging section of a nozzle. The area ratios are,
As /Athroat = 2, Ae /Athroat = 4 and A1 = Ae .
Figure 10.11: Converging-diverging nozzle with shock.
1) Determine M1 , Me , the Mach number just ahead of the shock Ma , and the Mach number
just behind the shock, Mb . Assume the gas is Air with γ = 1.4.
Solution
The area ratio from the throat to the shock is 2. One needs to solve
f (Ma ) = 1/2 (10.42)
for the supersonic root. The solution from tables or a calculator is Ma = 2.197. The
normal shock relation for the downstream Mach number is

γ−1
1+ 2 Ma2
Mb2 = (10.43)
γ−1
γMa2 − 2
which gives Mb = 0.547. At station 1 the area ratio to the throat is 4 and the Mach number
is subsonic. Solve for the subsonic root of
f (M1 ) = 1/4. (10.44)
The solution is M1 = 0.147. The area ratio from behind the shock to station e is two. If
we equate mass flows at both points and assume isentropic flow from station b to e we can
write
Ab f (Mb ) = Ae f (Me ) . (10.45)

Solve (10.45) for f (Me ).
Ab
f (Me ) = f (Mb ) = (1/2) (0.794) = 0.397 (10.46)
Ae
The exit Mach number is Me = 0.238. So far the structure of the flow is as shown
below.
Figure 10.12: Converging-diverging nozzle with Mach numbers labeled.
2) Determine the pressure ratio across the nozzle, Pe /Pt1 .

Solution
Since the exit flow is subsonic the exit pressure matches the ambient pressure. Use
(10.33).
Pe (Athroat /Ae )
= γ+1
Pt1
γ+1 2(γ−1)

γ−1

2
1/2
2 M e 1 + 2 M e
(10.47)
(1/4)
= 1/2 = 0.604
1 3
1 + (1/5) (0.238)2

2 (0.238)
3) What pressure ratio would be required to position the shock at station e?

Solution
The structure of the flow in this case would be as shown in 10.13. To determine the pressure
ratio that produces this flow use (10.37).
Figure 10.13: Converging-diverging nozzle with shock at the exit.

Pt Pt1
= =
Pambient exit⊥shock Pe exit⊥shock
γ+1 1
Ae γ+1 2(γ−1) γ−1 2
2
Me(behindshock) 1 + Me(behindshock) = (10.48)
Athroat 2 2

Pe 1
= 1 = 0.295
Pt1 exit⊥shock
4(1.2)3 (0.479) 1 + (1/5) (0.479)2
2
This is considerably lower than the pressure ratio determined in part 2.
10.4.2 Cold gas thruster
A cold gas thruster on a spacecraft uses Helium (atomic weight 4) as the working gas. The
gas exhausts through a large area ratio nozzle to the vacuum of space. Compare the energy
of a parcel of gas in the fully-expanded exhaust to the energy it had when it was in the
chamber.
Answer
In the chamber the energy per unit mass, neglecting kinetic energy is
Echamber = Cv Tchamber . (10.49)

Assume the expansion takes place adiabatically. Under that assumption, the stagnation
enthalpy is conserved.
1
Cp Tchamber = Cp T + U 2 = constant (10.50)
2
Since the area ratio is large the thermal energy of the exhaust gas is small compared to
the kinetic energy.
1 2 ∼ 1 2 ∼
Eexhaust = Cv Texhaust + Uexhaust = Uexhaust = Cp Tchamber (10.51)
2 2
Divide (10.51) by U1 2 . The result is
Eexhaust Cp Tchamber 5
= =γ= . (10.52)
Echamber Cv Texhaust 3
The energy gained by the fluid element during the expansion process is due to the pressure
forces that accelerate the element. In fact what is recovered is exactly the work required
to create the original pressurized state.
10.4.3 Gasdynamics of a double throat, starting and unstarting super-

sonic flow
One of the most important applications of the gas-dynamic tools we have been developing
is to a channel with multiple throats. Virtually all air-breathing propulsion systems utilize
at least two throats; one to decelerate the incoming flow and a second to accelerate the exit
flow. When a compressor and turbine are present several more throats may be involved.
The simplest application of of two throats is to the design of a supersonic wind tunnel.
Shown below is a supersonic wind tunnel that uses air as the working gas.
A very large plenum contains the gas at constant stagnation pressure and temperature, Pt ,
Tt . The flow exhausts to a large tank that is maintained at vacuum Pa = 0. The upstream
nozzle area ratio is A2 /A1 = 6 and the ratio of exit area to throat area is Ae /A1 = 2.
The test section has a constant area A3 = A2 . A shock wave is stabilized in the diverging
portion of the nozzle. The wall friction coefficient is very small.
1) Determine Pte /Pe .
Solution
The mass balance between stations 1 and e is
Figure 10.14: Supersonic wind tunnel with two throats.
ṁ1 = ṁe
(10.53)
P A P A
√ t1 1 f (M1 ) = √ te e f (Me ) .
γRTt1 γRTte
The flow exits to vacuum and so the large pressure ratio across the system essentially
guarantees that both throats must be choked, M1 = 1 and Me = 1. Assume the flow
is adiabatic and neglect wall friction. With these assumptions the mass balance (10.53).
reduces to
Pte A1
= = 0.5. (10.54)
Pt Ae
2) Determine the shock Mach number

Solution
From the relations for shock wave flow, the shock Mach number that reduces the stagnation
pressure by half for a gas with γ = 1.4 is Ms = 2.5.
3) Determine the Mach numbers at stations 2 and 3.
Solution
The Mach number at station 3 is determined by the area ratio from 3 to e and the fact
that the exit is choked.
Ae 1
= ⇒ M3 = 0.195 (10.55)
A3 3
Since the area of the test section is constant and friction is neglected the Mach number at
station 2 is the same M2 = 0.195.
4) Suppose Ae is reduced to the point where Ae = A1 . What happens to the shock?
Solution
Again use the mass flow equation (10.20) and equate mass flows at the two throats. In this
case (10.53) is
Pte A1
= = 1.0. (10.56)
Pt Ae
There is no shock and therefore there is no stagnation pressure loss between the two throats.
As Ae is reduced the shock moves upstream to lower Mach numbers till a point is reached
when the two areas are equal. At that point the shock has essentially weakened to the
point of disappearing altogether.
5) Suppose Ae is made smaller than A1 , what happens?
Solution
Since both the stagnation pressure and temperature are now constant along the channel
and the exit throat is choked the mass balance (10.53) becomes
Ae
f (M1 ) = . (10.57)
A1
The Mach number at the upstream throat becomes subsonic and satisfies (10.57) as the
area is further reduced.
6) Suppose Ae is increased above Ae /A1 = 2. What happens to the shock?
Solution
In this case the shock moves downstream to higher Mach numbers. The highest Mach
number that the shock can reach is at the end of the expansion section of the upstream
nozzle where the area ratio is A2 /A1 = 6. Equation (10.16) gives the Mach number of the
shock at that point as M2 = 3.368. The corresponding stagnation pressure ratio across
the shock is Pte /Pt = 0.2388. Using the mass balance again, the throat area ratio that
produces this condition is
Ae Pt
= = 4.188. (10.58)
A1 Pte
Throughout this process the exit is at Me = 1 and the flow in the test section is subsonic
due to the presence of the shock. In fact the Mach number in the test section from station
2 to 3 would be the Mach number behind a Mach 3.368 shock which is 0.4566. Note that
this is consistent with the area ratio A3 /Ae = 6/4.188 = 1.433 for which the subsonic
solution of (10.16) is 0.4566.
7) Now suppose Ae /A1 is increased just slightly above 4.188, what happens?
Solution
Again go back to the mass flow relation (10.53). Write (10.53) as
Pt1 A1 = Pte Ae f (Me ) . (10.59)
The upstream throat is choked and so the mass flow is fixed and the left-hand-side of
(10.59) is fixed. The shock is at the highest Mach number it can reach given the area ratio
of the upstream nozzle. So as Ae /A1 is increased above 4.188 there is no way for Pte /Pt to
decrease so as to maintain the equality (10.59) enforced by mass conservation. Instead an
event occurs and that event is that the shock is swallowed by the downstream throat and
supersonic flow is established in the test section. The supersonic wind tunnel is said to be
started. Since there is no shock present the flow throughout the system is isentropic and
the mass balance (10.59) becomes
A1 1
f (Me ) = = . (10.60)
Ae 4.188
The Mach number at the exit throat is now the supersonic root of (10.60), Me = 2.99. If
Ae /A1 is increased further the exit Mach number increases according to Equation (10.60). If
Ae /A1 is reduced below 4.188 the exit Mach number reduces below 2.99 until it approaches
one from above as Ae /A1 → 1 + ε. If Ae /A1 is reduced below one the wind tunnel unstarts
and the flow between 1 and the exit is all subsonic (no shock) with M1 = Me = 1.
10.5 Problems
Problem 1 - Consider the expression ρU n . The value n = 1 corresponds to the mass flux,
n = 2 corresponds to the momentum flux and n = 3 corresponds to the energy flux of a

compressible gas. Use the momentum equation
dP + ρU dU = 0 (10.61)
to determine the Mach number (as a function of n) at which ρU n ; n = 1 is a maximum in

steady flow.
Problem 2 - In the double-throat example above the flow exhausts into a vacuum chamber.
Suppose the pressure Pa is not zero. What is the maximum pressure ratio Pa /Pt that would
be required for the supersonic tunnel to start as described in the example? The exit area
can be varied as required.
Problem 3 - In the double-throat example above suppose the effect of wall friction is
included. How would the answers to the problem change? Would the various values
calculated in the problem increase, decrease or remain the same and why?
Problem 4 - Figure 10.15 shows a supersonic wind tunnel which uses helium as a working
gas. A very large plenum contains the gas at constant stagnation pressure and temperature
Pt , Tt . Supersonic flow is established in the test section and the flow exhausts to a large
tank at pressure Pa .
Figure 10.15: Supersonic wind tunnel with variable area exit.
The exit area Ae can be varied in order to change the flow conditions in the tunnel Initially
A2 /Ae = 4, and A2 /A1 = 8. The gas temperature in the plenum is Tt = 300K. Neglect
wall friction. Let Pt /Pa = 40.
1) Determine the Mach numbers at Ae , A1 and A2 .
2) Determine the velocity Ue and pressure ratio Pe /Pa .

3) Suppose Ae is reduced. Determine the value of Ae /A2 which would cause the Mach
number at Ae to approach one (from above). Suppose Ae is reduced slightly below this
value - what happens to the supersonic flow in the tunnel? Determine Pte /Pt and the Mach
numbers at A1 , A2 and Ae for this case.
Problem 5 - Figure 10.16 shows a supersonic wind tunnel which uses air as the working gas.
A very large plenum contains the gas at constant stagnation pressure and temperature, Pt ,
Tt . The flow exhausts to a large tank that is maintained at vacuum Pa = 0. The upstream
nozzle area ratio is A2 /A1 = 3. The downstream throat area Ae can be varied in order to
change the flow conditions in the tunnel. Initially, Ae = 0. Neglect wall friction. Assign
numerical values where appropriate.
Figure 10.16: Supersonic wind tunnel exiting to vacuum.
1) Suppose Ae /A1 is slowly increased from zero. Plot Pte /Pt as a function of Ae /A1 for
the range 0 ≤ Ae /A1 ≤ 3.
2) Now with Ae /A1 = 3 initially, let Ae be decreased back to zero. Plot Pte /Pt as a function
of Ae /A1 for this process.
Problem 6 - In Chapter 2 we looked at the blowdown through a small nozzle of a calorically
perfect gas from a large adiabatic pressure vessel at initial pressure Pi and temperature
Ti to the surroundings at pressure Pa and temperature Ta . I would like you to reconsider
that problem from the point of view of the conservation equations for mass and energy.
Use a control volume analysis to determine the relationship between the pressure, density
and temperature in the vessel as the mass is expelled. Show that the final temperature
derived from a control volume analysis is the same as that predicted by integrating the
Gibbs equation.
Problem 7 - Consider the inverse of Problem 6. A highly evacuated, thermally insulated
flask is placed in a room with air temperature Ta . The air is allowed to enter the flask
through a slightly opened stopcock until the pressure inside equals the pressure in the
room. Assume the air to be calorically perfect. State any other assumptions needed to
solve the problem.
(i)Use a control volume analysis to determine the relationship between the pressure, density
and temperature in the vessel as mass enters the vessel.
(ii)Determine the entropy change per unit mass during the process for the gas that enters
the vessel.
(iii)Determine the final temperature of the gas in the vessel.
May I suggest that you break the process into two parts. When the stopcock is first opened,
the opening is choked and the flow outside the flask is steady. But after a while the opening
un-chokes and the pressure at the opening increases with time. In the latter case the flow
outside the flask is unsteady and one needs to think of a reasonable model of the flow in
order to solve the problem.
Chapter 11
Area change, wall friction and heat

transfer
11.1 Control volume
Assume a calorically perfect gas, ( P = ρRT , Cp and Cv constant) is flowing in a channel

in the absence of body forces. Viscous friction acts along the wall and there may be heat
conduction through the wall but there is no mass addition. Consider a differential length
of the channel.
Figure 11.1: Differential changes along a channel.
The area-averaged equations of motion are
11-1
CHAPTER 11. AREA CHANGE, WALL FRICTION AND HEAT TRANSFER 11-2
d (ρU A) = 0

1 4Cf dx
dP + ρU dU = − ρU 2 (11.1)
2 D
Cp dT + U dU = δqw = Cp dTt .
For convenience let’s put these equations into fractional differential form.
dρ 1 dU 2 dA
+ 2
= −A
ρ 2 U
γM 2 dU 2 γM 2

dP 4Cf dx
+ = −
P 2 U2 2 D
(11.2)
dT (γ − 1) M 2 dU 2 dTt
+ 2
=
T 2 U T
dP dρ dT
− − =0
P ρ T
As noted when we developed the quasi-one-dimensional equations, the spirit of this analysis
is that A (x), Cf (x) and Tt (x) are specified input functions along the channel. The problem
is to use the equations of motion to determine how the various (output) properties of the
flow change along the channel particularly the Mach number.
11.2 Entropy and stagnation pressure
At any point in the flow the static and stagnation states of the gas are related by
γ
Pt Tt γ−1
= . (11.3)
P T
This relationship essentially defines Pt . If we take the log of this equation and differentiate,
the result is Gibbs equation in terms of stagnation quantities.

γ dTt dPt γ dT dP ds
− = − = (11.4)
γ−1 Tt Pt γ−1 T P R
Now use the fractional differential form of the area averaged equations to replace dT /T
and dP/P in Gibbs equation. The result is a differential form describing the effect of heat
transfer and friction on the stagnation pressure.
γM 2 γM 2

dPt dTt 4Cf dx
=− − (11.5)
Pt 2 Tt 2 D
Stagnation pressure losses are proportional to M 2 indicating that the losses due to both
heat transfer and friction are very small in a low Mach number flow and, conversely, very
large in a high Mach number flow.
The entropy relation is
γM 2 4Cf dx

ds γ γ−1 2 dTt
= 1+ M + (11.6)
R γ−1 2 Tt 2 D
which describes the 1-D variation of entropy due to viscous friction and heat conduction
effects.
It would appear from this result that one can increase the stagnation pressure of a flow by
cooling through the wall. Lets investigate this possibility. Consider the combined effects of
friction and heat transfer through the wall on the stagnation pressure. The heat transfer
process is characterized by the Stanton number defined as
Qw
St = . (11.7)
ρU Cp (Twall − Tt )
In (11.7)
Tt = the adiabatic wall recovery temperature, equal to the stagnation temperature of the
free stream for a recovery factor of one ( Pr = 1 ),
Twall = the actual wall temperature,
Qw = the heat transfer rate to the fluid (energy/area-sec).
The total heat per unit mass transferred to the fluid is
!
Qw πDdx πDdx 4St dx
dht = = St Cp (Twall − Tt ) πD2
= Cp (Twall − Tt ) . (11.8)
ρU A D
4
Now the stagnation pressure change is

γM 2

dPt 4Cf dx Twall − Tt 4St dx
=− + . (11.9)
Pt 2 D Tt D
A careful study of boundary layer flows with heat transfer leads to the recognition that
heat transfer and viscous friction are coupled. This is expressed in terms of the well known
Reynolds analogy discussed in Chapter 5 which states that
2St = Cf . (11.10)
Using (11.10), the fractional differential equation for the stagnation pressure becomes
γM 2

dPt Twall + Tt 4Cf dx
=− (11.11)
Pt 2 2Tt D
which is always positive, implying a loss of stagnation pressure regardless of whether Qw

is positive or negative. Cooling the wall does decrease the rate at which the stagnation
pressure drops in the direction of the flow.
11.3 Velocity, density, temperature and pressure
Now let’s examine the effects of dA, Cf dx and dTt on the other properties of the gas. In
equation (11.2) replace each term in the ideal gas law using the mass, momentum and
energy equations. The result is an equation for the differential change in velocity.
1 − M2 dU 2 dA γM 2

4Cf dx dTt
= − + + (11.12)
2 U2 A 2 D T
Now the velocity equation is used in each of the mass, momentum and energy equations
in turn to generate equations for the differential changes in density, temperature and pres-
sure.
1 − M2 M2 dA γM 2

dρ 4Cf dx 1 dTt
= − − (11.13)
2 ρ 2 A 4 D 2 T
1 − M2 1 − γM 2 dTt

dT γ−1 2 dAγ γ−1 4 4Cf dx
= M − M + (11.14)
2 T 2 A 2 2 D 2 T
1 − M2 γM 2 dA γM 2 γM 2 dTt

dP 4Cf dx
1 + (γ − 1) M 2

= − − (11.15)
2 P 2 A 4 D 2 T
These equations provide some insight into how various inputs affect the flow but they
are not particularly useful for solving problems because, as they stand they cannot be
integrated.
11.4 The Mach number
Lets determine the differential equation for the Mach number. Recall
U 2 = M 2 γRT. (11.16)
Take the derivative of (11.16).
dM 2 dU 2 dT
= − (11.17)
M2 U2 T
Replace the differentials on the right-hand-side of (11.17) using (11.12) and (11.14).
 
1 − M2 2 2 1 + γM 2

 dM = − dA + γM 4Cf dx dTt
 2
+ (11.18)
1+ γ−1
2 M 2 2M A 2 D 2 Tt
When only one effect occurs at a time, the above equation can be integrated exactly and
this is the basis for the gas tables relating Mach number to friction (Fanno line - Cf
assumed constant), heat addition (Rayleigh line) and area change. When two or more
effects are present, it is generally necessary to specify A (x), Tt (x) and Cf (x) in detail and
to integrate the equations numerically. Notice that when heat transfer and or friction are
present the sonic point does not have to occur at the minimum area point. In (11.18) let
M = 1. The relation becomes

dA γ 4Cf dx 1+γ dTt
= + . (11.19)
A 2 D 2 Tt
If we introduce the Stanton number


dA γ 4Cf dx 1+γ Twall − Tt 4St dx
= + (11.20)
A 2 D 2 Tt D
and the Reynolds analogy 2St = Cf , then the sonic point occurs where

dA (γ − 1) Tt + (γ + 1) Twall 4Cf dx
= . (11.21)
A 4Tt D
According to (11.21) in a real flow, the sonic point will always occur slightly downstream
of the throat in the diverging part of the nozzle where dA is slightly positive.
11.5 Mass flow, area-Mach-number function
Note that the heat addition and friction terms on the right hand side of (11.18) can be
rearranged with the use of (11.5) to yield
 
1 − M2 2
 dM = − dA − dPt + 1 dTt .
 2
(11.22)
1+ γ−1
2 M 2 2M A Pt 2 Tt
Equation (11.22) can be integrated to
r
Pt2 A2 Tt1 f (M1 )
= . (11.23)
Pt1 A1 Tt2 f (M2 )
where
γ+1
A∗

γ+1 2(γ−1) M
f (M ) = = γ+1 (11.24)
A 2
γ−1 2

2(γ−1)
1+ 2 M
is the area-Mach number function introduced in Chapter 9 and A∗ is a reference area where
M = 1. Between any two points in a channel in the presence of friction and heat transfer,
we can equate mass flows.
ṁ1 = ṁ2
(11.25)
Pt1 A1 Pt2 A2
√ f (M1 ) = √ f (M2 )
Tt1 Tt2
This simple equality, used so heavily in Chapter 9 to analyze a channel with two throats,
is one of the most important in all of gas dynamics. It is amazingly useful in solving
seemingly very complex problems as we shall see in several examples.
11.6 Integrated relations
11.6.1 1-D, adiabatic, constant area flow with friction (Fanno line flow)
The relationship between Mach number and friction from (11.18) is

!
4Cf dx 1 − M2 dM 2
= . (11.26)
D 1 + γ−1
2 M
2 γM 4
Assume the friction coefficient is constant (or equal to some mean value averaged over
the length of the duct in question). Let the final state correspond to the reference value
M = 1.
!
Lmax 1
1 − M2 dM 2
Z Z
4Cf dx
= (11.27)
0 D M2 1 + γ−1
2 M
2 γM 4
Carry out the integration.

 
M2 1) M 2

4Cf Lmax 1− γ+1 (γ +
= + ln   (11.28)
D γM 2 2γ 2 1 + γ−1 M 2
2
The length of duct L required for the flow to pass from a given initial Mach number M1
to a final Mach number M2 is found from

4Cf L 4Cf Lmax 4Cf Lmax
= − . (11.29)
D D M1 D M2
In each of the following

1 − M2 dU 2 γM 2

4Cf dx
2
=
2 U 2 D
1 − M2 γM 2

dρ 4Cf dx
=−
2 ρ 4 D
(11.30)
M2

1− dT γ γ−1 4Cf dx
=− M4
2 T 2 2 D
1 − M2 γM 2

dP 4Cf dx
1 + (γ − 1) M 2

=−
2 P 2 D
replace 4Cf dx/D with 1 − M 2 / 1 + ((γ − 1) /2) M 2 dM 2 / γM 4 and carry out the

integration. The result is the set of Fanno line relations.
!1/2
γ+1 2
U 2 M ρ∗
= =
U∗ 1 + γ−1
2 M
2 ρ
!
γ−1 2
T∗ 1+ 2 M
= γ+1
T 2
!1/2 (11.31)
γ−1 2
P∗ 1+ 2 M
=M γ+1
P 2
γ+1
Pt∗

γ+1 2(γ−1) M
= γ+1
Pt 2
γ−1 2

2(γ−1)
1+ 2 M
Recall that friction exerted by a no-slip wall at rest does not affect the stagnation enthalpy.
The various Fanno line relations (11.28) and (11.31) are plotted in Figure 11.2 for γ =
1.4.
11.6.2 Example - frictional flow in a long pipe
Figure 11.3 shows air at constant stagnation pressure and temperature, Pt , Tt exhausting
from a very large plenum to the surrounding atmosphere through a long straight pipe. The
gas temperature in the plenum is Tt = 300K and Pt /Pa = 10.
Figure 11.2: Fanno line relations.
Figure 11.3: Frictional flow in a long constant-area pipe.

The friction coefficient in the pipe is Cf = 0.01 and the pipe is 10 diameters long.
1) Determine the Mach number at the entrance to the pipe M1 .
Solution
Begin with the assumption that the flow is choked at the exit. Solve for M1 using (11.28).
The result is
4Cf Lmax
= 0.4 ⇒ M1 = 0.625. (11.32)
D
2) Determine Pte /Pt and Pe /Pa .

Solution
The Fanno line relations (11.31) give the stagnation pressure drop along the pipe.
Pte
= 0.862 (11.33)
Pt
Alternatively, we could get the same result using (11.25) evaluated at the entrance and
exit of the pipe where the Mach numbers are known. Thus at the pipe exit
Pte
= 10 × 0.862 = 8.62. (11.34)
Pa
Since the exit Mach number is one
γ
Pte γ+1 γ−1
= = 1.23.5 = 1.893 (11.35)
Pe 2
and the static pressure ratio at the exit is
Pe 8.62
= = 4.554. (11.36)
Pa 1.893
The fact that the exit static pressure comes out so much larger than the ambient, solidly
confirms our assumption that the exit is choked.
3) Suppose the pipe is lengthened. What value of L/D will produce Pe /Pa = 1?
Solution
At this condition the exit is choked, Me = 1 and the exit pressure is no longer above the
ambient pressure, that is Pe = Pa . Note the identity

Pe Pt Pte Pe
= . (11.37)
Pa Pa Pt Pte
The condition Pe /Pa = 1 implies that the stagnation pressure drop along the pipe must
be

Pte 1 Pte 1.893 1
1 = 10 ⇒ = = . (11.38)
Pt 1.893 Pt 10 5.28
From the Fanno line stagnation pressure relation in (11.31) we find that this stagnation
pressure ratio corresponds to M1 = 0.11. Using this entrance Mach number (11.28) gives
the length of the pipe.
4Cf Lmax
= 54.32 (11.39)
D
The value of L/D that unchokes the exit is
L 54.32
= = 1358. (11.40)
D 4 × 0.01
11.6.3 Integrated relations - 1-D frictionless, constant area flow with

heat transfer (Rayleigh line flow)
In the presence of heat transfer, the relationship between Mach number and stagnation
temperature from (11.18) is
!
1 − M2 dM 2

dTt 1
= γ−1 . (11.41)
Tt 1 + γM 2 1+ 2 M
2 M2
Let the final state correspond to M = 1.

!
Tt∗ 1
1 − M2 dM 2
Z Z
dTt 1
= γ−1 (11.42)
Tt Tt M2 1 + γM 2 1+ 2 M
2 M2
Carry out the integration indicated in (11.42).

2
Tt∗ 1 + γM 2
= (11.43)
Tt 2 (1 + γ) M 2 1 + γ−1 M 2
2
The stagnation temperature change required for the flow to pass from a given initial Mach
number M1 to a final Mach number, M2 is found from
Tt∗

Tt2 Tt
= . (11.44)
Tt1 Tt∗ M2 Tt M1
In each of the following relations
1 − M2 dU 2

dTt
2
=
2 U T
(11.45)
M2

1− dρ 1 dTt
=−
2 ρ 2 T
and
1 − M2 1 − γM 2

dT dTt
=
2 T 2 T
(11.46)
1 − M 2 dP γM 2 dTt

=−
2 P 2 T
replace dTt /Tt with 1 − M 2 / 1 + ((γ − 1) /2) M 2 dM 2 / γM 4 and carry out the in-

tegration. The result is the set of Rayleigh line relations.
U (γ + 1) M 2 ρ∗
= =
U∗ 1 + γM 2 ρ
2
T∗ 1 + γM 2

=
T (γ + 1) M
(11.47)
P∗ 1 + γM 2
=
P γ+1
! γ
γ+1
Pt∗ 1 + γM 2 γ−1

2
= γ−1
Pt γ+1 1+ 2 M
2
The various Rayleigh line relations (11.42) and (11.47) are plotted below for γ = 1.4.
Figure 11.4: Rayleigh line relations.
11.6.4 Example - double throat with heat addition
In Chapter 9 we studied the flow in a channel with two throats. Here we will revisit the
problem and add heat to the flow. Lets recall the flow situation. The figure below shows a
supersonic wind tunnel that uses Air as a working gas. A very large plenum contains the
gas at constant stagnation pressure and temperature, Pt , Tt . The flow exhausts to a large
tank that is maintained at vacuum Pa = 0. The upstream nozzle area ratio is A2 /A1 = 6
and the ratio of exit area to throat area is Ae /A1 = 2. The test section has a constant
area A3 = A2 . A shock wave is stabilized in the diverging portion of the nozzle. Neglect
wall friction.
Suppose heat is added to the flow between stations 2 and 3.
1) Does the shock Mach number increase, decrease or remain the same? Explain.
Solution
The mass balance between stations 1 and e gives
P t A1 Pte Ae
√ = √ = constant (11.48)
Tt Tte
which is considerably lower than the pressure ratio determined in part 2.

As heat ispadded, the shock must move upstream to a lower Mach number so that the
ratio Pte / Tte remains constant. This is essentially a constraint imposed by the fixed
Figure 11.5: Supersonic wind tunnel with normal shock and heat addition.
mass flow through the upstream throat. Through the shock movement upstream, the
stagnation pressure at the exit goes up as the temperature goes up. The mechanism for
stagnation pressure loss between stations 1 and e is a combination of shock loss and heat
addition.
2) What happens to the Mach numbers at stations 2 and 3 as the heat is added?
Solution
The Mach number at station 3 is fixed by choking at the exit and, because of the geometry
constraint, must remain constant during the heat addition process. With the Mach number
at 3 fixed, the heat addition between 2 and 3 must lead to a decrease in Mach number at
station 2. This is consistent with the upstream movement of the shock.
3) Suppose sufficient heat is added between 2 and 3 so that the throat at Station 1 just
barely becomes un-choked. Determine the Mach numbers at stations 2 and 3 for this
condition. How sensitive are these Mach number values to the choice of working gas? If
the gas were changed to Argon (atomic weight = 40) would the Mach numbers at 2 and 3
be much different? If not, why not?
Solution
As noted above, the Mach number at station 3 is fixed by choking at the exit. For Ae /A3 =
1/3 the Mach number at station 3 is M3 = 0.197. The Mach number at station one is very
close to one and the section from 1 to 2 is shock free thus the Mach number at station 2
for an area ratio A1 /A2 = 6 is M2 = 0.097.
Note that the subsonic branch of f (M ) is practically the same for various values of γ.
Thus the Mach numbers at 2 and 3 would not be expected to change much when the test
gas is changed from air γ = 7/5 to argon γ = 5/3. The factor ((γ + 1) /2)((γ+1)/(2(γ−1)))
out in front of f (M ) evaluates to 1.893 for γ = 1.4 versus 1.77 for γ = 1.66.
4) Recall the relationship between stagnation pressure loss and heat addition. Argue
whether the stagnation pressure loss between stations 2 and 3 is large or small?
Solution
At the point where the upstream throat barely chokes, the Mach number at station 2 is
M2 = 0.097. The stagnation pressure loss between stations 2 and 3 is proportional to
the Mach number squared and would be expected to be very small. To a pretty decent
approximation at this Mach number, Pte ∼
= Pt .
5) At approximately what ratio of Tte /Tt does the un-choking of the throat at station 1
occur?
Solution
Since there is no shock and the heat addition takes place at low Mach number and we
are asked for an approximate result, the stagnation temperature ratio at which the flow
un-chokes can be determined from the mass flow balance.
A A T
√ 1 = √ e ⇒ te = 4 (11.49)
Tt Tte Tt
11.7 Detonations and deflagrations
To complete this chapter, we will consider an important class of 1-D flows with heat
addition where the source of heat is a combustible gas mixture of fuel and oxidizer. The
1-D governing equations are
d (ρU ) = 0
d P + ρU 2 = 0

(11.50)

1 2
d Cp T + U = δq
2
The flows governed by these equations are characterized by waves that combine the features
of shock waves and heat addition. Consider a channel filled with a combustible mixture as
shown in 11.6.
Figure 11.6: Compression waves created by a source of ignition in a combustible mixture.
Upon ignition with a spark, the subsequent process of heat release and expansion of hot
gases is energetic enough to produce pressure disturbances that propagate along the tube
away from the source forming a self-sustaining combustion wave. One of two things may
happen.
1) A weak deflagration wave may occur in which heat is transferred to the upstream gas by
conduction. In this case the wave speed is generally very low, a few tens of centimeters per
second for hydrocarbon-air mixtures up to some tens of meters per second for a mixture
of hydrogen and oxygen. The speed of the wave cannot be determined from the heat of
reaction and jump conditions alone. It can only be determined by a model that treats the
details of the wave and includes the effects of heat conduction and viscosity. This is the
sort of combustion wave one observes when a gas flame on a stove is ignited.
2) If the reaction is sufficiently energetic and heat loss mechanisms such as radiation and
conduction to the channel walls are sufficiently small then the pressure waves preceding
the flame front can catch up to one another in a process of non-linear steepening due
to the dependence of acoustic speed on gas temperature. The steepening of the pressure
front quickly forms a shock wave. The shock preheats the unburned gas passing through
it causing the reaction to proceed at an even faster rate. Eventually the flow behind the
combustion wave becomes thermally choked and the shock-combustion wave combination,
called a detonation, moves along the pipe at an extreme speed on the order of thousands
of meters per second. In this case the wave speed can be determined from the heat of
reaction and the jump conditions across a region of heat addition.
Figure 11.7 somewhat artificially decomposes the detonation into a conventional shock
followed by a thermally choked region of heat addition called a strong deflagration. De-
pending on the speed of the reaction and the mean free path of the gas, the whole region
of change is extremely thin (a fraction of a millimeter) and the intermediate zone a is an
idealization of a state that only exists at one point in the wave if at all. The equations of
motion can be integrated to form the jump conditions
Figure 11.7: Chapman-Jouget detonation wave structure.
ρ1 U1 = ρ2 U2
P1 + ρ1 U1 2 = P2 + ρ2 U2 2
(11.51)
1 1
Cp T1 + U1 2 + ∆ht = Cp T2 + U2 2
2 2
P = ρRT
where ∆ht is the stagnation enthalpy change per unit mass across the wave. Now we need
to determine the possible states of motion that can occur and the Mach number of the
wave for a given enthalpy change. First combine the mass and momentum equation to
form the equation

P2 ρ1
−1 = γM12 1− . (11.52)
P1 ρ2
The energy equation can be rearranged to read

1 U12 T2 1 ρ1 2 U12

∆ht
+1+ = + (11.53)
2 Cp T1 Cp T1 T1 2 ρ2 Cp T1
or
2 !
∆ht γ−1 ρ1 P2 ρ1
1+ + M12 1− = . (11.54)
Cp T1 2 ρ2 P1 ρ2
Use (11.52) to eliminate M1 in (11.54). The result is
 
∆ht ρ1

P2

2γ  + 1 −
Cp T1ρ2
−1 = . (11.55)
P1 γ−1 γ+1 ρ1
− 1
γ−1 ρ2
Equations (11.52) and (11.55) are two functions relating pressure and density that are both
satisfied simultaneously and fully define the state of the gas.

P2 ρ1
−1=F γ, , M1
P1 ρ2
(11.56)
P2 ρ1 ∆ht
− 1 = G γ, ,
P1 ρ2 Cp T1
If ∆ht /Cp T1 = 0 (11.55) reduces to the Hugoniot relation for a normal shock. Equating
(11.52) and (11.55) leads to a quadratic equation for ρ1 /ρ2 the roots of which are
2 1/2
ρ1 1 + γM12 ± M12 − 1 − 2 (γ + 1) M12 C∆h t
p T1
= . (11.57)
ρ2 (γ + 1) M12
For ∆ht /Cp T1 = 0 the roots are ρ1 /ρ2 = 1, 2 + (γ − 1) M12 / (γ + 1) M12 which are

the normal shock wave values. Note that the upstream Mach number M1 may be either
subsonic or supersonic. A cross-plot of (11.52) and (11.55) will define the state of the gas
for a given ∆ht /Cp T1 . Several different types of combustion waves can occur.
1) A weak deflagration where M1 < 1 and M2 < 1.
2) A strong deflagration or thermally choked wave where M1 < 1 and M2 = 1.
3) A weak detonation where M1 > 1 and M2 < 1.
4) A thermally choked detonation where M1 > 1 and M2 = 1. This is also called a

Chapman-Jouget detonation. Virtually all stable detonations are of Chapman-Jouget type.
In this case the wave can be decomposed into a shock wave and a strong deflagration as
shown in figure 11.7.
The Mach number of a Chapman-Jouget wave for a given value of ∆ht /Cp T1 can be deter-
mined using the Rayleigh line equation for stagnation temperature change. Note that

∆ht Tt2 Tt1
= −1 . (11.58)
Cp T1 Tt1 T1
Now substitute (11.43) into (11.58) to get

 2 
1 + γM12

∆ht γ−1 2
= −1  1+ M1 (11.59)
Cp T1 2 (γ + 1) M12 1 + γ−1 M 2 2
2 1
or
2
∆ht 1 − M12
= . (11.60)
Cp T1 2 (γ + 1) M12
Solve Equation (11.60) for the Mach number.
2 !1/2
∆ht ∆ht
M12 = 1 + (γ + 1) ± 1 + (γ + 1) −1 (11.61)
Cp T1 Cp T1
The subsonic and supersonic roots of (11.61) are plotted below.
Figure 11.8: Mach number of a detonation wave determined by heat release.

11.7.1 Example - detonation in a mixture of fluorine and hydrogen di-

luted by nitrogen.
As an example, consider a stoichiometric mixture of 5% H2 and 5% F2 in90% N2 where

the percentages are mole fractions. Let the initial temperature of the nitrogen be 300K.
For fluorine reacting with hydrogen the heat of reaction is 1.35 × 105 calories per mole
or 1.36 × 107 m2 /sec2 per unit mass. The stagnation enthalpy change across the wave
compared to the enthalpy of the free stream gas is approximately
∆ht 1.36 × 107

= = 4.88. (11.62)
Cp T1 929 × 300 × 10
The extra factor of ten in the denominator comes from the 90% dilution of the reactants
with nitrogen. Using (11.61) and (11.62) with γ = 1.4 gives M1 = 5.0, Ma = 0.41. This
latter result is consistent with the normal shock value. Using normal shock and Rayleigh
relations the conditions through the detonation wave in the rest frame of the wave are
shown below.
Figure 11.9: Detonation wave example.
Note that the drift speed of the gas behind the detonation in a frame of reference at rest
with respect to the upstream gas is U2 − U1 = −700m/ sec. The high pressure and strong
wind generated by the detonation is responsible for the extreme destructive force of this
type of explosion.
11.8 Problems
Problem 1 - Frictional flow of a compressible gas in an adiabatic pipe is sketched in figure

11.10. The gas comes from a large plenum at rest at constant Pt and Tt .
Show that the maximum Mach number must occur at the pipe exit, that the maximum
possible Mach number is one and that supersonic flow cannot occur in the pipe.
Figure 11.10: Frictional flow in an adiabatic pipe.
Problem 2 - Natural gas, which is composed almost entirely of methane (molecular weight
16, γ = 1.3) is to be pumped through an insulated pipe one meter in diameter connecting
two compressor stations 60 kilometers apart. At the upstream station, the stagnation
pressure is 7 atmospheres and the stagnation temperature is 300 K. The mean friction
coefficient is Cf = 0.001. Calculate the maximum mass flow rate that can be pumped
through the pipe by the downstream station.
Problem 3 - In figure 11.11 Air flows from a large plenum through an adiabatic pipe
with L/D = 100. The pressure ratio is Pt /Pa = 1.5 and the friction coefficient is Cf =
0.01.
Figure 11.11: Frictional flow in an adiabatic pipe with given parameters.
Determine M1 , Me and Pte /Pt .

Problem 4 - In figure 11.12 a heater is used to increase the stagnation temperature
of a compressible flow of air exhausting from a large plenum at constant Pt and Tt to
vacuum.
Figure 11.12: Flow in a constant area pipe with heat addition.
1) When Tte /Tt is increased does M1 increase, decrease or remain the same ?
2) Consider the case where the increase in stagnation temperature is very large, Tte /Tt >>
1. What is the value of Pte /Pt in this case?
Problem 5 - Figure 11.13 shows air at constant stagnation pressure and temperature,
Pt , Tt exhausting from a very large plenum to the surrounding atmosphere through a long
straight pipe surrounded by a heater.
Figure 11.13: Flow in a constant area pipe with heat addition and friction neglected.
The gas temperature in the plenum is Tt = 300 K and Pt /P0 = 10. The heater increases
the stagnation temperature to Tte = 400 K. Neglect the effects of friction.
1) Determine the Mach number at the entrance to the pipe.
2) Determine Pte /Pt and Pe /P0 . Recall (11.25).
3) Suppose the amount of heat added is increased. What value of Tte , if any, will produce
Pe /P0 = 1?
Problem 6 - The figure below shows a supersonic wind tunnel supplied with Carbon
Dioxide ( γ = 4/3, Molecular weight = 44 ) from a very large plenum. The CO2 passes
through the test section and exhausts to a large vacuum chamber, Pa = 0. A high power,
infrared laser tuned to one of the absorption lines of CO2 is used to heat the gas between
stations 2 and 3.
Relevant areas are Ae /A1 = 3, A2 = A3 , A2 /A1 = 5. Neglect wall friction.
1) The laser is initially turned off and supersonic flow is established in the test section.
Determine M2 and Me .
2) The laser is turned on and the stagnation temperature is increased slightly between 2
and 3 so that Tt3 /Tt2 = 1.01. Determine M2 . Estimate Pt3 /Pt2 and Me .
3) How much stagnation temperature increase can occur between stations 2 and 3 before
the wind tunnel un-starts?
Figure 11.14: Supersonic wind tunnel with two throats and heat addition by laser absorption.
4) Suppose the stagnation temperature increase between stations 2 and 3 is such that
Tt3 /Tt2 = 4. Determine Pte /Pt .
Problem 7 - Include the effects of friction and heat addition in a problem similar to
Problem 2. The gas temperature in the plenum is Tt = 300 K and Pt /P0 = 10. The exit
stagnation temperature Tte is not specified, rather a model for the wall temperature will
be specified and Tte will be determined for a given length of pipe. Let Cf = 0.01 and
L/D = 10. The differential change in stagnation temperature in terms of the Stanton
number is

dTt Twall 4St dx
= −1 . (11.63)
Tt Tt D
To solve the problem you will need the constant area Mach number relation
 
1 − M2 2 2 1 + γM 2

 dM = γM 4Cf dx Twall 4St dx
 2
+ −1 .
1+ γ−1
2 M 2 2M 2 D 2 Tt D
(11.64)
Let the wall temperature be distributed such that
Twall
=3 (11.65)
Tt
all along the wall.

1) Use (11.63) to determine Tte .
2) Use the Reynolds analogy (11.10) and integrate (11.64) to determine the Mach number
at the entrance to the pipe.
3) Use mass conservation in the form of (11.25) to determine the stagnation pressure ratio
across the pipe and confirm that the exit flow is choked.
4) What value of L/D (if any) would cause the exit flow to un-choke?
Problem 8 - Consider the frictionless flow of a calorically perfect gas in a channel with
heat addition. The wall is shaped so as to keep the static pressure constant. The flow
enters at condition 1 and departs at condition 2. Let the ratio of stagnation temperature
between the entrance and exit be τ = Tt2 /Tt1 .
(a) Find an expression for M2 in terms of γ, M1 and τ .
(b) Show that
A2 γ−1 2
=τ+ M1 (τ − 1) . (11.66)
A1 2
(c) Show that
!γ/(γ−1)
Pt2 τ
= γ−1 . (11.67)
Pt1 τ + 2 M12 (τ − 1)
Problem 9 - Plot the temperature versus entropy of a calorically perfect gas in a constant
area flow with friction. This is done by creating a chart of temperature versus Mach number
and a chart of entropy versus Mach number. When the two variables are plotted versus
one another with the Mach number as a parameter along the trajectory (ranging, say, from
zero to 4), the resulting curve is called the Fanno line. Repeat the process for the case of
constant area flow with heat transfer to produce the Rayleigh line.
Problem 10 - Figure 11.15 shows a supersonic wind tunnel supplied with Air from a very
large plenum. The air passes through the test section and exhausts to a large vacuum
chamber, Pa = 0. A sphere is placed in the test section between stations 2 and 3. The
tunnel and sphere are adiabatic. The effect of wall friction on the flow can be neglected
compared to the effect of the drag of the sphere.
Relevant areas are A2 /A1 = 5, A2 = A3 and Ae /A1 = 3. A reasonable model for the
stagnation pressure drop due to the sphere is
γM22

Pt3 Asphere
=1− Cd . (11.68)
Pt2 2 Atunnel
Figure 11.15: Supersonic wind tunnel with two throats and a drag producing sphere in the
test section.
Let Cd = 1 and Asphere /Atunnel = 0.04.

1) Supersonic flow is established in the test section. Determine, M2 and M3 .
2) Suppose Ae is decreased. At what value of Ae /A1 does the wind tunnel un-start?
3) Is there a normal shock between 1 and 2 after the tunnel un-starts? If so, estimate the
Mach number of the normal shock.
4) With the tunnel unstarted suppose Ae is increased. Estimate the value of Ae /A1 at
which the wind tunnel starts.
Chapter 12
Steady waves in compressible

flow
12.1 Oblique shock waves
Figure 12.1 shows an oblique shock wave produced when a supersonic flow is deflected by
an angle θ.
Figure 12.1: Flow geometry near a plane oblique shock wave.
We can think of the deflection as caused by a planar ramp at this angle although it could be
generated by the blockage produced by a solid body placed some distance away in the flow.
In general, a 3-D shock wave will be curved, and will separate two regions of non-uniform
flow. However, at each point along the shock, the change in flow properties takes place in
a very thin region much thinner than the radius of curvature of the shock. If we consider a
small neighborhood of the point in question then within the small neighborhood, the shock
may be regarded as locally planar to any required level of accuracy and the flows on either
12-1
CHAPTER 12. STEADY WAVES IN COMPRESSIBLE FLOW 12-2
side can be regarded as uniform. With the proper orientation of axes the flow is locally
two-dimensional. Therefore it is quite general to consider a straight oblique shock wave in
a uniform parallel stream in two-dimensions as shown below.
Balancing mass, two components of momentum and energy across the indicated control
volume leads to the following oblique shock jump conditions.
ρ1 u1 = ρ2 u2
P1 + ρ1 u1 2 = P2 + ρ2 u2 2
(12.1)
ρ1 u1 v1 = ρ2 u2 v2
1 1
u1 2 + v1 2 = h2 + u2 2 + v1 2

h1 +
2 2
Since ρu is constant, v1 = v2 and the jump conditions become
ρ1 u1 = ρ2 u2
P1 + ρ1 u1 2 = P2 + ρ2 u2 2
(12.2)
v1 = v2
1 1
h1 + u1 2 = h2 + u2 2 .
2 2
When the ideal gas law P = ρRT is included, the system of equations (12.2) closes allowing
all the properties of the shock to be determined. Note that, with the exception of the
additional equation, v1 = v2 , the system is identical to the normal shock jump conditions.
The oblique shock acts like a normal shock to the flow perpendicular to it. Therefore
almost all of the normal shock relations can be converted to oblique shock relations with
the substitution
M1 → M1 Sin (β)
(12.3)
M2 → M2 Sin (β − θ) .
Recall the Rankine-Hugoniot relation


P2
ρ2 γ P1 + 1 + PP21 − 1
= (12.4)
ρ1 γ PP21 + 1 − PP21 − 1
plotted in figure 12.2.
Figure 12.2: Plot of the Hugoniot relation (12.4)
This shows the close relationship between the pressure rise across the wave (oblique or
normal) and the associated density rise. The jump conditions for oblique shocks lead to a
modified form of the very useful Prandtl relation

∗ 2 γ−1
u1 u2 = (a ) − v12 (12.5)
γ+1
where (a∗ )2 = γRT ∗ . From the conservation of total enthalpy, for a calorically perfect gas
in steady adiabatic flow
1 a2 1 γ+1
Cp Tt = Cp T + U 2 = + U2 = (a∗ )2 . (12.6)
2 γ−1 2 2 (γ − 1)
The Prandtl relation is extremely useful for easily deriving all the various normal and
oblique shock relations. The oblique shock relations generated using (12.3) are
P2 2γM1 2 Sin2 (β) − (γ − 1)

=
P1 γ+1
ρ2 u1 (γ + 1) M1 2 Sin2 (β)
= =
ρ1 u2 (γ − 1) M1 2 Sin2 (β) + 2
(12.7)
2γM1 2 Sin2 (β) − (γ − 1) (γ − 1) M1 2 Sin2 (β) + 2

T2
=
T1 (γ + 1)2 M1 2 Sin2 (β)
(γ − 1) M1 2 Sin2 (β) + 2
M2 2 Sin2 (β − θ) = .
2γM1 2 Sin2 (β) − (γ − 1)
The stagnation pressure ratio across the shock is
γ 1
(γ + 1) M1 2 Sin2 (β)
γ−1
Pt2 γ+1 γ−1
= . (12.8)
Pt1 (γ − 1) M1 2 Sin2 (β) + 2 2
2γM1 Sin2 (β) − (γ − 1)
Note that (12.8) can also be generated by the substitution (12.3).
12.1.1 Exceptional relations
One all new relation that has no normal shock counterpart is the equation for the absolute
velocity change across the shock.
2
M1 2 Sin2 (β) − 1 γM1 2 Sin2 (β) + 1

U2
=1−4 (12.9)
U1 (γ + 1)2 M1 4 Sin2 (β)
Exceptions to the substitution rule (12.3) are the relations involving the static and stag-
nation pressure, Pt2 /P1 and Pt1 /P2 across the wave. The reason for this is as follows.
Consider
γ
Pt2 Pt2 Pt1 Pt2 γ−1 2
γ−1
= = 1+ M1 . (12.10)
P1 Pt1 P1 Pt1 2
Similarly
γ
Pt1 Pt1 Pt2 Pt1 γ−1 2
γ−1
= = 1+ M2 . (12.11)
P2 Pt2 P2 Pt2 2
The stagnation to static pressure ratio in each region depends on the full Mach number,
not just the Mach number perpendicular to the shock wave.
12.1.2 Flow deflection versus shock angle
The most basic question connected with oblique shocks is: given the free stream Mach
number, M1 , and flow deflection, θ, what is the shock angle, β? The normal velocity ratio
is
u2 (γ − 1) M1 2 Sin2 β + 2 u2 v1
= 2 2
= . (12.12)
u1 (γ + 1) M1 Sin β u1 v2
From the velocity triangles in figure 12.1
u1
T an (β) =
v1
(12.13)
u2
T an (β − θ) = .
v2
Now
(γ − 1) M1 2 Sin2 (β) + 2

T an (β − θ) = T an (β) . (12.14)
(γ + 1) M1 2 Sin2 (β)
An alternative form of this relation is

 
2
M1 Sin2 (β) −1
T an (θ) = Cot (β)  . (12.15)
γ+1 2 2
1+ 2 M1 − M1 Sin2 (β)
The shock-angle-deflection-angle relation (12.15) is plotted in figure 12.3 for several values
of the Mach number.
Corresponding points in the supersonic flow past a circular cylinder sketched below are
indicated on the M1 = 1.5 contour. At point a the flow is perpendicular to the shock wave
and the properties of the flow are governed by the normal shock relations. In moving from
point a to b the shock weakens and the deflection of the flow behind the shock increases
until a point of maximum flow deflection is reached at b . The flow solution between a
and b is referred to as the strong solution in figure 12.3. Notice that the Mach number
behind the shock is subsonic up to point c where the Mach number just downstream of
Figure 12.3: Flow deflection versus shock angle for oblique shocks.
the shock is one. Between c and d the flow corresponds to the weak solutions indicated
in figure 12.3. If one continued along the shock to very large distances from the sphere
the shock will have a more and more oblique angle eventually reaching the Mach angle
β → µ = Sin−1 (1/M1 ) corresponding to an infinitesimally small disturbance.
Figure 12.4: Supersonic flow past a cylinder with shock structure shown.
Note that as the free-stream Mach number becomes large, the shock angle becomes inde-
pendent of the Mach number.
Cos (β) Sin (β)

lim T an (θ) = (12.16)
M1 →∞ γ+1
2 − Sin2 β
12.2 Weak oblique waves
In this section we will develop the differential equations that govern weak waves generated
by a small disturbance. The theory will be based on infinitesimal changes in the flow and
for this reason it is convenient to drop the subscript 0 10 on the flow variables upstream
of the wave. The sketch below depicts the case where the flow deflection is very small
dθ << 1. Note that M is not close to one.
Figure 12.5: Small deflection in supersonic flow.
In terms of figure 12.3 we are looking at the behavior of weak solutions close to the hori-
zontal axis of the figure. For a weak disturbance, the shock angle is very close to the Mach
angle Sin (µ) = 1/M . Let
1
Sin (β) = +ε (12.17)
M
and make the approximation
M 2 Sin2 (β) ∼
= 1 + 2M ε. (12.18)
Using (12.18) we can also develop the approximation
M3

1/2
Cot (β) ∼
= M2 − 1 1− ε . (12.19)
M2 − 1
Using (12.18) and (12.19) the (β, θ) relation (12.15) can be expanded to yield
2 − 1 1/2

4 M
T an (dθ) ∼ ∼
= dθ = ε. (12.20)
γ+1 M
The velocity change across the shock (??) is expanded as

2 2
1 1

U2 − U1
2 M2 M +ε − 1 γM 2 M +ε +1
+1 =1−4 2 . (12.21)
U1 (γ + 1)2 M 4 1
M +ε
Retaining only terms of order ε the fractional velocity change due to the small deflection
is
2
dU 8
+1 =1− ε. (12.22)
U (γ + 1) M
Equation (12.22) is approximated as
dU 4
=− ε. (12.23)
U (γ + 1) M
Write (12.23) in terms of the deflection angle

dU 4 4 γ+1 M
=− ε=− dθ (12.24)
U (γ + 1) M (γ + 1) M 4 (M 2 − 1)1/2
or
dU 1
=− dθ (12.25)
U (M 2 − 1)1/2
where dθ is measured in radians. Other small deflection relations are
dP γM 2
= dθ
P (M 2 − 1)1/2
dρ M2
= dθ (12.26)
ρ (M 2 − 1)1/2
dT (γ − 1) M 2
= dθ
T (M 2 − 1)1/2
and
dPt 2γ 3 16γM 3 3 ds
=− 2 M 2 Sin2 β − 1 = − 2ε = − R (12.27)
Pt 3(γ + 1) 3(γ + 1)
or using (12.20)
dPt γ (γ + 1) M 6 ds
=− 3/2
(dθ)3 = − . (12.28)
Pt 2
12(M − 1) R
Note that the entropy change across a weak oblique shock wave is extremely small; the
wave is nearly isentropic. The Mach number is determined from
dM 2 dU 2 dT 2 (γ − 1) M 2
= − = − dθ − dθ. (12.29)
M2 U2 T (M 2 − 1)1/2 (M 2 − 1)1/2
Adding terms

γ−1 2
dM 2 1+ 2 M
= −2 dθ. (12.30)
M2 (M 2 − 1)1/2
Eliminate dθ between (12.25) and (12.30) to get an integrable equation relating velocity
and Mach number changes.
dU 2 1 dM 2
2
= (12.31)
U 1 + γ−1 2 M2
2 M
The weak oblique shock relations (12.26) are, in terms of the velocity.
dP γ dU 2
= − M2 2
P 2 U
dU 2

dT γ−1
=− M2 2 (12.32)
T 2 U
dρ dU 2
= −M 2 2
ρ U
These last relations are precisely the same ones we developed for one dimensional flow
with area change in the absence of wall friction and heat transfer in chapter 9. From that
development we had
1 − M 2 dU 2 dA
2
=−
2 U A
(12.33)
1 − M2 dM 2 dA
=− .
2 1+ γ−1 2 M2 A
2 M
If we eliminate dA/A between these two relations, the result is
dU 2 1 dM 2
= (12.34)
U2

1 + γ−1 2 M2
2 M
which we just derived in the context of weak oblique shocks.
12.3 The Prandtl-Meyer expansion
The upshot of all this is that
dU 2 2
2
=− dθ (12.35)
U (M 2 − 1)1/2
is actually a general relationship valid for steady, isentropic flow. In particular it can be
applied to negative values of dθ. Consider flow over a corner.
Figure 12.6: Supersonic flow over a corner.

Express the angle in terms of the Mach number.
1/2
M2 − 1 dM 2
dθ = − (12.36)
2 1 + γ−1 2 M2
2 M
Now integrate the angle between the initial and final Mach numbers.
1/2
θ M2 M2 − 1 dM 2
Z Z
0
dθ = − (12.37)
2 1 + γ−1 2 M2
2 M
0 M1
Let ω be the angle change beginning at the reference mach number M1 = 1. The integral
(12.37) is
1/2 1/2 !
γ+1 γ−1 1/2 1/2
ω (M ) = T an−1 M2 − 1 − T an−1 M 2 − 1 . (12.38)
γ−1 γ+1
This expression provides a unique relationship between the local Mach number and the
angle required to accelerate the flow to that Mach number beginning at Mach one. The
straight lines in figure 12.5 are called characteristics and represent particular values of the
flow deflection. According to (12.38) the Mach number is the same at every point on a
given characteristic. This flow is called a Prandtl-Meyer expansion and (12.38) is called
the Prandtl-Meyer function, plotted below for several values of γ.
Figure 12.7: Prandtl-Meyer function for several values of γ.

Note that for a given γ there is a limiting angle at M 2 → ∞.
1/2 !
π γ+1
ωmax = −1 (12.39)
2 γ−1
π
For γ = 1.4, ωmax = 1.45 . The expansion angle can be greater than90◦ . If the
2
deflection is larger than this angle there will be a vacuum between ωmax and the wall.
12.3.1 Example - supersonic flow over a bump
Air flows past √

a symmetric 2-D bump at a Mach number of 3. The aspect ratio of the
bump is a/b = 3.
Figure 12.8: Supersonic flow over a bump.
Determine the drag coefficient of the bump assuming zero wall friction.
Drag f orce per unit span

Cd = 1 2 (12.40)
2 ρ1 U1 b
Solution
The ramp angle is 30◦ producing a 52◦ oblique shock with pressure ratio
P2
= 6.356. (12.41)
P1
The expansion angle is 60◦ producing a Mach number
M2 = 1.406
(12.42)
ω = 9.16◦ .
The stagnation pressure is constant through the expansion wave and so the pressure ratio
over the downstream face is
M3 = 4.268
(12.43)
ω = 69.16◦
and
 γ
 γ−1
γ−1 !3.5
P3  1+ 2 M2 2 1 + 0.2(1.406)2

1.395
3.5
=  = = = 0.0149 (12.44)
1 + 0.2(4.268)2

P2 1+ γ−1
M 2 4.643
2 3
and
P3 P3 P2
= = 0.0149 × 6.356 = 0.0945. (12.45)
P1 P2 P1
The drag coefficient becomes
P2 2b Sin (30) − P3 2b Sin (30) 6.356 − 0.0945

Cd = γ 2 = 1.4 = 0.994. (12.46)
2 M1 P 1 b 2 (9)
12.4 Problems
Problem 1 - Use the oblique shock jump conditions (12.2) to derive the oblique shock
Prandtl relation (12.5).
Problem 2 - Consider the supersonic flow past a bump discussed in the example above.
Carefully sketch the flow putting in the shock waves as well as the leading and trailing
characteristics of the expansion.
Problem 3 - Consider a streamline in compressible flow past a 2-D ramp with a very small
ramp angle.
Figure 12.9: Supersonic flow past a 2-D ramp.

Determine the ratio of the heights of the streamline above the wall before and after the
oblique shock in terms of M1 and dθ find the unknown coefficient in (12.47).
H2
= 1 + (???) dθ (12.47)
H2
Pay careful attention to signs.

Problem 4 - Consider a body in subsonic flow. As the free-stream Mach number is
increased there is a critical value, Mc , such that there is a point somewhere along the body
where the flow speed outside the boundary layer reaches the speed of sound. Figure 12.10
illustrates this phenomena for flow over a projectile.
Figure 12.10: Projectile in high subsonic flow.
In this figure 12.10 the critical Mach number is somewhere between 0.840 and 0.885 as
evidenced by the weak shocks that appear toward the back of the projectile in the middle
picture. The local pressure in the neighborhood of the body is expressed in terms of the
pressure coefficient.
P − P∞
CP = 1 2 (12.48)
2 ρ∞ U∞
Show that the value of the pressure coefficient at the point where sonic speed occurs is
γ
1+ γ−1 M∞c 2 γ−1
2
γ+1 −1
2
CPc = γ 2 . (12.49)
2 M∞c
State any assumptions needed to solve the problem.

Problem 5 - Consider frictionless (no wall friction) supersonic flow over a flat plate of
chord C at a small angle of attack as shown in figure 12.11.
Figure 12.11: Supersonic flow past a flat plate at a small angle of attack.
The circulation about the plate is defined as

I
Γ= U ds. (12.50)
where the integration is along any contour surrounding the plate.

1) Show that, to a good approximation, the circulation is given by
2U1 C
Γ= 1/2 α (12.51)
M1 2 − 1
where the integration is clockwise around the plate.

2) Show that, to the same approximation, Lif tperunitspan = ρ1 U1 Γ.
Problem 6 - Consider frictionless (no wall friction) flow of air at M = 2 over a flat plate
of chord C at 5◦ angle of attack as shown in figure 12.12.
Figure 12.12: Supersonic flow over a flat plate at 5◦ angle of attack.

Evaluate the drag coefficient of the plate. Compare with the value obtained using a weak
wave approximation.
Problem 7 - Consider frictionless (no wall friction) supersonic flow of Air over a flat plate
of chord C at an angle of attack of 15 degrees as shown in figure 12.13.
Figure 12.13: Supersonic flow over a flat plate at 15◦ angle of attack.
Determine the lift coefficient
L
CL = 1 2 (12.52)
2 ρ∞ U∞ C
where L is the lift force per unit span.

Problem 8 - Figure 12.14 shows a symmetrical, diamond shaped airfoil at a 5◦ angle of
attack in a supersonic flow of air.
Figure 12.14: Supersonic flow past a diamond shaped airfoil.
Determine the lift and drag coefficients of the airfoil.
Lif t per unit span

CL = 1 2
2 ρ∞ U∞ C
(12.53)
Drag per unit span
CD = 1 2
2 ρ∞ U∞ C
What happens to the flow over the airfoil if the free-stream Mach number is decreased to
1.5? Compare your result with the lift and drag of a thin flat plate at 5◦ angle of attack
and free-stream Mach number of 3.
Problem 9 - The figure below shows supersonic flow of Air over a 30◦ wedge followed by
a 10◦ wedge. The free stream Mach number is 3.
Figure 12.15: Supersonic over a wedge with a shoulder.
1) Determine M2 , M3 and the included angle of the expansion fan, φ.

2) Suppose the flow was turned through a single 10◦ wedge instead of the combination
shown above. Would the stagnation pressure after the turn be higher or lower than in the
case shown? Why?
Problem 10 - Figure 12.16 shows a smooth compression of a supersonic flow of air by a
concave surface. The free-stream Mach number is 1.96. The weak oblique shock at the
nose produces a Mach number of 1.932 at station 1. From station 1 to station 2 the flow
is turned 20 degrees.
Figure 12.16: Supersonic flow compressed by a concave surface.
1) Determine the Mach number at station 2.

2) Determine the pressure ratio P2 /P1 .
3) State any assumptions used.
Problem 11 - Figure 12.17 shows supersonic flow of air in a channel or duct at a Mach
number of three. The flow produces an oblique shock off a ramp at an angle of 16 degrees.
The shock reflects off the upper surface of the wind tunnel as shown below. Beyond the
ramp the channel height is the same at the height ahead of the ramp.
Figure 12.17: Mach 3 flow in a duct with a ramp.
1) Determine the Mach number in region 2.

3) Describe qualitatively how Pt and Tt vary between regions 1, 2 and 3.
4) Suppose the channel height is 10 cm . Precisely locate the shock reflections on the upper
and lower walls.
5) Suppose the walls were lengthened. At roughly what point would the Mach number
tend to one?
Problem 12 - Figure 12.18 shows supersonic flow of air turned through an angle of 30◦ .
The free stream Mach number is 3.
Figure 12.18: Supersonic flow turned 30◦ .
In case (a) the turning is accomplished by a single 30◦ wedge whereas in case (b) the
turning is accomplished by two 15◦ degree wedges in tandem. Determine the stagnation
pressure change in each case, (Pt2 /Pt1 )|(a) and (Pt3 /Pt1 )|(b) and comment on the relative
merit of one design over the other.
Problem 13 - Figure 12.19 shows the flow of helium from a supersonic over-expanded
round jet. If we restrict our attention to a small region near the intersection of the first
two oblique shocks and the so-called Mach disc as shown in the blow-up, then we can use
oblique shock theory to determine the flow properties near the shock intersection (despite
the generally non-uniform 3-D nature of the rest of the flow). The shock angles with respect
to the horizontal measured from the image are as shown.
Figure 12.19: Supersonic flow from an over expanded round jet.
1) Determine the jet exit Mach number. Hint, you will need to select a Mach number that
balances the pressures in regions 2 and 4 with a dividing streamline that is very nearly
horizontal as shown in the picture.
3) Determine the flow angles and Mach numbers in regions 3 and 4.
4) Determine P2 /P1 and P4 /P1 . How well do the static pressures match across the dividing
streamline (dashed line) between regions 2 and 4?
Problem 14 - Figure 12.20 shows the reflection of an expansion wave from the upper wall
of a 2-D, adiabatic, inviscid channel flow. The gas is helium at an incoming Mach number,
M1 = 1.5 and the deflection angle is 20◦ . The flow is turned to horizontal by the lower
wall which is designed to follow a streamline producing no reflected wave. Determine M2 ,
M3 and H/h.
Figure 12.20: Supersonic flow in an expansion.

Chapter 13
Unsteady waves in compressible

flow
13.1 Governing equations
One of the most important applications of compressible flow theory is to the analysis of
the generation and propagation of sound. In Chapter 6 we worked out the equations for
homentropic, unsteady flow that are the starting point for the derivation of the acoustic
equations. For ∇ × Ū = 0, and ∇s = 0, the equations of motion are
∂ρ ∂
+ (ρUk ) = 0
∂t ∂xk

∂Ui ∂ Uk Uk ∂P
ρ +ρ + =0 (13.1)
∂t ∂xi 2 ∂xi
γ
P ρ
= .
P0 ρ0
The role of the energy equation is played by the isentropic relation between pressure and
density.
13.2 The acoustic equations
Now consider an unsteady flow where disturbances in density and pressure are extremely
13-1
CHAPTER 13. UNSTEADY WAVES IN COMPRESSIBLE FLOW 13-2
small compared to constant background values.
ρ0
ρ = ρ0 + ρ0 ; << 1
ρ0
(13.2)
P0
P = P0 + P 0 ; << 1
P0
In addition, assume that the velocity field involves very small fluctuations about zero.
Linearize (13.1) by dropping higher order nonlinear terms.
∂ρ ∂Uk
+ ρ0 =0
∂t ∂xk
(13.3)
∂Ui ∂P
ρ0 + =0
∂t ∂xi
Now define the dimensionless density disturbance
ρ − ρ0
r= . (13.4)
ρ0
In terms of this variable the equations of motion become
∂r ∂Uk
+ =0
∂t ∂xk
∂Ui ∂P
ρ0 + =0 (13.5)
∂t ∂xi
P
= (1 + r)γ ∼
= 1 + γr.
P0
Slight rearrangement produces the acoustic equations
∂r ∂Uk
+ =0
∂t ∂xk
(13.6)
∂Ui ∂r
+ a0 2 =0
∂t ∂xi
where
P0
a0 2 = γ . (13.7)
ρ0
Now differentiate the linearized continuity equation with respect to time and the linearized
momentum equation with respect to space.
∂2r ∂ 2 Uk
+ =0
∂t2 ∂xk ∂t
(13.8)
∂ 2 Uk ∂2r
+ a0 2 =0
∂xk ∂t ∂xk ∂xk
Subtract one from the other in (13.8). The dimensionless density disturbance satisfies the
linear wave equation.
∂2r 2
2 ∂ r
− a 0 =0 (13.9)
∂t2 ∂xk ∂xk
Now repeat the process with the differentiation reversed. The result is
∂2r ∂ 2 Uk
+ =0
∂t∂xi ∂xk ∂xi
(13.10)
∂ 2 Ui 2
2 ∂ r
+ a 0 = 0.
∂t2 ∂xi ∂t
Again, subtract one equation from the other in (13.10). The velocity fluctuation satisfies
the equation
∂ 2 Ui 2
2 ∂ Uk
− a0 = 0. (13.11)
∂t2 ∂xk ∂xi
Recall the vector identity ∇ × ∇ × Ū = ∇ ∇ · Ū − ∇2 Ū Using this identity and the fact

that the flow is irrotational, (13.11) can be expressed as the linear wave equation
∂ 2 Ui 2
2 ∂ Ui
− a0 = 0. (13.12)
∂t2 ∂xk ∂xk
Similarly, the pressure satisfies the wave equation. Substitute


1 P − P0
r= (13.13)
γ P0
into (13.9)
∂2P 2
2 ∂ P
− a 0 = 0. (13.14)
∂t2 ∂xk ∂xk
Finally, the temperature also satisfies the wave equation.
∂2T 2
2 ∂ T
− a0 =0 (13.15)
∂t2 ∂xk ∂xk
13.3 Propagation of acoustic waves in one space dimension
In one dimensional flow the acoustic equations become
∂r ∂U
+ =0
∂t ∂x
(13.16)
∂Ui ∂r
+ a0 2 = 0.
∂t ∂x
The wave equation
∂2r 2
2∂ r
− a0 =0 (13.17)
∂t2 ∂x2
can be solved in a very general form
r (x, t) = F (x − a0 t) + G (x + a0 t) (13.18)
where F and G are arbitrary functions. The character of the solution can be understood
by considering first the case G = 0. Figure 13.1 illustrates the evolution of an initial
distribution of F propagating to the right in the x − t plane.
The disturbance moves to the right unchanged in shape. Acoustic waves are nondisper-
sive meaning that the wave speed does not depend on the wavelength. The figure below
illustrates the evolution of an initial distribution of G propagating to the left in the x − t
plane.
Figure 13.1: Right propagating wave.
Figure 13.2: Left propagating wave

In the absence of dispersion and dissipation, the waves simply translate while retaining their
form. The pressure disturbance produced by the wave is related to the density by

P − P0 ρ − ρ0
= γr = γ . (13.19)
P0 ρ0
In differential form
dP dρ
=γ . (13.20)
P ρ
The velocity induced by the acoustic disturbance can also be written in a very general
form.
U (x, t) = f (x − a0 t) + g (x + a0 t) (13.21)
Substitute the expressions for r and U into the acoustic equations.
∂ ∂
(F (x − a0 t) + G (x + a0 t)) + (f (x − a0 t) + g (x + a0 t)) = 0
∂t ∂x
(13.22)
∂ ∂
(f (x − a0 t) + g (x + a0 t)) + a0 2 (F (x − a0 t) + G (x + a0 t)) = 0
∂t ∂x
Let
ξ = x − a0 t
(13.23)
η = x + a0 t.
−a0 Fξ + a0 Gη + fξ + gη = 0
(13.24)
a0 Fξ + a0 Gη − fξ + gη = 0.
Add and subtract the relations in (13.24)
−a0 Fξ + fξ = 0
(13.25)
a0 Gη + gη = 0.
From (13.25) we can conclude that g = −a0 G; f = a0 F . The result (13.25) gives us the
relationship between density and velocity in left and right running waves.
U (x, t) = a0 F (x − a0 t) − a0 G (x + a0 t) (13.26)
Now
U F −G
= . (13.27)
a0 r F +G
In a right running wave
U
= 1. (13.28)
a0 r
In a left running wave
U
= −1. (13.29)
a0 r
In differential form (13.27) is equivalent to
dρ
dU = ±a0
ρ
(13.30)
dP dρ
=γ .
P ρ
13.4 Isentropic, finite amplitude waves
The speed of sound may vary from one point to another and in a general, 1-D, isentropic
flow.
γ−1
ρ 2
a = a0 (13.31)
ρ0
Locally, the particle disturbance velocity is given by (13.30) with the speed of sound re-
garded as a function of space and time.
dρ
dU = ±a (13.32)
ρ
Equation (13.32) is one of the most important equations in unsteady gas dynamics and is
often introduced as an intuitively obvious extension of equation (13.30). We can put (13.32)
on a firmer foundation by returning to the full nonlinear equations for one-dimensional,
isentropic flow. In one dimension (13.1) becomes
∂ρ ∂U ∂ρ
+ρ +U =0
∂t ∂x ∂x
(13.33)
γP0 ρ γ−2 ∂ρ

∂U ∂U
+U + 2 = 0.
∂t ∂x ρ0 ρ0 ∂x
Now assume that the density can be written as a function of the velocity.
ρ = ρ (U ) (13.34)
The derivatives of the density are
∂ρ dρ ∂U
=
∂t dU ∂t
(13.35)
∂ρ dρ ∂U
= .
∂x dU ∂x
Substitute (13.35) into (13.33).
dρ ∂U ∂U dρ ∂U
+ρ +U =0
dU ∂t ∂x dU ∂x
(13.36)
γP0 ρ γ−2 dρ ∂U

∂U ∂U
+U + 2 =0
∂t ∂x ρ0 ρ0 dU ∂x
Rearrange (13.36).
∂U ∂U ρ ∂U
+U + =0
∂t ∂x dρ/dU ∂x
γ−2 (13.37)
∂U ∂U γP0 ρ dρ ∂U
+U + 2 =0
∂t ∂x ρ0 ρ0 dU ∂x
Comparing the continuity and momentum equation in (13.37) we see that in order for
(13.34) to be a solution of (13.33) we must have
γ−2
ρ γP0 ρ dρ
= 2 . (13.38)
dρ/dU ρ0 ρ0 dU
Equation (13.38) can be rearranged to read
2 3−γ
ρ0 2

dρ ρ
= 2 . (13.39)
dU a0 ρ0
Take the square root of (13.39).
3−γ
dρ ρ0 ρ 2
=± (13.40)
dU a0 ρ0
Now substitute the isentropic relation between the speed of sound and density
γ−1
a ρ 2
= (13.41)
a0 ρ0
into (13.40). Equation (13.40) becomes
dρ
dU = ±a (13.42)
ρ
which confirms the general validity of (13.32). Using the isentropic assumption, (13.32)
can be integrated from some initial state to a final state.
Z ρ Z ρ γ−1
dρ ρ 2 dρ
U =± a = ±a1 (13.43)
ρ1 ρ ρ1 ρ1 ρ
Integrating,
γ−1 !
2a1 ρ 2 2
U =± −1 =± (a − a1 ) . (13.44)
γ−1 ρ1 γ−1
The local acoustic speed is


γ−1
a = a1 ± U (13.45)
2
and the wave speed at any point is

γ−1 γ+1
c = a ± U = a1 ± U ± U = a1 ± U. (13.46)
2 2
13.5 Centered expansion wave
Now consider the situation shown below where a piston is withdrawn from a fluid at
rest.
Figure 13.3: Withdrawal of a piston from a compressible gas.
The speed of the fluid in region (3) is equal to the piston speed.
2
Up = (a3 − a4 ) (13.47)
γ−1
Note that Up is negative. In effect (13.47) gives us the sound speed in region (3) in terms
of the known piston speed and the known speed of sound in region (4).

γ−1
a3 = a4 + Up (13.48)
2
From this we can see that the gas in region (3) is cooler than that in region (4), consistent
with what we would expect for an expansion. The front or leading characteristic of the
wave propagates to the right with wave speed equal to the speed of sound c4 = a4 . The
tail of the wave (terminating characteristic) moves to the right with wave speed

γ+1
c3 = a3 + Up = a4 + Up (13.49)
2
and the expansion zone widens at the rate (γ + 1) |Up | /2. It is useful to visualize this flow
in the x − t plane.
Figure 13.4: Trajectory of a piston and an expansion wave in the x-t plane.
The density ratio across the wave is given by the isentropic relation
2
ρ3 a3 γ−1
= (13.50)
ρ4 a4
or, using (13.48)
2
ρ3 γ − 1 Up γ−1
= 1+ . (13.51)
ρ4 2 a4
The pressure ratio is
2γ
P3 γ − 1 Up γ−1
= 1+ (13.52)
P4 2 a4
and the temperature ratio is
γ−1
γ − 1 Up 2

T3 P3 γ
= = 1+ . (13.53)
T4 P4 2 a4
In summary, our solution of the small disturbance wave equation from the previous section
has allowed us to determine all the properties of a finite expansion wave. Things worked
out this way because the expansion wave is isentropic to a high degree of accuracy. Note
the similarity between the 1-D unsteady expansion wave and the steady Prandtl-Meyer
expansion, and the presence of space-time characteristics in the x − t plane along which
the properties of the flow are constant.
13.6 Compression wave
Suppose the piston motion is to the right into the gas. None of the above relations change
but the physics does change and this can be seen from the x − t diagram.
Figure 13.5: Compression wave converging to a shock in the x-t plane.
The wave speed at the surface of the piston is
γ+1
c2 = a2 + Up = a1 + Up . (13.54)
2
In this case the characteristics tend to cross, at which point the isentropic assumption is no
longer valid. The system of compression waves catches up to itself (nonlinear steepening)
in a very short time collapsing to form a shock wave.
13.7 The shock tube
A very important device used to study shock waves in the laboratory is the shock tube
shown schematically in figure 13.6.
Figure 13.6: Shock tube flow.
Regions (1) and (2) of the shock tube are initially separated into high and low pressure
sections by a diaphragm. At t = 0 the diaphragm bursts and the high pressure driver gas
rushes into the low pressure test gas. The contact surface between the two surfaces acts like
a piston and the piston motion produces a shock wave moving at the speed cs . The shock
wave arises through a process of nonlinear steepening based on the fact that the speed of
a pressure wave increases with an increase in the temperature of the gas. Pressure waves
to the rear of the compression front move more rapidly than those near the front, quickly
catching up to form a very thin pressure jump as illustrated in figure 13.5. The important
gas-dynamic problem is to determine the pressure ratio across the shock P2 /P1 for a given
choice of gases (γ1 , a1 , γ4 , a4 ) and diaphragm pressure ratio P4 /P1 . In the following we will
combine the results for expansion waves with normal shock relations to derive the so-called
shock tube equation. The conditions at the contact surface are
P2 = P3
(13.55)
U2 = U3 = Up .
where U2 is the speed of the slug of gas set into motion by the opening of the diaphragm.
This is the effective piston speed of the fluid released by the diaphragm.
In a frame of reference moving with the shock wave the gas velocities are
U 0 1 = −cs
(13.56)
U 0 2 = −cs + Up
and the shock jump conditions are
U 02 1 + γ12−1 M1 2
=
U 01 γ1 +1
2 M1
2
(13.57)
2 γ1 −1
P2 γ 1 M1 − 2
= γ1 +1
P1 2
where M1 = cs /a1 = −U10 /a1 . Using the first relation in (13.57) we can write
U 02 − U 01 1 + γ12−1 M1 2 1 − M1 2
= − 1 = . (13.58)
U 01 γ1 +1
2 M1
2 γ1 +1
2 M1
2
The piston velocity is

  !
2
1 − M M1 2 − 1
Up = U20 − U10 = U10  1 0  = a1 γ1 +1 . (13.59)
γ1 +1
2 M1
U 1
2 M1 − a1
Note that Up is positive. Using the second relation in (13.57), equation (13.59) can be
expressed in terms of the shock pressure ratio as
 1/2

P2 2
Up = a1 −1   . (13.60)
P1 P2
γ1 (γ1 + 1) P1 + γ1 (γ1 − 1)
Equation (13.60) is the expression for the piston velocity derived using normal shock theory.
Now lets work out an expression for the piston velocity using isentropic expansion theory.
The velocity behind the expansion is
γ4 −1 !
2a4 P3 2γ4
U3 = Up = 1− . (13.61)
γ4 − 1 P4
Equate (13.60) and (13.61).
 1/2
γ4 −1 !
P2 2 2a4 P3 P2 P1 2γ4
a1 −1   = 1−
P1 P2
γ1 (γ1 + 1) P1 + γ1 (γ1 − 1) γ4 − 1 P2 P1 P4
(13.62)
Using the following identity
P3 P3 P2 P1
= (13.63)
P4 P2 P1 P4
and noting that P3 /P2 = 1 solve (13.61) for P4 /P1 = 1. The result is the basic shock tube
equation

 − 2γ4
γ4 −1
a1 P2
P4 P2  (γ4 − 1) a4 P1 −1
= 1 − (13.64)

P1 P1 1/2 
P2
4γ1 2 + 2γ1 (γ1 + 1) P1 −1
plotted in figure 13.7. Given (P4 /P1 , γ1 , γ4 , a1 , a4 ) we can determine the shock strength
P2 /P1 from (13.64).
Figure 13.7: Shock tube pressure ratio versus shock strength.
Once the shock strength is determined, the shock Mach number can be determined from

P2 2γ1 γ1 − 1
= Ms 2 − (13.65)
P1 γ1 + 1 γ1 + 1
or

2 γ1 + 1 P2 γ1 − 1
Ms = + . (13.66)
2γ1 P1 γ1 + 1
Notice that very large values of P4 /P1 are required to generate strong shocks. One way
to improve the performance of the shock tube is to select a driver gas with a large speed
of sound. Helium driving into nitrogen will produce a much stronger shock wave than say
Argon driving into nitrogen at a given P4 /P1 . In essence Helium provides a much lighter
piston that can translate much more rapidly into the test gas. The figure below provides
this comparison.
Figure 13.8: Shock Mach number for several driver gases.
Hydrogen as the driver gas produces the strongest shock waves however the fire and explo-
sion hazard involved has limited its use.
13.7.1 Example - flow induced by the shock in a shock tube
The flow in a shock tube is shown below.

The shock wave induces a gas velocity in the laboratory frame, Up = U2 = U3 where
Figure 13.9: Flow induced in a shock tube.
 1/2

P2 2
U2 = a1 −1   . (13.67)
P1 P2
γ1 (γ1 + 1) P1 + γ1 (γ1 − 1)
Suppose the gas in the driver section and test section is Air initially at 300 K. A shock
wave with a Mach number of 2 is produced in the tube.
1) Determine the stagnation temperature of the gas in region (2) in the laboratory frame.
2) Determine the stagnation temperature of the gas in region (3) in the laboratory frame.
Answer
The pressure ratio across a Mach number 2 shock in air is
P2
= 4.5 (13.68)
P1
and the temperature ratio is
T2
= 1.687. (13.69)
T1
p
The speed of sound in air is a1 = γRT = 347m/sec and the heat capacity at constant
pressure is, Cp = 1005m2 /sec2 − K. The piston velocity is
 1/2

P2 2
U2 = a1 −1   =
P1 P2
γ1 (γ1 + 1) P1 + γ1 (γ1 − 1) . (13.70)
1/2
2
347 × 3.5 × = 433.75 m/sec
15.68
The temperature in region (3) is obtained from (13.53).
2 2
T3 γ − 1 Up 433.75
= 1+ = 1 − 0.2 = 0.5625 (13.71)
T4 2 a4 347
Now, in region (2)
1 Up 2 433.752
Tt2 = T2 + = 1.687 × 300 + = 506.1 + 93.6 = 599.7 K (13.72)
2 Cp 2 × 1005
and in region (3)
1 Up 2 433.752
Tt3 = T3 + = 0.5625 × 300 + = 168.75 + 93.6 = 262.35 K. (13.73)
2 Cp 2 × 1005
13.8 Problems
Problem 1 - Consider the expression ρU n , n = 1 corresponds to the mass flux, n =

2 corresponds to the momentum flux and n = 3 corresponds to the energy flux of a
compressible gas. Use the acoustic relation
dρ
dU = ±a (13.74)
ρ
to determine the Mach number (as a function of n) at which ρU n is a maximum in a

one-dimensional, unsteady expansion wave. The steady case is considered in one of the
problems at the end of Chapter 9.
Problem 2 - In the shock tube example discussed above, determine the stagnation pressure
of the gas in regions (2) and (3). Determine the stagnation pressure in both the laboratory
frame and in the frame of reference moving with the shock wave.
Problem 3 - Figure 13.10 below shows a sphere moving over a flat plate in a ballistic
range. The sphere has been fired from a gun and is translating to the left at a Mach
number of 3. The static temperature of the air upstream of the sphere is 300 K and the
pressure is one atmosphere. On the plane of symmetry of the flow (the plane of the photo)
the shock intersects the plate at an angle of 25 degrees as indicated.
i) Determine the flow Mach number, speed and angle behind the shock in region 2 close to
the intersection with the plate. Work out your results in a frame of reference moving with
Figure 13.10: Sphere moving supersonically over a flat plate.
the sphere. In this frame the upstream air is moving to the right at 3 times the speed of
sound.
ii) What are the stream-wise and plate normal velocity components of the flow in region 2
referred to a frame of reference at rest with respect to the upstream gas.
iii) Determine the stagnation temperature and pressure of the flow in region 2 referred to
a frame of reference at rest with respect to the upstream gas.
Problem 4 - We normally think of the shock tube as a device that can be used to study
relatively strong shock waves. But shock tubes have also been used to study weak shocks
relevant to the sonic boom problem. Suppose the shock tube is used to generate weak
shock waves with P2 /P1 = 1 + ε. Show that for small ε the relationship between P2 /P1
and P4 /P1 is approximated by
P4
= 1 + Aε. (13.75)
P1
How does A depend on the properties of the gases in regions 1 and 4? Use the exact theory
to determine the strength of a shock wave generated in an air-air shock tube operated at
P4 /P1 = 1.2. Compare with the approximate result.
Problem 5 - The sketch below shows a one meter diameter tube filled with air and divided
into two volumes by a heavy piston of weight W . The piston is held in place by a mechanical
stop and the pressure and temperature are uniform throughout the tube at one atmosphere
and 300K. Body force effects in the gas may be neglected.
Figure 13.11: Gravity driven frictionless piston in a shock tube.
The piston is released and accelerates downward due to gravity. After a short transient
the piston reaches a constant velocity Up . The piston drives a shock wave ahead of it at
a wave speed cs equal to twice the sound speed in region 1. What is the weight of the
piston?
Problem 6 - Each time Stanford makes a touchdown an eight inch diameter, open ended
shock tube is used to celebrate the score. Suppose the shock wave developed in the tube
is required to have a pressure ratio of 2. What pressure is needed in the driver section?
Assume the driver gas is air. What is the shock Mach number? Suppose the shock tube is
mounted vertically on a cart as shown in figure 13.12 below. Estimate the force that the
cart must withstand when the tube fires.
Figure 13.12: Shock tube mounted on a cart.
Problem 7 - In figure 13.13 a moveable piston sits in the middle of a long tube filled with
air at one atmosphere and 300 K. At time zero the piston is moved impulsively to the right
at Up = 200 m/sec.
1) What is the pressure on the right face of the piston (region 2) in atmospheres?
Figure 13.13: Frictionless piston in a shock tube.
2) What is the pressure on the left face of the piston (region 3) in atmospheres?
Problem 8 - One of the most versatile instruments used in the study of fluid flow is the
shock tube. It can even be used as a wind tunnel as long as one is satisfied with short test
times. Figure 13.14 below illustrates the idea. An airfoil is placed in the shock tube and
after the passage of the shock it is subject to flow of the test gas at constant velocity Up
and temperature T2 . In a real experiment the contact surface is quite turbulent and so the
practical usefulness of the flow is restricted to the time after the arrival of the shock and
before the arrival of the contact surface, typically a millisecond or so.
Figure 13.14: Shock tube used as a wind tunnel.
Proponents of this idea point out that if the shock Mach number is very large the flow over
the body can be supersonic as suggested in the sketch above.
1) Show that this is the case.
2) For very large shock Mach number, which test gas would produce a higher Mach number
over the body, helium or air. Estimate the Mach number over the body for each gas.
Problem 9 - Figure 13.15 below shows a shock wave reflecting from the endwall of a shock
tube. The reflected shock moves to the left at a constant speed crs into the gas that was
compressed by the incident shock. The gas behind the reflected shock, labeled region (5),
is at rest and at a substantially higher temperature and pressure than it was in state (1)
before the arrival of the incident shock.
1) Suppose the gas in the driver and test sections is helium at an initial temperature of
300K prior to opening the diaphragm. The Mach number of the incident shock wave is 3.
Figure 13.15: Shock wave reflecting from an end wall.
Determine the Mach number of the reflected shock.

2) Determine T5 /T1 .
Problem 10 - One of the key variables in the design of a shock tube is the length needed
for a shock to develop from the initial compression process. Suppose a piston is used to
compress a gas initially at rest in a tube. During the startup transient 0 < t < t1 the piston
speed increases linearly with time as shown on the x − t diagram in figure 13.16.
Figure 13.16: Piston compressing air.
In a shock tube the startup time t1 is generally taken to be the time required for the
diaphragm to open. Let the gas be air at T1 = 300 K. Use t1 = 5 × 10−3 sec and A =
4 × 104 m/sec2 to estimate the distance needed for the shock wave to form.
Chapter 14
Thin airfoil theory
14.1 Compressible potential flow
14.1.1 The full potential equation
In compressible flow, both the lift and drag of a thin airfoil can be determined to a reason-
able level of accuracy from an inviscid, irrotational model of the flow. Recall the equations
developed in Chapter 6 governing steady, irrotational, homentropic (∇s = 0) flow in the
absence of body forces.

∇ ρŪ = 0

Ū · Ū ∇P
∇ + =0
2 ρ (14.1)
γ
P ρ
=
P0 ρ0
The gradient of the isentropic relation is
∇P = a2 ∇ρ. (14.2)
Recall from the development in Chapter 6 that

P γ−1 ∇P
∇ = . (14.3)
ρ γ ρ
14-1
CHAPTER 14. THIN AIRFOIL THEORY 14-2
Using (14.3) the momentum equation becomes

γ P Ū · Ū
∇ + = 0. (14.4)
γ−1 ρ 2
Substitute (14.2) into the continuity equation and use (14.3). The continuity equation
becomes
Ū · ∇a2 + (γ − 1) a2 ∇ · Ū = 0. (14.5)
Equate the Bernoulli integral to free stream conditions.
a2 a∞ 2 U∞ 2 a∞ 2

Ū · Ū γ−1 2
+ = + = 1+ M∞ = Cp Tt (14.6)
γ−1 2 γ−1 2 γ−1 2
Note that the momentum equation is essentially equivalent to the statement that the
stagnation temperature Tt is constant throughout the flow. Using (14.6) we can write
a2 Ū · Ū
= ht − . (14.7)
γ−1 2
The continuity equation finally becomes

Ū · Ū Ū · Ū
(γ − 1) ht − ∇ · Ū − Ū · ∇ = 0. (14.8)
2 2
The equations governing compressible, steady, inviscid, irrotational motion reduce to a

single equation for the velocity vector U. The irrotationality condition ∇ × Ū = 0 permits
the introduction of a velocity potential.
Ū = ∇Φ (14.9)
and (14.8) becomes

∇Φ · ∇Φ 2 ∇Φ · ∇Φ
(γ − 1) ht − ∇ Φ − ∇Φ · ∇ = 0. (14.10)
2 2
For complex body shapes numerical methods are normally used to solve for Φ . However
the equation is of relatively limited applicability. If the flow is over a thick airfoil or a bluff
body for instance then the equation only applies to the subsonic Mach number regime at
Mach numbers below the range where shocks begin to appear on the body. At high subsonic
and supersonic Mach numbers where there are shocks then the homentropic assumption
(14.2) breaks down. Equation (14.8) also applies to internal flows without shocks such as
fully expanded nozzle flow.
14.1.2 The nonlinear small disturbance approximation
In the case of a thin airfoil that only slightly disturbs the flow, equation (14.8) can be
simplified using small disturbance theory. Consider the flow past a thin 3-D airfoil shown
in Figure 14.1.
Figure 14.1: Flow past a thin 3-D airfoil
The velocity field consists of a free-stream flow plus a small disturbance
Ū = U∞ + u
V =v (14.11)
W =w
where
u
<< 1
U∞
v
<< 1 (14.12)
U∞
w
<< 1.
U∞
Similarly the state variables deviate only slightly from free-stream values
P = P∞ + P 0
T = T∞ + T 0 (14.13)
ρ = ρ∞ + ρ0
and
a = a∞ + a0 . (14.14)
This decomposition of variables is substituted into equation (14.8). Various terms are
Ū · Ū U∞ 2 u2 v 2 w2
= + uU∞ + + +
2 2 2 2 2 (14.15)
∇ · Ū = ux + vy + wz
and

Ū · Ū
∇ = (ux U∞ + uux + vvx + wwx , uy U∞ + uuy + vvy + wwy , uz U∞ + uuz + vvz + wwz )
2
(14.16)
as well as

Ū · Ū
(γ − 1) ht − ∇ · Ū =
2
U∞ 2 u2 v 2 w2

(γ − 1) ht − + uU∞ + + + ux +
2 2 2 2
(14.17)
U∞ 2 u2 v 2 w2

(γ − 1) ht − + uU∞ + + + uy +
2 2 2 2
U∞ 2 u2 v 2 w2

(γ − 1) ht − + uU∞ + + + uz
2 2 2 2
and, finally

Ū · Ū Ū · Ū ∼
(γ − 1) ht − ∇ · Ū − Ū · ∇ =
2 2
U∞ 2

(γ − 1) ht − + uU∞ ux +
2
U∞ 2

(γ − 1) ht − + uU∞ vy +
2 (14.18)
U∞ 2

(γ − 1) ht − + uU∞ wz −
2
ux U∞ 2 + uux U∞ + vvx U∞ + wwx U∞ −

uux U∞ − vuy U∞ − wuz U∞ .
Neglect terms in (14.17) and (14.18) that are of third order in the disturbance velocities.
Now

Ū · Ū Ū · Ū ∼
(γ − 1) ht − ∇ · Ū − Ū · ∇ =
2 2
U∞ 2

(γ − 1) ht − + uU∞ ux +
2
U∞ 2

(γ − 1) ht − + uU∞ vy +
2 (14.19)
U∞ 2

(γ − 1) ht − + uU∞ wz −
2

uux U∞ − vuy U∞ − wuz U∞
or

Ū · Ū Ū · Ū ∼
(γ − 1) ht − ∇ · Ū − Ū · ∇ =
2 2
(γ − 1) (h∞ − uU∞ ) (ux + vy + wz ) −

(14.20)

uux U∞ − vuy U∞ − wuz U∞ .
Recall that (γ − 1) h∞ = a∞ 2 . Equation (14.20) can be rearranged to read

Ū · Ū Ū · Ū ∼
(γ − 1) ht − ∇ · Ū − Ū · ∇ =
2 2
(14.21)
a∞ 2 (ux + vy + wz ) − ux U∞ 2 − (γ + 1) uux U∞ − (vvx U∞ + wwx U∞ ) −
(γ − 1) (uvy U∞ + uwz U∞ ) − vuy U∞ − wuz U∞ .
Divide through by a∞ 2 .

γ−1 Ū · Ū 1 Ū · Ū ∼
ht − ∇ · Ū − Ū · ∇ =
a∞ 2 2 a∞ 2 2
(γ + 1) M∞
1 − M∞ 2 ux + vy + wz − (14.22)

uux −
a∞
M∞
((γ − 1) (uvy + uwz ) + vuy + wuz + vvx + wwx )
a∞
Equation (14.22) contains both linear and quadratic terms in the velocity disturbances
and one might expect to be able to neglect the quadratic terms. But note that the first
term becomes very small near M∞ = 1. Thus in order to maintain the small disturbance
approximation at transonic Mach numbers the uux term must be retained. The remaining
quadratic terms are small at all Mach numbers and can be dropped. Finally the small
disturbance equation is
(γ + 1) M∞
1 − M∞ 2 ux + vy + wz −

uux = 0. (14.23)
a∞
The velocity potential is written in terms of a free-stream potential and a disturbance

potential.
Φ = U∞ x + φ (x, y, z) (14.24)
The small disturbance equation in terms of the disturbance potential is
(γ + 1) M∞
1 − M∞ 2 φxx + φyy + φzz =

φx φxx . (14.25)
a∞
Equation (14.25) is valid over the whole range of subsonic, transonic and supersonic Mach
numbers.
14.1.3 Linearized potential flow
If we restrict our attention to subsonic and supersonic flow, staying away from Mach
numbers close to one, the nonlinear term on the right side of (14.25) can be dropped and
the small disturbance potential equation reduces to the linear wave equation
β 2 φxx − (φyy + φzz ) = 0 (14.26)

q
where β = M∞ 2 − 1. In two dimensions
β 2 φxx − φyy = 0. (14.27)
The general solution of (14.27) can be expressed as a sum of two arbitrary functions
φ (x, y) = F (x − βy) + G (x + βy) . (14.28)
Note that if M∞ 2 < 1 the 2-D linearized potential equation (14.27) is an elliptic equation
that can be rescaled to form Laplace’s equation and (14.28) expresses the solution in terms
of conjugate complex variables. In this case the subsonic flow can be analyzed using the
methods of complex analysis. Presently we will restrict our attention to the supersonic
case. The subsonic case is treated later in the chapter.
If M∞ 2 > 1 then (14.27) is the 2-D wave equation and has solutions of hyperbolic type.
Supersonic flow is analyzed using the fact that the properties of the flow are constant along
the characteristic lines x ± βy = constant. Figure 14.2 illustrates supersonic flow past a
thin airfoil with several characteristics shown. Notice that in the linear approximation the
characteristics are all parallel to one another and lie at the Mach angle µ∞ of the free
stream. Information about the flow is carried in the value of the potential assigned to
a given characteristic and in the spacing between characteristics for a given flow change.
Right-leaning characteristics carry the information about the flow on the upper surface
of the wing and left-leaning characteristics carry information about the flow on the lower
surface.
Figure 14.2: Right and left leaning characteristics on a thin 2-D airfoil.
All properties of the flow, velocity, pressure, temperature, etc. are constant along the char-
acteristics. Since disturbances only propagate along downstream running characteristics
we can write the velocity potential for the upper and lower surfaces as
φ (x, y) = F (x − βy) y>0

(14.29)
φ (x, y) = G (x + βy) y < 0.
Let y = f (x) define the coordinates of the upper surface of the wing and y = g (x) define
the lower surface. The full nonlinear boundary condition on the upper surface is
v df
= . (14.30)
U y=f dx

In the spirit of the thin airfoil approximation this boundary condition can be approximated
by the linearized form
v df
= (14.31)
U y=0 dx

which we can write as


∂φ (x, y) df
= U∞ (14.32)
∂y
y=0 dx
or
U∞ df
F 0 (x) = − . (14.33)
β dx
On the lower surface the boundary condition is
U∞ dg
G0 (x) = . (14.34)
β dx
In the thin airfoil approximation the airfoil itself is, in effect, collapsed to a line along
the x -axis, the velocity potential is extended to the line y = 0 and the surface boundary
condition is applied at y = 0 instead of at the physical airfoil surface. The entire effect of
the airfoil on the flow is accounted for by the vertical velocity perturbation generated by
the local slope of the wing. The linearized boundary condition is valid on 2-D thin wings
and on 3-D wings of that are of ”thin planar form”.
Recall that (14.26) is only valid for subsonic and supersonic flow and not for transonic flow
where 1 − M∞ 2 << 1.
14.1.4 The pressure coefficient
Let’s work out the linearized pressure coefficient. The pressure coefficient is

P − P∞ 2 P
CP = 1 2 = 2 −1 . (14.35)
ρ U
2 ∞ ∞ γM ∞ P∞
The stagnation enthalpy is constant throughout the flow. Thus
T 1
U∞ 2 − U 2 + v 2 + w2 .

=1+ (14.36)
T∞ 2Cp T∞
Similarly the entropy is constant and so the pressure and temperature are related by
γ
P 1 γ−1
U∞ 2 − U 2 + v 2 + w2

= 1+ . (14.37)
P∞ 2Cp T∞
The pressure coefficient is
γ !
2 1 γ−1
U∞ 2 − U 2 + v 2 + w 2

CP = 1+ −1 . (14.38)
γM∞ 2 2Cp T∞
The velocity term in (14.37) is small.
U∞ 2 − U 2 + v 2 + w2 = − 2uU∞ + u2 + v 2 + w2

(14.39)
Now use the binomial expansion (1 − ε)n = ∼ 1 − nε + n (n − 1) ε2 /2 to expand the term in

parentheses in (14.39). Note that the expansion has to be carried out to second order. The
pressure coefficient is approximately
u2 v 2 + w2

2u
CP ∼
=− + 1 − M∞ 2 2 + . (14.40)
U∞ U∞ U∞ 2
Equation (14.40) is a valid approximation for small perturbations in subsonic or supersonic

flow.
For 2-D flows over planar bodies it is sufficient to retain only the first term in (14.40) and
we use the expression
u
CP ∼
= −2 . (14.41)
U∞
For 3-D flows over slender approximately axisymmetric, bodies we must retain the last
term so
v 2 + w2

2u
CP ∼
=− + . . (14.42)
U∞ U∞ 2
As was discussed above, if the airfoil is a 2-D shape defined by the function y = f (x) the
boundary condition at the surface is
df v
= = tan (θ) (14.43)
dx U∞ + u
where θ is the angle between the airfoil surface and the horizontal. For a thin airfoil this
is accurately approximated by
df ∼ φy ∼
= = θ. (14.44)
dx U∞
For a thin airfoil in supersonic flow the wall pressure coefficient is

2 df
CPwall = 1/2 (14.45)
M∞ 2 − 1 dx
q
2 2 2
where dU /U = − 2/ M∞ − 1 dθ, which is valid for M∞ ≥ 1 has been used. In the
thin airfoil approximation in supersonic flow the local pressure coefficient is determined by
the local slope of the wing. Figure 14.3 schematically shows the wall pressure coefficient
on a thin, symmetric biconvex wing.
Figure 14.3: Pressure distribution on a thin symmetric airfoil.
This airfoil will have shock waves at the leading and trailing edges and at first sight this
would seem to violate the isentropic assumption. But for small disturbances the shocks
are weak and the entropy changes are negligible.
14.1.5 Drag coefficient of a thin symmetric airfoil
A thin, 2-D, symmetric airfoil is situated in a supersonic stream at Mach number M∞ and
zero angle of attack. The y-coordinate of the upper surface of the airfoil is given by the
function
πx
y (x) = ASin (14.46)
C
where C is the airfoil chord. The airfoil thickness to chord ratio is small, 2A/C << 1.
Determine the drag coefficient of the airfoil.
Solution
The drag integral is
Z C
D=2 (P − P∞ )Sin (α) dx. (14.47)
0
where the factor of 2 accounts for the drag of both the upper and lower surfaces and α is
the local angle formed by the upper surface tangent to the airfoil and the x-axis. Since the
airfoil is thin the angle is small and we can write the drag coefficient as
!
1
P − P∞
Z
D x
CD = 1 2 =2 1 2 (α) d . (14.48)
2 ∞ U∞ C
ρ 0 2 ρ∞ U∞
C
The local tangent is determined by the local slope of the airfoil therefore T an (α) = dy/dx
and for small angles α = dy/dx. Now the drag coefficient is
!
1
P − P∞
Z
D dy x
CD = 1 2 =2 1 2 d . (14.49)
2 ρ∞ U∞ C 0 2 ρ∞ U∞
dx C
The pressure coefficient on the airfoil is given by thin airfoil theory (14.45) as

P − P∞ 2 dy
CP = 1 2 =
p (14.50)
2 ρ∞ U∞ M∞ 2 − 1 dx
and so the drag coefficient becomes
1 2 Z 1
4A2 π 2A2 π 2
Z
4 dy x πx πx
CD = p d = p Cos2 d = p .
M∞ 2 − 1 0 dx C C 2 M∞ 2 − 1 0 C C C 2 M∞ 2 − 1
(14.51)
The drag coefficient is proportional to the square of the wing thickness-to-chord ratio.
14.1.6 Thin airfoil with lift and camber at a small angle of attack
A thin, cambered, 2-D, airfoil is situated in a supersonic stream at Mach number M∞ and
a small angle of attack as shown in Figure 14.4.
The y-coordinate of the upper surface of the airfoil is given by the function
Figure 14.4: Trajectory of a piston and an expansion wave in the x-t plane.
x x x δx

f = Aτ + Bσ − (14.52)
C C C C
and the y-coordinate of the lower surface is
x x x δx

g = −Aτ + Bσ − (14.53)
C C C C
where 2A/C << 1 and δ/C << 1. The airfoil surface is defined by a dimensionless
thickness function
x
τ
C
(14.54)
τ (0) = τ (1) = 0
a dimensionless camber function

x
σ
C
(14.55)
σ (0) = σ (1) = 0
and the angle of attack
δ
T an (α) = − . (14.56)
C
Determine the lift and drag coefficients of the airfoil. Let
x
ξ= . (14.57)
C
Solution
The lift integral is
Z C Z C
L= (Plower − P∞ )Cos (αlower ) dx − (Pupper − P∞ )Cos (αupper ) dx. (14.58)
0 0
where α is the local angle formed by the tangent to the surface of the airfoil and the x-axis.
Since the angle is everywhere small we can use Cos (α) = 1. The lift coefficient is
Z 1 Z 1
D
CL = 1 2 = CPlower dξ − CPupper dξ. (14.59)
2 ρ∞ U∞ C 0 0
The pressure coefficient on the upper surface is given by thin airfoil theory as

2 df dτ 2
dσ
CPupper = p = p A +B −δ . (14.60)
C M∞ 2 − 1 dξ C M∞ 2 − 1 dξ dξ
On the lower surface the pressure coefficient is

2 dg 2 dτ dσ
CPlower = − p =− p −A +B −δ . (14.61)
C M∞ 2 − 1 dξ C M∞ 2 − 1 dξ dξ
Substitute into the lift integral
Z 1 1 Z −δ −δ
−2 −2
Z Z
df dg
CL = p dξ + dξ = p df + dg .
C M∞ 2 − 1 0 dξ 0 dξ C M∞ 2 − 1 0 0
(14.62)
In the thin airfoil approximation, the lift is independent of the airfoil thickness.

4 δ
CL = p (14.63)
M∞ 2 − 1 C
The drag integral is
Z C Z C
D= (Pupper − P∞ )Sin (αupper ) dx + (Plower − P∞ )Sin (−αlower ) dx. (14.64)
0 0
Since the airfoil is thin the angle is small, and we can write the drag coefficient as
! !
1 1
Pupper − P∞ Plower − P∞
Z Z
D ∼
CD = 1 2 = 1 2 (αupper ) dξ + 1 2 (−αlower ) dξ.
2 ρ∞ U∞ C 0 2 ρ∞ U∞ 0 2 ρ∞ U∞
(14.65)
The local tangent is determined by the local slope of the airfoil therefore T an (α) = dy/dx
and for small angles α = dy/dx. Now the drag coefficient is
Z 1 Z 1
D ∼ 1 dyupper 1 dylower
CD = 1 2 = CPupper dξ + CPlower − dξ. (14.66)
2 ρ ∞ U∞ C
C 0 dξ C 0 dξ
The drag coefficient becomes
2 2 !
Z 1 Z 1
2 dτ dσ dτ dσ
CD ∼
= p A +B −δ dξ + −A +B −δ dξ .
C2 M∞ 2 − 1 0 dξ dξ 0 dξ dξ
(14.67)
Note that most of the cross terms cancel. The drag coefficient breaks into several terms.
4
CD ∼
=p ×
M∞ 2 − 1
2 Z 1 2 2 Z 1 2 Z 1 2 Z 1 !
A dτ B dσ B δ dσ δ
dξ + dξ − dξ + dξ
C 0 dξ C 0 dξ C C 0 dξ C 0
(14.68)
The third term in (14.68) is
Z 1 Z 1
B δ dσ B δ B δ
dξ = dσ = (σ (1) − σ (0)) = 0. (14.69)
C C 0 dξ C C 0 C C
Finally
2 Z 2 2 Z 2 2 Z !
1 1 1
4 A dτ B dσ δ
CD ∼
=p dξ + dξ + dξ .
M∞ 2 − 1 C 0 dξ C 0 dξ C 0
(14.70)
In the thin airfoil approximation, the drag is a sum of the drag due to thickness, drag due
to camber, and drag due to lift.
14.2 Similarity rules for high speed flight
The Figure below shows the flow past a thin symmetric airfoil at zero angle of attack. The
fluid is assumed to be inviscid and the flow Mach number, U∞1 /a∞ where a∞ 2 = γP∞ /ρ∞
is the speed of sound, is assumed to be much less than one. The airfoil chord is c and the
maximum thickness is t1 . The subscript one is applied in anticipation of the fact that we
will shortly scale the airfoil to a new shape with subscript 2 and the same chord.
Figure 14.5: Pressure variation over a thin symmetric airfoil in low speed flow.
The surface pressure distribution is shown below the wing, expressed in terms of the pres-
sure coefficient.
Ps − P∞
CP1 = 1 2 (14.71)
2 ρ∞ U∞1
The pressure and flow speed throughout the flow satisfy the Bernoulli relation. Near the
airfoil surface
1 1
P∞ + ρ∞ U∞1 2 = Ps1 + ρ∞ Us1 2 . (14.72)
2 2
The pressure is high at the leading edge where the flow stagnates, then as the flow ac-
celerates about the body, the pressure falls rapidly at first, then more slowly reaching a
minimum at the point of maximum thickness. From there the surface velocity decreases
and the pressure increases continuously to the trailing edge. In the absence of viscosity,
the flow is irrotational.
∇ × u1 = 0 (14.73)
This permits the velocity to be described by a potential function.
u1 = ∇φ1 (14.74)
When this is combined with the condition of incompressibility, ∇ · ū1 = 0, the result is
Laplace’s equation.
∂ 2 φ1 ∂ 2 φ1
+ =0 (14.75)
∂x1 2 ∂y1 2
Let the shape of the airfoil surface in (x1 , y1 ) be given by
y1 x
1
= τ1 g . (14.76)
c c
where τ1 = t1 /c is the thickness to chord ratio of the airfoil. The boundary conditions that
the velocity potential must satisfy are

∂φ1 dy1 dg (x1 /c)
= U∞1 = U∞ τ1
∂y1 y1 =cτ1 g (
x1
) dx1 body d (x1 /c)
c
(14.77)

∂φ1
→ U∞1 .
∂x1 ∞
Any number of methods of solving for the velocity potential are available including the
use of complex variables. In the following we are going to restrict the airfoil to be thin,
τ1 << 1. In this context we will take the velocity potential to be a perturbation potential
so that
u1 = U∞1 + u1 0
(14.78)
v1 = v1 0
or
∂φ1
u1 = U∞1 +
∂x1
(14.79)
∂φ1
v1 = .
∂y1
The boundary conditions on the perturbation potential in the thin airfoil approximation
are

∂φ1 dy1 dg (x1 /c)
= U∞1 = U∞1 τ1
∂y1 y1 =0 dx1 body d (x1 /c)
(14.80)
∂φ1
→ 0.
∂x1 ∞
The surface pressure coefficient in the thin airfoil approximation is

2 ∂φ1
CP1 = − . (14.81)
U∞1 ∂x1 y1 =0
Note that the boundary condition on the vertical velocity is now applied on the line y1 = 0.
In effect the airfoil has been replaced with a line of volume sources whose strength is
proportional to the local slope of the actual airfoil. This sort of approximation is really
unnecessary in the low Mach number limit but it is essential when the Mach number is
increased and compressibility effects come in to play. Equally, it is essential in this example
where we will map a compressible flow to the incompressible case.
14.2.1 Subsonic flow M∞ < 1
Now imagine a second flow at a free-stream velocity, U∞2 in a new space (x2 , y2 ) over a new
airfoil of the same shape (defined by the function g (x/c) but with a new thickness ratio
τ2 = t2 /c << 1. Part of what we need to do is to determine how τ1 and τ1 are related to
one another. The boundary conditions that the new perturbation velocity potential must
satisfy are

∂φ2 dy2 dg (x2 /c)
= U∞2 = U∞2 τ2
∂y2 y2 =0 dx2 body d (x2 /c)
(14.82)
∂φ2
→ 0.
∂x2 ∞
In this second flow the Mach number has been increased to the point where compressibility
effects begin to occur: the density begins to vary significantly and the pressure distribution
begins to deviate from the incompressible case. As long as the Mach number is not too
large and shock waves do not form, the flow will be nearly isentropic. In this instance the
2-D steady compressible flow equations are
∂u2 ∂u2 1 ∂P
u2 + v2 + =0
∂x2 ∂y2 ρ ∂x2
∂v2 ∂v2 1 ∂P
u2 + v2 + =0
∂x2 ∂y2 ρ ∂y2
(14.83)
∂ρ ∂ρ ∂u2 ∂v2
u2 + v2 +ρ +ρ =0
∂x2 ∂y2 ∂x2 ∂y2
γ
P ρ
= .
P∞ ρ∞
Let
u2 = U∞2 + u2 0
v2 = v2
(14.84)
ρ = ρ∞ + ρ0
P = P∞ + P 0
where the primed quantities are assumed to be small compared to the free stream condi-
tions. When quadratic terms in the equations of motion are neglected the equations (14.83)
reduce to
ρ∞ U∞2 2 ∂u2 0 ∂v2 0

1− + = 0. (14.85)
γP∞ ∂x2 ∂y2
Introduce the perturbation velocity potential.
∂φ2
u2 0 =
∂x2
(14.86)
∂φ2
v2 =
∂y2
The equation governing the disturbance flow becomes
∂ 2 φ2 ∂ 2 φ2
1 − M∞ 2 2 + = 0. (14.87)
∂x2 2 ∂y2 2
Notice that (14.87) is valid for both sub and supersonic flow in the thin airfoil approxi-
mation. Since the flow is isentropic, the pressure and velocity disturbances are related to
lowest order by
P 0 + ρ∞ U∞2 u2 0 = 0 (14.88)
and the surface pressure coefficient retains the same basic form as in the incompressible
case.

2 ∂φ2
CP2 = − (14.89)
U∞2 ∂x2 y2 =0
Since we are at a finite Mach number, this last relation is valid only within the thin airfoil,
small disturbance approximation and therefore may be expected to be invalid near the
leading edge of the airfoil where the velocity change is of the order of the free stream
velocity. For example for the ”thin” airfoil depicted in Figure 14.5 which is actually not
all that thin, the pressure coefficient is within −0.2 < Cp < 0.2 except over a very narrow
portion of the chord near the leading edge.
Equation (14.87) can be transformed to Laplace’s equation, (14.75) using the following
change of variables
x2 = x1
1
y2 = q y1
1 − M∞2 2 (14.90)

1 U∞2
φ2 = φ1
A U∞1
where, at the moment, A is an arbitrary constant. The velocity potentials are related
by

1 U∞2
φ2 (x2 , y2 ) = φ1 (x1 , y1 ) (14.91)
A U∞1
or
 

U∞1 1
φ1 (x1 , y1 ) = A φ2  x1 , q y1  (14.92)
U∞2 1 − M∞ 2 2
and the boundary conditions transform as

 

∂φ1 Aτ2  dg (x1 /c)
= U∞1  q
∂y1 y1 =0 1 − M∞2 2 d (x1 /c)

∂φ1 (14.93)
→0
∂x1 ∞

∂φ2
→ 0.
∂x2 ∞
The transformation between flows one and two is completed by the correspondence
A
τ1 = q τ2 (14.94)
2
1 − M∞2
or

t1 A t2
=q . (14.95)
c 1 − M∞ 2 2 c
Finally the transformed pressure coefficient is
CP1 = ACP2 . (14.96)
These results may be stated as follows. The solution for incompressible flow over a thin
airfoil with shape g (x1 /c) and thickness ratio, t1 /c at velocity U1 is identical to the subsonic
compressible flow at velocity, U2 and Mach number M2 over an airfoil with a similar shape
but with the thickness ratio
q
t2 1 − M∞2 2 t1
= . (14.97)
c A c
The pressure coefficient for the compressible case is derived by adjusting the incompressible
coefficient using CP2 = CP1 /A. This result comprises several different similarity rules that
can be found in the aeronautical literature depending on the choice of the free constant A.
Perhaps the one of greatest interest is the Prandtl-Glauert rule that describes the variation
of pressure coefficient with Mach number for a body of a given shape and thickness ratio.
In this case we select
q
A = 1 − M∞2 2 (14.98)
so that the two bodies being compared in (14.97) have the same shape and thickness ratio.
The pressure coefficient for the compressible flow is
CP1
CP2 = q . (14.99)
2
1 − M∞ 2
Several scaled profiles are shown in Figure 14.6. Keep in mind the lack of validity of (14.99)
near the leading edge where the pressure coefficient is scaled to inaccurate values.
Figure 14.6: Pressure coefficient over the airfoil in Figure 14.5 at several Mach numbers
as estimated using the Prandtl-Glauert rule (14.99).
14.2.2 Supersonic similarity M∞ > 1
All the theory developed in the previous section can be extended to the supersonic case
by simply replacing 1 − M∞ 2 with M∞ 2 − 1. In this instance the mapping is between the
equation
∂ 2 φ2 ∂ 2 φ2
M∞2 2 − 1 − =0 (14.100)
∂x2 2 ∂y2 2
and the simple wave equation
∂ 2 φ1 ∂ 2 φ1
− = 0. (14.101)
∂x1 2 ∂y1 2
A generalized form of the pressure coefficient valid for subsonic and supersonic flow is
 
CP τ
= F  q  (14.102)
A A 1 − M∞ 2
where A is taken to be a function of 1 − M∞ 2 .

14.2.3 Transonic similarity M∞ ∼

=1
When the Mach number is close to one, the simple linearization used to obtain (14.87)
from (14.83) loses accuracy. In this case the equations (14.83) reduce to the nonlinear
equation
∂u1 0 ∂v1 0 (γ1 + 1) M∞1 2 0 ∂u1 0

1 − M∞ 1 2 + − u1 = 0. (14.103)
∂x1 ∂y1 U∞1 ∂x1
In terms of the perturbation potential
∂ 2 φ1 ∂ 2 φ1 (γ1 + 1) M∞1 2 ∂φ1 ∂ 2 φ1

1 − M∞1 2 + − = 0. (14.104)
∂x1 2 ∂y1 2 U∞1 ∂x1 ∂x1 2
This equation is invariant under the change of variables
x2 = x1
q
1 − M∞1 2
y2 = q y1
(14.105)
1 − M∞2 2

1 U∞2
φ2 = φ1
A U∞1
where
1 − M∞ 1 2 M∞ 2 2

1 + γ2
A= . (14.106)
1 + γ1 1 − M∞ 2 2 M∞ 1 2
Notice that, due to the nonlinearity of the transonic equation (14.104) the constant A is
no longer arbitrary. The pressure coefficient becomes
1 − M∞1 2 M∞2 2

1 + γ2
CP1 = CP2 (14.107)
1 + γ1 1 − M∞2 2 M∞1 2
and the thickness ratios are related by
23
1 − M∞ 2 2 M∞1 2

t2 1 + γ1 t1
= . (14.108)
c 1 + γ2 1 − M∞1 2 M∞2 2 c
In the transonic case, it is not possible to compare the same body at different Mach numbers
or bodies with different thickness ratios at the same Mach number except by selecting gases
with different γ. For a given gas it is only possible to map the pressure distribution for
one airfoil to an airfoil with a different thickness ratio at a different Mach number. A
generalized form of (14.102) valid from subsonic to sonic to supersonic Mach numbers
is
1  
CP (γ + 1) M∞ 2 3
1 − M∞ 2
2 =F 2
. (14.109)
τ 2
3
τ (γ + 1) M∞ 3
Prior to the advent of supercomputers capable of solving the equations of high speed flow,
similarity methods and wind tunnel correlations were the only tools available to the aircraft
designer and these methods played a key role in the early development of transonic and
supersonic flight.
14.3 Problems
Problem 1 - A thin, 2-D, airfoil is situated in a supersonic stream at Mach number M∞

and a small angle of attack as shown in Figure 14.7.
The y-coordinate of the upper surface of the airfoil is given by the function
x x x
f (x) = A 1− −δ (14.110)
C C C
CHAPTER 14. THIN AIRFOIL THEORY 25
Figure 14.7: Bi-convex airfoil at angle of attack
and the y-coordinate of the lower surface is
x x x
g (x) = −A 1− −δ (14.111)
C C C
where 2A/C << 1 and δ/C << 1. Determine the lift and drag coefficients of the air-
foil.
Appendix A
Some results from the kinetic

theory of gases
A.1 Distribution of molecular velocities in a gas
A.1.1 The distribution derived from the barometric formula
In Chapter 2 the variation of gas density in the atmosphere was derived. In the upper
atmosphere where the gas temperature is approximately constant
ρ (z + h) − γgh
= e a0 2 . (A.1)
ρ (z)
Rewrite (A.1) in terms of molecular parameters
n (z + h) mgh
= e− kT (A.2)
n (z)
where n (z) is the number of molecules per unit volume at the height z, m = Mw /N , is
the mass of each molecule (or the average mass in the case of a mixture like air). The
Boltzmann’s constant
k = 1.38 × 10−23 Joules/K (A.3)
is related to the gas constant and Avagadros number N by
A-1
APPENDIX A. SOME RESULTS FROM THE KINETIC THEORY OF GASES A-2
Ru
k= . (A.4)
N
Equation (A.2) expresses the distribution of molecules in any system at constant tempera-
ture subject to a body force. The fraction of molecules per unit area above a given altitude
z + h is also given by the distribution (A.2). That is
R∞
n (z) dz mgh
Rz+h
∞ = e− kT (A.5)
z n (z) dz
which is a basic property of an exponential distribution. The distribution of molecules

under the downward pull of gravity given by (A.5) can be used to infer how the velocities
of gas molecules are distributed. Consider molecules that pass through some reference
height z on their way to height z + h.
Figure A.1: Trajectory of an upward moving molecule
The upward velocity component uh of a molecule whose kinetic energy at z is equal to its
potential energy when it arrives at z + h satisfies
1
mgh = muh 2 . (A.6)
2
This is the minimum positive velocity needed for a molecule starting at z to reach the
height z + h.
Let Ju>uh (z) be the flux of molecules (number of molecules/area-sec) passing through the
plane z that possess a vertical velocity component greater than uh and let Ju>0 (z) be the
flux of molecules passing through the plane z with any positive upward velocity. The frac-
tion of molecules passing upward through z with the requisite velocity is Ju>uh (z) /Ju>0 (z).
This ratio should be the same as the fraction of molecules above the height z +h. Therefore
one can infer the equality
R∞
Ju>uh (z) n (z) dz
= Rz+h
∞ . (A.7)
Ju>0 (z) z n (z) dz
Using (A.5) we deduce that
Ju>uh (z) mgh muh 2

= e− kT = e− 2kT (A.8)
Ju>0 (z)
which is plotted in Figure A.2.
Figure A.2: Fraction of molecules at any reference height with positive vertical velocity
component exceeding uh .
Now use (A.7) to generate a probability density function (pdf) g (uh ) for the velocities.
The pdf is defined such that, at any height z , g (uh ) duh equals the fraction of molecules
per unit volume with velocity between uh and uh + duh . The pdf we are seeking would
look something like Figure A.3.
Figure A.3: Probability density function of one velocity component. The shaded area divided
by the whole area (which is one) is the fraction of molecules with velocity between uh and
uh + duh .
The desired pdf is normalized so that

Z ∞
g (uh )duh = 1. (A.9)
−∞
Z ∞ √
2
Note that e−x dx = π.
−∞
The number of molecules that pass through a unit area of the plane z in unit time (the flux
of molecules) with vertical velocity component uh is uh g (uh ) duh . In terms of the velocity
pdf
Z ∞ muh 2
Ju>uh (z)
= uh g (uh )duh = e− 2kT . (A.10)
Ju>0 (z) uh
Differentiate both sides of (A.10).
muh − muh 2
−uh g (uh ) = − e 2kT (A.11)
kT
Thus far
m − muh 2
g (uh ) = − e 2kT . (A.12)
kT
Apply the normalization condition (A.9) to (A.12).
∞ 1/2
m − muh 2
Z
2πm
e 2kT duh = (A.13)
−∞ kT kT
Finally, the normalized 1-dimensional velocity pdf is
m 1/2 muh 2
g (uh ) = e− 2kT . (A.14)
2πkT
We arrived at this result using a model of particles with random vertical velocity com-
ponents moving in a gravitational field. In a gas in thermodynamic equilibrium at a
temperature T , the molecules constantly undergo collisions leading to chaotic motion with
a wide range of molecular velocities in all three directions. The randomness is so strong
that at any point no one direction is actually preferred over another and one can expect
that the probability density function for a velocity component in any direction at a given
height z will have the same form as (A.14).
A.1.2 The Maxwell velocity distribution function
We can use a slightly different, more general, argument to arrive at the velocity distribution
function in three dimensions. The randomness produced by multiple collisions leads to a
molecular motion that is completely independent in all three coordinate directions and so
the probability of occurrence of a particular triad of velocities (ux , uy , uz ) is equal to the
product of three one-dimensional probability distributions.
f (ux , uy , uz ) = g (ux ) g (uy ) g (uz ) (A.15)
In addition, since no direction in velocity space (ux , uy , uz ) is preferred over any other, then
the three dimensional velocity probability distribution function must be invariant under
any rotation of axes in velocity space. This means f (ux , uy , uz ) can only depend on the
radius in velocity space.
f (ux , uy , uz ) = f ux 2 + uy 2 + uz 2

(A.16)
The only smooth function that satisfies both (A.15) and (A.16) and the requirement that
the probability goes to zero at infinity is the decaying exponential.
2 +u 2 +u 2
f (ux , uy , uz ) = Ae−B (ux y z ) (A.17)
In a homogeneous gas the number density of molecules with a given velocity is accurately
described by the Maxwellian velocity distribution function
m 3/2 m 2 2 2
f (u1 , u2 , u3 ) = e− 2kT (u1 +u2 +u3 ) (A.18)
2πkT
where the correspondence (ux , uy , uz ) → (u1 , u2 , u3 ) is used. The result (A.18) is consistent
with the use of (A.14) in all three directions. The dimensions of f are sec3 /m3 . The pdf
(A.18) is shown in figure A.4.
As with the 1-D pdf, the 3-D pdf is normalized so that the total probability over all velocity
components is one.
Z ∞ Z ∞ Z ∞
f du1 du2 du3 = 1 (A.19)
−∞ −∞ −∞
To understand the distribution (A.18) imagine a volume of gas molecules at some temper-
ature T .
Figure A.4: Maxwellian probability density function of molecular velocities.
Figure A.5: Colliding molecules in a box.
At a certain instant a snapshot of the motion is made and the velocity vector of every
molecule in the gas sample is measured as indicated in the sketch above. The velocity
components of each molecule define a point in the space of velocity coordinates. When
data for all the molecules is plotted, the result is a scatter plot with the densest distribution
of points occurring near the origin in velocity space as shown in figure A.6.
Figure A.6: Schematic of coordinates of a set of colliding molecules in velocity space.
The probability density function (A.18) can be thought of as the density of points in figure
A.4. The highest density occurs near the origin and falls off exponentially at large distances
from the origin. If the gas is at rest or moving with a uniform velocity the distribution is
spherically symmetric since no one velocity direction is preferred over another.

The one-dimensional probability distribution is generated by integrating the Maxwellian
in two of the three velocities.
Z ∞ Z ∞ m 3/2 m 2 2 2
f1D = e− 2kT (u1 +u2 +u3 ) du2 du3 (A.20)
−∞ −∞ 2πkT
m 3/2 mu1 2 Z ∞ mu2 2

Z ∞
mu3 2
− 2kT − 2kT
f1D = e e du2 e− 2kT du3 . (A.21)
2πkT −∞ −∞
Carrying out the integration in (A.21) the 1-D pdf is
m 1/2 mu1 2
f1D = e− 2kT (A.22)
2πkT
which is identical to (A.14).
A.2 Mean molecular velocity
Any statistical property of the gas can be determined by taking the appropriate moment
of the Maxwellian distribution. The mean velocity of molecules moving in the plus x1
direction is
Z ∞ Z ∞ Z ∞ m 3/2 m 2 2 2
û1 = u1 e− 2kT (u1 +u2 +u3 ) du1 du2 du3 . (A.23)
−∞ −∞ 0 2πkT
Note that the integration extends only over velocities in the plus x direction. When the
integral is evaluated the result is
1/2
kT
û1 = . (A.24)
2πm
Let’s work out the flux in the 1-direction of molecules crossing an imaginary surface in the
fluid shown schematically in Figure A.7. This is the number of molecules passing through
a surface of unit area in unit time.
The flux is simply
Figure A.7: Molecules moving through an imaginary surface normal to the x1 -direction.
1/2
kT
J1 = nû1 = n . (A.25)
2πm
At equilibrium, the number of molecules crossing per second in either direction is the same.
Note that the molecular flux through a surface is the same regardless of the orientation of
the surface.
The mean square molecular speed is defined as the average of the squared speed over all
molecules.
(u
d 2 ) = (u 2 +\
u2 2 + u3 2 ) (A.26)
1
From the Maxwellian
Z ∞ Z ∞ Z ∞ m 3/2 − m (u1 2 +u2 2 +u3 2 ) 3kT

c2 =
u u1 2 + u2 2 + u3 2 e 2kT du1 du2 du3 =
−∞ −∞ −∞ 2πkT m
(A.27)
where the integration is from minus infinity to plus infinity in all three directions. The
root-mean-square speed is
q r
urms = uc2 = 3kT . (A.28)
m
This result is consistent with the famous law of equipartition which states that, for a gas of
monatomic molecules, the mean kinetic energy per molecule (the energy of the translational
degrees of freedom) is
1 c2 3
E = mu = kT. (A.29)
2 2
A.3 Distribution of molecular speeds
The Maxwellian distribution of molecular speed is generated from the full distribution by
determining the number of molecules with speed in a differential spherical shell of thickness
du and radius u in velocity space. The volume of the shell is dVu = 4πu2 du and so the
number of molecules in the shell is f dVu = 4πu2 f du where u2 = u1 2 + u2 2 + u3 2 . The
Maxwellian probability distribution for the molecular speed is therefore
m 3/2 mu2
2
fu = 4πu f = 4π u2 e− 2kT (A.30)
2πkT
shown in Figure A.8.
Figure A.8: Maxwell distribution of molecular speeds.
The mean molecular speed is
r
Z ∞ Z ∞ m 3/2
3 − mu
2 8kT
û = ufu du = 4π u e 2kT du = . (A.31)
0 0 2πkT πm
In summary, the three relevant molecular speeds are
r
kT
û1 =
2πm
r
3kT (A.32)
ûrms =
m
r
8kT
û = .
πm
A.4 Pressure
A.4.1 Kinetic model of pressure
Before we use the Maxwellian distribution to relate the gas pressure to temperature it
is instructive to derive this relation using intuitive arguments. Consider a gas molecule
confined to a perfectly rigid box. The molecule is moving randomly and collides and
rebounds from the wall of the container preserving its momentum in the direction normal
to the wall. In doing so the molecule undergoes the change in momentum
∆p = −2mu1 (A.33)
and imparts to the wall A1 the momentum 2mu1 . Suppose the molecule reaches the
opposite wall without colliding with any other molecules. The time it takes for the molecule
to bounce off A2 and return to collide again with A1 is 2L/u1 and so the number of collisions
per second with A1 is u1 /2L.
Figure A.9: A rigid box containing an ideal gas.
The momentum per unit time that the molecule transfers to A1 is
mu1 2
F orce on wall by one molecule = . (A.34)
L
Let Np be the number of molecules in the box. The total force on A1 is the pressure of the
gas times the area and is equal to the sum of forces by the individual molecules
Np
2 1X
PL = mi u1i 2 (A.35)
L
i=1
where the index i refer to the ith molecule. If the molecules all have the same mass
 
Np
Nm  1 X
P = u1i 2  . (A.36)
V Np
i=1
The first factor in (A.36) is the gas density ρ = Np m/V . No one direction is preferred over
another and so we would expect the average of the squares in all three directions to be the
same.
Np Np Np
1 X 2 1 X 2 1 X 2
u1i = u2i = u3i (A.37)
Np Np Np
i=1 i=1 i=1
The pressure is now written
ρurms 2
P = (A.38)
3
where
 
Np Np Np
1  X X X
urms 2 = u1i 2 + u2i 2 + u3i 2  . (A.39)
Np
i=1 i=1 i=1
The root-mean-square velocity defined by a discrete sum in (A.39) is the same as that
derived by integrating the Maxwellian pdf in (A.28). Using (A.28) and (A.38) the pressure
and temperature are related by

ρ 3kT
P = = ρRT. (A.40)
3 m
We arrived at this result using a model that ignored collisions between molecules. The
model is really just a convenience for calculation. It works for two reasons; the time
spent during collisions is negligible compared to the time spent between collisions and,
in a purely elastic exchange of momentum between a very large number of molecules in
statistical equilibrium, there will always be a molecule colliding with A2 with momentum
−mu1 while a molecule leaves A1 with the same momentum.
A.4.2 Pressure directly from the Maxwellian pdf
The momentum flux at any point is
Z ∞ Z ∞ Z ∞ m 3/2 m 2 2 2
Πij = nmui uj e− 2kT (u1 +u2 +u3 ) du1 du2 du3 . (A.41)
−∞ −∞ −∞ 2πkT
For i 6= j the integral vanishes since the integrand is an odd function. For i = j the integral
becomes
Z ∞ mu1 2
Z ∞ mu2 2
Z ∞ mu3 2
mn 2kT 2 − −
Π11 = 3/2 u1 e 2kT du1 e 2kT du2 e− 2kT du3 = nkT (A.42)
π m −∞ −∞ −∞
similarly Π22 = nkT and Π33 = nkT . The fluid density is
ρ = nm (A.43)
and the gas constant is
Ru Mw Ru
k= = = mR (A.44)
N N Mw
where N is Avogadros number.
molecules
N = 6.023 × 1026 . (A.45)
kilogram − mole
The force per unit area produced by molecular collisions is

ρ
Π11 = Π22 = Π33 = nkT = mRT = ρRT = P. (A.46)
m
The normal stress derived by integrating the Maxwellian distribution over the molecular
flux of momentum is just the thermodynamic pressure.
 
P 0 0
Πij = P δij =  0 P 0  (A.47)
0 0 P
p
Recall the mean molecular speed û = 8kT /πm. The speed of sound for a gas is a2 =
γP/ρ. Accordingly, the speed of sound is directly proportional to the mean molecular
speed.
1/2 1/2
γP γnkT γπ 1/2
a= = = û (A.48)
ρ nm 8
A.5 The mean free path
A molecule of effective diameter σ sweeps out a collision path which is of diameter 2σ. The
collision volume swept out per second is πσ 2 û. With the number of molecules per unit
volume equalp to n, the number of collisions per second experienced by a given molecule
2
is νc = nπσ 8kT /πm. The average distance that a molecule moves between collisions
is
ûrel ûrel
λ= = (A.49)
νc nπσ 2 û
where the mean relative velocity between molecules is used. This accounts for the fact that
the other molecules in the volume are not static; they are in motion. When this motion is
taken into account, the mean free path is estimated as
1
λ= √ . (A.50)
2nπσ 2
Typical values for the mean free path of a gas at room temperature and one atmosphere
are on the order of 50 nanometers (roughly 10 times the mean molecular spacing).
A.6 Viscosity
Consider the flux of momentum associated with the net particle flux across the plane y = y0
depicted in figure A.10.
On the average, a particle passing through y0 in the positive or negative direction had its
last collision at y = y0 −λ or y = y0 +λ,where λ is the mean free path. Through the collision
process, the particle tends to acquire the mean velocity at that position. The particle flux
across y0 in either direction is J = n(kT /2πm)1/2 . The net flux of x1 - momentum in the
x2 -direction is
Figure A.10: Momentum exchange in a shear flow.
dU
Π12 = mJU (y0 − λ) − mJU (y0 + λ) ∼
= −2mJλ . (A.51)
dy
The viscosity derived from this model is
1/2
kT
µ = 2mJλ = 2ρλ . (A.52)
2πm
In terms of the mean molecular speed this can be written as
8kT 1/2 1

1
µ = ρλ = ρûλ. (A.53)
2 πm 2
The viscosity can be expressed in terms of the speed of sound as follows.
8nkT 1/2
1/2
1 2
µ = ρλ = ρaλ (A.54)
2 πnm πγ
This result can be used to relate the Reynolds number and Mach number of a flow

ρU L ∼ U L L
Re = = =M (A.55)
µ aλ λ
where L is a characteristic length of the flow and the factor (2/πγ)1/2 has been dropped.
The ratio of mean free path to characteristic length is called the Knudsen number of the
flow.
λ
Kn = (A.56)
L
A.7 Heat conductivity
The same heuristic model can be used to crudely estimate the net flux of internal energy
(per molecular mass) across y0 .
dT
Q2 = mJCv T (y0 − λ) − mJCv T (y0 + λ) = −2mJCv λ (A.57)
dy
The heat conductivity derived from this model is
1
κ = ρûλCv . (A.58)
2
The model indicates that the Prandtl number for a gas should be proportional to the ratio
of specific heats.
µCp Cp
Pr = = =γ (A.59)
κ Cv
A more precise theory gives
4γ
Pr = (A.60)
9γ − 5
which puts the Prandtl number for gases in the range 2/3 ≤ Pr < 1 depending on the
number of degrees of freedom of the molecular system.
Notice that we got to these results using an imprecise model and without using the
Maxwellian distribution function. In fact, a rigorous theory of the transport coefficients
in a gas must consider small deviations from the Maxwellian distribution that occur when
gradients of temperature or velocity are present. This is the so-called Chapman-Enskog
theory of transport coefficients in gases.
A.8 Specific heats, the law of equipartition
Classical statistical mechanics leads to a simple expression for Cp and Cv in terms of β,

the number of degrees of freedom of the appropriate molecular model.
β+2
Cp = R
2
β
Cp = R (A.61)
2
β+2
γ=
β
For a mass point, m, with three translational degrees of freedom, β = 3, the energy of the
particle is
1 1 1
E = mu2 + mv 2 + mw2 (A.62)
2 2 2
where (u, v, w) are the velocities in the three coordinate directions. The law of equipartition
of energy says that any term in E that is quadratic (proportional to a square) in either the
position or velocity contributes (1/2) kT to the thermal energy of a large collection of such
mass points. Thus the thermal energy (internal energy) per molecule of a gas composed of
a large collection of mass points is
3
ẽ = kT. (A.63)
2
For one mole of gas
3
N ẽ = Ru T (A.64)
2
where Ru = N k is the universal gas constant. On a per unit mass of gas basis the internal
energy is
3
e = RT (A.65)
2
where
Ru
R= . (A.66)
molecularweight
This is a good model of monatomic gases such as Helium, Argon, etc. Over a very wide
range of temperatures
5
Cp = R
2
(A.67)
3
Cv = R
2
from near condensation to ionization.
A.9 Diatomic gases
A.9.1 Rotational degrees of freedom
The simplest classical model for a diatomic molecule is a rigid dumb bell such as that
shown in figure A.11. In the classical theory for a solid object of this shape there is one
rotational degree of freedom about each coordinate axis shown in the figure.
Figure A.11: Rigid dumb bell model of a diatomic molecule.
The energy of the rotating body is
1 1 1
Esoliddumbell = I1 ω1 2 + I2 ω2 2 + I3 ω3 2 . (A.68)
2 2 2
The moment of inertia about the 2 or the 3 axis is
I2 = I3 = mr D2 (A.69)
where
m1 m2
mr = (A.70)
m1 + m2
is the reduced mass. The moment of inertia of the connecting shaft is neglected in (A.69)
where it is assumed that d1,2 << D so that the masses of the spheres can be assumed
to be concentrated at the center of each sphere. In classical theory, the energy (A.68) is
a continuous function of the three angular frequencies. But in the quantum world of a
diatomic molecule the energies in each rotational degree of freedom are discrete. As in the
classical case the energy varies inversely with the moment of inertia of the molecule about
a given axis. When the Schroedinger equation is solved for a rotating diatomic molecule
the energy in the two coordinate directions perpendicular to the internuclear axis is
2 2
h K (K + 1) h K (K + 1)
E2,3 diatomic molecule = = (A.71)
2π 2I2,3 2π 2 (mr D2 )
where h is Plancks constant
h = 6.626 × 10−34 J − sec . (A.72)
The atomic masses are assumed to be concentrated at the two atomic centers. The rota-
tional quantum number K is a positive integer greater than or equal to zero, (K = 0, 1, 2,
3, ...). The law of equipartition can be used with (A.71) to roughly equate the rotational
energy with an associated gas temperature. Let
1
kθr ∼
= E2,3 diatomic molecule . (A.73)
2
Equation (A.73) defines the characteristic temperature
2
h 2
θr = (A.74)
2π k (mr D2 )
for the onset of rotational excitation in the two degrees of freedom associated with the 2
and 3 coordinate axes. In terms of the characteristic temperature, the rotational energy
is
E2,3 diatomic molecule = K (K + 1) kθr . (A.75)
Rotational parameters for several common diatomic species are given in figure A.12.
Figure A.12: Rotational constants for several diatomic gases.
The characteristic rotational excitation temperatures for common diatomic molecules are
all cryogenic and, except for hydrogen, fall far below the temperatures at which the mate-
rials would liquefy. Thus for all practical purposes, with the exception of hydrogen at low
temperature, diatomic molecules in the gas phase have both rotational degrees of freedom
fully excited.
A.9.2 Why are only two rotational degrees of freedom excited?
The discretized energy for the degree of freedom associated with rotation about the inter-
nuclear axis is
2
h K (K + 1)
E1diatomic molecule = (A.76)
2π 2I1
and the characteristic excitation temperature is
2
h 1
θr1 = (A.77)
2π 2kI1
Almost all the mass of an atom is concentrated in the nucleus. This would suggest that the
appropriate length scale, d, to use in (A.76) might be some measure of the nucleus diameter.
Consider a solid sphere model for the two nuclear masses. For a homo-nuclear molecule the
characteristic rotational excitation temperature for the 1-axis degree of freedom is
2
h 1
θr1solid nuclear sphere = (A.78)
2π 2k (md2 /5)
The diameter of the nucleus is on the order of 10−14 m four orders of magnitude small than
the diameter of an atom. From this model, one would expect
θr1solid nuclear sphere ∼

= 108 θr . (A.79)
This is an astronomically high temperature that would never be reached in a practical

situation other than at the center of a star.
Another approach is to ignore the nuclear contribution to the moment of inertia along the
1-axis but include only the moment of inertia of the electron cloud. If we take all of the
mass of the electron cloud and concentrate it in a thin shell at the atomic diameter the
characteristic rotational temperature is
2
h 1
θr1electron cloud hollow sphere = . (A.80)
2π 2k (1/3) melectron cloud hollow sphere datom 2
Most atoms are on the order of 10−10 m in diameter which is comparable to the internuclear
distances shown in figure A.12. But the mass of the proton is 1835 times larger than the
electron. So the mass of the nucleus of a typical atom with the same number of neutrons
and protons is roughly 3670 times the mass of the electron cloud. In this model we would
expect
θr1electron cloud hollow sphere ∼

= 3670θr (A.81)
This puts the rotational excitation temperature along the 1-axis in the 8 to10, 000K range,
well into the range where dissociation occurs. The upshot of all this is that because
I1 <<< I2 and I3 the energy spacing of this degree of freedom is so large that no excitations
can occur unless the temperature is extraordinarily high, so high that the molecule is likely
to be dissociated. Thus only two of the rotational degrees of freedom are excited in the
range of temperatures one is likely to encounter with diatomic molecules. The energy of a
diatomic molecule at modest temperatures is
E = ET + ER
1 1 1
ET = mu2 + mv 2 + mw2 (A.82)
2 2 2
1 1
ER = I2 ω2 2 + I3 ω3 2 .
2 2
At room temperature
7
Cp = R
2
(A.83)
5
Cv = R.
2
From the quantum mechanical description of a diatomic molecule presented above we learn
that Cp can show some decrease below (7/2) R at very low temperatures. This is because
of the tendency of the rotational degrees of freedom to freeze out as the molecular kinetic
energy becomes comparable to the first excited rotational mode. More complex molecules
with three-dimensional structure can have all three rotational degrees of freedom excited
and so one might expect n = 6 for a complex molecule at room temperature.
A.9.3 Vibrational degrees of freedom
At high temperatures Cp can increase above (7/2) R because the atoms are not rigidly
bound but can vibrate around the mean internuclear distance much like two masses held
together by a spring as indicated in figure A.13.
Figure A.13: Diatomic molecule as a spring-mass system.
The energy of a classical harmonic oscillator is
1 1
Espring−mass = mr ẋ2 + κx2 (A.84)
2 2
where mr is the reduced mass, κ is the spring constant and x is the distance between the
mass centers.
In the microscopic world of a molecular oscillator the vibrational energy is quantized ac-
cording to

h 1
EV = ω0 j+ (A.85)
2π 2
where ω0 is the natural frequency of the oscillator and the quantum number j = 0, 1, 2, 3, ....
The natural frequency is related to the masses and effective spring constant by
r
κ
ω0 = . (A.86)
mr
For a molecule, the spring constant is called the bond strength. Notice that the vibrational
energy of a diatomic molecule can never be zero even at absolute zero temperature.
A characteristic vibrational temperature can be defined using the law of equipartition just
as was done earlier for rotation. Let
1
kθv ∼
= EV . (A.87)
2
Define
r
h ω0 h 1 κ
θv = = . (A.88)
2π k 2π k mr
Vibrational parameters for several common diatomic species are given in Figure A.14.
Figure A.14: Vibrational constants for several diatomic gases.
The characteristic vibrational excitation temperatures for common diatomic molecules are
all at high combustion temperatures. At high temperature seven degrees of freedom are
excited and the energy of a diatomic molecule is
E = ET + ER + EV
1 1 1
ET = mu2 + mv 2 + mw2
2 2 2
(A.89)
1 1
ER = I2 ω2 2 + I3 ω3 2
2 2
1 1
EV = mr ẋ2 + κx2
2 2
where the various energies are quantized as discussed above. At high temperatures the
heat capacities approach
9
Cp = R
2
(A.90)
7
Cv = R.
2
Quantum statistical mechanics can be used to develop a useful theory for the onset of
vibrational excitation. The specific heat of a diatomic gas over a wide range of temperatures
is accurately predicted to be
2
Cp 7 θv /2T
= + . (A.91)
R 2 Sinh (θv /2T )
The enthalpy change of a diatomic gas at and above temperatures where the rotational
degrees of freedom are fully excited is
Z T Z T 2 !
7 θv /2T
h (T ) − h (T1 ) = Cp dT = R + dT. (A.92)
T1 T1 2 Sinh (θv /2T )
With T1 set (somewhat artificially) to zero, this integrates to
h (T ) 7 θv /T
= + θ /T (A.93)
RT 2 e v −1
plotted in figure A.15.

Figure A.15: Enthalpy of a diatomic gas.
A.10 Energy levels in a box
We discussed the quantization of rotational and vibrational energy levels but what about
translation? In fact the translational energies for particle motion in the three coordinate
directions are quantized when the particle is confined to a box. The wave function for a
single atom of mass m contained inside a box with sides Lx , Ly , Lz satisfies the Schroedinger
equation
ih ∂Ψ (x̄, t) h2
+ 2 ∇2 Ψ (x̄, t) − V (x) Ψ (x̄, t) = 0 (A.94)
2π ∂t 8π m
where h = 6.626068 × 10−34 m2 kg/ sec is Plancks constant. The potential is
V (x̄) = 0 0 < x < L x , 0 < y < Ly , 0 < z < L z

(A.95)
V (x̄) = ∞ on and outside the box.
The solution must satisfy the condition Ψ = 0 on and outside the walls of the box. For
this potential (A.94) is solved by a system of standing waves.
2πE
1/2 |k̄|
8 −i h
t nx πx ny πy nz πz
Ψ (x, y, z, t) = e Sin Sin Sin (A.96)
Lx Ly Lz Lx Ly Lz
The quantized energy corresponding to the wave vector

nx π ny π nz π
k̄ = , , (A.97)
Lx Ly Lz
is
h2 h2 nx 2 ny 2 nz 2

kx 2 + ky 2 + kz 2 =

E|k̄| = 2 2 + 2 + . (A.98)
8π m 8m Lx Ly Lz 2
The corresponding frequency is
E|k̄|
ν= . (A.99)
h
The volume of the box is V = Lx Ly Lz . The wave function is zero everywhere outside the
box and must satisfy the normalization condition
Z Lx Z Ly Z Lz
Ψ (x̄, t) Ψ∗ (x̄, t)dxdydz = 1. (A.100)
0 0 0
on the total probability that the particle is inside the box. The implication of (A.100) and
the condition Ψ = 0 on the walls of the box, is that the quantum numbers (nx , ny , nz )
must be integers greater than zero. The energy of the particle in the box cannot be zero
similar to the case of the harmonic oscillator. For a box of reasonable size, the energy
levels are very closely spaced. A characteristic excitation temperature can be defined for
translational motion at the lowest possible energy. Let
1
kθT ≡ ET (A.101)
2
where
3h2
ET = . (A.102)
8mV 2/3
The translational excitation temperature is
3h2
θT = . (A.103)
4mkV 2/3
Take monatomic helium gas for example. The characteristic translational excitation tem-
perature is
2
3h2 3 6.626 × 10−34 3.57 × 10−14
θTHe = = = K. (A.104)
4 6.643 × 10−27 1.38 × 10−23 V 2/3

4mHe kV 2/3 V 2/3
In general, the characteristic translational temperature is far below any temperature at

which a material will solidify. Clearly translational degrees of freedom are fully excited for
essentially all temperatures of a gas in a box of reasonable size.
A.10.1 Counting energy states
For simplicity let Lx = Ly = Lz = L. In this case, the energy of a quantum state is
h2 2 2 2

E|k̄| = n x + n y + n z . (A.105)
8mV 2/3
Working out the number of possible states with translational energy of the monatomic gas
less than some value E amounts to counting all possible values of the quantum numbers,
nx , ny and nz that generate E|k| < E. Imagine a set of coordinate axes in the nx , ny
and nz directions. According to (A.98) a surface of constant energy is approximately a
sphere (with a stair-stepped surface) with its center at the origin of the nx , ny and nz
coordinates. The number of possible states with energy less than the radius of the sphere is
directly related to the volume of the sphere. Actually only 1/8th of the sphere is involved
corresponding to the positive values of the quantum numbers. Figure A.16 illustrates the
idea.
Figure A.16: Points represent quantum states in a box for energies that satisfy
8EmV 2/3 /h2 ≤ nx 2 + ny 2 + nz 2 . Three cases are shown where the maximum value of
the quantum numbers are 4, 16 and 32.
Following Boltzmanns notation, let W be the number of states.


1 4π 3/2 π 3/2
Woneatom (E, V ) = nx 2 + ny 2 + nz 2 = nx 2 + ny 2 + nz 2 (A.106)
8 3 6
In terms of the volume of the box and the gas kinetic energy the number of states with
energy E|k| < E is, according to (A.105) and (A.106)
π 8m 3/2

Woneatom (E, V ) = V E 3/2 . (A.107)
6 h2
Equation (A.107) gives the number of possible energy states with energy less than E for
a single atom. If we add a second atom to the box then the numberqof possible states lies
within the positive region of a six dimensional sphere of radius R = 8mV 2/3 E/h2 .
8mV 2/3 E
= n1x 2 + n1y 2 + n1z 2 + n2x 2 + n2y 2 + n2z 2

2
(A.108)
h
The volume of a sphere in six dimensions is π 3 /6 R6 . For a box with N molecules the

total number of possible energy states of the system lies within a sphere of 3N dimensions
with volume
π 3N/2 R3N
3N
. (A.109)
2 !
The volume of the positive region of this 3N dimensional sphere is
π 3N/2 R 3N

3N
. (A.110)
2 !
2
In terms of the energy, the number of distinct states with energy less than E is
π 3N/2 2m 3N/2 N 3N/2

W (E, V, N ) = V E . (A.111)
N ! 3N h2

2 !
States with the same energy must not be counted more than once. Since the atoms are
indistinguishable, the additional factor of N ! in the denominator is needed to account for
states that are repeated or degenerate.
If the number of molecules in the volume V is more than a few tens or so, the number of
states can be represented using the Stirling approximation for the factorial.
α! ∼
= αα e−α (A.112)
Now
N 3N/2
3N/2 5N/2 V 4mE
W (E, V, N ) = π e . (A.113)
N 3h2 N
The result (A.113) is presented in Becker’s Theory of Heat, page 135, equation 35.4. A
more accurate version of the Stirling approximation is
α! ∼ 1/2
= (2πα) αα e−α
which would give
N 3N/2
π 3N/2 e5N/2

V 4mE
W (E, V, N ) = √ . (A.114)
6πN N 3h2 N
Note that for values of N greater than a few hundred the number of states with energy less
than E is virtually the same as the number of states with energy equal to E because of
the extremely steep dependence of (A.114) on N . All of the energy states are concentrated
within a membrane thin spherical shell at the energy E.
A.10.2 Entropy of a monatomic gas in terms of the number of states
According to the Law of Equipartition and the development in the early part of this
appendix, the internal energy of a monatomic gas is
1 c2 = 3 kN T.
E = mN u (A.115)
2 2
The number of states can be expressed in terms of the temperature as
N 3N/2
3N/2 5N/2 V 2mkT
W (T, V, N ) = π e . (A.116)
N h2
Take the logarithm of (A.116).


V 3 3 2πmk 5
Log (W (T, V, N )) = N Log + Log (T ) + Log + (A.117)
N 2 2 h2 2
Differentiate (A.117) and multiply both sides by Boltzmanns constant, k .
dW 3kN dT d (V /N )
k = + kN (A.118)
W 2 T V /N
Recall Ru = kNa where Na is Avogadros number. Equation (A.118) now becomes
dW 3 dT d (V /N )
k = nRu + nRu . (A.119)
W 2 T V /N
where the number of moles of gas in the box is n = N/Na .

The model we are considering is that of a monatomic gas with only three degrees of freedom.
For such a gas the constant volume molar heat capacity is Cv = 3Ru /2. Equation (A.119)
now reads
dW dT d (V /n)
k = nCv + nRu . (A.120)
W T V /n
Recall the Gibbs equation
dE P
dS = + dV. (A.121)
T T
For an ideal gas with the equation of state P V = nRu T the Gibbs equation becomes
dT dV
dS = nCv + nRu . (A.122)
T V
Comparing (A.122) with (A.120) leads to the conclusion that
dW
dS = k . (A.123)
W
Integrating (A.123) leads to the famous Boltzmann relation for the entropy
S = kLog (W ) + α. (A.124)
Substutute (A.113) into (A.124). The entropy of the gas is
3/2 ! !
V E 5 3 4πm
S = kN Log + + Log . (A.125)
N N 2 2 3h2
The relation (A.125) is known as the Sackur-Tetrode equation after its discoverers in the
early 1900s. It can be used to derive the various equations that govern a gas. In terms of
the temperature, the entropy expression is equation (A.117) multiplied by k

V 3/2 5 3 2πmk
S = kN Log (T ) + + Log (A.126)
N 2 2 h2
which agrees with equation 35.8 in Beckers Theory of Heat, page 136.
The additive constant in (A.126) is not in disagreement with the third law nor is the fact
that (A.126) becomes singular at T = 0. The theory just developed is for a Boltzmann gas
that necessarily implies temperatures well above absolute zero where the material is not in
a condensed state and Boltzmann statistics apply. The constant is needed for the theory
of condensation. The Sackur-Tetrode equation provides a remarkably accurate value of
the entropy for monatomic gases. As an example, select helium at a pressure of 105 N/m2
and temperature of 298.15 K. At these conditions V /N = 4.11641 × 10−26 m3 /molecule.
The mass of the helium atom is m = 6.64648 × 10−27 kg using these values the entropy
evaluated from (A.126) is
3/2 !
S V 2πmkT
= Log e5/2 = 15.1727 (A.127)
kN N h2
The value provided in tabulations of helium properties is 15.17271576. Similar results can
be determined for other monatomic gases.
According to the third law α = 0 and
S = kLog (W ) (A.128)
is the absolute entropy of the system. Consider the state of a single atom contained in a
box in the limit T → 0. From the solution to the Schroedinger equation, (A.98), the least
energy that the atom can have is
3h2
E111 = . (A.129)
8mV 2/3
This is a tiny but finite number. In this model of a single atom gas, as the temperature
goes to zero the atom in the box falls into the lowest quantum state so the entropy is
S = kLog (1) = 0.
We derived (A.128) using the energy states in an ideal monatomic gas rather than follow
the more general proof of Boltzmann. The remarkable thing about (A.128) is that it applies
to a wide range of systems other than ideal gases including solids and liquids.
Appendix B
Equations in cylindrical and

spherical coords
B.1 Coordinate systems
B.1.1 Cylindrical coordinates
Figure B.1: right-handed cylindrical coordinate system
B-1
APPENDIX B. EQUATIONS IN CYLINDRICAL AND SPHERICAL COORDS B-2
x = rCos (θ)
y = rSin (θ)
z=z (B.1)
1/2
r = x2 + y 2
θ = ArcT an (y/x)
B.1.2 Spherical polar coordinates
Figure B.2: Right handed spherical polar coordinate system.

x = rSin (θ) Cos (φ)
y = rSin (θ) Sin (φ)
z = rCos (θ)
1/2 (B.2)
r = x2 + y 2 + z 2
1/2
θ = ArcT an x2 + y 2 /z
φ = ArcT an (y/x)
B.2 Transformation of vector components
Basic trigonometry can be used to show that the Cartesian and curvilinear components
are related as follows.
Ur = Ux Sin (θ) Cos (φ) + Uy Sin (θ) Sin (φ) + Uz Cos (θ)
Uθ = Ux Cos (θ) Cos (φ) + Uy Cos (θ) Sin (φ) − Uz Sin (θ)
Uφ = −Ux Sin (φ) + Uy Cos (φ)

(B.3)
Ux = Ur Sin (θ) Cos (φ) + Uθ Cos (θ) Cos (φ) − Uφ Sin (φ)
Uy = Ur Sin (θ) Sin (φ) + Uθ Cos (θ) Sin (φ) + Uφ Cos (φ)
Uz = Uz
Ur = Ux Sin (θ) Cos (φ) + Uy Sin (θ) Sin (φ) + Uz Cos (θ)
Uθ = Ux Cos (θ) Cos (φ) + Uy Cos (θ) Sin (φ) − Uz Sin (θ)
Uφ = −Ux Sin (φ) + Uy Cos (φ)

(B.4)
Ux = Ur Sin (θ) Cos (φ) + Uθ Cos (θ) Cos (φ) − Uφ Sin (φ)
Uy = Ur Sin (θ) Sin (φ) + Uθ Cos (θ) Sin (φ) + Uφ Cos (φ)
Uz = Ur Cos (θ) − Uθ Sin (θ)
B.3 Summary of differential operations
1 ∂ 1 ∂Uθ ∂Uz
∇ · Ū = (rUr ) + + (B.5)
r ∂r r ∂θ ∂z
1 ∂2ρ ∂2ρ

1 ∂ ∂ρ
∇2 ρ = r + + 2 (B.6)
r ∂r ∂r r2 ∂θ2 ∂z

∂Ur 1 ∂Uθ Ur ∂Uz
¯τ : ∇Ū = τrr + τθθ + + τzz +
∂r r ∂θ r ∂z
(B.7)
∂ Uθ 1 ∂Ur 1 ∂Uz ∂Uθ ∂Uz ∂Ur
τrθ r + + τθz + + τrz +
∂r r r ∂θ r ∂θ ∂z ∂r ∂z
∂ρ
∇ρ|r =
∂r
1 ∂ρ
∇ρ|θ = (B.8)
r ∂r
∂ρ
∇ρ|z =
∂z
1 ∂Uz ∂Uθ
∇ × Ū r = −
r ∂θ ∂z
∂Ur ∂Uz
∇ × Ū θ = − (B.9)
∂z ∂r
1 ∂ 1 ∂Ur
∇ × Ū z = (rUθ ) −
r ∂r r ∂θ
1 ∂ 1 ∂τrθ τθθ ∂τrz

∇ ·¯τ |r = (rτrr ) + − +
r ∂r r ∂θ r ∂z
1 ∂τθθ 1 ∂ ∂τθz
∇ ·¯τ |θ = + 2 r2 τrθ + (B.10)
r ∂θ r ∂r ∂z
1 ∂ 1 ∂τθz ∂τzz
∇ ·¯τ |z = (rτrz ) + +
r ∂r r ∂θ ∂z
2 ∂Uθ Ur
∇2 Ū r = ∇2 Ur − 2

− 2
r ∂θ r
2 ∂Ur Uθ (B.11)
∇2 Ū θ = ∇2 Uθ + 2

− 2
r ∂θ r
∇2 Ū z = ∇2 Uz

where
1 ∂ 2 () ∂ 2 ()

1 ∂ ∂ ()
∇2 () = r + + (B.12)
r ∂r ∂r r2 ∂θ2 ∂z 2
∂Ur Uθ ∂Ur Uθ 2 ∂Ur

Ū · ∇Ū r = Ur + − + Uz
∂r r ∂θ r ∂z
∂Uθ Uθ ∂Uθ Ur Uθ ∂Uθ (B.13)
Ū · ∇Ū θ = Ur + + + Uz
∂r r ∂θ r ∂z
∂Uz Uθ ∂Uz ∂Uz
Ū · ∇Ū z = Ur + + Uz
∂r r ∂θ ∂z
1 ∂ 1 ∂ 1 ∂Uφ
r2 Ur +

∇ · Ū = 2
(Sin (θ) Uθ ) + (B.14)
r ∂r rSin (θ) ∂θ rSin (θ) ∂φ
∂2ρ

2 1 ∂ 2 ∂ρ 1 ∂ ∂ρ 1
∇ ρ= 2 r + 2 Sin (θ) + 2 (B.15)
r ∂r ∂r r Sin (θ) ∂θ ∂θ r Sin (θ) ∂φ2

∂Ur 1 ∂Uθ Ur 1 ∂Uφ Ur Uθ Cot (θ)
¯τ : ∇Ū = τrr + τθθ + + τφφ + + +
∂r r ∂θ r rSin (θ) ∂φ r r

∂ Uθ 1 ∂Ur ∂Uφ 1 ∂Ur Uφ
τrθ r + + τrφ + − +
∂r r r ∂θ ∂r rSin (θ) ∂φ r

1 ∂Uφ 1 ∂Uθ Uφ Cot (θ)
τθφ + −
r ∂θ rSin (θ) ∂φ r
(B.16)
∂ρ
∇ρ|r =
∂r
1 ∂ρ
∇ρ|θ = (B.17)
r ∂θ
1 ∂ρ
∇ρ|φ =
rSin (θ) ∂φ
1 ∂ (Sin (θ) Uφ ) 1 ∂Uθ

∇ × Ū r = −
rSin (θ) ∂θ rSin (θ) ∂φ
1 ∂Ur 1 ∂
∇ × Ū θ = − (rUφ ) (B.18)
rSin (θ) ∂φ r ∂r
1 ∂ 1 ∂Ur
∇ × Ū φ = (rUθ ) −
r ∂r r ∂θ

1 ∂ 1 ∂ 1 ∂τrφ τθθ + τφφ
r2 τrr +

∇ ·¯τ |r = 2
(Sin (θ) τrθ ) + −
r ∂r rSin (θ) ∂θ rSin (θ) ∂φ r
1 ∂ 2
1 ∂ 1 ∂τθφ τrθ Cot (θ)
∇ ·¯τ |θ = 2
r τrθ + (Sin (θ) τθθ ) + + − τφφ
r ∂r rSin (θ) ∂θ rSin (θ) ∂φ r r
1 ∂ 2
1 ∂τθφ 1 ∂τφφ τrφ 2Cot (θ)
∇ ·¯τ |φ = r τrφ + + + + τθφ
r2 ∂r r ∂θ rSin (θ) ∂φ r r
(B.19)
2 ∂Uθ 2Uθ Cot (θ) 2 ∂Uφ 2Ur

∇2 Ū r = ∇2 Ur − 2

− − 2 − 2
r ∂θ r2 r Sin (θ) ∂φ r
2 ∂Ur Uθ 2Cos (θ) ∂Uφ

∇2 Ū θ = ∇2 Uθ + 2

− 2 − (B.20)
r ∂θ r Sin2 (θ) r2 Sin2 (θ) ∂φ
Uφ 2 ∂Ur 2Cos (θ) ∂Uθ

∇2 Ū φ = ∇2 Uφ −

+ + 2
r Sin2 (θ)
2 r2 Sin (θ) ∂φ r Sin2 (θ) ∂φ
∂ 2 ()

2 1 ∂ 2 ∂ () 1 ∂ ∂ () 1
∇ () = 2 r + 2 Sin (θ) + 2 (B.21)
r ∂r ∂r r Sin (θ) ∂θ ∂θ r Sin2 (θ) ∂φ2
where
Uθ 2 + Uφ 2

∂Ur Uθ ∂Ur Uφ ∂Ur
Ū · ∇Ū r = Ur + + −
∂r r ∂θ rSin (θ) ∂φ r
∂Uθ Uθ ∂Uθ Uφ ∂Uθ Ur Uθ Uφ 2 Cot (θ)

Ū · ∇Ū θ = Ur + + + − (B.22)
∂r r ∂θ rSin (θ) ∂φ r r
∂Uφ Uθ ∂Uφ Uφ ∂Uφ Ur Uφ Uθ Uφ Cot (θ)

Ū · ∇Ū φ = Ur + + + +
∂r r ∂θ rSin (θ) ∂φ r r
B.4 Continuity
∂ρ 1 ∂ 1 ∂ ∂
+ (ρrUr ) + (ρUθ ) + (ρUz ) = 0 (B.23)
∂t r ∂r r ∂θ ∂z
∂ρ 1 ∂ 1 ∂ 1 ∂
ρr2 Ur +

+ 2 (Sin (θ) ρUθ ) + (ρUφ ) = 0 (B.24)
∂t r ∂r rSin (θ) ∂θ rSin (θ) ∂φ
B.5 Momentum
r − component
Uθ 2

∂Ur ∂Ur Uθ ∂Ur ∂Ur ∂P
ρ + Ur + − + Uz + =
∂t ∂r r ∂θ r ∂z ∂r

(rτrr ) + − + + ρgr
θ − component

∂Uθ ∂Uθ Uθ ∂Uθ Ur Uθ ∂Uθ 1 ∂P
ρ + Ur + + + Uz + = (B.25)
∂t ∂r r ∂θ r ∂z r ∂θ
+ 2 r2 τrθ + + ρgθ
r ∂θ r ∂r ∂z
z − component

∂Uz ∂Uz Uθ ∂Uz ∂Uz ∂P
ρ + Ur + + Uz + =
∂t ∂r r ∂θ ∂z ∂z
(rτrz ) + + + ρgz
r ∂r r ∂θ ∂z
r − component
Uθ 2 + Uφ 2

∂Ur ∂Ur Uθ ∂Ur Uφ ∂Ur ∂P
ρ + Ur + + − + =
∂t ∂r r ∂θ rSin (θ) ∂φ r ∂r

1 ∂ 1 ∂ 1 ∂τrφ τθθ + τφφ
r2 τrr +

2
(Sin (θ) τrθ ) + − + ρgr
θ − component
Uφ 2 Cot (θ)

∂Uθ ∂Uθ Uθ ∂Uθ Uφ ∂Uθ Ur Uθ 1 ∂P
ρ + Ur + + + − + =
∂t ∂r r ∂θ rSin (θ) ∂φ r r r ∂θ
1 ∂ 2
1 ∂ 1 ∂τθφ τrθ Cot (θ)
r τrθ + (Sin (θ) τθθ ) + + − τφφ + ρgθ
r2 ∂r rSin (θ) ∂θ rSin (θ) ∂φ r r
φ − component

∂Uφ ∂Uφ Uθ ∂Uφ Uφ ∂Uφ Ur Uφ Uθ Uφ Cot (θ) 1 ∂P
ρ + Ur + + + + + =
∂t ∂r r ∂θ rSin (θ) ∂φ r r rSin (θ) ∂φ
1 ∂ 2
1 ∂τθφ 1 ∂τφφ τrφ 2Cot (θ)
r τrφ + + + + τθφ + ρgφ
(B.26)
B.6 Energy equation
Recall that the stagnation enthalpy is
P
ht = e + + k. (B.27)
ρ
∂ 1 ∂ 1 ∂ ∂
(ρ (e + k)) + (rρht Ur ) + (ρht Uθ ) + (ρht Uz ) +
∂t r ∂r r ∂θ ∂z

1 ∂ 1 ∂Qθ ∂Qz ∂Ur 1 ∂Uθ Ur ∂Uz
(rQr ) + + − τrr + τθθ + + τzz −
r ∂r r ∂θ ∂z ∂r r ∂θ r ∂z

∂ Uθ 1 ∂Ur ∂Uz ∂Ur 1 ∂Uz ∂Uθ
τrθ r + + τrz + + τθz + −
∂r r r ∂θ ∂r ∂z r ∂θ ∂z
(B.28)
Ur (rτrr ) + − + −

Uθ + 2 r2 τrθ + −
r ∂θ r ∂r ∂z

Uz (rτrz ) + + = {powersources}
r ∂θ r ∂r ∂z
∂
(ρ (e + k)) +
∂t
1 ∂ 1 ∂ 1 ∂
r2 ρht Ur +

2
(ρht Sin (θ) Uθ ) + (ρht Uφ ) +

1 ∂ 1 ∂ 1 ∂Qφ
r2 Qr +

2
(Sin (θ) Qθ ) + −

∂Ur 1 ∂Uθ Ur 1 ∂Uφ Ur Cot (θ) Uθ
τrr + τθθ + + τφφ + + −

∂ Uθ 1 ∂Ur ∂Uφ 1 ∂Ur Uφ
τrθ r + − τrφ + − −
∂r r r ∂θ ∂r rSin (θ) ∂φ r

1 ∂Uφ 1 ∂Uθ Cot (θ) Uφ
τθφ + − −
r ∂θ rSin (θ) ∂φ r

1 ∂ 2
1 ∂ 1 ∂τrφ τθθ + τφφ
Ur 2
r τrr + (Sin (θ) τrθ ) + − −

1 ∂ 2
1 ∂ 1 ∂τθφ τrθ Cot (θ) τφφ
Uθ r τrθ + (Sin (θ) τθθ ) + + − −
r2 ∂r rSin (θ) ∂θ rSin (θ) ∂φ r r

1 ∂ 2
1 ∂τθφ 1 ∂τφφ τrφ 2Cot (θ) τθφ
Uφ r τrφ + + + + =
{powersources}
(B.29)
B.7 Components of the stress tensor

∂Ur 2
τrr = 2µ − µ − µv ∇ · Ū
∂r 3

1 ∂Uθ Ur 2
τθθ = 2µ + − µ − µv ∇ · Ū
r ∂θ r 3

∂Uz 2
τzz = 2µ − µ − µv ∇ · Ū
∂z 3
(B.30)
∂ Uθ 1 ∂Ur
τrθ =µ r +
∂r r r ∂θ

∂Ur ∂Uz
τrz = µ +
∂z ∂r

∂Uθ 1 ∂Uz
τθz = µ +
∂z r ∂θ

∂Ur 2
τrr = 2µ − µ − µv ∇ · Ū
∂r 3

1 ∂Uθ Ur 2
τθθ = 2µ + − µ − µv ∇ · Ū
r ∂θ r 3

1 ∂Uφ Ur Cot (θ) Uθ 2
τφφ = 2µ + + − µ − µv ∇ · Ū
rSin (θ) ∂φ r r 3
(B.31)
∂ Uθ 1 ∂Ur
τrθ =µ r +
∂r r r ∂θ

1 ∂Ur ∂ Uφ
τrφ = µ +r
rSin (θ) ∂φ ∂r r

Sin (θ) ∂ Uφ 1 ∂Uθ
τθφ = µ +
r ∂θ Sin (θ) rSin (θ) ∂φ
APPENDIX B. EQUATIONS IN CYLINDRICAL AND SPHERICAL COORDS 13
B.8 Energy dissipation function Φ

2 2 2 !
∂Ur 1 ∂Uθ Ur ∂Uz
Φ = 2µ + + + +
∂r r ∂θ r ∂z
1 ∂Ur 2

∂ Uθ (B.32)
µ r + +
∂r r r ∂θ
2 2
∂Ur ∂Uz ∂Uθ 1 ∂Uz 2 2
µ + +µ + − µ − µν ∇ · Ū
∂z ∂r ∂z r ∂θ 3

2 2 2 !
∂Ur 1 ∂Uθ Ur 1 ∂Uφ Ur Cot (θ) Uθ
Φ = 2µ + + + + + +
1 ∂Ur 2

∂ Uθ
µ r + +
∂r r r ∂θ
2
1 ∂Ur ∂ Uφ
µ +r +
rSin (θ) ∂φ ∂r r
2
Sin (θ) ∂ Uφ 1 ∂Uθ 2 2
µ + − µ − µν ∇ · Ū
r ∂θ Sin (θ) rSin (θ) ∂φ 3
(B.33)

Cantwell - Fundamentals of Compressible Flow PDF

Uploaded by

Copyright:

Available Formats

Cantwell - Fundamentals of Compressible Flow PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cantwell - Fundamentals of Compressible Flow PDF

Uploaded by

Copyright:

Available Formats

AA210A

Fundamentals of Compressible Flow

Department of Aeronautics and Astronautics

1 Introduction to fluid flow 1-1

2 Thermodynamics of dilute gases 2-1

2.2.1 Temperature and the zeroth law . . . . . . . . . . . . . . . . . . . . 2-1

3 Control volumes, vector calculus 3-1

4 Kinematics of fluid motion 4-1

4.1.4 Incompressible flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-6

5 The conservation equations 5-1

6 Several forms of the equations of motion 6-1

6.5.1 Unsteady potential flow, the unsteady Bernoulli integral . . . . . . . 6-8

7 Entropy generation and transport 7-1

8 Viscous flow along a wall 8-1

8.9.2 Non-adiabatic wall with dP/dx = 0 and Pr = 1 . . . . . . . . . . . . 8-47

9 Quasi-one-dimensional flow 9-1

10 Gasdynamics of nozzle flow 10-1

10.4.3 Gasdynamics of a double throat, starting and unstarting supersonic

11 Area change, wall friction and heat transfer 11-1

12 Steady waves in compressible flow 12-1

13 Unsteady waves in compressible flow 13-1

14 Thin airfoil theory 14-1

A Some results from the kinetic theory of gases A-1

B Equations in cylindrical and spherical coords B-1

B.2.2 Spherical polar coordinates . . . . . . . . . . . . . . . . . . . . . . . B-4

Introduction to fluid flow

1.2 Conservation of mass

which can be rearranged to read

Divide (1.3) through by the infinitesimal volume ∆x∆y∆z .

∂ρ ρ U |x+∆x − ρ U |x ρ V |y+∆y − ρ V |y ρ W |z+∆z − ρ W |z

Let ∆x → 0, ∆y → 0, ∆z → 0. In this limit (1.4) becomes

∂ρ ∂ρU ∂ρV ∂ρW

1.2.1 Incompressible flow

1.2.2 Index notation and the Einstein convention

In index notation equation (1.5) is concisely written in the form

where the subscript refers to the ith vector component.

1.3 Particle paths, streamlines and streaklines in 2-D steady

or, in terms of components

When (1.14) is substituted into (1.13) the result is Laplace’s equation.

Figure 1.3: Computed streamlines over a wing flap.

1.3.1 Analysis of particle paths and streamlines

Figure 1.4: Particle trajectory in a 2-D steady flow field.

The equations that determine the trajectory are:

This is the situation depicted schematically in Figure 1.5 .

The differential of Ψ (x, y) is

If we use (1.17) to replace the differentials dx and dy in (1.23) the result is

−V (x, y) dx + U (x, y) dy = 0. (1.26)

1.3.2 The integrating factor

On a line of constant ψ = ψ0 the differential dψ = 0 and

In order to convert equation (1.26) to a perfect differential it must be multiplied by an

dψ = −M (x, y) V (x, y) dx + M (x, y) U (x, y) dy (1.31)

and the partial derivatives are

1.3.3 Sample problem - the integrating factor and reconstruction of a

Solve the first order ordinary differential equation

where H (x, y) is an arbitrary smooth function. Rearrange (1.33) to read

−yH (xy) dx + xdy = 0 (1.34)

This differential expression fails the cross derivative test

∂ (−yH (xy)) ∂ (x)

Multiplying (1.34) by (1.36) converts it to a perfect differential.