Dynamics of Compressive Fluid Flow PDF

The Dynamics
and Thermodynamics of
COMPRESSIBLE FLUID
FLOW
by
ASCHER H. SHAPIRO
Volume I
JOHN WlllFY S SONS

The Dynamics and Thermodynamics of Compressible Fluid Flow : a f ~
Ascher H. Shapiro : o u - g i
j L >€%I i j g h-
9 djgoT &I+ 0-93
Awo.+-r : ir41,-
AnSo*t : fiu
\qYqo-Y011 : &.JJ&
[email protected] : & 9 F f v-
The Dynamics
and Thermodynamics of
COMPRESSIBLE FLUID
FLOW
BY
ASCHER H . SHAPIRO
Professor of Mechanical Engiiredng
Massachusetts Institute of Technology
w o VOLUMES
VOLUME I
JOHN WILEY & SONS

New York Chichester Brisbane .Toronto. Singapore
Copyright @ 1953 by J o h n W i e y & Sons, Inc
All Rights Reserved

Reproduction or translation o f any part o f this work beyond that
permitted by Sections 107 or 108 o f the 1976 United States Copy-
right Act without the permission o f the copyright owner i s unlaw-
ful. Requests for permission or further information should be
addressed to the Permissions Department, John Wiley & Sons, Inc.
I S B N 0 471 06691-5
Library of Congress Catalog Card Number: 53-8869

P K l N T E D IN TEE UNITED BTATEB # I F AMEIIICA
10 9
TO THE MEMORY OF MY FATHEH.
BERNARD SHAPIRO
PREFACE
During the past two decades a rapid growth of interest in the motion
of compressible fluids has accompanied developments in high-speed
flight, jet engines, rockets, ballistics, combustion, gas turbines, ram jets
and other novel propulsive mechanisms, heat transfer a t high speeds,
and blast-wave phenomena. My purpose in writing this book is to make
available to students, engineers, and applied physicists a work on com-
pressible fluid motion which would be suitable as an introductory text
in the subject as well as a reference work for some of its more advanced
phases. The choice of subject matter has not been dictated by any
particular field of engineering, but rather includes topics of interest to
aeronautical engineers, mechanical engineers, chemical engineers, ap-
plied mechanicians, and applied physicists.
I n selecting material from the vast literature of the field the basic
objective has been to make the book of practical value for engineering
purposes. To achieve this aim, I have foilowed the philosophy that the
most practical approach to the subject of compressible fluid mechanics
is one which combines theoretical analysis, clear physical reasoning, and
empirical results, each leaning on the other for mutual support and ad-
vancement, and the whole being greater than the sum of the parts.
The analytical developments of this book comprise two types of treat-
ments: those leading to design methods and those leading to exemplary
methods. The design methods are direct and rapid, and easily applied
to a variety of problems. Therefore, they are suited for use in the engi-
neering office. The discussions of these design methods are detailed and
illustrative examples are often given. The exemplary methods, on the
other hand, comprise those theoretical analyses which are time consum-
ing, which generally require mathematical invention, and which are not
easily applied to a variety of problems. Such methods are primarily of
value for ~ieldingdetailed answers to a small number of typical prob-
lems. Although they are not in themselves suitable for the engineering
office, the examples which they permit to be worked out often provide
important information about the behavior of fluids in typical situations.
Thus they serve as guides to the designer in solving the many complex
problems where even the so-called design met,hods are not sufficient.
The treatment of exemplary methods in this book usually consists of a
brief outline of the method, together with a presentation of those results
obtained by the method which illuminate significant questions concern-
V
vi PREFACE
ing fluid motion and which help to form the vital "feel" so desired by
designers.
In keeping with the spirit of the several foregoing remarks, all
the important results of the book have been reduced to the form
of convenient charts and tables. Unless otherwise specified, the
charts and tables are for a perfect gas with a ratio of specific heats (k)
of 1.4.
In those parts of the book dealing with fundamentals, emphasis is
placed on the introduction of new concepts in an unambiguous manner,
on securing a clear physical understanding before the undertaking of an
analysis, on the rigorous application of physical laws, and on showing
fruitful avenues of approach in analytical thinking. The remaining part
of the work proceeds a t a more rapid pace befitting the technical ma-
turity of advanced students and professionals.
The work is organized in eight parts. Part I sets forth the basic con-
cepts and principles of fluid dynamics and thermodynamics from which
the remainder of the book proceeds and also introduces some funda-
mental concepts peculiar to compressible flows. I n Part I1 is a discus-
sion of problems accessible by the most simple picture of fluid motion
-the one-dimensional analysis. Part I11 constitutes a summary of the
basic ideas and concepts necessary for the succeeding chapters on two-
and three-dimensional flow. Parts IV, V, and VI then present in order
comprehensive surveys of subsonic flows, of supersonic flows (including
hypersonic flow), and of mixed subsonic-supersonic flows. In Part VII
is an exposition of unsteady one-dimensional flows. Part VlII is an
examination of the viscous and heat conduction effects in laminar and
turbulent boundary layers, and of the interaction between shock waves
and boundary layers. For those readers not already familiar with it,
the mathematical theory of characteristic curves is briefly developed in
Appendix A. Appendix B is a collection of tables which facilitate the
numerical solution of problems.
The "References and Selected Bibliography" a t the end of each c h a p
ter will, it is hoped, be a helpful guide for further study of the volumi-
nous subject. Apart from specific references cited in each chapter, the
lists include general references appropriate to the subject matter of each
chapter. The choice of references has been based primarily on clarity,
on completeness, and on the desirability of an English text, rather than
on historical priority.
My first acknowledgment is to Professor Joseph H. Keenan, to whom
1 owe my first interest in the subject, and who, as teacher, friend, and
colleague, has been a source of inspiration and encouragement.
In an intangible yet real way I am indebted to my students, who have
made teaching a satisfying experience, and to my friends and colleagues
PREFACE tril
at the Massachusetts Institute of Technology who contributed the cli-
mate of constructive criticism so conducive to creative effort.
Many individuals and organizations have been cooperative in supply-
ing me with helpful material and I hope that I have not failed to ac-
knowledge any of these a t the appropriate place in the text. The Na-
tional Advisory Committee for Aeronautics and the M.I.T. Gas Turbine
Laboratory have been especially helpful along these lines.
I was fortunate in being able to place responsibility for the important
work of the drawings in the competent hands of Mr. Percy H. Lund,
who, with Miss Prudence Santoro, has been most cooperative in this
regard.
For help with the final revision and checking of the manuscript I wish
to give thanks to Dr. Bruce D. Gavril and Dr. Ralph A. Burton.
Finally, but by no means least, I must express a word of appreciation
to Sylvia, and to young Peter, Mardi, and Bunny, who, one and all,
made it possible for me to escape from the office into the somewhat less
trying atmosphere of the home, and there to carry this work forward
to its completion.
ASCRERH. SHAPIRO
Arlington, Mass.
April 4, 1953
CONTENTS
VOLUME I
Part I . Background
CHAPTER PAGE
1 FOUNDATIONS OF FLUID DYNAMICS 3
Properties of the Continuum. Systems and Control Volumes.
Conservation of Mass. Momentum Theorem. Theorem of Mo-
ment of Momentum. Units and Dimensions.
2 FOUNDATIONS OF THERMODYNAMICS 23
The First Law of Thermodynamics. The Second Law of Thermo-
dynamics. Thermodynamic Properties of the Continuum. The
First Law for a Control Volume., The Second Law fot a Control
Volume. The Perfect Gas.
The Velocity of Sound. Physical Differences Between Incompres-

sible, Subsonic, and Supersonic Flows. The Mach Number and
Mach Angle. Similarity Parameters. Domain of the Continuum.
Classification of Compressible Flows. Optical Methods of Inves-
tigation.
Part II. One-Dzmensional Flow

4 ISENTROPIC
FLOW
General Features of Isentropic Flow. Adiabatic Flow of a Perfect
Gas. Isentropic Flow of a Perfect Gas. Working Charts a.nd
Tables for Isentropic Flow. Choking in Isentropic Flow. Opera-
tion of Nozzles Under Varying Pressure Ratios. Special Relations
for Low Mach Numbers. Deviations from Perfect Gas Laws.
Performance of Real Nozzles. Some Applications of Isentropic
Flow.
5 NORMAL SHOCK WAVES 112

Governing Relations of the Normal Shock. Normal Shock in a
Perfect Gas. Working Formulas, Curves, and Tables. Weak
Shock Waves. Formation of Shock Waves. Thickness of Shock
Waves. Normal Shocks in Ducts. Moving Shock Waves. Oper-
ating Characteristics of Converging-Diverging Nozzle. One-Di-
mensional Supersonic Diffusers. Supersonic Pitot Tube.
ix
x CONTENTS
CHAPTER PAGE
6 FLOW IN CONSTANT-AREA DUCTS WITH FRICTION 159
Adiabatic, Constant-Area Flow of a Perfect Gas. Performance of
Long Ducts a t Various Pressure Ratios. Isothermal Flow in Long
Ducts. Experimental Friction Coefficients.
Simple-Heating Relations for a Perfect Gas. Choking Effects in

Simple TO-Change. Shock Waves with Changes in Stagnation
Temperature. The Recovery Factor. The Coefficient of Heat
Transfer.
8 GENERALIZED ONE-DIMENSIONAL CONTINUOUS FLOW 219

Physical Equations and Definitions. Working Equations and
Tables of Influence Coefficients. Flow with Constant Specific
Heat and Molecular Weight. General Features of Flow Patterns.
General Method of Solution. Simple Types of Flow. Example
of Combined Friction and Area Change. Examples of Combined
Friction and Heat Transfer. Special Conditions at the Sonic
Point.
Part 111. Introduction to Flow in Two and Three Dimensions
The Physical Significance of Irrotational Motion. Euler's Equa-

tions of Motion. Kelvin's Theorem. The Connection Between
the Rotation and the Thermodynamic Properties of the Flow.
The Equation of Continuity. The Laws of Thermodynamics.
Differential Equations in Terms of the Velocity Potential. Dif-
ferential Equations in Terms of the Stream Function. Relations
Between the Velocity Potential and the Stream Function.
Part N. Subsonic Flow

10 TWO-DIMENSIONAL, SUBSONIC FLOW WITH SMALL PERTURBA-
TIONS 303
Linearization of the Potential Equation. Linearization of the
Pressure Coefficient. Flow Past a Wave-Shaped Wall. Gothert's
Rule. The Prandtl-Glauert Rule. Experimental Results for Thin
Profiles. Wind Tunnel Corrections. Flow Inside Two-Dimen-
sional Passages.
11 HODOGRAPH
METHOD FOR TWO-DIMENSIONAL, SUBSONIC
FLOW 336
Derivation of the Hodograph Equations. The Tangent-Gas Ap-
proximation. The Karman-Tsien Pressure Correction Formula.
Calculation of Profile Shape Correction. Extension of Karman-
Tsien Method. Miscellaneous Examples.
CONTENTS
CHAIT E R
12 MISCELLANEOUS METHODS
A N D RESULTS FOR TPio-DIMEN-
SIONAL, SUBSONIC FLOW
The Rayleigh-Janzen Method of Expansion in Series of the Mach
Number. The Prandtl-Glauert Method of Expansion in Series of
a Shape Parameter. Relaxation Method. Some Measured Effects
of Compressibility in Subsonic Flow. The Streamline Curvature
Method.
THREE-DIMENSIONAL, SUBSONIC FLOW
Gothert's Rule for Uniform Flow with Small Perturbations. Flow
Past Ellipsoids. Bodies of Revolution. Spheres. Wings of Finite
Span. Sweptback Wings. Sweptback Wings of Finite Span.
Part V . Supersonic Flow

TWO-DIMENSIONAL, SUPERSONIC FLOW WITH SMALL PER-
TURBATIONS
Linearization of the Equations. The General Solution for Linear-
ized Supersonic Flow. Geometrical Interpretation of the General
Solution. Flow Past a Wave-Shaped Wall. Supersonic Airfoils.
Reflection and Intersection of Waves.
METHOD OF CHARACTERISTICS FOR TWO-DIMENSIONAL, SU-
PERSONIC FLOW
Flow with Waves of One Family by Extension of Linear Theory.
Flow with Waves of Both Families by Extension of Linear Theory.
Application of Theory of Characteristics. Simple Waves by
Theory of Characteristics. Field Method Versus Lattice-Point
Method. Unit Processes. Graphical Versus Numerical Method.
Some Special Features of Supersonic Flow. Applications of
Method of Characteristics. Design of Supersonic Wind Tunnel
Nozzles. Adiabatic, Nonviscous Flow with Rotation.
O BLIQUE S HOCKS
Oblique Shock Equations. Shock Geometry. Shock Polars.
Some Special Aspects of Oblique Shocks. Very Weak Shocks.
Reflection and Interaction of Shocks. Curved Shocks. Explicit
Solutions by Series Expansions. Examples of Two-Dimensional
Flows Containing Shocks. Two-Dimensional Profiles. Interac-
tion of Shock Waves with Boundary Layer.
Appendix
THEORY OF CHARACTERISTICS
The Characteristic Curves. Method of Constructing Character-
istic Curves. Simple Waves.
VOLUME II
Part V . Supersonic Flow (Continued)
CHAPTER
17 AXIALLY SYMMETRIC SUPERSONIC FLOW
Exact Solution for Flow Past a Cone. Linear Theory for Slender
Bodies of Revolution. Method of Characteristics. Miscellaneous
Experimental Results.
18 SUPERSONIC FLOW PAST WINGS OF FINITE SPAN

Preliminary Considerations of Finite Wings. Sweptback Wings.
Similarity Rule for Supersonic Wings. The Method of Super-
sonic Source and Doublet Distributions. The Method of Conical
Fields. Typical Theoretical Results for Finite Wings. Compari-
son of Theory with Experiment.
19 HYPERSONIC FLOW
Similarity Laws for Hypersonic Flow. Oblique Shock Relations
for Hypersonic Flow. Simple-Wave Expansion Relations for Hy-
personic Flow. Hypersonic Performance of Two-Dimensional
Profiles. Hypersonic Performance of Bodies of Revolution. Ex-
perimental Results.
Part VI. Mixed Flow

20 THE HODOGRAPH
METHOD FOR MIXED SUBSONIC-SUPERSONIC
FLOW
Equations of the Hodograph Method. Source-Vortex Flow.
Compressible Flow with 180" Turn. The Limit Line. Solution
of Hodograph Equations by Hypergeometric Functions.
21 TRANSONIC FLOW
The Transonic Similarity Law. Applications of the Transonic
Similarity Law. Flow in Throat of Converging-Diverging Nozzle.
Relaxation Method. Transonic Flow Past a Wavy Wall. Flow
a t Mach Number Unity. Slopes of Force Coefficients a t M, = 1.
Transonic Flow Past Wedge Nose.
22 DRAG AND LIIT AT TRANSONIC SPEEDS

Experimental Validity of Transonic Similarity Law. Character-
istics of Wing Profiles. Characteristics of Wings. Transonic
Drag of Bodies of Revolutioh. Detached Shocks. Theoretical
Consideration of Transonic Flow Without Shocks. Interaction
Between Boundary lay^ and Shock Wave.
xii
CONTENTS xiii
Part VIZ. Unsteady Motion in One Dimension

CHAPTER
23 UNSTEADY WAVE MOTION OF SMALL AMPLITUDE
Equations of Motion. Waves of Small Amplitude. Simplified
Physical Analysis of Pressure Pulse. Characteristic Curves.
Application of Theory. Development of Wave Form. Effects
of Gradual Changes in Area.
24 UNSTEADY, ONE-DIMENSIONAL, CONTINUOUS FLOW
Extension of Linearized Theory. Method of Characteristics.
Simple Waves. Waves of Both Families. Unit Operations and
Boundary Ccnditions. Unsteady, One-Dimensional Flow. Re-
marks on Details of Working Out the Method of Characteristics.
Some Examples.
25 UNSTEADY, ONE-DIMENSIONAL SHOCK WAVES
Analysis in Terms of Stationary Shock Formulas. Analysis of
Moving Shocks. The Shock Tube-Riemann's Problem. Weak
Shock Waves. Modified Calculation Procedure for Weak Shocks.
End Conditions and Interaction Effects for Strong Shocks. Com-
parison Between Experimental and Theoretical Results.
Part VZZZ. Flow of Real Gases with Viscosity and

Heat Conductivity
26 THE LAMINAR BOUNDARY LAYER
Differential Equations of the Laminar Boundary Layer. Flow
With Prandtl Number Unity. Flow With Arbitrary Prandtl
Number. Integral Equations of the Laminar Boundary Layer.
Laminar Boundary Layer for Axi-Symmetric Flow. Experimental
Results for Laminar Boundary Layers. Stability of the Laminar
Boundary Layer.
27 THE TURBULENT BOUNDARY LAYER
Differential Equations of the Turbulent Boundary Layer. In&
gral Equations of the Turbulent Boundary Layer. Analyses of
Recovery Factor, Skin Friction, and Heat Transfer for Turbulent
Flow Past a Flat Plate with Turbulent Prandtl Number of Unity.
Theoretical and Experimental Results for Skin Friction on Flat
Plates. Recovery Factor for Turbulent Flow. Turbulent Boun-
dary Layer on Bodies of Revolution.
28 BOUNDARY LAYERS IN TUBES AND IN THE PRESENCE OF SHOCK
WAVES
Flow in Tubes. Shock-Boundary Layer Interactions in Super-
sonic Flow. Shock-Boundary Layer Interactions in Transonic
Flow. Normal Shocks in Ducts. Boundary-Layer Separation
Produced by Shock Waves.
INDEX FOR VOLUMES I AND I1
PART I
BACKGROUND
Chapter 1
FOUNDATIONS OF FLUID DYNAMICS
1.1. Introductory Remarks

An engineering science like fluid dynamics rests on foundations com-
prising both theory and experiment. With fluid dynamics, progress has
been especially dependent upon an intimate cross-fertilization between
the analytical and empirical branches; the experimental results being
most fruitfully interpreted in terms of theoretical reasoning, and the
analyses in turn suggesting critical and illuminating experiments which
further amplify and strengthen the theory.
The analytical branch of a science is constructed from concepts, defini-
tions, and the statements of physical laws. The latter are in terms of
the concepts and definitions and are in conformity with experimental
observations. All analyses concerning the motion of compressible fluids
must necessarily begin, either directly or indirectly, with the statements
of the four basic physical laws governing such motions. These laws,
which are independent of the nature of the particular fluid, are
(i) the law of conservation of mass
(ii) Newton's second law of motion
(iii) the first law of thermodynamics
(iv) the second law of thermodynamics
In addition to these fundamental principles, it is usually necessary to
bring into an analysis certain subsidiary 1a.w~relating to the particular
fluid or class of fluids in question. Examples are the equation of state
of a perfect gas, the proportionality law between shear stress and rate
of shear deformation in a Newtonian fluid, the Fourier law of heat
conduction, etc.
I n this book emphasis is placed on the manner in which important
conclusions spring from analyses growing out of the four basic laws men-
tioned above. For this reason the first two chapters are devoted to a
review of these principles and the associated concepts and definitions.
In this way the general point of view and phraseology of the book will
be established.
Definition of a Fluid. The rigorous classification of substances in
various ways is usually thwarted because certain substances behave so
3
4 FOUNDATIONS O F F L U I D DYNAMICS Ch. 1
anomalously as to defy being neatly placed in a pigeonhole. For our

present purpose, we wish principally to distinguish between the numer-
ous common substances we call fluids and such other substances as solids
and plastics.
We shall define a jluid as a substance which deforms continuously
under the action of shearing forces. When shearing forces are applied to
a solid, the latter undergoes a certain deformation which does not change
so long as the applied forces are maintained constant. A fluid, however,
whether viscous or nonviscous, exhibits relative motion between its
elementary parts so long as shearing forces are applied. Thus we say
that "a fluid cannot withstand shearing stresses."
An important corollary which follows from the definition of a fluid is
the observation that if there is no relative motion within the fluid, i.e.,
if fluid particles are not deformed, then there can be no shear stresses
acting on such particles.
LIQUIDS vs. GASES. The usual methods of attempting to distinguish
rigorously between a liquid and a gas, both of which are fluids, are futile
and indeed not of any practical use. That this is so may be seen by con-
sidering that a mass of "water" a t 1 atm and 100°C in a glass cylinder
closed by a piston may, by suitable heat transfers and motions of the
piston, be transformed to a mass of "steam" a t 1 atm and 100°C, with-
out a meniscus once being observed! For most practical purposes the
words "liquid" and "gas" are of value insofar as the former denotes a
fluid which generally exhibits only small percentage changes in density.
The subject matter of this book relates for the most part to highly
compressible fluids, and so we shall generally speak of gases.
The Concept of a Continuum. Matter, while seemingly continuous, is
composed of myriads of molecules in constant motion and collision.
The most fundamental approach in analyzing the motion of matter in
the aggregate is, therefore, to set down the laws of motion for each indi-
vidual molecule and to trace the history of each molecule, or of statistical
groups of molecules, subsequent to some initial state of affairs. This
approach, which usually goes under the name of kinetic theory or statis-
tical mechanics, has obvious merits, but, on the other hand, is often too
cumbersome for practical calculations.
In most engineering problems our primary interest lies not in the
motions of molecules, but rather in the gross behavior of the fluid thought
of as a continuous material. Although the postulate of a continuous
fluid is nothing but a convenient fiction, it is fortunately a valid approach
to many practical problems where only macroscopic or phenomeno-
logical information is of interest. The treatment of fluids as continua
may be said to be valid whenever the smallest volume of fluid of interest
contains so many molecules as to make statistical averages meaningful.
Art. 1.2 PROPERTIES OF THE CONTINUUM 5
The great simplification afforded by the concept of a continuum is

that instead of dealing with instantaneous states of innumerable mole-
cules, we deal instead with certain macroscopic properties describing
the gross behavior of the substance. In the motion of compressible
fluids the relevant properties are density, pressure, shear stress, velocity,
coefficient of viscosity, temperature, internal energy, entropy, and
coefficient of thermal conductivity. These are defined in Arts. 1.3
and 2.4.
This book concerns the motion of compressible fluids which may be
treated as continua. To avoid the impression that the methods and
results of this book are universally valid, i t seems well at this point to
rnention that the macroscopic approach fails whenever the mean free
path of the molecules is of comparable size with the smallest significant
dimension of the problem. Thus, whenever we deal with highly rarefied
gases (as in rocket flight a t extreme altitudes, high vacuum technology,
or electronic tubes), the continuum approach of classical fluid mechanics
and thermodynamics must be abandoned in favor of the microscopic
approach of kinetic theory.
NOMENCLATURE
acceleration u component of velocity in x-
area direction
area vector ZJ component of velocity in y-
force direction
force vector '0 volume
magnitude of body force per V speed
unit mass V velocity
constant of proportionality in w mass rate of flow
Newton's second law x, y, z Cartesian coordinates
mass
moment of a force, or torque
normal force per unit area, or Y angle
pressure '/ coefficient of viscosity
magnitude of radius vector P mass density at a point
radius vector T tangential force per unit area,
time or shear stress
1.2. Properties of the Continuum

We discuss here those continuum properties relevant t o the laws of
motion.
Density at a Point. Consider the mass of fluid 6m in a volume KJ
surrounding the point P in a continuous fluid (Fig. l.la). The ratio
6m/6'0 is called the average mass density of the fluid within the volume
8 FOUNDATIONS OF FLUID DYNAMICS Ch. 1
6'U. Now suppose that a t first 6'U is rather large, and that it is subse-
quently shrunk about the point P. Then a plot of 6m/6'U versus 621
would be typified by Fig. l.lb. At first the average density tends to
approach an asymptote as the volume encloses fluid more and more
homogeneous in nature. However, when 62, becomes so small as to
contain relatively few molecules, the average density fluctuates sub-
stantially with time as molecules pass into and out of the volume, and
so i t is impossible to speak of a definite value for 6m/6W. We may then
(4 (b)
FIG.1.1. Definition of density at a point.
imagine the smallest volume which can be regarded as continuous t o

be 6'Uf1and then define the density at a point as
6m
p = lim -
au + auf 6'U
This definition illuminates the idea of a continuum and shows the true
nature of a continuum property "at a point 1 as a fictitious but highly
J
useful concept.
Fluid Velocity at a Point. The fluid velocity a t a point is quite inde-
pendent of the instantaneous velocity of the molecule nearest that point.
Rather we consider the motion of the center of gravity of the volume
62,' (Fig. l.lb) instantaneously surrounding that point, and define the
jluid velocity at the point P as the instantaneous velocity of this center of
gravity. Thus the fluid velocity a t a point is the instantaneous velocity
of the fluid particle which a t that moment is passing through the point.
By jluid particle we mean here a small mass of fluid of fixed identity and
of size comparable with 62,'.
Whereas density a t a point is a scalar quantity, fluid velocity a t
a point is a vector. After the introduction of a coordinate system,
it is therefore possible to resolve the vector velocity into three scalar
components.
Art. 1.2 PROPERTIES OF THE CONTINUUM 7
STREAMLINES. At any instant, every point of a fluid continuum has
a corresponding fluid velocity vector. The curves which are everywhere
tangent to the velocity vector are known as the instantaneous streamlines
of the flow, and comprise one of the most common and most useful graph-
ical representations of the flow. When we speak of the flow pattern, we
often mean the streamline picture.
In unsteady flows the streamline pattern changes from instant to
instant. When the flow is steady, however, the streamlines are constant in
time and represent also the path lines, or the trajectories of fluid particles.
STREAM TUBE. Consider an elementary area dA normal to the veloc-
ity vector a t some point. The streamlines passing through the circum-
ference of d A form a surface which is called a stream tube. Since there
is no velocity component normal to the streamline, the walls of a stream
tube may be thought of as being impervious to flow.
Stress at a Point. Consider a plane passing in a given direction
through point P of the continuous medium of Fig. 1.2, and imagine a
I Domain of I
(4 (b)
FIG.1.2. Definition of normal stress at a point.
circular area 6A in this plane surrounding P. It is a matter of experience

that the fluid on one side of this surface exerts a force on the fluid on the
other side; and, by the third law of Newton, the fluid on the latter
side exerts an equal but opposite force on the fluid on the first side.
Such forces are called surface forces.
NORMAL STRESS. The surface force acting on the area 6A may be re-
solved into a component normal to 6A and a component lying in the
plane of 6A. Considering first the normal force component, and referring
to Fig. 1.2b, we define thefluid pressure at the point P in the given direc-
tion as the limit of the normal component of force per unit area:
SHEAR STRESS. Similarly, the shear stress 7 is defined in terms of the
limit of the tangential component of force per unit area acting on 6A:
Both the pressure and the shear stress are, of course, vector quantities.
The area 6A' and volume 6V1, i t should be noted, have comparable
dimensions.
HYDROSTATIC PRESSURE. We now raise the question whether a t the
point P the pressure, or normal stress, depends upon the orientation of
the plane 6A.
To answer this question, let us first consider an inviscid fluid, that is,
a fluid in which no viscous stresses (tangential or shear stresses) exist
(1 -
even though there is relative motion within the fluid. Suppose for
I
I
I
I
I
I
\
_.
,------
Connituum-l
&
\ pxdy
/
dy
--- \
- 7--J
pydx
\
\
\'
\I
I
I
mathematical simplicity that only
two dimensions are considered, and
let us then investigate the dynamics
of an infinitesimally small triangu-
lar piece of fluid of unit depth sur-
rounding P in Fig. 1.3. Then, since
shear stresses are assumed absent,
only normal stresses, or pressures,
act on the three faces. The forces
'.
.----
\
_---__----/' x
FIG.1.3. Normal stress forces acting
exerted by the fluid outside the
triangle on the fluid within the
at a point. triangle are shown in the sketch,
where p, denotes the pressure a t
point P in the x-direction, p, denotes the pressure a t point P in the
y-direction, and p, denotes the pressure a t point P in the 7-direction.
Let g, and g, be the components of body force (e.g., gravity force)
acting on the fluid per unit mass. Now we may write Newton's second
law in the x-direction, as follows:
where a, is the acceleration in the x-direction. From the geometry of

the triangle, we see that
and then, noting that dx dy is negligible compared with dy, we obtair~

the result that p, = p,.
4rt. 1.2 PROPERTIES OF THE CONTINUUM 9
By similarly applying Newton's second law in the y-direction, we

get p, = py. Thus we have finally that
and, since the direction -y was chosen without prejudice, we conclude

that in a n inuiscid fluid, whether in motion or not, the $fluid pressure at a
point i s the same in all directions. Such a condition is termed a hydro-
static state of stress.
Consider next a viscous fluid a t rest. Shear stresses are again absent
because of the lack of relative motion, and a similar analysis yields the
same result of a hydrostatic state of stress.
STATE OF STRESS I N VISCOUS FLUID. Turning next to the general case
of a viscous fluid in motion, shearing stresses must be inserted in Fig. 1.3.
By writing the laws of motion, i t is soon found that p,, p,, and p, are
all different, and indeed the state of stress is similar to that existing in
a solid, as described graphically by Mohr's circle of stress.
In this case, which is in fact the situation usually obtaining in prac-
tice, i t is meaningless to speak of the pressure a t a point, but rather
it is necessary to speak in terms of six stress components a t a point.(')
Nevertheless, we often do speak of the pressure a t a point in viscous
fluids, and this is justified by the fact that in such cases (specifically,
when the Reynolds Number is large compared with unity) the shear
stresses are small compared with the normal stresses, and consequently
the variation of normal stress with orientation is correspondingly small.
Coefficient of Viscosity. I n all real fluids, experimental observations
indicate that a shearing deformation is always accompanied by a shearing
stress. Most common engineering
fluids are Newtonian in nature,
which means that the shearing stress
is proportional to the rate of shear-
ing deformation.
Consider in two dimensions a fluid
particle of fixed identity which a t
some instant of time t is rectangular
and which at the same t has one
corner a t point P in space (Fig. 1.4).
Let the shear stresses acting on the
two faces passing through P be de-
noted by T,, and T,. X
As in the analysis
- connected with FIG.1.4. Shear deformation in a fluid.
Fig. 1.3, the inertia forces and body
forces are of higher order than the surface forces. Then, applying New-
ton's second law for moments of forces, i t follows that T,, = 7,.
After a time interval dt, the particle has moved to the position shown
by the dashed lines, and the corner a t P has moved to P'. Furthermore,
the included angle a t P has changed by the amount dr. A Newtonian
&id is now defined as one exhibiting the proportionality
The constant of proportionality is called the coeficient of viscosity,, p

and is defined by
In Chapter 9 a study of the kinematics of two-dimensional motion

yields an expression for dr/dt leading to
where u and v are the respective velocity components in the x- and

y-directions.
PERFECT FLUIDS. A perfect fluid is defined as one having a zero coef-
ficient of viscosity, which means physically that shear stresses are absent
despite shearing deformations in the fluid. No real fluids are inviscid,
but the concept of a perfect fluid is useful inasmuch as it provides a
simple model which a t the same time approximates real fluids in many
situations.
An i n v i s d fluid differs from a viscous fluid in one other important
respect. All real, and hence viscous, fluids stick to solid boundaries
with which they are in contact; that is, a t solid boundaries the fluid
has the same velocity as the boundary. A perfect fluid, however, may
glide past solid boundaries with complete freedom.
BOUNDARY LAYER. Since many real engineering fluids have small
viscosity, but nevertheless stick to solid boundaries, the concept of a
boundary layer has proved immensely useful. According to this con-
cept, viscous forces in the fluid may be ignored in comparison with
inertia forces a t some distance from the boundary. In a thin boundary
layer near the solid boundary, however, shear forces must necessarily
be significant because of the large shearing deformations resulting from
the fluid's being a t rest a t the solid boundary. This division of the field
of flow into two distinct parts is often of great practical service in making
problems simple enough for analytical treatment.
Mathematical Description of Continuum. There are two possible
viewpoints in describing the motion of fluids. First, there is the method
Art. 1.3 SYSTEMS AND CONTROL VOLUMES 11
of Lagrange, usually used in the dynamics of discrete particles, wherein

the history of individual fluid particles is described; that is, a t each
instant of time the location, density, state of stress, etc., are specified
for a certain fluid particle of fixed identity.
The method of Eukr, on the other hand, focuses attention on a fixed
point of space, and specifies a t each instant of time the density, pressure,
~ t c . of
, the fluid particle which happens to occupy that point a t that
instant.
I n most problems of fluid motion Euler's description proves far more
convenient, and it is the viewpoint adopted in this book.
Treating the density as a typical continuum property defined a t a
point, we say in the Eulerian method that the density is a function of
location and time. Mathematically, with x, y, z as Cartesian coor-
dinates, we indicate this by writing
Then for arbitrary increments in x, y, z,and t, the corresponding incre-

ment in p may be written in terms of the partial differential quotients:
Relations like Eqs. 1.6 and 1.7 may be written for all continuum
properties thus far defined.
STEADY FLOW. Many engineering problems refer to a steady condi-
tion of operation. Specifically we define steady$ow as a condition where
at, each point of space there is no variation of any property with respect
to time. Using the density for illustration, we may then write
1.3. Systems and Control Volumes

The System. The four basic principles listed in Art. 1.1 are always
stated in the first instance in terms of a "system." A system, which is
sometimes called a free body or isolated body, is defined as an arbitrary
collection of matter of fixed identity. Everything external to the system
is called the surroundings. The boundary oj a system is defined as the
imaginary surface which separates the system from its surroundings.
Through the use of the system concept one focuses attention on the
body or substance of immediate interest and then observes interactions
between the system and the surroundings.
12 FOUNDATIONS OF FLUID DYNAMICS Ch. I
To illustrate, consider Newton's second law, F = ma. Here F is the
resultant vector force exerted by all the surroundings on the system; m
is the mass of the system; and a is the vector acceleration experienced
by the center oj mass of the system.
The first step in applying the laws of conservation of mass, con-
servation of momentum, conservation of energy, and the second law
of thermodynamics in their elementary forms is necessarily the defini-
tion of a system. Without this step i t is useless to speak of such things
as force, mass, heat, work, etc., for these terms are ambiguous until
the system is specifically and rigorously defined.
To begin an analysis in terms of a system is never incorrect, but some-
times it is an inconvenient starting point when dealing with fluids.
Fluids are extremely mobile, and it is therefore difficult to identify the
boundaries of a fluid system for any appreciable length of time. This is
particularly true in the interior of turbomachines, where complex
processes occur and where different particles of fluid passing through
the machine experience diierent histories. With fluids in motion i t is,
therefore, simpler to think in terms of a given volume of space through
which fluid flows than i t is to think in terms of a particular mass of fluid
of fixed identity.
The Control Volume. With this in mind, we define a control volume as
an arbitrary volume, fixed in space, and through which fluid flows. The
identity of the fluid occupying the control volume changes from instant
to instant. The surface which bounds the control volume is called the
control surjace; i t is always a closed surface, but may be either singly
or multiply connected. For some purposes, useful results are obtained
by considering infinitesimal control volumes; for others, by considering
control volumes of finite size.
To make practical use of the control volume concept, it is first neces-
sary to cast the four basic principles into a form where they apply to the
control volume rather than to the system. Then these forms may be
used as starting points for analysis. The reformulation of the physical
laws into control volume form is effected in Arts. 1.4, 1.5, 2.5, and 2.6
'1,
1.4. Conservation of Mass

The principle of conservation of mass, when referred to a system of
fixed identity, simply states that the mass of the system is constant.
This statement is a concise summary of experimental observation,
relativity and nuclear effects being of course absent.
Consider an arbitrary control volume through which fluid streams
(Fig. 1.5a). We wish to derive the form of the law of conservation of
mass as it applies to this control volume. However, in order to apply
Art. 1.4 CONSERVATION OF MASS 13
the law, we must begin with a system of fixed identity, and so we define
our system as the fluid which a t some instant t occupies the control
volume.
Next, we consider what happens during the succeeding time interval
dt. By definition, the control volume remains fixed in space, but the
(b)
Fro. 1.5. Flow through a control volume.
system moves in the general direction of the streamlines. The two

positions of the system are shown in Fig. 1.5a by dashed lines.
For convenience in analysis, we consider three regions of space,
denoted by I, 11, I11 in Fig. 1.5a. At time t the system occupies spaces
+
I and 111, and a t time t dt it occupies spaces I and 11. Thus, since
the mass of the system is conserved, we write
where ml, means the mass of fluid in space I a t time t, and so on. A
simple rearrangement then gives
The first term represents the time rate of change of mass within space
I. But, as dt goes to zero, space I coincides with the control volume, and
so, in the limit,
where m,.,, denotes the instantaneous mass within the control volume.
The third term may be written
where 6mII,+,, represents the amount of mass crossing the elementary

surface dAout during the time dt. The ratio GmII,,,/dt is called the out-
going jlux of mass across the area dAOut,or the mass rate of flow, and is
S
denoted for convenience by dwout. The integral sign in dwoutsignifies
that the mass rate of flow is to be summed up for all elements of control
surface area dAout through which fluid leaves the control volume.
Similar reasoning yields
and so the conservation law may now be expressed as
which states that the rate of accumulation of mass within the control
volume is equal to the excess of the incoming rate of flow over the out-
going rate of flow.
Continuity Equation. For detailed computations, we note that at any
instant
where & is an element of control volume, p is the local mass density

Art. 1.4 CONSERVATION OF MASS l5
of that element, and the integral is to be taken over the entire control
volume. Furthermore,
With the help of Fig. 1.5b, we may express the mass rate of flow in
where p is the local instantaneous mass density in the neighborhood of

dAout and V , is the corresponding local instantaneous component of
velocity normal to dAout.
With the foregoing expressions, Eq. 1.8 may now be written
a form which is usually called the epwction of continuity, since it is based

on the assumption of a continuous medium. In vector form i t is written
where the last integral is a surface integral summed up over the entire
control surface.
Steady Flow. When the flow is steady, the identity of the fluid within
the control volume changes continuously, but the total mass remains
constant. Or, mathematically, ap,Qt is zero for each element of control
volume. Then Eq. 1.9 states that the incoming and outgoing mass
rates of flow are identical:
It is interesting to observe that the same result applies to incompress-

ible, unsteady flows.
One-Dimensional Steady Flow. Consider the steady flow through a
duct or stream tube (Fig. 1.6) and
assume the flow to be one-dimen-
sional. Then since p and V are Control
Surface
constant over each cross section,
Eq. 1.11 yields
I PZVZAZ (1-12)
PIVIA=
where A~ represents the cross-sec- FIG.1.6. Onedimensional flow
tional area a t section 1, measured in a plane normal to the axis of the

tube. This equation is of great practical utility for simplified calcula-
tions.
1.5. Momentum Theorem
The fundamental principle of dynamics is Newton's second law of
motion. Following the approach used in Art. 1.4, we now derive the
form of this law appropriate to a control volume.
Fig. 1.7 shows the arbitrary control volume, together with the position
+
of the system a t times t and t dt, the system being defined as the mate-
FIG.1.7. Flow through a control volume.
rial occupying the control volume a t time t. For the sake of generality
there is also shown protruding into the control volume some sort of
obstacle, such as a strut, screen, flameholder, guide vane, or turbo-
machine blade. For the control volume chosen, that part of the obstacle
lying within the control volume is part of the system whose motion dur-
ing the time dt is being studied.
Newton's second law is a vector relation. Considering the x-direction,
we write, for the system
where the left-hand side represents the algebraic sum of the x-forces
Art. 1.5 MOMENTUM THEOREM 17
acting on the system during the time interval dt, and the right-hand side
represents the time rate of change of the total x-momentum of the
system. The right-hand side may, however, be expressed as
The combination of terms
represents the time rate of change of x-momentum within space I. In

the limit as dt goes to zero, space I coincides with the control volume,
and so the term in question becomes the time rate of change of x-momen-
tum within the control volume.
We now form the expression
and note that a similar expression applies to the incoming flow. The
term
S V, dw which appears here is called the flux of x-momentum.
Now, with the expressions developed above, the dynamic relation
may be written
In the limit, as dt vanishes, we note that Z F , becomes the algebraic

net x-force acting instantaneously on the fluid within the control volume.
Eq. 1.13 is usually called the momentum theorem and states that the
net force acting instantaneously on the fluid within the control volume
ie equal to the time rate of change of momentum within the control
volume plus the excess of outgoing momentum flux over incoming
momentum flux.
Working Form of Momentum Theorem. A useful working form of
the momentum theorem is obtained by observing that
where p and V, are the local density and x-velocity, respectively, of the
fluid element of volume d o within the control volume. Then, using the
expression for dw developed in Art. 1.4, the momentum theorem is
written
ZF, = J
c.v.
'
at
bvz)do +Jpvnvz dA0.t -JpvnvZ dAin (1-14)
Two similar relations apply to the y- and z-directions. By using vector
notation, the momentum theorem for a control volume may be repre-
sented by a single equation,
a
2F =-
at J,..PV dl) +$ dV.dA)V
C.8.
(1.15)
which is the fundamental principle in the dynamics of fluid motion.

Discussion of Forces. The term ZF in Eq. 1.15 (or XF, in Eq. 1.14
for the x-direction) is the algebraic sum of all forces exerted by the
surroundings on the material instantaneously occupying the control
volume. Here it is well to recall that forces may be divided into two
classes, (i) body forces, and (ii) surface forces.
BODY FORCES. Body forces are forces which are proportional either
to the volume or mass of the body, and comprise those forces involving
action a t a distance. Such forces arise from force fields, and include
such examples as the force of gravitational attraction, magnetic forces,
and electrodynamic forces. For accelerating coordinate systems, inertia
forces such as centrifugal and Coriolis forces would also be included.
SURFACE FORCES. Surface forces are those forces which are exerted a t
the control surface by the material outside the control volume on the
material inside the control volume. Such forces are exerted in the form
of surface stresses. Often it is useful to divide surface forces into two
types: (i) those arising from normal stresses, or pressures, acting on the
control surface, and (ii) those arising from shear stresses, or viscous
stresses, acting on the control surface. Where interfaces between phases
are involved, surface tensions also form a special type surface force,
but they are not of interest for our present purpose.
Referring to Fig. 1.7, the force acting on the surface of intersection
between the obstacle and control surface must be included in ZF, and
generally involves shear stresses and either tensile or compressive stresses
in the metal. However, the fluid pressures and shear stresses acting
on the part of the obstacle within the control volume do not enter into
YF because the surfaces on which they act do not form part of the con-
trol surface.
Moving Reference Systems. The acceleration term appearing in the
law of motion is the acceleration with reference to the fixed stars. In
most engineering problems, the acceleration of the earth's surface with
respect to the fixed stars is negligibly small. In practice, therefore, we
usually evaluate Newton's law with reference to the earth's surface as
a coordinate system.
If, on the other hand, we employ an accelerating coordinate system,
as for example the rotor of a turbomachine, Newton's law in its usual
Art. 1.6 THEOREM OF MOMENT OF MOMENTUM 19
form is not valid since there must be inserted additional terms to take
account of the centrifugal and Coriolis accelerations. Since the mo-
mentum theorem derives from the law of motion, a similar conclusion
applies, namely that in Eq. 1.15 the velocities must be measured with
respect to the earth, or with reference to a coordinate system moving
a t constant velocity with respect to the earth.
1.6. Theorem of Moment of Momentum

In dynamics useful information is often obtained by employing New-
ton's law of motion in the form where it applies to the moments of the
forces. This form of the law is especially valuable for the analysis of
turbomachines, where we may be more concerned with moments of
forces (or torques) than with the forces themselves.
For a system, the sum of the moments of the external forces is equal
to the time rate of change of moment of momentum. I n vector notation
this is written
d
ZFxr = -
at izmV r,
I(&
where r is the radius vector from an arbitrary origin of moments. In
Cartesian notation, and referring
Y
to Fig. 1.8, this is written
d
M, = ZrFt =- (ZmrVJ
at
where r is the radius in the xy-plane /
to the element of mass m, Ft is the x

component normal to r of the vec- l a ~ & l ~ ; ~ ~ , " ~ ~
Newton'8~ecdid
~ t i n g
tor force projected in the xy-plane,
V t is the component normal to r of the vector velocity projected iu tne
xy-plane, and M , is the net moment of forces about the z-axis.
Working Form of Theorem of Moment of Momentum. By following
the methods employed in Arts. 1.4 and 1.5 for converting frorn a system
to a control volume, the theorem of moment oj momentum may be derived
for a control volume. In vector form, it is written
I n Cartesian form, and considerirg moments about, say, the z-axis,

Eq. 1.16 is written as
20 FOUNDATIONS O F FLUID DYNAMICS Ch. 1
which states that the algebraic sum of the moments is equal to the time
rate of change of moment of momentum within the control volume plus
the excess of outgoing flux of moment of momentum over the correspond-
ing incoming flux.
The remarks in Art. 1.5 concerning forces are evidently applicable
also to the moments of the forces.
1.7. Units and Dimensions

It seems in order a t this point to explain the system of dimensions in
use throughout the book.
Newton's dynamic formulation is stated in terms of four completely
independent physical quantities: force, mass, length, and time. Before
the statement of the law of motion, the units in which these quantities
-
could be measured were totally unrelated, and so Newton could state
his law only as a proportionality, F ma, or
where go is an experimentally determined constant whose magnitude

depends only upon the units of measure of the four primary quantities.
However, once the proportionality law is known, i t is evidently pos-
sible to redefine one of the units of measure in terms of the other three,
thus reducing by one the number of primary quantities.
These ideas are illustrated by the following tabulation of three sys-
tems of units in engineering use
-
Mass Length Time
- Force
-
Ibm ft sec lbf 32.174 lbm f t2
lbf sec
Ibm ft
Ibm ft sec poundal 1
poundal sec2
I slug ft
slug ft sec Ibf
lbf sec2
The last system, which is the one employed in this book, is one where
go has the numerical magnitude unity, and thus one for which unit force
gives unit mass unit acceleration. For such a system of units, we may
eliminate go from the mathematical statement of the law, which is
tantamount to eliminating one of the primary units of measure and
defining i t instead in terms of the remaining three. Thus, writing as
we do,
F = m
Art. 1.7 PROBLEMS
Since we usually employ the slug a s mass unit for numerical calcula-
tions, but often refer t o results in terms of pound masses, it is convenient
t o remember t h a t
1 slug = 32.174 ibm
REFERENCES AND SELECTED BIBLIOGRAPHY

1. L A MB , H . Hydrodynamics. 6th ed. New York: Dover Press, 1945.
2. PRANDTL, L., and TIETJENS, 0. G. Fundamentals of Hydro- and Aeromechanies.
New York: McGraw-Hill Book Co., Inc., 1934.
3. HUNSAILER, J. C., and RIGHTMIRE, B. G. Engineering Applications of Fluid Me-
chanics. New York: McGraw-Hill Book Co., Inc., 1947.
PROBLEMS
1.1. Consider the frictionless, steady flow of a compressible fluid in an in-
finitesimal stream tube.
(a) Demonstrate by the continuity and momentum theorems that
- + - + dA
dp
- = o dV
P A V
dp+pVdV+pgdz= 0
(b) Determine the integrated forms of these equations for an incompressible
fluid.
v AtdA
P Gmvity 2 zrdz
P
1.2. An incompressible fluid flows in a pipe of radius R. At the inlet, sec-

tion 1, the velocity is uniform over the cross section, with a value VI. At sec-
tion 2, where the flow is laminar and fully developed, the velocity varies with
radius according to the relation
(a) Demonstrate that V/V,, = s.

(b) If 7, is the average wall shearing stress retarding the flow between sec-
-
tions 1 and 2, find the pressure drop ( p , pz) in terms of V, p, L, r and TW.
1.3. The sketch shows a jet pump (ejector or injector) in which a p r i m
stream of high velocity liquid at section 1 entrains a secondary stream of the
same liquid at low velocity at section 2. At the end of the constantdiameter
mixing-tube, i.e., a t section 3, the streams are thoroughly mixed and uniform
in velocity, as the result of friction between the streams.
For the purpose of this analysis, assume that a t sections 1 and 2 both streams
have the same static pressure and that shearing stresses a t the walls of the mixing
tube are negligible.
Assuming that A1 = 0.1 ft 2, A3 = 1 ft2, Vl = 100 ft/sec, V p = 10 ft/sec,
and p = 64.4 lbm/ft3
(a) Calculate V3 (ft/sec).
(b) Calculate ps - pl (lbf/in2).
I Conslont -Diameter I
WV
Nozzle -
1.4. The sketch shows a vane with a turning angle P which moves with a
steady speed U. The vane receives a jet which leaves a fixed nozzle with speed V.
(a) Assuming that the vane is mounted on rails as shown in the sketch, show
that the work done against the restraining force is a maximum when U/V = M.
(b) Assuming that there are a large number of such vanes attached to a rotat-
ing wheel moving with peripheral speed U, show that the work delivered to the
wheel is a maximum when U/V = x.
1.6. In an experiment to determine drag, a circular cylinder of diameter d
--
=-- "0
----- "0
C
---+ --
"0
-
was immersed in a steady, twodimen-
sional, incompressible flow. Measure-
ments of velocity and pressure were
a made at the boundaries of the control
avo B i d 4d
surface shown. The pressure was found
to be uniform over the entire control
t;
--
surface. The x-component of velocity a t
I ,
."I -- ----+ --
Control Surface
Vo "a
1
Vo
-- the control surface boundary was
approximately as indicated by the
sketch.
PROB. 1.5. From the measured data, calculate
the drag coefficient of the cylinder,
based on the projected area and on the free stream dynamic head, y$Vo2.
Drag Force per Unit Length
CD =
$pV&
Chapter 2
FOUNDATIONS OF THERMODYNAMICS

In Art. 1.1 i t was pointed out that the four fundamental principles
governing the motion of compressible fluids are the law of conservation
of mass, Newton's laws of motion, and the two laws of thermodynamics.
The basic concepts, definitions, and analytical statements relating to
the first two of these four principles were reviewed in Chapter 1, and we
now proceed to a similar brief review of the foundations of thermo-
dynamics. It is not the intent here to treat this subject with the fullness
and detail found in well-known thermodynamics textbooks,(~~2) but
rather to set out the general point of view, method of procedure, and
necessary relations employed in this book, insofar as thermodynamics
is concerned.
Much of the preliminary discussion contained in Arts. 1.1 and 1.2 is
relevant also to thermodynamics. The definition of a fluid, the concept
of a continuum, and the definition of a continuum property "at a point"
are implicit in classical thermodynamics. Likewise, the definitions of
the system and of the control volume given in Art. 1.3, together with
the remarks concerning their relative utility, remain applicable.
NOMENCLATURE
area 4' vector heat flux per unit area
specific heat a t constant pressure 2 rate of heat transfer
specific heat a t constant volume R gas constant
internal energy per unit mass 6i universal gas constant
internal energy of a system in s entropy per unit mass
general S entropy
local acceleration of gravity T absolute temperature
enthalpy per unit mass u same as U, except for a unit
ratio of specific heats mass
mass U that part of the internal energy
pressure of a system independent of
power motion, gravity, electricity,
heat capillarity, and magnetism
heat transfer per unit time per v specific volume
unit area in the x-direction 2, volume
23
24 FOUNDATIONS OF TITERMODYNAMICS Ch.2
NOMENCLATURE--Continued
V speed thermal effieicncy
w mass rate of flow X coefficient of thermal conduc-
W work; molecular weight tivity
z elevation above given datum p der~sity
2.2. The First Law of Thermodynamics

In this article we shall outline the structure of the First. Law as it
relates to a system of fixed identity, and in Art. 2.5 we shall develop the
appropriate form of the First Law for a control volume.
Defmitions and Concepts. STATE. The state of a system is its wn-
dition or con$guration, described in sufficient detail so that one state
may be distinguished from all other states.
PROPERTY. A property of a system is any observable characteristic of a
system. 'The properties we shall deal with are all mensurable in terms
of numbers and units of measurement and include such physical quan-
tities as location, speed, direction, pressure, density, etc. A listing of a
sufficient number of independent properties constitutes a complete
definition of the state of a system.
PROCESS. A process is a change of state and is described in part by the
series of states passed through by the system. Often, but not always,
some sort of interaction between the system and surroundings occurs
during a process; the specification of this interaction completes the
description of the process.
CYCLE. A cycle is a process wherein the initial and final states of the
system are identical. For this to be true it is only necessary that the
initial and final states have identical values of all respective properties.
EQUALITY OF TEMPERATURE. It is a common fact of experience that
when two bodies, one of which "feels hot" to the touch, the second of
which "feels cold," are brought together, some sort of interaction occurs,
as evidenced by the changes in properties which occur in both. After
some time, however, no further changes are observed, and we then say
that the two bodies are equal in temperature. Formally, we define
equality of temperature between two bodies by saying that this equality
exists when no interaction occurs upon the two bodies being brought
into communication with each other.
TEMPERATURE. The concept of inequality of temperature derives by
implication from the concept of equality of temperature. Furthermore,
Art. 2.2 THE FIRST LAW OF THERMODYNAMICS 25
since we distinguish inequality of temperat,ure by observing changes
of properties of the two bodies concerned, it follows that an arbitrary
scale of temperature may be defined in terms of some convenient p r o p
erty of a standard body called a thermometer. A simple and familiar
temperature scale, for example, is the length of a mercury thread in a
glass column, with numerical values established by a precisely defined,
but nonetheless arbitrary, system of markings on the glass.
THE ZEROTH LAW OF THERMODYNAMICS. Further thought concerning
the possible utility of a thermometer as defined above reveals that the
practical applications of thermometry hinge on the assumption that
two bodies respectively equal in temperature to a third body must also
be equal in temperature to each other. Fortunately, this assumption is
amply verified by innumerable experiments. It is sometimes called
the Zeroth Law of Thermodynamics.
EQUAGTEMPERATURE PROCESS. The definition of equality of tem-
perature is based on the cognizance of some sort of interaction between
a system and its surroundings which occurs when equality of tempera-
ture does not prevail. Later we shall identify this type of interaction
as heat. For the present, however, we observe only that if a process
occurs in which the system and surroundings are at each step equal in
temperature, then the particular type of interaction associated with
temperature inequality is absent from the process. This in no way
implies that the temperature of the system need be constant during
the process.
The use of constant-temperature baths and of evacuated jackets
(such as Dewar flasks) are common laboratory techniques for effecting
equal-temperature processes.
Work. The laws of thermodynamics deal with interactions between
a system and its surroundings as they respectively pass through equilib-
rium states. These interactions may be divided into two classes: (i)
work and (ii) heat interactions.
In mechanics work is defined as an effect produced by a system on
its surroundings when the system moves the surroundings in the direc-
tion of a force exerted by the system on its surroundings. The magnitude
of the effect is measured by the product of the distance moved and the
component of force in the direction of motion.
Thermodynamics deals with phenomena considerably more complex
than does mechanics, and so it is necessary to generalize the foregoing
definition of work.
RECOGNITION OF WORK DONE. We say that work is done by a system
on its surroundings if some other process can be found in which the sys-
tem passes through the same series of states as in the original process,
26 FOUNDATIONS OF THERMODYNAMICS Ch. 2
but in which the sole effect in the surroundings is the rise of a

weight.
This means that "motion of a force through a distar~ce"is not essential
to a work intei-action. Instead, i t is only necexsa.ry that there be some
alternative surroundings which might engage in the process in such a
way that the motion of a force through a distance would be the only
effect external to the system.
To illustrate, suppose a storage battery, which we consider as the
system, discharges while lighting an electric light bulb. If the bulb
were replaced by an electric motor having very large conductors and a
pulley on which is wound a string suspending a weight, then the storage
battery could pass through the same series of states with no net outside
effect except the rise of the weight. By the criterion for a work inter-
action, it follows that the storage battery did work in the original process.
MEASUREMENT OF WORK. The amount of work done by a system is
measured by the number of standard weights which may be raised from
one prescribed level to another in the alternative surroundings previously
used for the recognition of work.
For example, the amount of work in ft-lb is the same as the number of
pound masses which can be raised from a certain location on the earth
to an elevation one foot higher.
When the system does work on the surroundings, i t is conventional
to call the work positive.
WORK RECEIVED. Thus far we have spoken only of work done by a
system. To complete the definition of work, we say that if a system
does work on the surroundings, then the surroundings receive work
of the same amount from the system. To recognize work received
by a system, it is then necessary to interchange temporarily the system
and surroundings.
When work is received by a system, the magnitude of the work is
taken to be negative.
With these conventions it follows that if a work interaction occurs
between bodies A and B, then W A = -WB, where W A denotes the
work done by body A and WBdenotes the work done by body B. This
rule is analogous to Newton's third law for forces: action and reaction
Ire equal.
The First Law. The principal experimental foundations of the First
Law are the remarkable investigations of Joule,(3)who carried out proc-
esses from a given initial state to a given final state by a variety of
processes, some involving electrical work and others involving mechani-
cal work with various mechanical arrangements. These different
orocesses involving several different kinds of work were, moreover,
Art. 2.2 THE FIRST LAW OF THERMODYNAMICS 27
carried out on a variety of systems, comprising several pure substances,

mixtures, and substances undergoing chemical reaction. The nature
of the experiments were such that the processes were carried out with
substantial temperature equality between the system and surroundings.
These experiments revealed that, for a given system passing from a
given initial state to a given final state by different equal-temperature
processes, the amount of work done was the same for all such processes.
This result has such important practical implications that i t is called
the First Law of Thermodynamics, and may be stated formally as
follows:
The amount of work done by any system in going from one state to
another does not depend on the course of states passed through by the system
nor on the manner of work interaction, so long as the system and surround-
ings are equal in temperature at each step of the process.
INTERNAL ENERGY. From this law follows the important corollary

that the amount of work done during a n equal-temperature process depends
only on the end states of the process and not on the intermediate series of
states. Since the work in an equal-temperature process depends on the
end states, this suggests that we define a new property or state function,
the decrease in which represents the work during an equal-temperature
process, namely,
El - E2 (we.t.>l -+ z (2.1)
where E is the newly defined property, and ( W e J l - 2 denotes the work
done by the system during an equal-temperature process while going
from state 1 to state 2.
Thus, assuming that for a given system an arbitrary value of E is
assigned to some specified reference state, corresponding values of E
a t all other states may be found by measuring the work in equal-temper-
ature processes and applying Eq. 2.1.
The property E defined by Eq. 2.1 is usually called the energy or the
internal energy of the system.
Heat. When temperature equality does not prevail during a process,

some type of interaction occurs which is different in nature from the
work interaxtion, and the work done is not equal to the decrease in the
property E. The type of interaction owing to temperature inequality
is called a heat interaction, and the amount of heat is measured by the
difference between the work done during the actual process and the work
which would be done during an equal-temperature process between the
same end states. Thus, denoting the heat "received" by Q, and the
actual work done by W, we define
But is given by the decrease in the property El according to Eq.

2.1, and so we have
Q-E2-E1+W (2.3)
Although Eq. 2.3 is nothing but a definition, it is often called the
First Law of Thermodynamics. This is in fact an acceptable statement
if one adds that E is a property defined by Eq. 2.1.
In differential form for a small change of state, Eq. 2.3 is written
where the symbols 6Q and 6W represent small amounts of quantities

which are not properties, whereas d E denotes an increment of a property.
I t is a matter of experience that when 6Q as defined by Eq. 2.4 is
positive, the system is a t a lower temperature (on all conventional
thermometric scales) than the surroundings, and vice versa. This
observation is, in fact, connected with the Second Law of Thermo-
dynamics. When 6Q is positive we say that heat is "received by the
system," or that heat is "transferred to the system."
It may be demonstrated that when heat is defined by Eq. 2.3, the
heat for the system is equal in magnitude to and opposite in sign to the
heat for the surroundings. Thus, if bodies A and B exchange heat only
with each other, then QA= -QB, which, again, is like Newton's third
law of motion.
JOULE'S CONSTANT. Thermodynamics as we now know it was pre-
ceded by calorimetry, which was based on the caloric theory of heat.
We find in the literature independent definitions of work and heat (for
example, the ft-lb and the Btu) together with the Joulean "mechanical
equivalent of heat."
Within the framework of the First Law as stated here, however, and
entirely in accord with Joule's interpretation, we may regard the Btu
as a measure of the change in El and define it as the change in E of a
pound mass of water between a state a t 1 atm and 59.5"F and a state
a t 1 atm and 60.5"F. Taking this as the definition of the Btu, experi-
ments substantially like Joule's, but more precise, may be interpreted
according to Eq. 2.1 t o mean that
1 Btu = 778.2 f t Ibf
The First Law for a Cycle. Since E is a property, the net change in E
for a cyclic process is exactly zero, and so from Eq. 2.4 it follows that
which states that the algebraic net heat received by the system during a
Art. 2.3 rHE SECOND LAW OF THERMODYNAMICS 29
cycle is equal to the algebraic net work done by the system during the
cycle.
Adiabatic Process. It is a matter of experience that when barriers,
called heat insulators, are placed between a system and surroundings
a t different temperatures, the heat Q reckoned from Eq. 2.3 becomes
very small, and it is not difficult to extrapolate in one's mind to the
condition where the insulation is so effective as to make the heat zero.
Such a process is said to be adiabatic. Since Q is also zero in an equal-
temperature process, it follows that an equal-temperature process ir
also an adiabatic process.
2.3. The Second Law of Thermodynamics

In this Article we shall briefly outline the structure of the Second Law,
following closely the detailed development of Keenan.(')
Heat Engines. A heat engine is defined as a system of fixed identity
which undergoes a cyclic process during which there are work and heat
interactions with the surroundings. The most familiar example of a
heat engine is the steam power plant, and the closed-cycle gas turbine
is another. Neither open-cycle gas turbines nor internal combustion
engines, on the other hand, are heat engines according to this definition.
THERMAL EFFICIENCY OF HEAT ENGINES. That part of the surround-
ings from which the system receives heat during the cycle is called a
source of heat, and the amount of heat is denoted by ql. Similarly, q2 rep-
resents the amount of heat rejected by the system to the sink during the
cycle. The First Law for a cyclic process, Eq. 2.5, requires that the net
work of the engine during a complete cycle be given by
The useful output of an engine is the mechanical work, We,,, whereas

the factor most closely connected with the cost of operation is the heat
received, q l . Consequently, the thermal eficiency of an engine is taken
PERPETUAL MOTION OF THE SECOXD KIKD. A heat engine having a

thermal efficiency of 100 per cent is called a perpetual motion machine
of the second kind. I t may be demonstrated that, if such a machine could
be constructed, it would be possible to obtain mechanical power without
the use of any fuel whatsoever, the heat q, being taken either from the
atmosphere, or, better still, from some low-temperature body which it
is desired to refrigerate.
The Second Law of Thermodynamics. Experience teaches that per-
petual motion of the second kind, although greatly t o be desired, is
impossible to achieve. This statement has such important practical
implications that i t is called the Second Law of Thermodynamics, which
may now be stated as follows:
N o system can pass through a complete cycle of states and deliver positive
work to the surroundings while exchanging heat with only a single source
of heat at uniform temperature.
The confidence placed in this law derives not alone from the failure
of efforts to construct perpetual motion machines of the second kind,
but even more from innumerable experiments which serve only t o con-
firm the many corollaries and consequences which stem from the Second
Law.
Reversibility and Irreversibility. Our personal experiences that nat-
ural events tend to proceed in one direction only, that is, that real
processes are irreversible, are intimately connected with the Second
Law.
REVERSIBILITT.We say that a process is reversible if it is possible for
the effects of the process t o be entirely effaced-that is, if there exists
some known way by which the system and all parts of the surroundings
can be restored to their respective initial states. A reversible process
"does not make history," for it may be completely undone, and in this
sense it is equivalent to the absence of a process.
No real processes are reversible, but often real processes can be refined
to the point where they approximate reversible processes. For this
reason the reversible process is a useful standard of comparison against
which real processes may be evaluated.
IRREVERSIBILITY. The definition of an irreversible process is implicit
in the definition of a reversible process. That is, we say that a process is
irreversible if there is no known way by means of which the system
and all parts of the surroundings can be restored to their respective
initial states.
For the types of phenomena of interest in this book, the irreversibilities
can always be traced to three basic causes: (i) viscosity, (ii) heat con-
duction, and (iii) mass diffusion. Note that all these are phenomeno-
logical manifestations of molecular actions.
The connection between irreversibility and the Second Law stems
from the proposition that if a way were found to undo any irreversible
process it would then be possible to construct a perpetual motion machine
of the second kind.
For example, suppose that in Fig. 2.la the heat passes by conduction
through the metal bar from the high-tem~eraturereservoir TI to the
Art. 2.3 T H E SECOND LAW OF THERMODYNAMICS 31
low-temperature reservoir Tz. If this process could be undone, then

heat ql would be transferred to reservoir TI while the same amount of
heat was lost from reservoir T2. However, by operating a heat engine
as shown in Fig. 2.lb, the cyclically operating system enclosed by the
dashed lines becomes a perpetual motion machine of the second kind,
inasmuch as it delivers positive work We,, to the surroundings while
receiving heat ql - qz from the single uniform-temperature source T2.
Metal
(a) (b)
FIG. 2.1. Conduction of heat from low temperature to high temperature permits
perpetual motion of second kind.
Thus we have proved that our oft perceived observation of the irre-
versibility of heat conduction from a high-temperature to a low-tem-
perature body is in fact a logical consequence of the Second Law. The
proposition stated above may be similarly proved for irreversible
processes in general, and this suggests an alternative definition of
irreversibility :
A process is irreversible i f the undoing of all the efects of the process
would make possible the construction of a perpetual motion machine of the
second kind.
Corollaries of the Second Law. The following corollaries of the Second
Law, comprising the most important results for its practical utilization,
stem by logical processes from its statement. Again we follow closely
the development of Keenan.(')
COROLLARY 1. NO heat engine operating between two heat reservoirs
of fixed and uniform temperature can have a greater thermal efficiency
than a reversible heat engine which operates between the same two
reservoirs, for it can be shown that otherwise perpetual motion of the
second kind would be possible.
Here it is well to remark that a reversible engine is reversible in each

element of its cycle, so that the reversal of a reversible engine leads to
the reversal of the net cyclic work and heat interchanges with the sur-
roundings. Friction must be absent from reversible engines, and they
can exchange heat with the surroundings only when the temperature
difference between the working fluid and the surroundings is vanishingly
small.
COROLLARY 2. From Corollary 1 it follows that all reversible heat
engines operating between the same two reservoirs of fixed and uniform
temperature have the same thermal efficiency.
COROLLARY 3. THE ABSOLUTE T EMPERATURE SCALE. From Corol-
lary 2 it may be shown that a temperature scale defined in terms of the
efficiency of a reversible engine operating between two reservoirs of
fixed and uniform temperature depends only on the temperatures of
these reservoirs and not on the nature of the engine nor of the thermo-
metric fluid.
Thus, by using a reversible engine as a thermometer, it is possible to
avoid a basic difficulty of thermometers constructed as outlined in Art.
2.2-namely, that different thermometers agree only a t one or two fixed
points but generally differ a t all other points.
The Kelvin scale of absolute temperature is defined by
where T I is the temperature of the reservoir supplying the heat q1 to

the engine, and T2 is the temperature of the reservoir receiving the heat
q2 rejected from the engine.
The ice point (O°C, 32OF) has a temperature of 273.2 on the Kelvin,
or absolute centigrade scale; and a temperature of 491.7 on the Rankine,
or absolute Fahrenheit scale. In this book absolute temperatures will
be reported on the Rankine scale (e.g., 530°R). The Rankine tempera-
ture is obtained from the Fahrenheit temperature by adding 459.7,
often rounded off as 460, to the latter.
It follows further from Corollary 3 that it is impossible for systems
to have zero or negative absolute temperature.
COROLLARY 4. THE INEQUALITY OF CLAUSIUS. I t may be shown that
when a system passes through a complete cycle, perpetual motion of the
second kind is possible unless
where 6& is a small amount of heat received a t a part of the boundary

Art. 2.3 THE SECOND LAW OF THERMODYNAMICS 33
of the system during an elementary part of the cycle, and T is the corre-
sponding absolute temperature a t that part of the boundary. The
integral is to be summed for all parts of the boundary and over the
entire cycle.
COROLLARY 5. THE ENTROPY. From Corollary 4 i t may be shown

that, in a cycle made up of reversible steps,
From this it follows that in going from a given state 1 t o a given state 2
by two different processes, both of which are reversible, the integral of
6Q/T will be the same for both processes. More generally, we may say
that the
S
(6Q/T),,, depends only on the end states and not on the
intermediate series of states. The quantity 6Q/T for a reversible in-
finitesimal process is, therefore, an exact differential, and is the differ-
ential of a thermodynamic property. This property, which is called the
and is most useful for practical calculations.

Thus, just as the First Law led to the definition of a new property
(the internal energy), so the Second Law also leads to the definition of
a property-the entropy.
Although the entropy is defined in terms of reversible processes and
can only be reckoned from Eq. 2.11 with the aid of imaginary reversible
processes, this in no way diminishes its utility for irreversible processes,
inasmuch as it is a property having a particular value (with respect t o
some reference condition) for each equilibrium state of a system. It
will be recalled that a similar situation exists for the internal energy,
which although defined only for equal temperature processes, has mean-
ing nonetheless for processes which are not a t equal temperature.
For reversible processes the heat received by a system in passing from
state 1 to state 2 is given, according to Eq. 2.11, by
and this may be represented by the area under the curve representing
the process on a thermodynamic diagram of temperature versus entropy
(Fig. 2.2).
t Reversible
Adiabatic
Irreversible f'
Adiabatic
Irreversible
Adiabatic
FIG. 2.2. Reversible heat transfer on FIG. 2.3. Reversible and irreversible
temperature-entropy diagram. adiabatics.
COROLLARY 6. THE S ECOND LAW I N ANALYTICAL FORM. From Corol-

laries 4 and 5, together with the definition of the entropy, it may be
shown that, in general, for an infinitesimal process of any system,
where 6Q and T have the same meaning as in Eq. 2.9. This relation is
the most convenient analytical form of the Second Law.
When a system is isolated from all heat exchange with the surround-
ings, Eq. 2.13 shows that
(dS)adiabatic >
-0 (2.14)
which is the well known principle of the increase of entropy.
Since the entropy is defined by Eq. 2.11, it follows that the equality
sign necessarily applies in Eqs. 2.13 and 2.14 for reversible processes.
Furthermore, it may be shown that the inequality sign always applies
to irreversible processes.
Fig. 2.3 illustrates the paths of reversible and irreversible adiabatics
on the temperature-entropy diagram.
RELATION BETWEEN FIRST A N D S ECOND LAW PROPERTIES FOR A PURE

SUBSTANCE. A system whose chemical composition during a given proc-
ess is both homogeneous and invariable is called a pure substance. For
a pure substance defined in this way, and in the absence of electrical,
magnetic, and capillary forces, it may be shown that
where 2, is the volume of the system and U is that part of the internal
Art. 2.4 THERMODYNAMIC PROPERTIES OF CONTINUUM 35
energy E which is independent of the system's motion and of gravita-

tional, electrical, magnetic, and capillary forces.
Eq. 2.15 refers to changes in properties between equilibrium states of
a system and is, therefore, the most practical way of measuring changes
in entropy, since the change in U can be evaluated from measurements
of work in processes which are not necessarily reversible.
2.4. Thermodynamic Properties of the Continuum

The continuum properties density and pressure, defined in the pre-
ceding chapter, are important also in thermodynamics. In addition,
we have in this chapter introduced several additional properties more
uniquely related to thermodynamics, namely, the temperature, the
internal energy, and the entropy.
Regarding the latter two, it is convenient to work with the "specific"
values of the properties, i.e., the values per unit mass. Hence we define
where e is the specific internal energy, or internal energy per unit mass,
E is the internal energy for the entire system, m is the mass of the sys-
tem, and so on. Small letters denote values per unit mass, and capital
letters represent values for the entire system. The specific volume is
the volume per unit mass and is, therefore, the inverse of the density,
The values of such properties as e, s, v, etc., "at a point," or for a

fluid particle, are defined according to the train of thought illustrated
by Fig. 1.1.
The Internal Energy U for a Pure Substance. Capillary, electrical,
and magnetic forces are absent or negligible in most engineering prob-
lems involving fluids, but the system may undergo accelerations and may
be subject to gravitational forces. In addition, the syetern may undergo
other changes in state owing to heat transfers or to motions of the
boundary against external forces, these changes being made manifest
by changes in pressure, density, and temperature
Often i t is convenient to employ an expression for e which separates
the effects mentioned above. Since e is a property, it may be evaluated
for any chosen process betwccn a selected pair of end states. Let us
select a simple process comprising three steps: (i) an equal-temperature
change of speed of the system in which there are no changes in p, p, and
z, where z denotes elevation above the earth's surface, (ii) an equal-
temperature change of elevation of the system in which there are no
changes in p, p, and V, where V denotes the speed of the system, and

(iii) a process involving changes in p, p, and T, but no changes in z and V.
For the first step, it follows from Eq. 2.1 and Newton's second law
that e2 - el = (Vz2 - V12)/2, the right-hand side usually being called
the increase in kinetic energy.
For the second step, it follows again from Eq. 2.1 and Newton's
second law that e2 - el = g(zz - zl) where g is the local acceleration
of gravity. The right-hand side is usually called the increase in potential
energy.
For the third step, by definition, we set ea - el = u2 - ul, in which
the right-hand side is the change of internal energy of a pure substance
in the absence of changes in speed and of body and capillary forces.
For the entire process, therefore, we have
The Enthalpy. In most flow problems the terms u and pv appear as a

sum, and so i t is convenient to define a shorthand symbol for this sum,
called the enthalpy :
P
h=u+pv=u+- (2.17)
P
Specific Heats. Two "specific heats" are in common use, namely,
cp and c,. As will be seen from their definitions, the word "heat1' in
their titles is a misnomer carried over from caloric theory:
Coefficient of Thermal Conductivity. Fourier's law of heat conduc-

tion states that the heat flux per unit area in a given direction is propor-
tional to the temperature gradient in the same direction, the factor of
proportionality depending only on the units of measurement and on the
local thermodynamic state. The proportionality constant is called the
coejicient of t h e m 1 conductivity A, and is defined by
where q, is the heat transfer per unit time per unit area in the x-direction,
and aT/ax is the temperature gradient in the x-direction. The minus
sign is inserted as the heat always flows from high to low temperatures.
In vector language, A is defined by
where q is the vector heat flux Der unit area. and vT is the vector gradient
of the temperature.
Art. 2.5 THE.FIRST LAW FOR A CONTROL VOLUME 37
Units and Dimensions. We shall adopt as standard units the ft, sec,
Ibf, slug, and degree Rankine. The units of some typical thermo-
dynamic quantities are then as follows:
Q f t ibf
u f t lbf/slug
c, ft lbf/slug R
s ft lbf/slug R
A lbf/sec R
2.5. The First Law for a Control Volume

Just as it was desirable to put the law of conservation of mass and
Newton's second law of motion into forms applicable to the flow through
Eoundory of
System at
Time l + d l
FIG.2.4. Flow through control volume.
a control volume, so now it is desirable to do the same for the two laws
of thermodynamics. In this article we shall do so for the First Law,
following identically the pattern of procedure employed in Chapter 1.
Consider the flow through the control volume of Fig. 2.4, with the
system defined as the material occupying the control volume a t time t.
We consider what happens during the time interval dt. Passing through
the control surface are a stationary strut and a rotating shaft attached
to a turbomachine, perhaps a compressor or turbine.
Eq. 2.4 may here be written, for the system, as
in terms of the rate of heat transfer, etc. Now we put the three terms
of this expression into more convenient form.
Rate of Heat Transfer. If 2 is the rate of heat transfer (ft lbf/sec)
through the control surface from the surroundings, then, as the interval
dt becomes vanishingly small, we may write
Rate of Change of E. Referring to Fig. 2.4, we write
Since space I coincides with the control volume as dt vanishes, this

becomes
dE
-dt = ( )
C.V. + J e
- J e dwi,,
Rate of Work Done. Omitting from our considerations capillary,

magnetic, and electrical forces, the work done during the process is the
result of normal and shear stresses a t the moving boundaries of the
system.
WORKDONE BY NORMAL STRESSES. Taking the normal stress a t the
boundary of the system as the hydrostatic pressure, the work done by the
system owing to normal forces at an element of area dAout is p dAoUtdx,
where dx is the component of distance moved normal to dAOut. But
dAmt dx is the volume of the mass element SmI1,+,,, which volume may
be written as v 6mII,,d,. The total rate of work done by normal stresses
during the process may now be set down, with the aid of the foregoing, as
\VORK D ONE BY SHEAR STRESSES. This work may be conveniently

cli;.ided into two categories: (i) the work done by the part of the shaft
inside the system on the part outside the system, owing to the torque in
Art. 2.5 THE FIRST LAW FOR A CONTROL VOLUME 39
the rotating shaft resulting from the shear stresses in the plane cut by
the boundary of the system; (ii) the shear work done a t the boundaries of
the system on adjacent fluid which is in motion. These are respectively
denoted by and @&ear, the power delivered by means of a rotating
shaft and the power delivered by other shear forces. Note that shearing
forces a t the strut do not contribute to @shear,since the strut is not in
motion. Likewise, if the control surface coincides with the stationary
wall of a duct or machine casing, @shear is zero a t such points even though
there may be appreciable shear stresses. On the other hand, P8he,
plays an important role for a control surface lying within the viscous
boundary layer.
In the limit, as dt vanishes, Pahaft and @,hear of course refer t o the
material instantaneously occupying the control volume.
The First Law. Combining the expressions developed above, and
wing the expression for e given by Eq. 2.16 and the expression for h
given by Eq. 2.17, we finally obtain the following expression applicable
to the control volume:
where the integrals involve such terms as the "flux of enthalpy," "flux
of kinetic energy," etc.
We may rewrite Eq. 2.20 somewhat more compactly with the help of
vector notation:
Steady-Flow Energy Equation. I n many practical engineering prob-

lems, the flow is steady and may also be re-
garded as one-dimensional. If the control sur- twx
face also coincides with the inner casing of a
machine, Pahear is zero. Eq. 2.21 is then sim-
plified greatly. Let section 1 represent the
incoming flow through the control surface Wrr
(Fig. 2.5), and section 2 the outgoing flow.
The mass flow w is then alike a t both sections, v,
and so i t is convenient to define Q = 2/w as pd pt
p, P.
the heat received per unit mass of fluid passing s,
through, and W , = PShaft/was the shaft work F I ~ 2.5.
. One-dimen-
delivered per unit mass of fluid passing through. si0nal flow.
40 FOUNDATIONS OF THEXMODYNAMICS Ch. 2
With these assumptions, Eq. 2.21 is reduced to the form
which is the familiar form found in most textbooks on engineering

thermodynamics.
Nonequilibrium States. Thermodynamics deals with equilibrium

states of a system, so that in this respect the science would be more
appropriately named 'Lthermostatics.'l How, then, can we apply the
usual methods of thermodynamics to flow problems where fluid particles
are undergoing changes of state a t a finite rate, and so cannot be passing
through equilibrium states? Investigations of this point reveal that in
most practical problems, even when supersonic speeds are involved,
the departure from thermodynamic equilibrium is negligibly small, and
we are, therefore, justified in proceeding in the present manner.
On the other hand, i t should be realized that there are a few problems,
notably the flow through a shock wave, wherein the departure from
thermodynamic equilibrium is very great.
2.6. The Second Law for a Control Volume

Employing the same control volume, the same system, and the same
process as in Fig. 2.4, Eq. 2.13 may be put into either of the following
forms applicable to the flow through a control volume:
where d Z is the incoming rate of heat transfer through an elementary

area d A of the control surface, and T is the corresponding temperature
of the control surface a t that point.
For steady, one-dimensional flow (Fig. 2.5), Eq. 2.23 reduces to
and, if heat transfers across the control surface are negligible, there
results
82 >
sl (2.25)
Art. 2.7 THE PERFECT GAS
2.7. The Perfect Gas

For most problems in gas dynamics, the assumption of the perfect
gas laws is sufficiently in accord with the properties of real gases as to be
acceptable. We shall therefore set down here for later reference the
special thermodynamic relations which apply to a perfect gas.
Equation of State. The definition of a perfect gas is in two parts, the
first of which is the pressure-density-temperature relation, or equation
of state, given by
where T is the absolute temperature, R is the gas constant of the particu-

lar gas, is the universal gas constant, and W is the molecular weight.
In the usual engineering units,
ft lbf f t lbf
rSI = 1545.4 = (1545.4)(32.174)
R Ibm-mole R slug-mole
Real gases appear to obey Eq. 2.26 exactly a t zero pressure or a t
infinite temperature. For temperatures not low compared with the
critical temperature, Eq. 2.26 is an accurate approximation to real gases
provided that the pressure is low compared with the critical pressure.
Internal Energy and Enthalpy Changes. From well-known thermo-
dynamic relations i t may be shown that for any pure substance obeying
Eq. 2.26 i t is also true that
which is equivalent to saying that the internal energy of a perfect gas

does not depend on density, and hence depends on temperature alone.
Since we are dealing with a pure substance, we may write with the
aid of the foregoing result,
so that the change in internal energy between any pair of end states is
given by
From the definition of the enthalpy it follows that h is also a function

of temperature only, inasmuch as both u and pv are functions only of
temperature. Hence
(ah/ap)~ =0
Gas Constants. When a substance obeys the perfect gas law there is
a simple relation between c,, c,, and R. From the definitions of c,
and h, the perfect gas law, and the knowledge that h depends on T alone,
we form
dh d du d
- - (u 4- pv) = - + - RT = c, + R
dT dT
(2.29)
The ratio of specific heats, k, plays a n important role in isentropic

processes. It is defined by
Of the four quantities c,, c,, R, and k, only two are algebraically
independent. With Eqs. 2.29 and 2.30, we may derive, for example,
k R
c, = -R ; c,=-- , etc.
k-1 k-1
For the most simple molecular model, the kinetic theory of gases shows
that
k=-n + 2
n
where n is the number of degrees of freedom of the molecule. Thus,
monatomic gases have n = 3, k = x. Diatomic gases such as oxygen,
nitrogen, etc., have n = 5 , k = 76. Extremely complex molecules,
such as freon and gaseous compounds of uranium have large values of
n and consequently values of k only slightly larger than unity.
Gas with Constant Specific Heats. The second part of the definition
of a perfect gas is that c,, which has already been shown to be dependent
on T only, is assumed to be constant. This restriction is not so good
an approximation to real gases as is the equation of state, and so some-
times we speak of a "semi-perfect gas," that is, one for which pv = RT
but having specific heats variable with temperature.
Art. 2.7 THE PERFECT GAS 43
Considering now a perfect gas, i t follows from Eq. 2.29 that .cp is also
a constant, and then Eqs. 2.27 and 2.28 yield
Changes of Entropy. Applying the special relations of a perfect gas

to the general relation between s, u, and v, we get
du pdv dT dv
&=--+-- -c,-+R-
T T T v
and, upon integration,
Alternatively, we may eliminate either T or v from this expression with

the aid of pv = R T , and so obtain
allowing the entropy change to be simply calculated in terms of the

end properties.
The Isentropic. Often the isentropic process is taken as a model or as
a limit for real adiabatic processes. If the entropy is constant at each
step of the process, it follows from Eqs. 2.32 that T and v, p and v, and
T and p are connected with each other during the orocess by the follow-
ing laws:
Tvk-' = const.
pvk = p l p k = const.
k
T R / p = const.
For isentropic flow processes the enthalpy change is important. I t
is calculated in terms of the initial temperature and the pressure ratio
as follows:
FOUNDATIONS OF THERMODYNAMICS Ch.2
REFERENCES AND SELECTED BIBLIOGRAPHI

1. KEENAN, J. H. Thermodynamics. New York: John Wiley & Sons, Inc., 1941.
2. ZEMANSKY, M. W. Heat and Thermodl~namies. New York: McGraw-Hill Book
Co., Inc., 1937.
3. JOULE, J. P. The ScientiJicPapers of James Prescott Joule. (Phys-Soc. of London.)
London: Taylor & Francis, 1884.
Chapter 3
INTRODUCTORY CONCEPTS TO COMPRESSIBLE FLOW
3.1. The Velocity of Sound

The term compressible flow implies variations in density throughout
the field of flow. These variations are, in many cases, the result princi-
pally of pressure changes from one point to a.nother. The rate of change
of density with respect to pressure is, therefore, an important parameter
in the analysis of compressible flow, and, as we shall see, it is closely
connected with the velocity of propagation of small pressure disturb-
ances, i.e., with the velocity of sound.
NOMENCLATURE
mean molecular velocity T absolute temperature
area V velocity
velocity of sound w mass rate of flow
specific heat at constant pressure W molecular weight
ratio of specific heats
mean free molecular path a Mach angle
length P bulk modulus of compression
Mach Number 6 boundary layer thickness
pressure X coefficientof thermal conductiv-
gas constant ity
universal gas constant P coefficient of viscosity
entropy per unit mass P density
Velocity of Propagation of a Plane Pressure Pulse. Let us calculate

the velocity of sound for a plane, infinitesimal pressure wave proceeding
along a pipe of uniform cross section. Such a wave might have been
initiated, for example, by a slight inward motion of a piston a t the left-
hand end of the pipe.
In Fig. 3.la the wave front is assumed t o propagate steadily to the
right with a velocity c . The fluid through which the wave front has
passed is a t a pressure p + d p , has a density p +
dp, and moves to the
right with a velocity d V . The fluid on the right, into which the wave
front is moving, has a pressure p and a density p, and is motionless.
To simplify the analysis, we reduce the process to one of steady motion
by imagining that the observer travels with the steady speed of the wave
45
46 INTRODUCTORY CONCEPTS T O COMPRESSIBLE FLOW Ch. 3
front. Fig. 3.lb shows the appearance of the process to such an obsemer
who is moving to the right with the velocity c. Fluid flows steadily
from right to left, and, as i t passes through the wave front, the velocity
is reduced from c to ( c - d V ) . At the same time, the pressure rises
+
from p to p d p , and the density from
(0) (b)
Wove Front Moving Stotionory Control
With Velocity c Wove Front
- dV
Stotionory C
FIG.3.1. Propagation of an infinitesimal pressure pulse.

(a) Observer a t rest.
(b) Observer moving with wave front, which is equivalent to superposing a left-
ward velocity c on the flow of part (a).
MOMENTUM EQUATION. Considering a stationary control surface

surrounding the stationary wave front of Fig. 3.1b1 the shear forces on
this control volume are negligibly small compared with the pressure
forces. The momentum equation may be written, +.herefore,as
where A is the cross-sectional area. Simplifying, and noting that from

the continuity equation,
w = pAc
we get
d p = pcdv (3.1)
This equation, it might be noted, is Euler's equation for a steady motion
(note that d V is the decrease in velocity in the direction of motion).
CONTINUITY EQUATION. The continuity equation is then written for
the fluid on both d e s of the wave front. Noting that the area is un-
changed, we get,
PC = ( P +
dp)(c - d V )
uhich reduces tc
Art. 3.1 THE VELOCITY OF SOUND 47
Combination of Eq. 3.1 with Eq. 3.2 then yields
The ratio dp/dp is written in Eq. 3.3 as a partial derivative a t constant

entropy because the variations in pressure and temperature are vanish-
ingly small, and consequently, the process is nearly reversible. More-
over, the comparative rapidity of the process, together with the small-
ness of the temperature variations, makes the process nearly adiabatic.
I n the limit, for an infinitesimal wave, the process may be considered
both reversible and adiabatic, and, therefore, isentropic.
Identical results for the velocity of propagation of a small disturbance
are obtained for a cylindrical wave spreading from a line source and for
a spherical wave spreading from a point source.
Velocity of Sound in a Perfect Gas. For a perfect gas the relation
between pressure and density in an isentropic process is given by
p/pk = constant (3.4)
Putting this into logarithmic form, differentiating, and noting that
p = pRT for a perfect gas, we obtain
In p - k In p = constant
Thus, we get for the velocity of sound in a perfect gas,
where W is the molecular weight.

EFFECT OF MOLECULAR WEIGHT. Since k varies only between narrow
limits, we see that, in general, gases with small molecular weights have
large sound velocities and vice versa. At normal atmospheric tempera-
ture, the speed of sound in air is of the order of 1100 ft/sec; in hydrogen,
of the order of 4200 ftjsec; and in freon refrigerating gases or uranium
hexafluoride, of the order of 300 ft/sec.
An interesting conclusion concerning turbomachinery can be drawn
from these figures. To avoid excessive stresses, the tip speeds of rotors
in turbomachines must not exceed a figure in the neighborhood of 1000
to 1500 ft/sec. It is also a matter of experience that losses in efficiency
mount rapidly as the flow velocity relative to the blades approaches the
sonic velocity. Thus, in air compressors, the limiting design factor on
rotative speed may be either stress considerations or compressibility con-
siderations. In hydrogen compressors, on the other hand, the compress-
48 INTRODUCTORY CONCEPTS TO COMPRESSIBLE FLOW Ch. 3
ibility limitation is never a factor. At the opposite end of the scale,

compressibility effects represent the principal limiting factor on the rota-
tive speed for freon and uranium hexafluoride compressors.
Velocity of Sound in Air. In the case of air at normal pressure and
temperature, Eq. 3.5 becomes
c = 49.02fi (3.6)
where c is in ft/sec and T is in degrees R, or degrees F abs.
VELOCITY OF SHOCK WAVE. A wave of finite strength, usually called
a shock wave, always propagates with a greater velocity than that indi-
cated in Eq. 3.3, the velocity of propagation increasing with the wave
strength. Shock waves are discussed in Chapter 5 and in Volume 11,
Chapter 25.
Velocity of Sound in Incompressible Fluid. Since an incompressible
fluid is one which, by definition, cannot experience changes in density,
Eq. 3.3 shows that the speed of sound is infinite in such a fluid. Pressure
pulses emitted anywhere in the fluid are thus felt simultaneously a t all
other points.
No fluids are truly incompressible, although liquids show little change
in density. The speed of sound in common liquids is of the order of
5000 ft/sec, which is enormously in excess of fluid velocities which can
be produced in a liquid.
Often we speak of the "incompressible flow of a gas," in the sense that
the fractional changes in density are so small as to be negligible.
Realms of Fluid Motion. Some of the different fields of fluid motion
may now be classified as follows:
ACOUSTICS.The fluid velocities are extremely small compared with
the velocity of sound, but the fractional variations in pressure, tempera-
ture, and density are of significant magnitude.
INCOMPRESSIBLE FLUID MECHANICS. The fluid velocities are small
compared with the velocity of sound, and the fractional variation in
density is insignificant; however, the fractional variations in pressure
and temperature may be very large. It will be shown later that the
error produced by neglecting compressibility in the computation of pres-
sure variations is of the order of one-fourth the square of the ratio of
the stream velocity to the sound velocity. Consequently, this ratio may
be as great as 0.2 (corresponding to a velocity of about 200 ft/sec for
air a t normal atmospheric temperature) before the computed error in
the pressure variation exceeds one per cent. For many problems in the
flow of gases, therefore, the flow may with little error be treated as
incompressible.
Art. 3.2 INCOMPRESSIBLE, SUBSONIC, AND SUPERSONIC FLOWS 49
COWPRESSIBLE FLUID MOTION. The velocities are appreciable com-

ge,rd with the velocity of sound, and the fractional variations in pres-
sure, temperature, and density are all of significant magnitude. This
realm, often called Gas Dynamics, is the principal subject of this book.
The science of Ballistics is a special branch of Gas Dynamics, pertain-
ing to the free flight of projectiles a t velocities greater than that of sound.
Although the term Gas Dynamics is used, it is well to remember that
the science of thermodynamics plays a role equally important to that of
the science of dynamics in the motion of compressible fluids.
3.2. Physical Differences Between Incompressible,

Subsonic, and Supersonic Flows
The influence of compressibility of the medium can lead to marked

qualitative differences between the physical natures of incompressible
and compressible flows. And, in compressible flows, there are aston-
ishing differences in flow pattern and general behavior, depending on
whether the fluid velocity is greater or less than the local speed of
sound. Indeed, it is safe to say that one's "instincts" or past experience
with subsonic flows are completely useless when dealing with supersonic
flows and are apt to lead to false conclusions.
Pressure Field Created by a Moving Point Disturbance. To illustrate
this in a simple way, i t is useful to adopt momentarily a point of view
which is somewhat oversimplified yet a t the same time brings out the
salient physical phenomena. When a body moves through a fluid, or
when a fluid flows past a body or within the walls of a duct, each element
of solid surface tends to divert the fluid from the course which i t might
otherwise take. For example, in the case of a projectile moving through
air, each element of the projectile's surface area pushes the neighboring
air out of the way, and this local disturbance creates a pressure pulse
which propagates into the exterior air, much as the impulsive motion of
the piston in Fig. 3.1 results in the propagation of a pressure wave down
the duct. At any instant of time, therefore, the pressure perturbation
a t a point away from the projectile will be some sort of sum (not neces-
sarily linear) of those pressure pulses which were initiated earlier a t each
element of projectile surface and which have arrived a t the point in ques-
tion after spreading spherically from the individual points of emission.
Following this line of thought, let us consider the pressure field cre-
ated by the most elementary type of a moving disturbance-namely, a
point source of disturbance moving a t uniform linear speed through a
compressible medium. At each instant of time the point source may be
imagined to emit an infinitesimal pressure wave which spreads spheri-
cally from the point of emission with the speed of sound relative to the
fluid. The pressure pattern which exists at any instant is then found by
superposition of all the pressure pulses which were previously emittod.
Fig. 3.2 shows several pressure pulse patterns for different values of
che speed of the source compared with the speed of sound in the fluid.
F I ~3.2.
. Pressure field produced by a point source of disturbance moving a t uni-
farm speed leftwards.
(a) Incompressible fluid (V/c = 0).
(b) Subsonic motion (V/c = x).
(c) Transonic motion (V/c = 1).
(d) Supersonic motion, illustrating Karman's three rules of supersonic flow
(Vlc = 2).
In each pattern the point 0 represents the present location of the dis-
turbance, the point - 1 represents the location one unit of time pre-
viously, and so on. For each of these previous locations there is drawn
a concentric circle showing the distance to which the corresponding wave
has spread. For example, to find the present location of the wave which
was emitted a t time -3, a circle is drawn with - 3 as a center and with
a radius 3 4 where t is the unit of time. The distance between point -3
and point 0 is then given by 3Vt where V is the velocity of the point
disturbance with respect t o the medium.
INCOMPRESSIBLE FLOW. When the medium is incompressible (Fig.

3.2a), or when the speed of the moving point disturbance is small com-
pared with the speed of sound, the pressure pulse spreads uniformly in
all directions.
SUBSONIC FLOW. When the source moves at subsonic speeds, Fig.
3.2b, the pressure disturbance is felt in all directions and a t all points
in space (neglecting dissipation due to viscosity), but the pressure pat-
tern i,s no longer symmetrical.
SUPERSONIC F LOW. For supersonic speeds Fig. 3.2d indicates that
the phenomena are entirely different from those a t subsonic speeds. All
the pressure disturbances are included in a cone which has the point
source a t its apex, and the effect of the disturbance is not felt upstream
of the source of disturbance. The cone within which the disturbances
are confined is called the Mach cone. Fig. 3 . 2 shows
~ the pressure pat-
tern a t the boundary between subsonic and supersonic flow, that is, for
the case where the stream velocity is identical with the sonic velocity;
here the wave front is a plane.
Karman's Rules of Supersonic Flow. Fig. 3.2d illustrates the three
rules of supersonic flow proposed by von Karman.'') These rules apply
exactly only for small disturbances, but are usually qualitatively appli-
cable for large disturbances.
T HE RULE OF FORBIDDEN SIGNALS. The effect of pressure changes
produced by a body moving a t a speed faster than sound cannot reach
points ahead of the body.
T HE ZONE OF ACTION AND THE ZONE OF SILENCE. A stationary point
source in a supersonic stream produces effects only on points that lie on
or inside the Mach cone extending downstream from the point source.
Conversely, the pressure and velocity a t an arbitrary point of the stream
can be influenced only by disturbances acting a t points that lie on or
inside a cone extending upstream from the point considered and having
the same vertex angle as the Mach cone.
T HE R ULE OF CONCENTRATED ACTION. The proximity of the circles
representing the different pressure impulses in Fig. 3.2 is a measure of
the intensity of the pressure disturbance a t each point in the field of
flow. Thus, for the stationary source, the intensity of the disturbance
is symmetrical about the source. For the subsonic source, the intensity
is unsymmetrical. In the case of the supersonic source, we have the
rule of concentrated action: the pressure disturbance is largely concen-
trated in the neighborhood of the Mach cone that forms the outer limit
of the zone of action.
52 INTRODUCTORY CONCEPTS TO COMPRESSIBLE PLOW Ch. 3
These rules explain why a projectile or rocket moving a t supersonic

speed cannot be heard until the wave attached to the nose of the body
FIG.3.3. Shadowgraph of supersonic bullet passing through cylindrical tube, show-

ing Mach cone and spherically spreading wave fronts (after Aclceret).
(a) Single tube.
(b) Several short tubes in series, showing individual wavelets which passed
through the open space between the tubes.
passes over the ear of the observer; and why, when the latter does occur,
the noise is concentrated in a "crack."
Configurations like those of Fig. 3.2 may easily be observed in the
form of gravity waves on a free water surface when a sharp-pointed
-
object is drawn through the water a t varvine
speeds. The bow wave of a surface ship re-
" -
sembles the Mach cone of Fig. 3.2d.
Patterns like those of Fig. 3.2 can be made
visible in gas flows by means of shadow,
schlieren, or interferometer techniques, which
are optical methods for demonstrating density
variations in a gas. Fig. 3.3a, for example, is
a shadowgraph of a bullet which has just
passed through a cylindrical tube while travel-
ing a t supersonic speed. Fig. 3.3b is a schlie-
ren photograph of a similar arrangement, ex-
cept that slots are cut in the tube to allow
F I G. 3.4. Schlieren pho- only a selected number of wavelets to pass out
tograph of bullet traveling
at near the speed of sound of the tube into the field of view.
(after Aclceret). Fig. 3.4 shows a bullet traveling a t nearly
the speed of sound and having a wave front
similar to that of Fig. 3 . 2 ~ . At low subsonic speeds, corresponding to
Fig. 3.2b, no wave front appears.
(b)
FIG.3.5. Cone-cylinder projectile traveling a t supersonic speed (Ballistic Research
Laboratory, Aberdeen).
(a) Schlieren photograph.
(b) Interferometer photograph.
The Mach cones produced in a supersonic wind tunnel by the nose,

fins, and tail of a model of a guided missile are shown in Fig. 3.6.
FIG. 3.6. Schlieren photograph of a model of a guided missile in a supersonic wind

tunnel. Mach cones from nose, fins, and tail are visible (NACA).
3.3. The Mach Number and Mach Angle

In the preceding article it was shown that the nature of the flow pat-
tern depends on the comparative magnitudes of the stream velocity and
the sonic velocity. The ratio of these two velocities is called the Mach
Number. Thus,
The speed of sound in Eq. 3.7 is to be taken a t the local temperature

and pressure of the stream, and, of course, it varies from point to point
in the flow field.
The semi-angle of the Mach cone (Fig. 3.2d) is related to the Mach
Number by
sin a = 1/M (3.8)
Note that the Mach angle is imaginary for subsonic flow.
We have found the Mach Number to be a criterion of the type of
flow pattern. Later i t will be shown that it is also a parameter which
almost always appears in the equations of motion. I n the next article
we shall see that it is a dimensionless parameter important for model
testing.
3.4. Similarity Parameters
Consider a prototype flow pattern and a model which is geometrically
similar (as regards solid boundaries) to the prototype. We inquire as to
the conditions which must be met in order that the flow pattern for the
model be similar to that for the prototype.
Art. 3.4 SIMILARITY PARAMETERS 55
Let L be a characteristic length for the prototype, and let jrJ. be the
corresponding characteristic length for the model, where fL is the scale
factor for length. Similarly, we can summarize the other properties of
the prototype flow and model flow as follows:
Prototype Model
fvV
fop
~ P P
~ L L
fcc
fMM
fkk
Ignoring viscosity for the present, Euler's equation of motion must be

satisfied for both the prototype and model. Thus:
Since the scale factors are constant, both of these equations can be satis-
fied only if
f p = fPfv2 (3.9a)
From Eq. 3.5 we obtain in similar fashion
c = dkp/p
and
so that
fcc = 4iix s
fc = m
Likewise, from Eq. 3.7, we have:
and
so that
Combining Eqs. 3.9a, 3.9b, and 3.9c, we finally get:

By examining the energy equation, taking into account the further

effects of viscous work and heat conductivity, it may be shown that the
ratio of specific heats, k , must be the same for both model and proto-
type, i.e., fk = 1. Thus, we conclude that
or the Mach Number must be the same for the model and prototype if
the flows are to be similar.
When viscosity is present, a similar analysis applied to the inertia and
viscous terms in the Navier-Stokes equation leads to the criterion that
the Reynolds Number must be the same for similarity in flow patterns.
Dimensionless Groups Governing Compressible Flows. By consider-
ing all the physical equations which govern the flow, namely, the Navier-
Stokes equation, the continuity equation, the energy equation, and the
equation of state, i t is possible to arrive a t four dimensionless param-
eters which must be the same in order for two flow patterns to be sim-
ilar. These are
(i) The Mach Number
(ii) The Reynolds Number
(iii) The ratio of specific heats, k
(iv) The Prandtl Number, cPw/X,where c, denotes the specific heat
a t constant pressure, rc, the coefficient of viscosity, and X the
thermal conductivity.
POTENTIAL FLOW OUTSIDE BOUNDARY LAYER. For those regions of
the flow outside the boundary layer, where viscous effects and heat con-
duction effects are relatively unimportant, it is usually necessary that
only M and k be alike in order to have similarity. Of the two, similarity
in M is by far the more important, sincek has a relatively weak influence
on the flow pattern.
BOUNDARY-LAYER FLOW. In the boundary layer or in the interior of
shock waves, viscous and heat conduction effects are all-important, and
so the Reynolds Number and Prandtl Number must be included in the
similarity conditions. Fortunately, the Prandtl Number is nearly the
same for all gases and varies only slowly with temperature.
3.5. Domain of the Continuum
The concept of the continuum has already been discussed in Chapter

1, where it was pointed out that the criterion for continuum mechanics
is that the mean free molecular path be small compared with the small-
est significant body dimension.
Art. 3.5 DOMAIN OF THE CONTINUUM 57
In order to determine when this condition is valid, let us consider a

steady flow and make some approximate calculations of an order-of-
magnitude nature. ,-
Now kinetic theory (z)shows that, in respect to orders of magnitude,
and
where a is the mean molecular velocity and I is the mean free molecular
path.
Using the foregoing relations, we now express the Reynolds Number
where L is a characteristic body dimension on which Rey is based. This

relation may be rearranged to give
L -
1- M
Rey
thus showing that the ratio of Reynolds Number to Mach Number is a

dimensionless parameter indicative of whether or not a given problem
i s amenable to the continuum hypothesis. It is seen that the contin-
uum concept is likely to fail either a t very high Mach Numbers or a t
extremely low Reynolds Numbers.
Rules for determining the validity of the continuum concept in terms
of Rey and M cannot be stated generally since they depend on the par-
ticular type of problem. However, we may illustrate the approach to
this problem by supposing that, in a given problem, the smallest sig-
nificant body dimension is of the order of the boundary-layer thickness,
6. If the Reynolds Number is large compared with unity, and if the
boundary-layer flow is also laminar, then the boundary-layer relations
for a flat plate show that
6 1
which, when combined with Eq. 3.10, yields
For this case Tsien (3) suggests that the realm of continuum gas dynam-
ics be limited to instances where the boundary-layer thickness is a t least
100 times the mean free path. That is,
Fig. 3.7 shows the Reynolds Number per unit length as a function of
flight Mach Number for various altitudes, based on the standard atmos-
0 1.0 2.0 3.0 4.0

Mach number,M
FIG.3.7. Reynolds Number per foot versus Mach Number, based on standard at-
mosphere (NACA Tech. Note, No. 1428).
3.6. Classification of Compressible Flows

When dealing with problems of steady motion, it is sometimes con-
venient to subdivide the subsonic and supersonic categories in classify-
ing different types of flow.
The Adiabatic, Steady-Flow Ellipse. Consider a stream tube in which
the flow does not exchange heat with the fluid in neighboring stream
tubes. The steady-flow energy equation for the flow in such a tube is
then
v 2
- + h = constant
2
Now, for a perfect gas, Ah = c, AT, and T = c 2 /kR. Introducing these,
together with the relations between the gas constants, the energy equa-
tion may be written as
2 2
V2 +-
k-1
c2 = constant = -co2 = V
k-1
2
max (3.11)
where co is the speed of sound a t the stagnation condition (where V is

zero) and Vma,is the maximum possible velocity in the fluid (where the
absolute temperature is zero).
Art. 3.7 OPTICAL METHODS OF INVESTIGATION 5!4
According to Eq. 3.11, the possible st,a,t.~sin the stream tube are rep-
resented in a diagram of c versus V by the steady-flow ellipse (Fig. 3.8).
Different parts of this ellipse represent schematically different realms of
compressible flow having significantly different physical characteristics.
Incompressible Flow. The velocity is small compared with the sonic
speed. Changes in c are very small compared with changes in V.
Subsonic Compressible Flow. The velocity and sonic speed a w of
comparable magnitudes, but the former is less than ithe latter. Changes
in Mach Number occur primarily Subsonic
because of changes in V, and only
secondarily through changes in c. Incorn-
press~ble '.,
C V.C Transonic
Transonic Flow. The di$erence
between V and c is small compared V>C
with either V or c. Changes in V v
"m,.
and c are of comparable magnitude.
FIG.3.8. Steady-flow adiabatic ellipse,
Supersonic Flow- The velocity showing realms of compressible flow.
and sonic speed are of comparable
magnitude,-but the former is larger than the latter. Changes in Mach
Number take place through substantial variations in both V and c.
Hypersonic Flow. The velocity is very large compared with the sonic
speed. Changes in velocity are very small, and thus variations in Mach
Number are almost exclusively the result of changes in c.
3.7. Optical Methods of Investigation

Apart from the conventional methods of experimentally investigating
Row patterns by means of pressure and velocity surveys, compressible
gas flows lend themselves particularly well to optical methods of inves-
tigation.
Comparison of Methods. Fundamentally, the optical methods in
common use (the interferometer, the schlieren, and the shadowgraph)
depend on one of two physical phenomena: (i) the speed of light depends
on the index of refraction of the medium through which it passes, and
the index of refraction of a gas in turn depends upon its density; and,
as a consequence of this first phenomenon, (ii) light passing through a
density gradient in a gas (and therefore through a gradient in index of
refraction) is deflected in the same manner as though i t were passing
through a prism. In a high-speed gas flow the density changes are
sufficiently large to make these phenomena sizable enough for optical
observation.
60 INTRODUCTORY CONCEPTS T O COMPRESSIBLE FLOW Ch. 3
The interferometer, based on phenomenon (i), measures directly
changes of density, and is primarily suited for quantitative determina-
tion of the density field.
The schlieren method, based on phenomenon (ii), measures density
gradients. Although it is theoretically adaptable to quantitative use,
it is inferior to the interferometer in this respect, and its greatest utility
is in giving an easily interpretable picture of the flow field together with
a rough picture of the density variations in the flow.
The shadowgraph method, also based on phenomenon (ii), measures
the second derivative of the density (i.e., the first derivative of the
density gradient). Therefore it makes visible only those parts of the
flow where the density gradients change very rapidly, and i t has found
its greatest utility in the study of shock waves.
Of the three methods mentioned, the interferometer yields the most
information, and the shadowgraph the least. On the other hand, the
interferometer is the most costly and the most difficult to operate,
whereas the shadowgraph is the least costly and the easiest to operate.
Each method therefore has its own useful niche in experimental work,
and the choice of method depends on the nature of the investigation.
In order to assist the reader in the interpretation of photographs taken
by the three methods, the basic principles of operation will be outlined,
without attention to the numerous optical and mechanical refinements
required. Details of these refinements may be found in References 4
to 12.
The Interferometer. Fig. 3.9 shows the essential parts of the Mach-
Zehnder interferometer, the type which seems to be the most useful for
gas-flow investigations.
FORMATION OF INTERFERENCE PATTERNS. Referring to Fig. 3.9, light
from the source first passes through a collimating lens which renders
the light parallel, and then passes through a monochromatic filter. I t
then passes through the first "splitting plate," which is a plane half-
silvered mirror that passes half the light and reflects half the light. The
light which is reflected is changed back to its original direction by the
lower mirror and then passes through the test section. The light which
passes through the first splitting plate passes through a pair of glass
compensating plates which duplicate in thickness and quality the glass
side walls of the test section, and is then directed to the lower splitting
plate by means of the upper mirror. The two beams which divided at
the upper splitting plate are recombined a t the lower splitting plate and
are then focussed by a lens system on either a ground glass screen or on
a photographic plate.
We shall assume for this discussion that the flow in the test section is
normal to the paper and that the flow is two-dimensional in planes nor-
Art. 3.7 OPTICAL METHODS OF INVESTIGATION 61
ma1 t o the paper and normal to the light beam. Thus, except for slight
refraction effects, each ray of light passing through the test section tra-
verses gas of constant density.
The chief difference between the two beams of light in Fig. 3.9 is that
the lower beam has passed through the test section. Since the different
rays of the lower beam are retarded by different amounts as they pass
through portions of the test section of different density, these rays will
COMPENSATOR
CHAMBER
!loUNT
TRANSLATOR
ROTATION MOUNT
COMPENSATING
HALF SILVERED
INTERFEROMETER
SCHEMATIC DIAGRAM - I NTERFEROMETER

FIG.3.9. Mach-Zehnder type of interferometer (M.I.T. Gas Turbine Laboratory).
have various phase differences with the corresponding rays of the upper
beam when they are recombined, and thus an interference pattern will
be formed a t the screen.
INFINITE-FRINGE INTERFEROGRAM. Suppose to begin with that all

the mirrors and splitting plates are exactly parallel and in line, that there
is no flow, and that the air in the test section has exactly the same den-
sity as the reference air of the upper light beam. Then the two light
beams will be exactly in phase when they recombine, and the screen
will appear uniformly bright. Now imagine that the pressure in the
test section is uniformly and gradually increased. Since the entire lower
light beam is thus retarded by the same amount, the screen will remain
uniform in light intensity, but will gradually darken as the test-section
density is increased. After the density has reached the point where the
lower beam is out of pha.se with the upper by one-half wave length, the
screen will be uniformly dark. Further increases of density will produce
a lightening of the screen until, with a phase difference of one wave
length, the screen will again be a t maximum brightness. As the density
continues to increase, the cycle of darkening and brightening will be
repeated.
With the same mirror geometry as before, consider now the situation
with flow through the test section. The optics are the same as above,
except that the screen is no longer uniform in brightness, but has on it
-
contours of brightness. These contours represent contours of constant
index of refraction, and therefore
represent contours of constant
density in the flow field. Fig. 3.10
shows such an "infinite-fringe" in-
terferogram of flow through a
supersonic nozzle, t h e t e r m
"infinite-fringe"
- signifying
- . -that the
FIG. 3.10. Infinite-fringe interfero- light field is uniform in the ab-
gram of flow through supersonic nozzle
(M.I.T. Gas Turbine Laboratory). sence of flow through the test sec-
tion. The dark bands in Fig. 3.10
are loci of points where there is complete interference, and represent
contours of constant density.
QUANTITATIVE EVALUATION. The change in density from one dark
band to the next in an interferogram like that of Fig. 3.10 may be eval-
uated in the following manner. On the dark bands the light waves
passing through the test section are out of phase with those which pass
through the room air in the compensating chamber by s,lx,254, . . -
room-air wave lengths. Therefore, the light beams passing through ad-
jacent dark bands of the test section are out of phase by one room-air
wave length. Hence, if a represents the fluid lying in one dark band,
and b the fluid in an adjacent dark band, the difference in time for a
light beam to pass through a as compared with that to pass through b
is given by
where X and V denote respectively the wave length of the light and the
speed of light. This difference in traversal time is also rclatcd to the
difference in speed of light in the test section. Thus,
where L ik the length of test section along the light direction, and V b
and Va denote respectively the speeds of lisht in fluid b and a.
Now the frequency j of a given monochromatic light is constant.
The velocity of light in a given medium is related to the velocity of

light in vacuo through the index of refraction n, defined by
-
Eliminatinp. t- "h - t.. from
-u -- ---- Ens.
-1-
3.12 and 3.13, and introducing Eqs. 3.14
and 3.15, we may obtain for t hie difference in index of refraction between
adjacent 1
na = X v m u o I L (3.16)
It remains to connect the index of refraction with the gas density.
This is done through the empirical Gladstone-Dale equation, which
states that
n-1
- --KG-D (3.17)
P
where KG-D,the Gladstone-Dale constant, is constant for a given gas.

Eliminating n in Eq. 3.16 in favor of p by means of Eq. 3.17, we obtain
finally
The right-hand side of this equation is easily computed from the di-
mension of the test section, the color of the monochromatic light em-
ployed, and the value of KG.D for air. Referring again t o Fig. 3.10,
the density in the low-speed flow upstream of t8hethroat is found by
measuring the temperature and pressure in this low-speed region. With
this as a reference, the density on each dark band in the nozzle may be
computed from Eq. 3.18, although the accuracy of this procedure using
the infinite-fringe interferogram is not high unless the optical compo-
nents are extraordinarily accurate.
FRINGE-DISPLACEMENT METHOD. For a more accurate quantitative
evaluation, the method described above is modified as now described.
Let us return to the situation where room air is in the test section of
Fig. 3.9. If the second splitting plate is rotated through a small angle
with respect to the first splitting plate, two "coherent" beams of light
which were in phase a t the first splitter will, through the change in
lengths of light paths, be out of phase a t the screen. If the splitters
are rotated about axes normal t o the paper, there will be formed on the
screen successive light and dark fringes, uniformly spaced, with each
fringe lying parallel to the axis of rotation (see, for example, Fig. 3.11a).
The spacing of successive dark fringes may be shown to be equal to

2/6,X, where 6 is the difference in angles of rotation between the two
splitters, and thus the number of fringes in the flow field is controlled
by adjusting the geometry of the optical elements.
Now suppose that the air density in the test section is uniformly in-
creased. This will produce a uniform displacement of all the wave fronts
passing through the test section. This displacement in turn will cause
FIG.3.11. Fringe-shift interferograms for supersonic nozzle (M.I.T. Gas Turbine

Laboratory).
(a) No flow.
(b) Supersonic flow in test section.
(c) Superposition of (a) and (b).
the interference bands on the screen to shift in a direction normal to

the bands, although the bands will remain parallel and uniformly spaced.
The fringe shift is a measure of the change in density in the test section.
By an analysis similar to that given previously, it may be shown that
where pz is the density in the test section, p l is the density a t the initial
reference condition, d is the distance between dark fringes in the refer-
ence condition, and I is the distance shifted by a dark fringe in passing
from condition 1 to condition 2.
When there is flow in the test section, with corresponding nonuniform
density changes, similar fringe shifts will occur, but, as they are no longer
uniform, the resultant fringes will be curved (Fig. 3.11b). Eq. 3.19 may
Art. 3.7 OPTICAL METHODS O F INVESTIGATION 65
then be applied a t each point in the flow. If both a flow and a no-flow
photograph are taken, Eq. 3.19 may be used to determine a t each point
the density change referred to the no-flow density. Or, from the flow
photograph alone, Eq. 3.19 may be used for finding the density differ-
ences between two points of the flow.
SUPERPOSITION METHOD. If the no-flow negative (Fig. 3.11a) is super-
imposed on the flow negative (Fig. 3.11b), the light passing through the
combination produces the light pattern of Fig. 3.11~. This is a particu-
larly convenient way of determining the contours of constant density,
for a consideration of the geometry of the two fringe patterns shows
that the fuzzy dark bands of Fig. 3 . 1 1 ~ are lines of constant fringe shift,
and, therefore, that they are lines of constant density. Indeed, the pat-
tern of Fig. 3 . 1 1 ~is like that of Fig. 3.10, and Eq. 3.18 may be applied
to either.
The Schlieren Method. Fig. 3.12 shows one of the numerous types of
schlieren (striae) arrangements. Light from a uniformly illuminated line
FL-I
Two- Dim.
f
Finite
Line Lens
source Screen or
Knife Photographic
Gloss Edge Plote
W0llS
z
Fra. 3.12. Schlieren system.
source of small but finite width is collimated by the first lens and then
passes through the test section. It is then brought to a focus by the
second lens and projected on the screen. At the focal point, where there
exists an image of the source, there is introduced a knife-edge which
cuts off part of the light. With no flow in the test section the knife-
edge is usually adjusted so as to intercept about half the light, and the
screen is uniformly illuminated by the portion of the light escaping the
knife-edge. When the flow is established in the test section (assumed
here, for simplicity, to be two-dimensional, with each light ray passing
through a path of constant air density) any light ray passing through a
region in which there is a density gradient normal to the light direction
will be deflected as though it had passed through a prism. Therefore,
depending on the orientation of the knife-edge with respect to the den-
sity gradirnt, and on t,he sign of the density gradient, more or less of
the light passing through each part of the test section d l escape the
knifeedge and illuminate -the screen.
Thus the schlieren system makes density gradients visible in terms of
intensity cf illumination. A photographic plate at the viewing screen
records density gradients in the test section as different shades of gray.
THEORY OF LIGHT DEFLECTION. Let us assume that the flow in the
test section is parallel and in the x,y-plane, and that the light passes
through the test section in the z-direction.
Since the speed of a wave front of light varies inversely with the index
of refraction n of the medium through which the light travels, it follows
that a given wave front will rotate as it passes through a gradient in n.
Accordingly, the normal to the wave front will follow a curved path;
since this normal is what we mean by the light ray, the latter is refracted
as it passes through the density gradient. Noting that n is nearly unity
for gases, i t may be shown that
where R is the radius of curvature of the light ray. The total angular
deflection E of the ray in passing through the test section of width L is
therefore given by 7
Resolving this into Cartesian components, and taking note of Eq. 3.17,
we get for the angular deflections of the light in the x- and y-directions
If the knife-edge is aligned in the y-direction (i.e., normal to the flow),

as in Fig. 3.12, only deflections E , will influence the light passing the
knife-edge. Hence density gradients in the x-direction will be made
visible, but gradients in the y-direction will not be visible. Similarly,
if the knife-edge is parallel to x, only gradients in the y-direction will
become visible.
In interpreting a schlieren photograph of a two-dimensional flow, it
is convenient to imagine that the photograph represents a plan view of
a relief map of the flow field in which the vertical elevation is the gas
density. If the map is illuminated by glancing side light nearly parallel
to x, the pattern of shadows and brightness will correspond to a schlieren
photograph with knifeedge in the y-direction. Similarly, illumination
in the y-direction corresponds to placing the schlieren knife-edge parallel
to the x-direction.
Figs. 3.13a and 3.13b are schlieren photographs of the supersonic
nozzle of Figs. 3.10 and 3.11, and show the different effects obtainable
FIG.3.13. Sehlieren photographs of flow through nozzle of Figs. 3.10 and 3.11
(M.I.T. Gas Turbine Laboratory).
(a) Knife-edge horizontal.
(b) Knife-edge vertical.
with different orientations of the knife-edge. In Fig. 3.13a (knife-edge

horizontal) light areas on the upper half correspond to dark areas on
the lower half, and vice versa; the boundary layers on the walls are
clearly visible, since with this knife-edge orientation the density gra-
dients normal to the flow are made visible. I n Fig. 3.13b (knife-edge
vln
vertical) the upper and lower halves look alike; the boundary layers are
not visible except on the inclined walls of the nozzle.
The Shadowgraph. A shadow system (Fig. 3.14) comprises simply a
small, bright source, a collimating lens, and a viewing screen or photo-
graphic plate. If the source is far
from the test section the collimat-
ing lens is unnecessary.
4
Assume a t first that the test sec- w l n t
Source
tion has stagnant air in it and that
Viewing
the intensity of illumination on the Flow Screen or
Photographic
screen is uniform. When flow is Piate
established in the test section the F IG. 3.14. Shadowgraph system.
light beam will be refracted wher-
ever there is a density gradient. However, if thc- density gradient were
constant, every light ray would be deflected by the same amount, and
there would be no change in illumination on the screen. Only if there
is a gradient in density gradient will there be any tendency for the light
rays to diverge or converge. From this it is evident that variations in
illumination of the screen are proportional to the second derivative of
the density gradient, i.e., to the term
aZP aZp
-+-
ax2 ay2
assuming two-dimensional flow in the x,y-plane.
The shadowgraph is therefore suited to flows with rapidly varying den-

sity gradients, and is insensitive to flows with gently varying gradients.
It is especially convenient and simple for making shock waves visible.
Fig. 3.15a is a typical shadowgraph of a detached shock wave. Across
such a shock the density increases from one side to the other in an
S-shaped curve. Hence the r8te of change of density gradient is posi-
tive on the upstream side of the shock and is negative on the downstream
(a) (b)
FIG.3.15. Detached shock in front of blunt body (Ordnance Aerophysics Labora-
tory).
(a) Shadowgraph.
(b) Schlieren, knife-edge horizontal.
side. In the shadow picture the shock therefore shows as a dark line
followed by a bright line. For comparison, Fig. 3.15b shows a schlieren
photograph of a similar shock.
R E F E R E N C E S A N D S E L E C T E D BIBLIOGRAPW
1. VON KLRMLN,T H. Supersonic Aer~dy~nlics-l'rinciplcs and Applications,
Jour. Aero. Sci., Val. 14, No. 7 (1947), p. 373.
2. JEANS, SIR JAMES. An Introduction to the Kinetic Theory of Gases. Cambridge:
Cambridge University Press, 1946.
3. TSIEN, H. S. Superaerodynamics, Mechanics of Rarefied Gases, Jour. Aero. Sci.,
Vol. 13, No. 2 (1946), p. 653.
4. SCHARDIN, H. Dm Toeplersche Schlierenverfahren, Forschungsheft V.D.I., 367,
Ausgabe B, Band 5 (.July-August, 1934).
5. ZOBEL, TH. Development and Construction of an Interferometer for Optical
Measurements of Density Fields, NACA Tech. Memo., No. 1184 (1947).
6. B ARNES, N. F., and B ELLINGER , S. L. Schlieren and Shadowgraph Equipment
for Air Flow Analysis, Jour. Opt. Soc. Amer., Val. 35, No. 8 (1945), p. 497.
7. L ADENBURG , R., W INCKLER, J., and V AN VOORHIS,C. C. Interferametric
Studies of Faster Than Sound Phenomena, Phys. Rev., Vol. 73, No. 11 (1948),
p. 1359.
8. ASHKENAS, H. I., and BRYSON, A. E. Design and Performance of a Simple Inter-
ferometer for Wind-Tunnel Measurements, Jour. Aero. Sci., Val. 18, No. 2
(1951). n 82.
Art. 3.7 PROBLEMS 69
9. DEFRATE, L. A., BARRY, F. W., and BAILEY, D. Z. A Portable Mach-Zehnder
Interferometer, Meteor Report, No. 51, M . I. T., Cambridge, Mass. (1950).
10. SCHARDIN, H. Theorie und Anwendung des Mach-Zehnderschen Interferenz-
Refraktometers, Zeilschrift jur Instrumentenkunde, Vol. 53 (1933), pp. 396,
424. Also see R.A.E. Translation, No. 79.
11. LIEPMANN, H. W.,and PUCKEIT, A. E. Introduction to Aerodynamics of a Com-
pressible Fluid. New York: John Wiley & Sons, Inc., 1947.
12. PANKHURST, R. C., and HOLDER, D. W. Wind- Tunnel Technique. London:
Sir Isaac Pitman & Sons, 1952.
PROBLEMS
3.1. The compressibility of a liquid is usually expressed in terms of the bulk
modulus of compression,
8 = P-d p
dp
Show that
3.2. Calculate the velocity of sound a t 70°F in the following media:

(a) Air, (b) hydrogen, (c) uranium hexafluoride, (d) mercury vapor, (e)
water vapor, (f) liquid water a t 14.7 psia.
3.3. To what pressure must liquid water be compressed in order that it
leave a nozzle a t atmospheric pressure with a jet velocity equal to the sonic
velocity? Assume a constant bulk modulus of compression of 300,000 psi.
3.4. Plot and compare curves of the sonic velocity in air versus absolute
temperature (a semi-log plot is suggested) using
(a) a value of k of 1.4 for all temperatures, and (b) the actual value of k
for each temperature.
3.6. A projectile in flight carries with i t a more or less conical-shaped shock
front. From physical reasoning it appears that a t great distances from the
projectile this shock wave becomes truly conical and changes in velocity and
density across the shock become vanishingly small.
Photographs of a bullet in flight show that a t a great distance from the bullet
the total included angle of the cone is 50.3". The pressure and temperature of
the undisturbed air are 14.62 psia and 73"F, respectively.
Calculate the velocity of the bullet, in ft/sec, and the Mach Number of the
bullet relative to the undisturbed air.
3.6. Show that, for a perfect gas, the fractional change in pressure across a
small pressure pulse is given by
d p - dV
- - k-
P c
and that the fractional change in absolute temperature is given by
3.7. (a) Demonstrate that a compression wave (i.e., a pressure pulse which
increases the density of the fluid over which it passes) which moves rightward
imparts a rightward velocity to the fluid.
(b) Derive similar rules for a rightward-moving rarefaction wave and for
leftward-moving compression and rarefaction waves.
(c) The rightward-moving compression wave of Fig. 3.la strikes a stationary
wall closing the right-hand end of the duct. Demonstrate that it is necessary
for a reflected wave to travel leftward, and determine whether the reflected
wave is a compression or rarefaction. Compare the pressure change across the
incident wave with that across the reflected wave.
3.8. A compression pulse changes the velocity of the fluid over which it
passes by 10 ft/sec. Calculate the pressure rise (psi) across the pulse for (a)
water, and (b) air a t 14.7 psia and 70°F.
P A R T I1
ONE-DIMENSIONAL FLOW
Chapter 4
ISENTROPIC FLOW

The one-dimensional, sbeady-flow treatment of isentropic flow finds
important applications in two kinds of problems, namely, (i) flow in
ducts and (ii) flow in stream tubes.
Flow in Ducts. The flow in pipes and ducts is very often adiabatic.
When the duct is short, as it is in nozzles and diffusers, the frictional
effects are comparatively small, and the flow may as a first approxima-
tion be considered reversible, and, therefore, isentropic. Furthermore,
since the function of nozzles and diffusers is to accelerate or decelerate
a stream as efficiently as possible, the isentropic process provides a use-
ful standard of comparison for actual nozzles and diffusers.
Flow in a Stream Tube. In virtually all problems involving flow
around bodies, and in many involving flow through passages, there are
elementary stream tubes which lie entirely outside the boundary layer.
Viscous and heat conduction effects for such stream tubes, it appears
from experiment, are negligible. Hence the equations of isentropic flow
may be considered exact, unless discontinuities such as shock waves
appear.
The One-Dimensional Approximation. By a one-dimensional flow we
mean a flow in which all fluid properties are uniform over any cross sec-
tion of the duct. Or, more strictly, we mean a flow in which the rate of
change of fluid properties normal to the streamline direction is negligibly
small compared with the rate of change along the streamline.
No approximation whatsoever is involved in the case of a stream tube,
for the flow through an infinitesimal stream tube is, in the limit, exactly
one-dimensional.
When applying the one-dimensional assumption to the flow in ducts,
where i t is well known that the properties vary over each cross section,
we in effect deal with certain kinds of average properties for each cross
section. The errors in predicting the rate of change of properties along
the duct axis may be expect,ed to be small then, if
(i) The fractional rate of change of area with respect to distance
along the axis is small ( d A / A dx << I).
73
74 ISENTROPIC FLOW Ch.4
(ii) The radius of curvature of the duct axis is large compared with
the passage diameter.
(iii) The shapes of the velocity and temperature profiles are approx-
imately unchanged from section to section along the axis of the
duct.
The great virtue of the one-dimensional approximation is the marvel-
lous simplicity i t affords, leading as i t does t o rapid calculation methods
for a great variety of practical engineering problems. Indeed, the in-
formation resulting from the one-dimensional point of view is, when care-
fully interpreted, so useful and so reliable that this method is one of the
most powerful tools in the armory of the engineer.
At the same time, it is well to remember that the one-dimensional
approach, by its very nature, supplies information only as to the way
in which the average fluid properties over the duct cross section vary
with distance along the axis of the duct, and is completely silent as to
the variation of properties normal t o the streamlines. For many prac-
tical problems the latter is of the essence, and then the one-dimensional
treatment must be supplemented by a two- or even three-dimensional
analysis.
NOMENCLATURE
cross-sectional area s entropy per unit mass

speed of sound T absolute temperature
specific heat at constant pres- 3 thrust
sure V velocity
specific heat at constant volume w mass rate of flow
pressure coefficient W molecular weight
nozzle discharge coefficient
impulse function 7 nozzle efficiency
mass velocity, w / A p viscosity
enthalpy per unit mass p density
Mach Number ( )O signifies stagnation state
dimensionless speed, V/c* ( )* signifies state at which the
pressure Mach Number is unity; does
gas constant not apply to M*
universal gas constant () signifies free stream conditions
4.2. General Features of Isentropic Flow

Let us first consider the isentropic flow of any fluid whatsoever through
a passage of varying cross section (Fig 4.1). All possible states lie on a
line of constant entropy, as shown in Fig. 4.2. One state on this isen-
tropic corresponds to zero velocity. The pressure a t this state, po, is
Art. 4.2 GENERAL FEATURES O F ISENTROPIC FLOW 75
usually called the isentropic stagnation pressure, and is sometimes called

the total pressure. The value of the stagnation enthalpy, ho, is independent
of whether or not entropy changes occur, since it has the same value for
all states which are reachable from it adiabatically.
Stagnation
Sectipn Confrol
////,/,/,,/,
To 1 w'-C Flow
Ti
I
/ / / / / I / / /
, /
FIG. 4.1. Flow between stagnation sec- FIG. 4.2. Isentropic ac-
tion and any other section. celeration or deceleration on
Mollier chart, showing stag-
nation enthalpy and isen-
tropic stagnation pressure.
Governing Physical Equations. The following physical equations may

be written for a control surface extending between the stagnation section
and any other section in the channel:
First Law of Thermodynamics.
v2
ho=h+-
2
Second Law of Thermodynamics.
s = so
Equation of Continuity.
G = w / A = pV
Equation of State. This may be expressed in the form of charts, tables,
or algebraic equations. We may write implicitly that
Definition of Mach Number.
Performance Curves. For a given stagnation condition, i.e., for fixed

values of so and po, Eqs. 4.1 may be used for constructing the perform-
ance curves shown in Fig. 4.3. For example, suppose an arbitrary value
of p less than po is selected. Then the corresponding values of h and p
may be found from Eq. 4.ld, inasmuch as s is known from Eq. 4.lb; the
corresponding value of V may next be found from Eq. 4.la; and, finally,
the corresponding values of w / A
and M may be reckoned from Eqs.
4.lc and 4.le.
The curves shown in Fig. 4.3 are
( a typical of gases and vapors but'
are somewhat different for liquids
and liquid-vapor mixtures.
Critical Pressure Ratio. The
most interesting feature of Fig. 4.3
PO 'Const. +Critical is the maximum in the curve of
flow per unit area, which indicates
0 PIPo I
that an accelerating stream start-
increase in cross section. The pressure ratio, p/po, where the flow per
unit area is a maximum, is called the critical pressure ratio, and has a
value, for all real gases and vapors, of approximately one-half.
Pressure ratios greater than the critical correspond to subsonic flow,
and pressure ratios less than the critical correspond to supersonic flow.
This will now be demonstrated in a completely general manner.
Distinction Between Subsonic and Supersonic Flow. We first write
the steady-flow energy equation in differential form for two cross sec-
tions infinitesimally distant from each other. Thus
From the thermodynamic relation, T ds = dh - dp/p, and the con-

dition of constant entropy, we have
so that
This will be recognized as Euler's equation of motion for an inviscid

fluid. This is not surprising, as the kinetic-energy term in the steady-
flow energy equation was originally obtained with the help of Newton's
second law of motion.
Next we introduce the equation of continuity in logarithmic diieren-
tial form,
giving
Art. 4.2 GENERAL FEATURES OF ISENTROPIC FLOW 77
Substituting Eq. 4 . 2 ~into E q . 4.2d and rearranging, we get
Since the process is isentropic,
~ P D P= (apiap), = c2
so that
-- 1 - M~
dp (4.2e)
pV2
Now, from Eq. 4.2c, which is the dynamic equation for a frictionles:.
fluid, it is seen that the pressure always decreases in an accelerating flow.
and increases in a decelerating flow. In other words,
Using this result in conjunction with Eq. 4.2e, we arrive a t the follow-
ing conclusions of practical significance:
(i) For subsonicspeeds (M < l),
(ii) For supersonic speeds ( M > l),
(iii) For sonic speeds (M = l),
Thus, we have the astonishing result that the effects of an area change,
say an increase in area, are exactly opposite for subsonic and supersonic
flow.
The possible types of flow, according to this tabulation, are summa-
rized schematically in Fig. 4.4.
A t Mach Number unity the area goes through a minimum. This
important conclusion is valid irrespective of the type of fluid considered,
whether gaseous or liquid.
In constructing the curves of Fig. 4.3, the equation of state, Eq. 4.ld,
was employed. Since the equation of state for real gases and vapors can
seldom be put into simple algebraic form, the curves of Fig. 4.3 cannot,
in general, be formulated analytically, but instead are found through
78 ISENTROPIC FLOW Ch. 4
direct computation. If, on the other hand, it is assumed that the perfect
gas laws are valid, analytical results are obtainable, and the numerical
computation of problems is greatly simplified. For many engineering
gases, particularly air a t moderate pressures and temperatures, the
deviations from the perfect gas laws are negligible; hence, most calcula-
tions are based on these simple relations.
(M>I1 V increases
SUBSONIC SUPERSONIC
O Z Z LE
\
-Flow
-
Flow
-
p decreases p increases
(M<I) V increases V decreases
/--
- T H ROAT
Flow
( M I1
dp50
dMe0
FIG. 4.4. Effects of area change on pressure and velocity in subsonic and super-
sonic flow.
4.3. Adiabatic Flow of a Perfect Gas

Before restricting the discussion to isentropic flow, certain relations
obtainable from the energy equation alone will be derived. These rela-
tions are valid for any adiabatic flow of a perfect gas, whether reversible
or not.
For a perfect gas we have
Ah = c, AT
Art. 4.3 ADIABATIC FLOW O F A P E R F E C T GAS 79
Using Eq. 4.3a and 4.3d, the steady-flow energy equation (Eq. 4.la)
becomes
2k
V = 42/2c,(~o- T ) = d/--R(To - T) (4.4)
k-1
From this we see that for a fixed stagnation temperature To, (sometimes
called total temperature), all states with the same temperature have the
same velocity. Referring to the temperature-entropy diagram of Fig.
4.5, lines of constant velocity are horizontal, and the vertical distance
between To and T is proportional to the square of the velocity.
l rreversible
T Deceleration
T*
I C I
S S
(4 (b)
FIG.4.5. Flow processes on temperature-entropy diagram.
(a) Isentropic acceleration or deceleration.
(b) Irreversible adiabatic deceleration.
Three Reference Speeds. Since negative temperatures on the abso-

lute scale are not attainable, it is evident from Eq. 4.4 that there is a
maximum velocity corresponding to a given stagnation temperature.
This maximum velocity, which is often used for reference purposes, is
given by
(4.5a)
Another useful reference velocity is the speed of sound a t the stagna

tion temperature,
(4.5b)
Still a third convenient reference velocity is the critical speed, i.e.,
the velocity a t Mach Number unity. Using an asterisk to denote con-
ditions a t M = 1, we have, by definition,
w--,--
or, using Eq. 4.4 and the equation for the sound speed in a perfect gas,
which gives, after rearrangement

T* -
2
To k+1
substituting this value of T* for T i n Eq. 4.4, we get
fir-
Now, using Eqs. 4.5, we get the following relations between the three
reference velocities, together with the numerical values for k = 1.4:
Stagnation-Temperature Ratio. Eq. 4.4 may be written
Since cp = kR/(k - 1) and cZ = kRT, this takes the simple and con-
which shows that the stagnation-temperature ratio depends only on the

gas constant k and Mach Number M.
The Energy Equation in Kinematic Form. The energy equation for
adiabatic flow of a perfect gas is
V2 + 2cpT = constant
Moreover, for a perfect gas,
Art. 4.3 ADIABATIC FLOW OF A PERFECT GAS 81
Combining these relations, and evaluating the constant a t the three

reference conditions of (i) zero speed, (ii) zero temperature, and (iii)
sonic speed, we obtain three alternate and useful forms of the energy
equation involving only the local velocity, local sound speed, and k:
Note that Eqs. 4.5 through 4.7 may be obtained rather easily from
Eq. 4.8.
The Dimensionless Velocity M*. As a dimensionless parameter the
Mach Number is very convenient, but it has two disadvantages: (i) i t
is not proportional to the velocity alone and, (ii) a t high speeds it tends
towards infinity. Often, therefore, i t is useful to work with a dimension-
less velocity obtained through dividing the flow velocity V by one of the
three reference velocities of Eq. 4.8. Generally the most useful of these
ratios is defined by
M* = V/c* = V/V*
It should be noted immediately that although in general the asterisk
denotes the value of a property a t Mach Number unity, this convention
is not followed in the definition of M*. The latter is not the value of M
at the local sonic condition, but is rather defined as given above.
There is a unique relation between M and M* for adiabatic flow. From
the definitions of M* and M,
Furthermore, the first and last parts of Eq. 4.8 may be divided by c*' to
give
L
Eliminating c2/c*' from this, pair of equations, and rearranging, we get
the useful formulas
and
82 ISENTROPIC FLOW Ch. i
The value of M* is a simple index of when the flow is subsonic and

when supersonic, for we find from Eqs. 4.9 and 4.10 that
when M < 1, then M* <1
when M > 1, then M* > 1
when M = 1, then M * = 1
when M = 0, then M* = 0
Fig. 4.3 illustrates the relative magnitudes of M and M* in subsonic

and supersonic flow.
Flow per Unit Area. Next we will derive a useful relation between the
flow per unit area, stagnation temperature, static pressure and Mach
Number. Starting with the equation of continuity, we make the follow-
ing rearrangements:
This is best arranged in the form of a mass flow parameter involving

To, p, and the molecular weight W. The parameter itself then depends
only on k and M according to the
relation
The parameter is plotted in Fig.

4.6 for k = 1.4. With this chart
i t is easy to solve the common
FIG. 4.6. Mass flow parameter versus problem of computing the local
M and M*, Units: w (sludsec); A (ft2); Mach Number from given values
To(deg. Rankine); p (lbf/ft2); W (Ib/lb
mol, or slug/slug mol). of w, A, To, and p.
Art. 4.4 ISENTROPIC FLOW OF A PERFECT GAS
4.4. Isentropic Flow of a Perfect Gas

All the relations of the preceding section are valid for both isentropic
and nonisentropic flows. Referring to Fig. 4.5, they are applicable to
any sort of change in the vertical (T) coordinate, irrespective of changes
in the horizontal (s) coordinate, provided that all states have the same
stagnation temperature.
We now further restrict the analysis to the isentropic case. All states
along the channel or stream tube lie on a line of constant entropy and
have the same stagnation temperature. The state of zero velocity is
called the isentropic stagnaticm state, and the state with M = 1 is called
the critical state. All states with the same entropy and the same stagna-
tion temperature have the same isentropic stagnation state and the same
critical state. When a stream with a given pressure, temperature, and
velocity (Fig. 4,5b) is decelerated to zero velocity, the final pressure will
be less than the isentropic stagnation pressure if the decleration is irre-
versible, but the final temperature will be equal to the adiabatic stagna-
tion temperature for either reversible or irreversible deceleration.
The relations between pressure, temperature, and density for an
isentropic process of a perfect gas are
Also, the pressure-temperature density relation of a perfect gas is
Temperature, Pressure, and Density Ratios as Functions of Mach

Number. Substitution of Eqs. 4.12 into Eq. 4.7 now yields the impor-
The particular values of the temperature, pressure, and density ratios

a t the critical state (i.e., a t the minimum area) are found by setting
M = 1 in the above expressions. The resulting formulas, together with

the numerical values for k = 1.4, are as follows:
Thus, the temperature a t the throat is only about 17 per cent less
than the stagnation temperature, whereas the throat pressure is only
about half the isentropic stagnation pressure.
The critical pressure ratio, p*/po, is of the same order of magnitude
for all gases. It varies almost linearly with k from 0.6065 for k = 1 to
0.4867 for k = 1.67.
Mass Flow Relations in Terms of Mach Number. From Eqs. 4.14
and 4.10 it is clear that either T/To, p/po, M or M* may be taken
as an independent parameter, and that the remaining quantities would
then depend uniquely on the value of the chosen independent parameter.
By and large, the variable M has been found to be the most convenient
choice as far as simplicity of practical calculations is concerned. We
shall, therefore, follow the practice in this and succeeding chapters of
deriving all the working formulas in terms of M as the independent
variable.
To find a convenient formula for the mass flow per unit area in terms
of M, we eliminate p in the equation preceding Eq. 4.11 by means of the
isentropic law (Eq. 4.14b). Thus we obtain
This shows that, for a given Mach Number, the flow is proportional to
the stagnation pressure and inversely proportional t o the square root
of the stagnation temperature. For this reason, flow test data on com-
pressors and turbines, or indeed on any flow passage which operates
over a wide range of pressure and temperature levels, are usually plot-
ted with w f i / p o as the flow variable. In this way the results of a
given test become applicable to operation a t levels of temperature and
Dressure different from the original test conditions.
Art. 4.4 ISENTROPIC FLOW OF A PERFECT GAS 89
Now it is evident that if M were eliminated from Eq. 4.16 with the
help of Eq. 4.14b, we would have w/A in terms of k, R, po, To, and p/po.
The resulting expression is then the algebraic formula, for a perfect gas,
corresponding to the curve of w/A versus p/po in Fig. 4.3.
MAXIMUM FLOW PER UNIT AREA. T O find the condition of maximum
flow per unit area, we could compute the derivative d(w/A)/d(p/po)
and set this derivative equal to zero. From this we would find that the
critical pressure ratio is given by Eq. 4.15b.
An equivalent procedure would be to differentiate Eq. 4.16 with re-
spect to M and set this derivative equal to zero. At this condition, we
would find that M = 1.
However, neither of these procedures is necessary inasmuch as we
have proved quite generally in Art. 4.2 that the cross-sectional area for
isentropic flow passes through a minimum a t Mach Number unity.
Therefore, to find (w/A),,,, we need only set M = 1 in Eq. 4.16.
Thus we find
For a given gas, therefore, the maximum flow per unit area depends
only on the ratio For given values of the stagnation pressure
and stagnation temperature and for a passage with a given minimum
area, Eq. 4.17 shows that the maximum flow which can be passed is
relatively large for gases of high molecular weight and relatively small
for gases of low molecular weight. Doubling the pressure level doubles
the maximum flow, whereas doubling the absolute temperature level
reduces the maximum flow by about 29 per cent.
FLIEGNER'S
FORMULA. Using the values k = 1.4 and R = 53.3
O
ft lbf/lbm R, corresponding to air, we obtain from Eq. 4.17
for air, where w is in lbm/sec, A* in ft2, To in OR, and po in lbf/ft2.

This formula, which we have derived on purely analytical grounds,
was discovered empirically by Fliegner nearly a century ago a t a time
when the theoretical considerations outlined here were scarcely under-
stood! Fliegner's experiments, which were conducted on a simple con-
verging nozzle, gave a value of the constant within about 1 per cent of
the value in Eq. 4.18.
THE AREA F ~ T I O .Just as we have found it convenient to work with
the dimensionless ratios p/po, etc., so it is convenient to introduce a
dimensionless area ratio. Obviously the appropriate reference area is
A*, and so we compute from Eqs. 4.16 and 4.17 the formula
The area ratio is always greater than unity, and for any given value of
A/A* there always correspond two values of M-one for subsonic flow,
and the other for supersonic flow.
The Impulse Function. For problems involving jet propulsion it is
sometimes convenient to employ a quantity called the impulse function,
defined by
F = pA pAV2+ (4.20)
FIG. 4.7. Illustrating use of impulse function.
Applying the momentum equation to the flow through the control sur-
face of Fig. 4.7, i t is seen that
where 3 is the net "thrust" produced by the stream between sections 1

and 2. The term thrust, as used here, is defined as the net force exerted
by the stream on the internal solid surfaces which the fluid wets, acting
in the direction opposite to the direction of flow. It includes theforces
due to pressure and viscous stresses on the duct walls as well as the total
drag of any stationary obstacles in the stream. This is true whether the
flow is adiabatic or nonadiabatic, and whether the flow is reversible or
irreversible.
For a perfect gas,
and so
For isentropic flow, a dimensionless impulse function is formed by

writing
F P A
-=-a
POA* P O A *
(1 kM2) + (4.23)
Art. 4.5 WORKING CHARTS AND TABLES 87
where p/p0 and A/A* are functions of M given respectively by Eqs. 4.14b
and 4.19.
Another way of forming a dimensionless impulse function is by evalu-
ating F* a t M = 1, and setting
substituting p/po, p ~ / p * ,and A/A* from Eqs. 4.14b, 4.15b, and 4.19,
respectively, there is obtained after simplification,
4.6. Working Charts and Tables for Isentropic Flow

Since the formulas thus far derived lead to tedious numerical calcu-
lations, often of a trial-and-error nature, practical computations are
greatly facilitated by working charts and tables
Chart for Isentropic Flow. Fig. 4.8 represents in graphical form the
various dimensionle.% ratios for isentropic flow with M as independent
variable. Since changes of fluid
properties in isentropic flow are
brought about through changes in
cross-sectional area, the key curve
on this chart is that of A/A*. The
effects of changes in area on other
properties may easily be found by
tracing the curve of A/A*, keeping
in mind that A*, p0, F*, etc. are
all constant reference values for a
given problem. For example, an
increase in area a t subsonic speed
produces a decrease in velocity, an
increase in F , and increases in p,
T,and p.
It may be seen from Fig. 4.8 that F IG. 4.8. Working chart for isentropic
up to a Mach Number of about 0.3 flow, with k = 1.4.
(corresponding to about 300 ft/sec
for air a t normal conditions), changes in density are almost negligible
for engineering calculations. This explains why in so many instances
air is treated as though i t were an incompressible fluid.
Working Tables. For accurate or extensive calculations, Table B.2

lists the various isentropic flow functions for k = 1.4, with Mach Number
as independent argument.
Illustrative Example. T h e use of the compressible flow functions is
best explained by a n example.
PROBLEM. A supersonic wind-tunnel nozzle is to be designed for M = 2,
with a throat section 1 sq ft in area. The supply pressure and temperature
at the nozzle inlet, where the velocity is negligible, are 10 psia and 10O0F,
respectively. The preliminary design is to be based on the assumptions that the
flow is isentropic, with k = 1.4, and that the flow is onedimensional at the
throat and test section. It is desired to compute the mass flow, the test-section
area, and the fluid properties at the throat and test section.
SOLUTION. Table B.2 is to be used. First we work out the reference stagna-
tion properties. We are given p, = 10 psia and To = 459.7 100 = 559.7'R. +
From these we compute
Next, we find the properties at the throat by selecting values from Table
B.2 at M = 1:
p*/po = 0.528; :. p* = lO(0.528) = 5.28 psia
Entering Table B.2 at M = 2, we now calculate properties in the test section:
p/po = 0.1278; . ' p = 0.1278(10) = 1.278 psia
Finally, we compute the mass flow from Eq. 4.18:
Alternatively, the mass flow may be computed from the continuity equation
at the throat or test section. For example,
.%I%. 4.6 CHOKING I N ISENTROPIC FLOW
4.6. Choking in Isentropic Flow

The fact that the curve of mass flow per unit area has a maximum is
connected with an interesting and important effect called choking.
Let us consider two sections of a stream tube having a ratio of areas
A2/A1, and let us specify all flow properties a t section 1, such as pl, TI,
M1, etc. From the tables or graphs we can then solve for the properties
a t section 2, except as discussed later. For example, corresponding to
M1 we may find in the tables (p/po)l, (T/To)l, and (A/A*)l. Then,
since A* is constant during the process, we may write
and so we may compute (A/A*)2. Returning to the tables, we then

obtain a t this value of (A/A*)2 the corresponding values of M2, (p/poj2,
(T/T,J2, etc. Since po and To are constant, this allows us to compute
p, and T2 from
-p2 -
- (P/PO)Z. _T2= (T/To)z
PI (P/PO)I' TI (T/To)I
Now, let us consider a passage with a given area ratio A2/A1 and
compute in the manner outlined above the values of Mz corresponding
to several values of MI. The re-
sults may then be plotted as in
Fig. 4.9. Examination of this chart
indicates two peculiarities:
(i) For a given initial Mach
Number M1 and a given area ratio
A2/A1, there are either two solu-
tions for the final state M2 or none
a t all. When there are two soh-
tions, one of the two is subsonic
and the other is supersonic. Which
one of the two occurs depends in
part on whether a throat inter-
venes between sections 1 and 2, FIG. 4.9. Typical curves of Mz versus
for we have demonstrated that in MI for fixed values of area ratio A2/A1.
order to go from supersonic speed
to subsonic speed, or vice versa, it is necessary for the flow to pass
through a throat at M = 1. For example, if M1 is subsonic and t h e
passage is converging, then MZ must also be subsonic. On the other
hand, if MI is subsonic and the passage is convergingdiverging (i.e.,
has a minimum area between sections 1 and 2) the flow at section 2
may be either subsonic, as in a conventional venturi, or supersonic, as
in a supersonic nozzle; which of these two situations prevails depends
on the pressures imposed a t the inlet and exit of the passage, as dis-
cussed more fully in Art. 4.7.
(ii) When, for selected values of M1 and A2/Al, there is no solution in
Fig. 4.9, the solution is imaginary in the mathematical sense. This
occurs only when Az is smaller than Al. Physically, this result signifies
that for given conditions a t section 1, there is a maximum contraction
which is possible; this maximum contraction corresponds to sonic veloc-
ity at section 2. Or, put quite simply, if conditions a t section 1 are
specified, the mass flow is accordingly determined, and there is then a
minimum cross-sectional area required to pass this flow. This phenom-
enon is called choking, and may be summarized by saying that for a given
area reduction. there is in subsonic flow a maximum initial Mach Num-
ber which can be maintained steadily; and in supersonic flow a minimum
initial Mach Number which can be maintained steadily. At either of
these limiting conditions, the flow a t section 2
I
4 +-----+
I Flow
is sonic, and is said to be choked.
pr
T,
-1
I
---+ I
To illustrate further the phenomenon of
s/ 6
Ad]ustoble Walls
choking, let us suppose that a t a section 1 in a
duct there is a subsonic flow with certain values
FIG. 4.10. Illustrates of MI, pl, TI, and A1. These parameters fix the
choking of flow. flow rate, w. Let us imagine further that, a t
a section 2 downstream, the walls are flexible;
thus the area AZ is adjustable, as shown in Fig. 4.10. If A2 is equal to
Al, all conditions a t section 2 will be identical with the corresponding
conditions a t section 1. A slight reduction in A2 will produce certain
effects a t section 2, which, according to Fig. 4.8, will comprise an in-
crease in Ma, a decrease in pz, and a decrease in T2. This slight reduc-
tion in A2 without a change in conditions a t section 1 must, therefore,
be accompanied by a reduction in the back pressure, pz, according to
the requirements of Fig. 4.8. Further reductions in Az may be made in
the same way until the value of Mz reaches unity.
After this point has been reached, there is no way of reducing the area
further without a simultaneous change in the steady-state conditions
a t section 1. If, for example, the pressure and temperature a t section 1
are maintained constant, a reduction in A2/A1 beyond its limiting value
will, after a transient period of wave propagation, result in a reduced
steady-state MI, which in turn means that the flow rate will be decreased.
The maximum possible value of M1 (which will correspond to the maxi-
mum possible flow rate) is obtained when Mz = 1. To obtain this
limiting flow, the back pressure, p2, must of course be adjusted accord-
ingly. Fig. 4.8 shows that any area reduction whatsoever may be made
if the initial Mach Number is sufficiently low or sufficiently high.
Art. 4.7 OPERATION OF NOZZLES 91
4.7. Operation of Nozzles Under Varying

Pressure Ratios
The phenomenon of choking discussed above may be manifested in
several different ways. To illustrate still another aspect of choking, let
us discuss the practical problem of nozzles operating under varying pres-
sure ratios.
Converging Nozzles. Suppose, for the sake of concreteness, that a
converging passage (Fig. 4.11a) with a large entrance area a t section 0
fvariablel
p,= conFt
Exhauster
Valve
Regime
0
Distance A l o n g Nozzle
(4
0 PB'Po I 0 Pe'Po I
(b) (c)
Fro. 4.11. Operation of converging nozzle at various back pressures.
discharges into a region where the back pressure, p ~ is, controllable by

means of a valve. The values of po and To will be maintained constant,
and the experiment will involve variations in p ~ . If p~ denotes the
pressure in the exit plane of the nozzle, we inquire as to the effects of
variations in back pressure on the distribution of pressure in the passage,
on the flow rate, and on the exit-plane pressure. These effects are por-
trayed graphically in Figs. 4.11a1 b, and c, respectively.
92 ISENTROPIC now ch. 4
To begin with, suppose that pB/po = 1, shown as condition (i) in
Fig. 4.11a. The pressure is then constant through the nozzle, and there
is no flow.
If p~ is now reduced to a value slightly less than p,, as shown by
condition (ii), there will be flow with a constantly decreasing pressure
through the nozzle. Because the exit flow is subsonic, the exit-plane
pressure p~ must be the same as the back pressure p ~except , for minor
secondary circulation effects in the exhaust space. That this must be
so can be seen by supposing for a moment that p , ~is substantially larger
than p ~ .If this were so the stream would expand laterally upon leaving
the nozzle; however, such an area increase at subsonic speeds causes the
stream pressure to rise even further. Since the back pressure is, by
definition, the pressure which the stream ultimately achieves in the
exhaust space, i t follows that p~ cannot be larger than p ~ .A similar
argument leads to the conclusion that p~ cannot be substantially less
than p ~ .
A further reduction in p~ to condition (iii) acts to increase the flow
rate and to change the pressure distribution, but there is no qualitative
change in performance.
Similar considerations apply until condition (iv) is reached, a t which
point equals the critical pressure ratio and the value of ME eq~lnls
unity.
Further reductions in p ~ / p o ,say to condition (v), cannot produce
further changes in conditions within the nozzle, for the value of pE/po
cannot be made less than the critical pressure ratio unless there is a
throat upstream of the exit section (it is assumed here that the stream
fills the passage). Consequently, a t condition (v), the pressure distribu-
tion within the nozzle, the value of p ~ / p o ,and the flow rate, are all
identical with the corresponding quantities for condition (iv). The
pressure distribution outside the nozzle cannot be predicted on one-
dimensional grounds, and for the present is shown in Fig. 4.11a as a
wavy curve.
To summarize the preceding discussion, the two different types of
flow will be denoted as regime I and regime 11. These regimes may be
compared as follows:
Regime I Regime I1
w 6
dependent on pB/po
wd?& independent of p ~ / p ,
AEPO AEPO
Art. 4.7 OPERATION O F NOZZLES 93
In regime I the values of p~ and PB are virtually identical. Hence,

except for a constant multiplier, the flow curve in regime I of Fig. 4.11b
is identical with the curve of w/A in the subsonic part of Fig. 4.3.
A simple converging nozzle of the type discussed often serves as a
flow nozzle. It is particularly useful when %/Po is less than the critical
pressure ratio, for then the flow rate is given by Eq. 4.18, and measure-
ments of only PO, TO, and AE are necessary for computation of the flow
rate.
For accurate measurements, the effects of boundary layer and of
departures from one-dimensionality require that the nozzle be calibrated.
Discharge coefficients for rounded-entrance nozzles are usually of the
order 0.98 to 0.99, except for very low Reynolds Numbers where they
may be considerably less.
The converghg nozzle may occasionally be used to advantage as a
simple flow regulator because of the fact that the flow rate is independent
of back pressure when the latter is less than about half the supply pres-
sure.
Converging-Diverging Nozzles. Consider an experiment similar to

the one described above, except that a converging-diverging nozzle is to
be used (Fig. 4.12).
With PB less than po by only a small amount, the flow is similar to
that through a venturi passage, and it may be treated approximately as
incompressible. The correspond-
ing pressure distribution is shown
vo a ;
:
by curves (i) and (ii) in Fig. 4.12.
When p ~ / p ois reduced to the
value corresponding to curve (iii),
the Mach Number a t the throat is
unity, and no further reductions in I
pr/po are possible if the stream
W-
fills the passage. -P
We consider next the operation Po P/P,
when the flow is entirely super- rvl-

sonic, corresponding to Curve (iv). Distance Along Nozzle
The of PB'P' for curve (iv) FIG. 4.12. Operation of converging-
corresponds exactly to the area diverging nozzle at various back pressures.
ratio of the nozzle, AE/AT, as
given by the isentropic tables (in this case AT = A*, since MT = 1).
This is often called the "design" pressure ratio of the nozzle.
No flow pattern fulfilling the conditions of isentropic and one-dimen-
sional flow can be found which will correspond to values of pe/po be-
tween those of curves (iii) and (iv) in Fig. 4.12. One method of finding
solutions for these boundary conditions is to suppose that irreversible
discontinuities involving entropy increases occur somewhere within the

passage. The analysis of such discontinuities, called shock fronts, is
the subject of Chapter 5 . A complete discussion of the converging-
diverging nozzle will, therefore, be postponed until the shock wave
analysis has been presented.
4.8. Special Relations for Low Mach Numbers

In many flow problems the Mach Numbers are comparatively small,
but compressibility effects cannot be entirely ignored. Using binomial
expansions, the formulas of the preceding articles may be put into
simple algebraic forms which are accurate and convenient for such cases.
Correction to Incompressible Pitot-Tube Formula. For example, sup-
pose that it is desired to examine the error incurred in the computation
of pressure variations when the gas is assumed incompressible. From
Eqs. 4.8 and 4.14b, we find that
Expanding the right-hand side of this expression by the binomial

theorem and rearranging, we get, if only terms up to ( V / C are
~ )included,
~
If the fluid were taken as incompressible, the right-hand side of Eq.

4.25 would reduce to unity, and the equation would be identical with
Bernoulli's theorem. The departure from unity is then a measure of the
error incurred in ignoring compressibility. This error in the calculation
of pressure changes is shown in the following table for several values of
v/G3:
Suppose that the incompressible formula were used for interpreting

the reading of a pitot tube, based on the density a t the stagnation prec
sure; a t what air speed would this formula be in error by 1 per cent? In
Art. 4.9 DEVIATIONS FROM PERFECT GAS LAWS 95
this case p is the static pressure and p, is the pressure measured a t the
mouth of the tube. If V is in error by 1 per cent, then V2 is in error by
2 per cent. Hence we set
( =0.02; v
o m which - r 0.28
4 co co
The latter figure corresponds to an air speed a t normal temperatures of
about 300 ft/sec. At higher speeds the error increases quite rapidly.
Isentropic Formulas in Powers of Mach Number. Expanding Eqs.
4.9, 4.14, 4.16, and 4.19 in powers of M2 by means of the binomial the-
orem, the following convenient formulas for low-speed isentropic flow,
valid up to orders of M4, may be found:
4.9. Deviations from Perfect Gas Laws

Thus far all the isentropic formulas have been based on the two as-
sumptions of a perfect gas, namely, (i) p = pRT, and (ii) c, = constant.
In practice, either of these assumptions may be weak to some extent.
For example, if the process occurs a t very high temperatures but a t
moderate pressures, as in the case of ram jets, there may be appreciable
variations in specific heat. On the other hand, if the process occurs a t
moderate temperatures, but is carried out a t an extremely high pressure
level, as in hypersonic wind tunnels, there may be significant deviations
from the law p = pRT.
These two effects have been studied ('1 for the isentropic flow of air.
Since the analysis is a lengthy one, only the main results of practical
significance are summarized here
Effect of Variable Specific Heat when p = pRT. The equation of

state of a perfect gas is retained at first, but the specific heat is expressed
from quantum mechanics as
where 0 = 5526OR for air. The results are embodied in Fig. 4.13, in
which are shown the per cent errors in the pressure, temperature, and
density ratios incurred through use of constant rather than variable
specific heats. At stagnation tem-
peratures of 1000°R or less, the -16
error is seen to be small for en-
gineering purposes; but, a t tem- -12
peratures greater than 2000°R, the
error can be substantial, especially
a t supersonic Mach Numbers. 2
-8
..
-
L
W
-2 -4
"..
C
a
L m
0 -I
L
a
0
-
W
C
Ql
U 0
a,
4
a
1 8
0 2 4 6 8 10 0 2 4 6 8 10
M M
(4 (d)
FIG.4.13. Error incurred through assumption of constant specific heats for air,
nith p = pRT (after Donaldson).
Effect of Deviations from Perfect-Gas Equation of State. Based on

the use of constant specific heats but the van der Waals equation of state
for air, the isentropic pressure ratio and area ratio are plotted versus M
in Fig. 4.14 for several stagnation pressures. It is seen that the devia-
tions from results obtained with the perfect-gas equation of state are
negligible up to about 50 atmospheres, but are appreciable a t 200 atmos-
pheres and above.
Combined Effect of Variations in Specific Heat and Deviations from
Perfect-Gas Equation of State. The simultaneous effects on pressure
ratio and area ratio of both high temperature level (i.e., variations in
specific heat) and high pressure level (i.e., deviations from p = pRT)
are illustrated in Fig. 4.15. There is an interesting anomaly here in
that the effects of pressure level are in one direction a t low pressures and
in the other direction a t high pressurw.
Art. 4.9 DEVIATIONS FROM PERFECT GAS LAWS 97
Fro. 4.14. Effect of pressure level on isentropic flow functions, using van der Waals
equation for air (after Donaldson).
FIG. 4.15. Simultaneous effects of high prcssure and high temperature on pressure
ratio and area ratio (after Donaldson).
4.10. Performance of Real Nozzles

Because of frictional effects, the performance of real nozzles differs
slightly from that computed with the isentropic flow relations. Since
the departures from isentropic flow are usually small, the usual design
procedure is based on the use of the isentropic flow formulas modified
by two types of empirically determined coefficients-the nozzle efficiency
and the coefficient of discharge.
Nozzle Efficiency. The term nozzle efficiency is employed primarily

in turbine design where it is important to estimate accurately the
average velocity leaving a nozzle
- ve
Pe
passage. Consider a nozzle (Fig.
4.16) supplied with gas a t low
velocity and a t pressure pi and
temperature Ti. The gas expands
adiabatically, but with increasing
T~ entropy, to state e. If it had ex-
T
panded without friction to the
same final pressure, the end state
would have been s. We now define
Tc
the nozzle efficiency as the ratio of
the exit kinetic energy to the exit
Pe
kinetic energy which would be ob-
--
S
tained in a frictionless nozzle ex-
Fro. 4.16. Illustrating definition of
nozzle efficiency. panding the gas to the same final
pressure. With the help of the
steady-flow energy equation the efficiency may be written
This may be rearranged further to give
which is convenient for reckoning the exit velocity when the efficiency,
pressure ratio, and stagnation temperature are all known.
Art. 4.10 PERFORMANCE OF REAL NOZZLES 99
Occasionally the term "velocity coefficient" is used, denoting- the

square root of the nozzle efficiency.
Frictional effects in nozzles are usually confined to thin boundary
layers on the walls. Since the boundary layer thickness depends prima-
rily on the Reynolds Number (based on some equivalent nozzle length)
and on the pressuredistance curve in the nozzle, no simple expression
for nozzle efficiency can be given which is applicable to all noziles. In
general, the nozzle efficiency becomes nearly unity for extremely large
nozzles because the boundary layer thickness is so small compared with
the size of the passage. With very small nozzles, however, the boundary
layer may nearly fill the passage, and then the nozzle efficiency may
drop drastically.
When well-designed nozzles with straight axes are operated a t their
design pressure ratio and a t high Reynolds Numbers, they are found to
have efficiencies ranging from 94 to 99 per cent, and even higher for
sizable wind tunnel nozzles.
Well-designed turbine nozzles with curved axes have efficiencies of
the order of 90 to 95 per cent when operated with suitable pressure ratios
a t high Reynolds Numbers.
Nozzle Discharge Coefficient. The nozzle discharge coefficient, C,,

is defined as the ratio of the actual nozzle flow to the flow calculated from
the isentropic laws for the initial and final pressures of the actual nozzle.
w
C, =
Isentropic Flow Rate
If the over-all pressure ratio of the nozzle is such that the velocity a t
the minimum section is subsonic, then the "isentropic flow" is reckoned
in terms of the exit conditions of the nozzle. However, if the pressure
ratio is such that sonic velocity prevails a t the minimum section, then
the "isentropic flow" is reckoned by using the formula for choking flow
a t the throat. These specifications apply to both converging and con-
verging-diverging passages.
The remarks made previously concerning the factors influencing
nozzle efficiency pertain also to the discharge coefficient.
For well-designed nozzles with straight axes having "pipe" Reynolds
Numbers measured at the minimum area of lo6 or more, the discharge
coefficient is of the order of 0.99, but it may be considerably less for
low Reynolds Numbers.
Neither the discharge coefficient nor the velocity coefficient of rounded-
entrance nozzles suitably designed for the operating pressure ratio are
significantly dependent upon the leaving Mach Number.
Sharp-Edged Orifice Meter. The deviation from unity of the dis-
-
charge coefficient for a sharp-edged orifice meter is due primarily to the
contraction ( v e n a contracts) in the
/ / / / / / / / / / / / / / / / stream following the orifice. The
PI "2 contraction in turn is due to three-
//////,,/,,,/// dimensional effects. The coefficient
of contraction increases substan-
tially as the result of compressi-
0.9
bility effects (Fig. 4.17). It should
0.8
be noted that the isentropic flow
Cw
on which the discharge coefficient
0.7 is based is reckoned as though a
rounded-entrance converging noz-
0.6 zle, having the same exit area as
I 0.6 0.4 0.2 O the orifice, were supplied with gas
PZ /PI a t stagnation pressure pl and dis-
FIG. 4.17. Discharge coefficient of h d to a region having the
sharp-edged orifice meters with zero ve- c arge
locity of approach (after J. A. Perry). pressure pz.
4.11. Some Applications of Isentropic Flow

Thrust of Rocket Motor. Consider a rocket motor (Fig. 4.18a) which
generates gas steadily a t po and To. The nozzle has a throat area At,
an exit area A,, and discharges to an atmosphere a t pressure pa. Experi-
mental data verify that the isentropic flow equations predict, within a
few per cent, the thrust produced by such a rocket.
FIG.4.18. Isentropic flow in rocket motor.

(a) Diagrammatic sketch.
(b) Pressure distributions on internal and external surfaces of diverging portion
of nozzle.
Since rocket motors generate gas a t about 500 psia and operate in
atmospheres a t 14.7 psia or less, a converging-diverging nozzle is usually
used. Except under operating conditions far from the design point,
sonic conditions occur at the throat, and the flow to the nozzle exit is
shock-free. We shall assume these conditions for the present analysis.
Art. 4.1 1 SOME APPLICATIONS O F ISENTROPIC FLOW 101
In the following chapter, however, it is pointed out that a nozzle having

a given area ratio and operating a t certain ratios of back pressure to
supply pressure cannot be isentropic because shocks are present. Such
cases are not covered by the equations about to be derived.
Under the assumed conditions, the pressure ratio pe/po is fixed by the
area ratio, and so the exit-plane pressure in general differs from the sur-
rounding atmospheric pressure. Applying the momentum equation to
the control volume of Fig. 4.18a, we find the thrust 3 to be given by
&hich is then put into dimensionless form through division by poAt:
From Eq. 4.17, for choking flow,
and, from the energy equation and isentropic law,
Substituting these into the thrust equation, and rearranging, there

results
3
Since the pressure ratio p,/po depends only on the area ratio, Eq. 4.33
indicates that the thrust for a nozzle of given size and geometry depends
only on po and the ratio pa/po, and is independent of the temperature To.
EFFECT O F A RE A RATIO. We now ask, for given values of At, pol and
p,, what exit area should be used in order to obtain maximum thrust?
By applying the calculus to Eq. 4.33 it may be shown after a laborious
calculation that 3 is a maximum when the area ratio is chosen i n such a
way as to make the pressure in the exit plane exactly equal to pa. How-
ever, this result may more easily be obtained with the simplest of phys-
ical reasoning. The net thrust on the rocket is the resultant of static
preseures acting on all the surfaces of the motor. Suppose, as in Fig.
4.18b, that there is a certain exit area for which p, = pa. If the nozzle
is continued beyond this point, the pressure in the nozzle will drop
further, and the added piece of divergent nozzle will have negative thrust
FIG. 4.19. Performance characteristics of rocket nozzle, k = 1.2 (after Malina).
because the internal pressure on this added piece is less than the external
pressure. By similar reasoning, i t follows that cutting off a piece of
nozzle upstream of the plane where p, = p, would also act to reduce
the thrust. Hence we conclude that the thrust is a maximum when
p, = pa. Applying this criterion to Eq. 4.33; we get
If the nozzle were a simple converging nozzle, A, would equal At, and
p,/po would be the critical pressure ratio. Making these substitutions
Art. 4.1 1 SOME APPLICATIONS OF ISENTROPIC FLOW 103
in Eq. 4.33 and simplifying, there is obtained
To illustrate the effect of area ratio on nozzle thrust, we form the ratio
--
( 2 ) (l - -A)++ -& ( p e- - -Pa)
k+l
k-1 k + l At Po
- (4.36)
This ratio is plotted against area ratio in Fig. 4.19 for various values
of p,/po, with k = 1.2 (typical value for rocket gas). It is seen that the
curves are quite flat near their maxima, so that the area ratio need not
be exactly adjusted in order to obtain substantially maximum thrust. I n
practice, rocket nozzles are usually designed with p, greater than pa,
since this reduces the size of the nozzle without materially reducing the
thrust.
Reynolds Number for Supersonic Wind Tunnel. Fig. 4.20 shows a
supersonic wind tunnel with a test section having a Mach Number MI
FIG. 4.20. Supersonic wind tunnel.
and in which is inserted a model of length L. We shall derive a con-

venient relation for the Reynolds Number of the model, based on fluid
properties in the test section and on the length L. With the aid of the
perfect gas laws and the isentropic flow relations, we form the expression
Now TI, and accordingly the viscosity pl, are determined by the values
of To and MI. Thus the Reynolds Number per unit length and per unit
stagnation pressure depends, for a given gas, only on the test-section
Mach Number and on the stagnation temperature. A convenient chart
representing this relation for air flow is shown in Fig. 4.21.
With Friction
FIG. 4.21. Reynolds Number for su- FIG. 4.22. Flow in duct with friction.
personic wind tunnel (NACA Tech. Note,
No. 1428).
Supersonic Flow in Duct with Friction. When a duct is supplied with

a supersonic flow by a converging-diverging nozzle, and when the flow
in the duct is adiabatic but ,not frictionless, a useful expression may be
obtained for determining the local Mach Number in terms of the local
area and static pressure. The nomenclature is shown in Fig. 4.22.
Since the nozzle throat is choked, the mass flow through the system
is given by Eq. 4.17 when modified by the discharge coefficient C,:
Furthermore, Eq. 4.11 yields
Dividing one of these by the other, it may be shown by direct com-

parison that
where (A*/A)i,,, is the function of k and M given by Eq. 4.19, and

(p/po)iSe, is the function of 1c and M given by Eq. 4.14b.
Since the type of problem discussed here arises frequently in experi-
mental work, the quantity on the right-hand side of Eq. 4.38 is listed in
the isentropic flow tables. Given the areas At and A and the measured
Art. 4.11 PROBLEMS 10.5
pressures pi and p, together with the discharge coefficient C,, the local
Mach Number M may be found from these tables by a quick computa-
tion. An illustrative example is given in Chapter 6.
Eq. 4.38 typifies a technique which we shall find useful from time t o
time-namely, t o use the tabulated functions of k and M in Appendix B
not only for the particular types of flow underlyi~gthe construction of
the tables, but for such other problems where identical functions of k
and M appear.

1. DONALDSON, C. DTTP.Note on the Importance of Imperfect-Gas Effects and
Variation of Ifeat Capacities on the Isentropic Flow of Gases, NACA R.M.,
No. L8J14 (1948).
PROBLEMS
4.1. Consider the reversible adiabatic flow of steam through a passage of
variable cross section. At the section where the velocity is zero the pressure
and temperature are po = 50 psia and TO = 800°F. Denoting the pressure a t
any other point in the stream by p, plot against p/po the values of specific vol-
ume (ft3/lb), velocity (ft/sec), and mass velocity (lb/ft2 sec), for the following
conditions:
(a) The properties of steam are taken from the Steam Tables of Keenan
and Keyes.
(b) The steam is considered as a perfect gas, with a value of 1.3 for k.
(c) The steam is considered as incompressible with a density equal to the
density corresponding to po and To.
In the above calculations choose for the lowest value of p/po the value of p
which corresponds to the first appearance of moisture in part (a).
4.2. -consider the reversible, adiabatic flow of a perfect gas. Plot the values
of p*/po, T*/To, p*/po, c*/co, Vm,/c*, and Vm,/co, all versus k, for values of
the latter between 1 and 2.
4.3. An airplane flies at an altitude of 40,000 ft (temperature = -67.0°F,
pressure = 2.72 psia) with a speed of 400 mph. Neglecting frictional effects,
(a) Calculate the critical velocity of the air relative to the aircraft.
(b) Calculate the maximum possible velocity of the air relative to the
aircraft.
4.4. Sketch a curve of pressure (p/p*) versus velocity (V/V*) for isentropic
flow, paying special attention to zero or infinite slopes, direction of curvature,
and points of inflection. Indicate the values of p/p* and V/V* a t their maxi-
mum and minimum points, at points of zero or infinite slope, and a t points of
inflection. What is the physical significance of the tangent to the curve of p/p*
versus V/V*?
4.5. A stream of air flows in a duct of 4 inches diameter a t a rate of 2.20
Ib/sec. The stagnation temperature is 100°F. At one section of the duct the
static pressure is 6 psia.
Calculate the Mach Number, velocity, and stagnation pressure a t this section.
4.6. A perfect gas (k = 1.4, R = 100 f t lbf/lbmOR) is supplied to a converg-
ing nozzle a t low velocity and a t 100 psia and 540°F. The nozzle discharges
to atmospheric pressure, 14.7 psia. Assuming frictionless adiabatic flow, and a
mass rate of flow of 1 lbm/sec, calculate
(a) The pressure in the exit plane, in psia
(b) The velocity in the exit plane, in ft/sec
(c) The cross-sectional area of the exit plane, in square feet
4.7. Show that for isentropic flow of a perfect gas, the pressure, temperature,
and density, when made dimensionless with respect to the corresponding critical
values, are given by
4.8. Derive simplified, approximate versions of the isentropic flow relations

for a perfect gas valid a t Mach Numbers large compared with unity.
4.9. A pitot-static tube records a static pressure of 5.20 psig and a difference
between impact pressure and static pressure of 19.42 inches of mercury. The
barometer reads 29.73 inches Hg, and the stagnation temperature of the air
stream is 80°F. Compute the air velocity, (a) assuming the air incompressible,
and (b) assuming the air compressible.
4.10. A stream of air flowing in a duct is a t a pressure of 20 psia, has a Mach
Number of 0.6, and flows a t a rate of 0.5 lb/sec. The cross-sectional area of the
duct is one square inch.
(a) Compute the stagnation temperature of the stream in degrees F.
(b) What is the maximum percentage reduction in area which could be intro-
duced without reducing the flow rate of the stream?
(c) For the maximum area reduction of part (b), find the velocity and pressure
a t the minimum area, assuming no friction and no heat transfer.
4.11. A converging nozzle with an exit area of one square inch is supplied
with air a t low velocity and a t a pressure and temperature of 100 psia and
200°F, respectively.
Plot the mass rate of flow through the nozzle versus back pressure, assuming
the flow to be isentropic.
Ch.4 PROBLEMS 107
4.12. A rocket combustion chamber is supplied with 24 lb/sec of hydrogen

and 76 lb/sec of oxygen. Before entering the nozzle all the oxygen is consumed,
the pressure is 23 atmospheres, and the temperature is 4960°F. Neglecting
dissociation and friction, find the throat area of the nozzle required. Assume
k = 1.25.
4.13. At a certain point in a stream tube, air flows with a velocity of 500
ft/sec and has a pressure and temperature of 10 psia and 40°F, respectively.
(a) Calculate the following quantities a t a point further downstream in the
stream tube where the cross-sectional area is 15 per cent smaller than a t the
upstream section: the stagnation pressure and temperature, the stream pressure
and temperature, the velocity, the Mach Number, and the value of M * .
(b) Compute the maximum possible reduction in area of the stream tube.
For the section with the minimum area, compute the quantities listed in part (a).
4.14. When a body is placed in a stream which at infinite distance upstream

is in uniform flow with free-stream conditions V w , p,, M,, etc., the local pressures
in the neighborhood of the body are usually reported in terms of the dimen-
sionless pressure coefficient, CD:
(a) Show that the value of the pressure coefficient corresponding to the
appearance of the critical velocity is given by
(b) Plot log (-Cp*) versus log M , for k = 1.4 and for values of M, between
0.1 and 1.0.
(c) Suppose that an airplane is flying a t sea level with a velocity of 500 mph.
What is the maximum pressure coefficient which may be attained on the wings
without the speed becoming anywhere supersonic?
4.15. Pressure coefficients, lift coefficients, drag coefficients, etc., of airfoils
which are in a free stream with conditions p,, &, etc., are usually expressed in
terms of the dynamic head of the free stream. Thus
Alternate definitions for compressible flow, not usually employed, are as

follows:
cpr -= PoP -- pPmm .' cL,=PoL/A
- Pw
; etc.
where p o is the isentropic stagnation pressure corresponding to pw and hL,.

Derive an expression for Cp.'/Cp in terms of M, and k. Plot CP1/Cp versus
r& for k = 1.4 and for values of M, between 0 and w.
4.16. (a) Show, for a source-type flow (either two- or three-dimensional)
of a compressible fluid, that either supersonic or subsonic flow may subsist,
but that both types may not exist together.
(b) Show that for a finite flow rate from the source there is a minimum radius
within which a source-type flow pattern is impossible. What is the Mach
Number at this minimum radius? Find expressions for r/r,,,i, as functions of
M and k for the line source (two-dimensional) and the point source (three-
dimensional), where T is the radius at M and r,i, is the minimum radius.
(c) For the four possible types of flow outside the minimum radius, depending
on whether the flow is outwards (source) or inwards (sink) and on whether it is
subsonic or supersonic, specify the directions of the pressure gradient and of
the fluid acceleration.
4.17. Consider the vortex motion of a perfect gas in which all streamlines
have the same entropy and the same stagnation-enthalpy. The equation of
this motion is V r = I'/27r, where V is the tangential velocity, r is the radius of
the streamline, and 'l is a constant called the circulation.
(a) Show that there is a minimum radius inside of which the vortex motion
may not exist, and that this radius is given by
(b) Show that the field of flow outside the minimum radius includes all Mach
Numbers from zero to infinity, and that the radius corresponding to the critical
velocity is given by
4.18. From schlieren photographs of the flow of air through a converging-

diverging nozzle it is found that the average Mach angle over the exit cross
section is 40". The measured static pressure a t the exit cross section is 0.198
atm, while the pressure upstream of the nozzle, where the velocity is small,
is 1.000 atm.
Calculate the ratio of the average exit kinetic energy per unit mass of the
stream to the exit kinetic energy corresponding to isentropic expansion to the
measured exit pressure,
(a) Using the assumption that air is a perfect gas, with k = 1.4
(b) Using the Air Tables of Keenan and Kaye and a measured value for
the stagnation temperature, To,of 2400°F abs
4.19. During a reaction-stand test of a turbojet engine, measurements
indicate a thrust of IS45 lb when the flow rate is 30 lb/sec. The temperature
at the entrance to the thrust nozzle, where the velocity is 300 ft/sec, is 1400°F.
The nozzle has no diverging section, so that the stream reaches atmospheric
pressure, 14.7 psia, somewhere outside the nozzle.
Assuming no heat loss from the gas, that the direction of the air stream enter-
ing the engine is a t right angles to the direction of thrust, and that the nozzle
is frictionless, estimate the pressure in the exit plane of the nozzle.
Ch. 4 PROBLEMS 109
4.20. Derive the following expressions for isentropic flow with the pressure
ratio p/po as a parameter:
V = pk -%l E d =
4.21. Derive relations between M and V/co and between M and V/V-,
for adiabatic flow of a perfect gas.
4.22. Derive a relation between M* and the mass flow parameter
applicable to adiabatic flow of a perfect gas.
4.23. By expanding Eq. 4.11 in a power series of M with the aid of the
binomial theorem, show that for low Mach Numbers the mass flow parameter
may be approximated by
4.24. Derive Eqs. 4.14 without use of the steady-flow energy equation, by
employing Euler's equation for frictionless flow, dp = -pV dV, and the perfect
gas relations p = pRT, c2 = kRT, and plpk = constant.
4.25. Consider a perfect gas flowing in a constantarea duct adiabatically

and without friction. Changes in state come about as the result of changes in
elevation in the earth's gravity field. The z-direction is away from the center
of the earth, and hence gravity acts in the negative z-direction.
(a) Starting from first principles, determine by analysis the direction of
change (increase or decrease) of the Mach Number, gas speed, sound speed,
density, pressure, stagnation temperature, and isentropic stagnation pressrrre,
all for a positive increase in z.
(i) For subsonic speeds
(i) For supersonic speeds
(b) Is choking possible for this type of flow? Justify your answer.
(c) Considering frictionless, adiabatic gas flows for aircraft, fluid machinery,
and yentilating systems, would you expect gravity effects to be significant for
(i) Speeds negligible compared with the speed of sound?
(ii) Subsonic speeds of the order of Mach Number 0.5?
(iii) Speeds very close to the local speed of sound?
(iv) Supersonic speeds of the order of Mach Number 2.0?
(v) Supersonic speeds of the order of Mach Number 10.0?
In analyzing this problem it is suggested that the governing equations be written

in differential form.
4.26. Consider a nozzle with an efficiency q between the inlet and any station
downstream.
(a) Derive the expression
(b) Derive a corresponding expression for
(c) Show that Mach Number unity does not occur at the minimum area, and
find the Mach Number a t the throat.
4.27. Consider a supersonic nozzle constructed with a ratio of exit to throat
area of 2.0. The nozzle is supplied with air at low speed at 100 psia and 140°F.
The over-all nozzle efficiency from inlet to exit is 90 per cent, but the flow is
isentropic .to the throat.
Calculate the pressure, velocity, and Mach Number at exit, and compare
with. the corresponding values for isentropic flow.
4.28. A tank having a volume of 100 ft3 is initially filled with air at 100 psia
and 140°F. Suddenly the air is allowed to escape to the atmosphere (14.7 psia)
through a frictionless converging nozzle of one-inch diameter. It is agreed to
assume that the flow is quasi-steady, i.e., that the steady flow equations may be
applied to the nozzle at any instant of time. Furthermore, the tank is to be
considered as insulated perfectly against heat conduction and as having no
heat capacity.
Plot the pressure in the tank versus elapsed time.
4.29. A supersonic nozzle with a throat area of 1 sq in. discharges air into a
duct having a cross-sectional area of 2 sq in. The supply pressure is 100 psia,
and the nozzle has a discharge coefficient of 0.98. At one section of the duct
the pressure is 14.2 psia. Calculate the Mach Number and isentropic stag-
nation pressure at this section.
Ch. 4 PROBLEMS 111
4.30. Consider the isentropic flow of a highly compressible liquid having a

pressuredensity relation given by
where 8 is a constant.
8= r (g)
(a) Show that
pa - p* = @ I n 2
where po is the stagnation pressure and p* is the critical pressure.
(b) Derive expressions for p / p ~and A/A* in terms of M and 8.
4.31. A large main is connected to an evacuated tank with a volume of 10
3
ft by means of a rounded-entrance,
/'//''''/'''/
- ,,E
converging nozzle having a diameter
of 0.01 in. Initially, a diaphragm " ' / / / / ' / '
over the orifice seals the tank from Main rank
the main. The air in the main is a t ,,,////////,,
100 psia and 70°F. The diaphragm
,
3 ,,/,/,///////,/
is suddenly broken and air rushes P ROB. 4.31.
into the tank. Estimate the time
required for the pressure in the tank to reach 25 psia, based on the following
assumptions:
(i) The flow is quasi-static.
(ii) There is no heat conduction from the tank to the air.
(iii) The pressure and temperature in the main are constant.
Under what circumstances will these assumptions lead to accurate results?
4.32. Show that the coefficient of contraction for a Borda re-entrant orifice
where M is the Mach Number of the jet.

Note that the coefficient goes from 0.50 a t M = O to approximately 0.64 at
M = 1, and that the percentage change is of the same order of magnitude a.
that for ordinary sharp-edged orifices.
Chapter 5
NORMAL SHOCK WAVES

It was shown in Chapter 4 that for certain combinations of initial and
final pressures there are no completely isentropic solutions to the prob-
lem of one-dimensional flow through a converging-diverging passage.
In order to find solutions fitting these boundary conditions, some new
phenomenon must be introduced. A clue to the nature of this phe-
nomenon is given by schlieren (striation) photographs of the flow in
nozzles and around projectiles. These photographs show that a t super-
sonic speeds there sometimes occur very rapid changes in the velocity
and pressure of the stream.
In a real fluid discontinuities in fluid properties cannot exist for finite
periods of time, for viscous, heat conduction, and mass diffusion effects
tend to smooth out such discontinuities. However, photographs of super-
sonic projectiles (Figs. 3.5 and 3.6) show "conical shock waves" which
are so thin that the changes in pressure and velocity which occur across
the wave are effected in an extraordinarily short distance. Further evi-
dence is adduced in Fig. 5.1, showing a normal shock wave in a super-
sonic wind tunnel. Here the pressure increases across the shock by five-
fold, and the velocity decreases by a factor of three; yet the changes
occur in a distance too small to be measured on a photographic plate.
Indeed, theoretical calculations and experimental measurements show
that the thickness of a normal shock is of the order of lop5 inches and
is comparable with the mean free path of the molecules.
For practically all engineering calculations we are concerned primarily
with the net changes in fluid properties across the shock, and have no
interest in the complex viscous and heat conduction phenomena in the
~nteriorof the shock wave. It is profitable, therefore, to ignore the in-
terior details of the shock wave, and to employ a simplified model,
namely, a pure discontinuity.
We therefore focus our attention on the relations between the stream
properties on the two sides of such a discontinuity, and on the question
of whether all such discontinuities are possible. Although experiments
show that these discontinuities are often oblique t o the direction of flow
(as in Fig. 3.5) we will consider now only one-dimensional motion, i.e.,
112
.Art. 5.1 INTRODUCTORY REMARKS 113
only discontinuities normal to the direction of flow (as i n Fig. 5.1). Ob-
lique discontinuities will be treated i n C h a p t e r 16.
Additional material relevant to the s u b j e c t m a t t e r of the present chap-
ter may be found i n Volume 11, C h a p t e r s 25 and 28.
FIG. 5.1. Schlieren photograph of

"normal shock" in supersonic wind tun-
=
nel a t M 2, with flow from left to right.
The straight vertical line is a reference
wire outside the tunnel, and the light par-
allel l;nes are defects in the glass walls
US the tunnel. Theshockitself is not quite
straight. A t each end of the normal
shock is a pair of forked oblique shocks
which arise from the interaction of the
normal shock with the boundary layers
on the walls of the tunnel (M.I.T. Gas
Turbine Laboratory).
NOMENCLATURE
area w mass rate of flow

speed of sound 2, Y Cartesian coordinates
specific heat a t constant pres-
sure P bulk modulus of compressia 3
enthalpy per unit mass 6 shock thickness
ratio of specific heats 'ID diffuser efficiency
mean free molecular path LC coefficient of viscosity
Mach ~ u m b e r P density
V/c*
pressure ( 10 signifies stagnation state
strength of a shock wave ( ) * signifies state at which M = 1
heat ( )m signifies free-stream conditions
gas constant ( )Z signifies conditions upstream c,f
entropy per unit mass normal shock
time ( )Y signifies conditions downstream
absolute temperature of normal shock
specific volume signifies a quantity associated
velocity with a pressure wave
114 NORMAL SHOCK WAVES Ch. 5
5.2. Governing Relations of the Normal Shock

Physical Equations. Suppose that conditions on the upstream and
downstream sides of the discontinuity are denoted respectively by sub-
scripts x and y (Fig. 5.2). Then, since there
Disconli- Control
volume is no heat transfer across the control surface,
the steady-flow energy equation may be written
in the form
h,+?=
vz2 v,2
h v t Z - - ho (5.1)
FIG. 5.2. Control sur-where ho is the stagnation enthalpy on both

face around a normal dis-
sides of the shock.
continuity.
The cross-sectional areas must be the same
m both sides of the discontinuity. From the equation of continuity,
therefore,
W
- = pzvz = PvVv (5.2)
A
Application of the momentum theorem to the flow through the dis-
continuity yields
W
PJ: - P, = -(I.', - VZ) (5.3)
A
Combination of Eqs. 5.2 and 5.3 then gives
Pz +
pzVz2 = P1/ pyvya + (5.4)
The equation of state of the fluid may be written implicitly in the form
h = h(s, P) (5.5a)
s = S(P, P) (5.5b)
The Fanno L i e . Let us fix all conditions a t section x and inquire as
to the conditions a t section y.
The energy equation (Eq. 5.1), the A---------- ' ' o , ~ ~ o ~
continuity equation (Eq. 5.2), and
b
the equations of state (Eq. 5.5)
together define a locus of states
passing through point x, a locus
which in the h-s diagram of Fig. h
5.3 is called the Fanno line.
For example, if a particular
value of Vvis chosen, then p, may
be computed from Eq. 5.2, h, may
he computed from Eq. 5.1, and s,
may be found from Eq. 5.5a. By S
*
repeating this calculation for vari- F I ~5.3.
. Fanno line and Rayleigh line.
Art. 5.2 GOVERNING RELATIONS OF THE NORMAL SHOCK 115
ous values of V,, the Fanno line may easily be constructed. Since the
momentum equation has not yet been introduced, the Fanno line rep-
resents states with the same flow per unit area and the same stagnation
enthalpy, but not necessarily with the same value of the impulse func-
tion. Thus, in general, frictional effects are required to pass continu-
ously along the Fanno line from state x to any other state on the line.
The Rayleigh Line. We next consider the locus of states passing

through point x which are defined by the momentum equation (Eq. 5.4),
the continuity equation (Eq. 5.2), and the equation of state (Eq. 5.5).
To compute this locus, we first select a particular value of V,; then p,
may be computed from Eq. 5.2, p, may be computed from Eq. 5.4, and
s, may be found from Eq. 5.5. By repeating these calculations for var-
ious values of V,, the locus of possible states reachable from state x,
while satisfying Eqs. 5.4, 5.2 and 5.5, may be plotted on the diagram.
The resulting curve is called the Rayleigh line. Since states on the
Rayleigh line have, in general, different stagnation enthalpies, they are
reachable from each other by continuous changes along the Rayleigh line
only through heat transfer effects.
At this point it is well to state that the introduction of the Rayleigh
and Fanno lines in the present discussion is not meant to imply that
the shock process occurs along either or both of these curves. Rather
they are introduced because they help to explain graphically certain
peculiar features of normal shocks, and also because we shall later find
them of use in discussing a number of other one-dimensional flow
processes.
The normal shock itself must satisfy Eqs. 5.1, 5.2, 5.4, and 5.5 simul-
taneously. For a given state x, therefore, the end state of the normal
shock must lie on both the Fanno line and Rayleigh line passing through
x. Hence the intersection of the two lines a t point y represents the
conditions at the downstream side of the discontinuity corresponding
to the prescribed conditions a t section x.
Consider now an infinitesimal process in the neighborhood of the
point of maximum entropy (point a) on the Fanno line. For such a
process, Eq. 5.1 becomes
Similarly, the continuity equation becomes

d(pV) = 0
DdV+ Vdp = 0
From the thermodynamic relation

T ds = dh - dp/p
we see that at point a,
dh = dp/p
Combining Eqs. 5.6, 5.7, and 5.8, and noting that there is no entropy
change for the infinitesimal process under consideration, we arrive a t
The right-hand side represents the local sound velocity; consequently,

we conclude that the Mach Number is unity a t point a.
In similar fashion i t may be demonstrated that M = 1 a t the point
of maximum entropy, point b, on the Rayleigh line. It may also be
shown that the upper branches of the Fanno and Rayleigh lines repre-
sent subsonic speeds, and the lower branches supersonic speeds.
Direction of Shock Wave. The analysis thus far given places no re-
striction on the direction of the process, that is, whether from x to y,
or from y to x. For all fluids thus far investigated, however, point y on
Fig. 5.3 lies to the right of point x. According to the Second Law of
Thermodynamics, the entropy may not decrease during an adiabatic
change. Hence, only the change from x to y is possible. In other words,
the normal discontinuity always involves a change from supersonic to
subsonic speed with a consequent pressure rise, and never the reverse.
This conclusion will, in the following article, be demonstrated rigorously
for a perfect gas.
6.3. Normal Shock in a Perfect Gas

For a given fluid and for a given state x upstream of the shock, Eqs
5.1, 5.2, 5.4, and 5.5 suffice to permit the calculation of fluid properties
a t state y. However, simple relations connecting the properties on the
two sides of the shock cannot be derived unless a simple equation of
state in algebraic form is employed. We turn now, therefore, to the
special case of a perfect gas.
Governing Equations. The energy equation, Eq. 5.1, may be written
in the case of a perfect gas as
Consequently, all the adiabatic formulas of Art. 4.3 are applicable to

states x and y. Furthermore, states x and y have the same adiabatic
Art. 5.3 NORMAL SHOCK IN A PERFECT GAS 117
values of T* and therefore the same adiabatic values of c* and of co.

Using the adiabatic relation between To, T, and M, Eq. 5.10b takes the
form
From the continuity expression, Eq. 5.2, and the relation p = pRT,
This equation is now modified by introducing the definition of M and

the formula for the sound speed in a perfect gas:
from which it follows that
Combining Eqs. 5.11 and 5.12, we find the equation of the Fanno line
in terms of p and M,
For a perfect gas, pV" k p ~ 2 .Hence the combined momentum and

continuity relation, Eq. 5.4, yields the following form of the Rayleigh
line equation in terms of p and M :
Finally, after eliminating pulpz from Eqs. 5.13 and 5.14, we get a
<
relation between M , and M,,

118 NORMAL SHOCK WAVES Ch.5
Working Formulas. Upon squaring and algebraic reamngement,

Eq. 5.15 may be solved explicitly for M,. Two solutions are obtained,
namely,
Mu = M, (5.16a)
and
The first of these solutions is trivial since it expresses the obvious fact
that conditions may be identical a t sections x and y. Inspection of Eqs.
5.11 and 5.13 shows that, for the trivial solution, p, = p, and T, = T,.
The second solution indicates the connection which must subsist be-
tween the states on the two sides of the discontinuity. In the subse-
quent discussion attention is given to the discontinuity solution.
Substituting for M, and M, from Eq. 4.10 gives the following simple
relation between M,* and M,* known as Prandtl's equation:
M,*M,* = 1
or, since c,* = c,*,
v,v, = c*~
From this i t is evident that if the flow a t x is supersonic, the flow a t y

must be subsonic, and vice versa.
To find the ratio of pressures fore and aft of the shock, we substitute
Eq. 5.16b into Eq. 5.14, and thus obtain
An alternative form is
Substituting the value of M; from Eq. 5.16b into Eq. 5.11, we get
for the temperature ratio,
k - 1 ME2) (- 2k M,' - 1)
Tu -
- - k-1 (5.19)
T* (k +
112
MZ2
2(k - 1)
Art. 5.3 NORMAL SHOCK IN A PERFECT GAS 119
The density ratio may be found in terms of Mz from Eqs. 5.18a and
5.19, and the perfect gas law,
and the velocity ratio follows from Eq. 5.2,
The ratio of stagnation pressures is a measure of the irreversibility in

the shock process. It may be found by observing that
Po,- Po, P, Pz
-
P O ~ P, Pz P O ~
Now p,/p, is given by Eq. 5.18a, and po,/p, and pZ/poz may b e found
from Eq. 4.14b. Using Eq. 5.16b for the value of M,, we get, after
algebraic simplification,
To evaluate the entropy change across the shock, we employ the

jwfect-gas formula,
Also, since
and
we get an alternative expression for the entropy change in terms of stag-

mtion temperatures and isentropic stagnation pressures,
In the present case, To, = To,, SO that
or, after substituting the value of po,/poZ given in Eq. 5.21,
Impossibility of a Rarefaction Shock. Careful study of Eq. 5.24 indi-

cates that for gases with 1 < k < 1.67 the entropy change is always
positive when M, .is greater than unity, and is always negative when M,
FIG.5.4. Entropy change across nor- FIG. 5.5. Illustrating the impossibility
mal of a rarefaction shock.
is less than unity. The general form of Eq. 5.24 is shown graphically
in Fig. 5.4. It is thus proven rigorously that for a perfect gas only the
shock from supersonic to subsonic speed is possible.
The irreversibility of the normal shock for perfect gases can also be
demonstrated with the help of the temperature-entropy diagram, Fig.
5.5. Since, for a perfect gas, the enthalpy is a linear function of tem-
perature, the Fanno and Rayleigh curves of Fig. 5.5 are similar to the
corresponding curves of Fig. 5.3. Continuous processes which follow
Art. 5.3 NORMAL SHOCK I N A PERFECT GAS 121
the Rayleigh line involve reversible heat transfer. Imagine that state y
is reached reversibly from state x by adding heat a t constant area along
the Rayleigh line from x to b and then rejecting heat from b to y. Since
the process is reversible, the area under the curve on the T-s diagram
represents the net heat transfer, i.e.,
Now, since the stagnation temperatures are the same a t x and y, the net
heat transfer for the process considered must be zero, and, consequently,
the net area under the curve xby must be zero. The heating from x to b
is at low temperatures, and the cooling from b to y is a t high tempera-
tures. Hence state y must have a greater entropy than state x.
Since s, must exceed s,, the equation preceding Eq. 5.24 leads to the
important conclusion that there is alvays a decrease in isentropic stag-
nation pressure across a shock wave.
The Rankine-Hugoniot Relation. An interesting relation connecting
the pressure ratio and density ratio is known as the Rankine-Hugoniot
equation. Substituting the value of M; from Eq. 5.16b into Eq. 5.11,
we obtain, after algebraic rearrangement,
Then, using Eq. 5.20, we find that
Pz
-+--kk -+1l
P,
PZ
or, solving for p,/pz,
--
Pz k+l P,
-
This relation is plotted in Fig. 5.6, together with the isentropic rela-
tion between pressure and density (p pk). It is seen that for diatomic
gases the density can at most increase by a factor of six, whereas the
pressure ratio may reach infinity. The curves indicate another impor-
tant feature, namely that weak shocks are nearly isentropic; by "weak
shock" we mean a shock in which the percentage rise in pressure is small.
py4 Mx
F I G . 5.6. The Rankine-Hugoniot FIG.5.7. Working curves for normal
mrve, for k = 1.4. shock, for k = 1.4.
5.4. Working Formulas, Curves, and Tables

As in the case of isentropic flow, it is convenient to regard the Mach
Number M, as the foremost parameter of the normal shock equations.
Then, for any chosen value of M,, the corresponding values of M,, pulp,,
Tu/Tz, p,/p,, V,/Vz, and po,/poz are easily computed from the equa-
tions of Art. 5.3.
Art. 5.5 WEAK SHOCK WAVES 128
Fig. 5.7 gives the important shock relations in graphical form for a
gas with k = 1.4.
To facilitate extensive or accurate numerical calculations, the normal
shock functions are tabulated in Table B.3 for k = 1.4.
Illustrative Example. The use of the tables is best explained by means

of a sample problem.
PROBLEM. A stream of air with a Mach Number of 2, pressure of 10 psia, and
temperature of 30°F passes through a normal shock. Calculate the final velocity
and stagnation pressure.
SOLUTION. We are given
From Table B.2 (isentropic flow) we find, at M, = 2
(E) = 0.1278; .
I
10
p , = -- 78.4 psia
0.1278
Also, since
c, = 4 9 . 1 =~49.1- = 1087 ft/sec
it follows that
V, = M&, = 2(1087) = 2174 ft/sec
Now, from Table B.3, at M, = 2.00,
V,/Vv = 2.667; p@/po, = 0.7209
whence
Vv = 2174/2.667 = 815 ft/sec
p~ = (78.4)(0.7209) = 56.5 psia
5.5. Weak Shock Waves

Certain important features of shock waves may be illuminated by
working out special relations for shocks of small strength.
Strength of Shock Wave. Let us define the strength of a shock wave as
the ratio of the pressure increase to the initial pressure, i.e.,
Rearrangement of Eq. 5.18a yields

Also, from the Rankine-Hugoniot relation, Eq. 5.27,
From Eq. 5.29 it appears that the term ( ~ -21) is directly propor-
tional to the shock strength. If the shock is relatively weak, Eq. 5.30
indicates that the ratio of density increase to initial density,
is also approximately proportional to the shock strength. Hence

(M,' - 1) and the percentage increase in density may be chosen as
alternative measures of shock strength.
Shock of Vanishing Strength. Suppose that the shock strength ap-
proaches zero. Then, from Eq. 5.29, the value of M, approaches unity.
Furthermore, from Eqs. 5.16b and 5.30, Mu approaches unity, and the
density change across the shock approaches zero. A shock of infinitesi-
mal strength is therefore identical with a sound wave. Since M, is
unity for such a shock, i t follows that a very weak pressure disturbance
travels with the speed of sound with respect to the fluid in which it is
propagating. This same conclusion was reached in Chapter 3. An
infinitesimal plane pressure wave is, therefore, the limiting case of a
normal shock. The shock itself travels with a speed greater than the
speed of sound in the fluid into which it is propagating, the ratio of the
two speeds being M,.
Irreversibility of Weak Shock. It is of interest to examine the irre-
versibility associated with a weak shock of finite strength. The irre-
versibility will be measured in terms of the entropy increase in passing
through the shock.
Rearrangement of Eq. 5.26 yields
Now, from Eq. 2.3213,

Art. 5.6 FORMATION OF SHOCK WAVES 116
Inserting the previous expression for pu/pZ, we find
(5.31;
Now, for weak shocks, each of the terms on the right-hand side of
Eq. 5.31 is of the form
In (1 +
E)
where E is much smaller than unity. We therefore employ for each term
the series exvansion
Carrying out these operations and simplifying, we find that the first-
order and second-order terms in the shock strength vanish, leaving
In terms of (MZ2- l), we find by substituting Eq. 5.29 into Eq. 5.32b
that the entropy increase is given by
Thus, for weak shocks, the entropy increase is of the third order with
respect to the shock strength. As a first approximation, therefore, the
irreversibility connected with weak shocks may be ignored, and the
isentropic relation between pressure and density may be used for con-
necting the states on both sides of the shock.
This result explains why in Fig. 5.6 the isentropic curve and Rankine-
Hugoniot curve have the same slope and the same radius of curvature
a t the lower left-hand corner of the chart. It explains also why in Fig.
5.7 there is scarcely any loss in stagnation pressure across the shock
for initial Mach Numbers less than about 1.25.
5.6. Formation of Shock Waves

We have seen from the Second Law of Thermodynamics that a com-
pression shock involves an entropy increase (or a loss in stagnation
pressure), and that rarefaction shocks are impossible. Two questions
come naturally to mind: (i) what are the mechanical reasons underlying
the appearance in nature of compression shocks and the impossibility

of rarefaction shocks? and (ii) what are the dissipative phenomena in a
shock leading to an entropy increase? The first of these questions is
answered in this article, and the second in Art. 5.7.
To explain in the simplest manner the physical mechanism leading to
compression shocks, we shall adopt the point of view of Art. 3.2. There
we regarded each element of surface bounding a fluid stream as pushing
the neighboring fluid out of its way and thereby propagating a pressure
pulse into the fluid. It is also necessary to
=
+ point out that a steady-state flow is always
arrived a t through a series of transient flows
which asymptotically approach the steady-
"k ,=I-
PL state condition after a long period of time.
:
t=2
Therefore, in order to explain the existence of
shock waves in a duct or attached to a moving
m:4;
a : .
body, i t is necessary to examine the way in
which the steady-state condition is achieved
from a condition of no motion.
ph; :. !:
Physical Description of Wave Development.
The simplest physical process incorporating the
elementary ideas expressed above is illustrated
p k X * in Fig. 5.8. The piston accelerates rightward
i . e,f into gas which is initially motionless in the
0 ib:=:di
I
duct. For purposes of analysis we will sup-
FIG. 5.8. Wave fronts
formedat successive times Pose that the gradual acceleration is approxi-
as piston accelerates mated by a series of instantaneous increases of
rightward by a series of
spaced impulses. speed occurring a t equal intervals of time, with
the piston traveling a t uniform speed between
each pair of impulsive accelerations. Each successive pressure wave
produced follows the laws derived in Art. 3.2.
In Fig. 5.8 the pressure distribution in the duct is shown schematically
at various times after the beginning of motion. The position of the
vertical axis on each chart indicates the instantaneous location of the
piston. After the first impulsive acceleration of the piston, i.e., a t t = 1,
the pressure wave (shown as a vertical front) has moved down the duct
a short distance and influenced the mass of gas labeled "a." The gas
mass "a" is therefore a t a slightly increased pressure and is moving to
the right with the velocity of the piston. Between t = 1 and t = 2,
the mass l Ia ,7 receives a further increase in pressure and velocity from
the second acceleration of the piston, and the original wave travels
along the duct and imparts a pressure and velocity pulse to mass 'lb."
This process continues as each of the pressure waves travels downstream
and as fresh waves are initiated by successive piston accelerations.
3
Now each of these pressure waves travels with the local speed of sound
relative to the fluid through which it is passing. But the masses nearer
the piston have forward motions greater than those further from the
piston. Moreover, since the process is isentropic, the masses nearer
the piston have greater sound velocities by virtue of the higher tempera-
tures associated with their greater pressures. Consequently, the pressure
waves nearer the piston tend to overtake those further from the piston.
For example, referring to t = 3, the wave in mass "a" travels faster than
the wave in mass "b," because "a" is moving faster to the right and
also because the velocity of sound in "a" is greater than the velocity
of sound in "b."
The net result of this process is that the wave profile becomes steeper
and steeper, until, a t t = 6, the pressure gradient becomes infinite.
Thus, a t t = 6, a small compression shock has been formed, a shock
which grows in strength as the process continues.
A similar analysis applied to a leftward motion of the piston, i.e., for
an expansion wave, indicates that expansion waves become less steep
with the passage of time. Thus, because of the nonlinearity of finite
wave motion, compression waves become steeper and ultimately form
a discontinuity, while expansion waves spread out and thus are unable
to support a discontinuity.
These results are summarized in Fig. 5.9, showing successive appear-
ances of a wave form of constant total pressure change, corresponding
(a) (b)
FIG.5.9. Development of wave form of constant strength.
(a) Compression wave.
(b) Expansion wave.
to the case where the piston accelerates to a constant speed and then
moves a t constant velocity; this is analogous to the case of a projectile
starting from rest and accelerating to a constant final speed. The wave
form a t a giveri instant is defined as the curve of some fluid property
(say p, p, or V ) p~dttedagainst distance 2.
STEEPENING OF COMPRESSION WAVES. The compression wave, Fig.
5.9a, becomes ever steeper as it propagates rightward, until, between
t = 4 and t = 5, i t tends to become infinitely steep a t some point. The
present analysis then ceases t o be valid because viscous and heat con-
duction effects are no longer negligible in the face of such extraordinary
gradients in velocity and temperature. Indeed, if the isentropic analysis
were continued t o t = 5, the wave would "topple over" to the form shown
by the dashed line; this is physically absurd, as i t means that a t the
same time and location the fluid has simultaneously three diierent values
of pressure, velocity, and density. Viscous and heat conduction effects
intervene soon after t = 4 and act to produce a stationary shock wave
of unchanging form.
SPREADING OF RAREFACTION WAVES. The rarefaction wave, Fig.
5.9b, becomes ever less steep as it propagates rightward, and so no dis-
continuity effects are observed. Indeed, if a rarefaction shock were
momentarily established, the dynamic effects outlined here would cause
the shock to decay immediately into a continuous expansion wave.
T HE SURF ANALOGY. A similar phenomenon may be seen a t ocean
beaches. As an incoming gravity wave approaches the shore, the front
part of the wave (corresponding to compression) is seen to steepen, and
the rear part (corresponding to
rarefaction) is seen to spread out.
In this case, however, the geome-
try of the flow permits the for-
ward part to tumble over, as in
Fig. 5.9a, t = 5; this tumbling
over produces the well-known
ocean "breakers. "
Analysis of Wave Develop-
ment. The foregoing descriptive
arguments may be put into
simple analytic form.
Consider a rightward-moving
plane compression wave, the in-
stantaneous form of which is
shown in Fig. 5.10 by plots of
pressure and velocity versus dis-
FIG. 5.10. Illustrates analysis on de-
velopment of wave form. tance. At two adjacent points
the fluid properties differ i n
magnitude by dV, dp, dp, dc, etc. The respective parts of tlw
wave passing through these same points differ in wave speed by t,hc
amount dV,. Since a finite wave may be thought of as a succession of
infinitesimal pressure pulses, each element of the wave may be analyzed
as in Art. 3.2. As long as the velocity and temperature gradients are
moderate, experience confirms the assumption that longitudinal viscous
and heat conduction effects are negligible. Consequently each ele-
mentary part of the wave travels a t the local speed of sound with respect
to the fluid in which i t is propagating. The propagation velocity of a
part of the wave with respect to fixed coordinates is, therefore,
and the propagation velocity of an adjacent part of the wave is
whence i t follows that

dl', = dV + dc
We are interested in whether adjacent parts of the wave travel a t the
same or different speeds. Therefore we form the ratio
From the analysis of Art. 3.2 we obtain, for a rightward-moving wave,
Furthermore, since each particle of fluid undergoes isentropic changes,

and we assume that the entire fluid was originally a t rest with uniform
pressure and temperature, the increments in pressure and density b e
tween adjacent fluid particles obey the relation
Differentiating with respect to p, we have
Substituting Eqs. 5.35 and 5.36 into Eq. 5.3413, we obtain
By definition, p = l / v , where v is the specific volume. Hence
Applying this transformation to Eq. 5.37a, we obtain

For a fluid to be thermodynamically stable (i.e., if it is not to collapse
or expand catastrophically), (dpldv), must be negative. Hence, the
isentrope must have a negative slope on the pressure-volume diagram.
Consequently, the sign of dVw/dp depends exclusively on the sign of
(dZp/dv2),, that is, on whether the isentrope on the pressure-volume
diagram is concave upwards or concave downwards.
CRITERIA FOR C HANGE I N WAVE SHAPE. NOW, if dVw/dp is positive,

the higher-pressure parts of the wave overtake the lower-pressure parts,
and a compression wave steepens as it progresses; similarly, a rarefaction
wave becomes less steep. But if dVw/dp is negative, a compression
wave becomes less steep and a
rarefaction wave steepens into a
compression shock. From these
Steepen considerations we arrive a t the
following criteria:
(i) Compression waves steepen
P P=A + BV and expansion waves flatten when
Wave Form
Unchanged
(d2p/dv2)s> 0 (5.38a)
(ii) Compression waves flatten
and expansion waves steepen
Compressions Flatten
when
Rarefactions Steepen
V ' l/P
- (dZp/dv2) < 0 (5.38b)
F I ~5.11.
. Criteria for steepening or The shapes of the isentropes
flattening of compression and rarefaction corresponding to these two cri-
waves in terms of curvature of p-u isen-
trope. teria are illustrated in Fig. 5.11.
GAS WITH UNCHANGING WAVE SHAPE. When the isentropic pressure-

density relation has a special form, waves of finite amplitude propagate
through the gas with unchanged shape. According to Eq. 5.38, this
occurs when
Integrating twice, we obtain
as the unique isentropic relation for unchanging wave shape, where A

and B are constants. This p-v isentrope is a straight line on the p-u
diagram of Fig. 5.11 and marks the dividing line in curvature between
gases in which compression waves steepen and gases in which they
flatten.
Art. 5.7 THICKNESS O F SHOCK WAVES 131
PERFECT GAS. I n a perfect gas, is constant during an isentropic

process. By differentiating twice i t is found that
Since k is positive for all known gases, it follows from Eq. 5.38a that
in a perfect gas compression waves steepen and rarefaction waves flatten.
LIQUIDS. The isentropic relation for liquids may usually be repre-
sented a t moderate pressures by
where fi is positive and nearly constant. Differentiating, we get
Comparing this with Eq. 5.38a, we conclude that, in liquids obeying the
law of constant p, compression waves steepen and expansion waves
flatten.
OTHER FLUIDS. NO example has yet been adduced of a real fluid obey-
ing Eq. 5.3813. However, there seems to be no basic principle forbidding
the existence of such a fluid. Possibly certain fluids in the neighborhood
of the critical pressure and temperature can exhibit stationary rare-
faction shocks.
5.7. Thickness of Shock Waves

We now turn to the question of how dissipative effects account for
the entropy increase in a shock wave.
In the foregoing article we saw that, in the absence of viscosity and
heat conduction, compression waves tend to become infinitely steep.
The analysis underlying this result was based on a form of the equation
of motion containing only pressure and inertia forces. However, as the
wave becomes very steep, viscous stresses must inevitably become
appreciable, no matter how small the coefficient of viscosity. Moreover,
heat conduction effects must inevitably become substantial, no matter
how small the coefficient of thermal conductivity, and so a particle of
fluid undergoes nonadiabatic effects.
Since viscous and heat conduction effects tend to wipe out discon-
tinuities in velocity and temperature, they tend also to resist the s t e e p
ening of a compression wave. From this it follows that the thickness of
a shock wave is in part controlled by viscous snd heat conduction effects.
In short, when a compression wave has reached a stationary form, the
spreading influences of viscosity and heat conduction just balance the
steepening influence of pressure and inertia forces.
Order-of-Magnitude Considerations. The ideas outlined above lead
to a simple estimate of the order of magnitude of the shock thickness.
In a shock of stationary form, the pressure, viscous, and inertia terms
1%-
in the equation of motion must all
-
6 --T---vx be of comparable magnitude.
Physical considerations dictate
that the stationary shock have a
form like that of Fig. 5.12. The
v ------L-- m velocity curve approaches the end
values V z and V , asymptotically,
Fro. 5.12. Definition of shock thickness. and so the thickness is actually
infinite. However, virtually the
entire change in velocity occurs in a very short distance. It is con-
venient, therefore, to define a characteristic shock thickness 6 as shown
Now, since the longitudinal viscous stress must be of the same order
of magnitude as the inertia stress,
or, evaluating the derivatives in order-of-magnitude fashion, and
from which we get
which says that the Reynolds Number of the shock, based on the thick-
ness 6 and on the fluid properties a t T* is of the order of magnitude of
unity.
By introducing certain further relations from the kinetic theory of
gases, it may be shown that, in respect to order of magnitude,
where I is the mean free molecular path. Thus it follows that the shock
thickness is of the order of the mean free path. Thermodynamic equilib-
Art. 5.7 TH1,CKNESS OF SHOCK WAVES 133
rium does not prevail within a shock, therefore, and the analysis of a
shock from continuum considerations a t best gives approximate results
for the shock structure. Kinetic theory or even quantum mechanics
may be necessary for an adequate analysis.
Shock Thickness in Perrect Gas. The structure of the shock wave
may be investigated more fully by solving the exact Navier-Stokes
equation and the exact energy equation for such a dissipative, one-
dimensional region. The results of such studies are more accurate than
the order-of-magnitude result reached above but are nonetheless subject
to the same weakness of being based on the assumption of a continuum.
A formula showing a shock thickness Reynolds Number as a function
of Mach Number, Prandtl Number, specific-heat ratio, and viscosity-
temperature variation has been derived by Shapiro and Kline.") It
has the form
where
4 2k
D=-+--- k + 1 M , * ~+ 1
3 Pr* 2Pr* M,*
of the form p -
and where n is the exponent in a viscosity-temperature relationship
Tn (from kinetic theory, n = for perfect gases; for
air, experimental values yield n = 0.768 a t ordinary temperatures).
The symbol Pr* denotes the value of the Prandtl Number a t the tem-
perature T*. When D is positive, the plus sign in Eq. 5.39a is used;
when D is negative, the minus sign.
VERY W EAK SHOCKS. When M,* approaches unity, Eq. 5.39a may
be approximated by
Hence the thickness of weak shocks is inversely proportional to the

strength of the shock if we choose the strength to be measured by
( M *- 1 Even for extremely weak shocks, however, the shock thick-
ness a t normal pressure and temperature is extremely small compared
with the dimensions of physical objects, as may be seen from Fig. 5.14.
TABLE OF SHOCK THICKNESSES. Using constants corresponding to
air (k = 1.4, Pr* = x, and n = 0.768), Eqs. 539a and b yield the
values shown in the following table for the shock thickness parameter.
Also shown are order-of-magnitude

- values for the ratio of shock thickness
to the mean free molecular path a t state x.
In the range where measurements of shock thickness have been made

(up to about M, = l.5), Eq. 5.39a is in excellent agreement with experi-
A
Tox 'Toy ----
/.-.
, C
%=
4
-.
','
./.
T
Poth of
V,
2 3 0 11200
V
(Ft/Sec)
- -EjNorrnol Shock
M,=I.I, T,=59F, pX*14.7psia
P/P, 1.3
1.2
P 'P,
Norrnol
Shock
/
A\@
8
'
1100 T/T, T/T,
T ~ f1 .l1 ~ #'
VY
(1054r'--q -3 -2 - 1 0 1 2 3 4 1.0
X r lo5, Inches S
+
(a) (b)
FIG.5.13. Variation of fluid properties in a normal shock.
mental observation for monatomic gases such as argon, and under- -
estimates the shock thickness by about 25%

for diatomic gases like nitrogen.
Fig. 5.13a shows typical velocity, tempera-
ture, and pressure distributions in a weak shock
wave. The corresponding course of thermo-
dynamic states is shown in the temperature-
entropy diagram of Fig. 5.13b, which indicates
that entropy decreases occur near the end of the
wave. The thickness of a compression shock a t
lMx-l) normal atmospheric pressure and temperature
F I ~5.14.
. Thickness of (Fig. 5.14) is extraordinarily small, thus justi-
a normal shock a t stand- fying the treatment of the shock as a discon-
ard atmospheric pressure
and temperature. tinuitv for most practical computations.
Art. 5.8 NORMAL SHOCKS IN DUCTS 135
5.8. Normal Shocks in Ducts

When the pressures a t the entrance and exit of a duct in which a gas
flows supersonically are adjusted so that a compression shock appears in
the duct, the latter is found to be radically different from a plane dis-
continuity. This differencearises as the result of an interaction between
the shock wave and the boundary layer on the walls of the duct.
The phenomenon in its details is most complicated, but the outstand-
ing physical features can be readily explained. The normal shock in the
core of the stream imposes an extraordinarily large adverse pressure
gradient on the stream as a whole. As the boundary layer flow has been
retarded by wall friction, however, it does not have sufficient momentum
to negotiate this adverse pressure gradient. Consequently, apart from
the thickening of the boundary layer produced by the adverse pressure
gradient, there is often a back flow in the boundary layer near the shock.
This back flow produces a separation of the stream from the walls.
The boundary layer effects described above lead to oblique shock waves
in the pattern, and consequently the flow is far from one-dimensional.
Supersonic Subsonic
Oblique Shocks Region of Separation

(4 (b)
FIG.5.15. Typical shock-boundary layer interactions in a duct.
(a) Thin boundary layer and weak shock.
(b) Thick boundary layer and strong shock.
Typical Experimental Results. Typical flow patterns are shown in

Fig. 5.15. When the shock is weak and the boundary layer thin, as
might be the case slightly downstream of the throats of supersonic
FIG. 5.16. Schlieren photographs q f compression shock in a duct, with flow from
left to right (M.I.T. Gas Turbine Laboratory).
(a) Moderately thick boundary layer. The light, diamond-shaped regions rep-
resent regions of acceleration.
(b) Very thick boundary layer.
no~zles,the normal shock extends over most of the stream and a thick-
ening of the boundary layer occurs (Fig. 5.15a). When the shock is
strong and the boundary layer thick, as in long ducts fed by supersonic
nozzles, back flow and separation occur (Fig. 5.15b). The main stream
136 NORMAL SHOCK WAWS Ch.5
separates from the walls, alternately passes through a series of accelera-
tions and shocks, and, finally, after reaching subsonic speeds, diverges
and fills the passage again.
(a) (bj
FIG.5.17. Schlieren photographs showing the effect of boundary-layer removal on
the compression shock in a duct, with flow from left to right (after A. Weise).
(a) Boundary layer present.
(b) Boundary layer removed through suction slits.
Typical schlieren photographs of these phenomena are shown in Figs.

5.1 and 5.16. That the boundary layer is indeed respohsible for the
complex flow pattern with re-
0.035 I I I peated shocks is seen from the
schlieren photographs of Fig. 5.17,
0.030 - where it is evident that the simple
normal shock can be made to
occur by removing the boundary
0.025 - layer from the walls of the pas-
sage. This figure also indicates the
0.020 - possibility of eliminating complex
and lengthy shock patterns by
P
means of boundary layer suction.
Fig. 5.18 illustrates a typical
static pressure distribution meas-
ured a t the wall of a duct contain-
ing a compression shock.(2) I n the
particular case shown, for which
the Mach Number a t the begin-
ning of the shock was 2.8, about
ten diameters of duct were re-
quired for the pressure rise to be
Distance From Tube Inlet attained. In other tests of similar
Tube Diameter nature the length of the "shock"
FIG. 5.18. Measured static pressure was found to vary 'between 14
distribution a t the wall of a duct contain-
ing a compression shock like that of Fig. diameters for a Mach Number of
5.16b (after E. G. Newberg, Jr.). 4.2, and 9 diameters for a Mach
Number of 1.8. These figures are
doubtless a function of the Reynolds Number and of the boundary-layer
thickness. Measurements of the maximum pressure rise for the steeply
Art. 5.9 MOVING SHOCK WAVES 137
ascending portions of the pressure-distance curves indicated that the

measured value of p u l p , was about 6 per cent less than the value corre-
sponding to the normal shock equations and thc measured value of M,.
This agreement is good, considering the fact that a t neither section x nor
section y is the flow strictly one-dimensional. The good agreement
between the measured and calculated pressure rises is explained by the
fact that the wall shearing forces in the region of separation in Fig. 5.15b
are extremely small, and hence the normal shock equations are approxi-
mately applicable between sections x and y. From this we reach the
important conclusion that, although the compression shock in a duct is
far from a plane discontinuity, the normal shock equations are a good
guide to the changes in fluid properties across such a shock region. It is
important to remember, however, that a substantial distance is required
for the net change in state to be effected.
Additional information on shocks in ducts is presented in Volume 11,
Chapter 28.
5.9. Moving Shock Waves
In a number of practical problems, such as those relating to explosion
waves, the V-1 "buzz bomb" engine, and the Comprex, i t is necessary
to deal with compression shocks which are not stationary. If in such
cases the shock wave travels a t constant speed, the problem may be
reduced to that of a stationary shock merely by employing a moving
coordinate system relative to which the shock is a t rest; the shock rela-
tions already derived are then applicable to the fluid properties in this
moving coordinate system. The same result is true to a very good
approximation even when the shock strength is changing with time or
when the shock speed is not constant; this is so because the shock thick-
ness is minute, from which it follows that the time rates of change of
mass, momentum, and energy within a control surface surrounding the
shock are negligible compared with the changes in the respective fluxes
of these quantities passing through the control surface.
Transformation of Moving Shock to Stationary Shock. To illustrate

the general approach by which the stationary shock relations can be
made applicable to a moving shock, consider a moving shock wave such
as mighthave been created by an explosion in a gas a t rest (Fig. 5.19b).
By superposing on the pattern of Fig. 5.19b a uniform velocity to the
right, the stationary shock pattern of Fig. 5.19a is obtained. In Fig.
5.19a, gas approaches the shock with the speed V z and pressure p,,
and with the supersonic Mach Number M,. It leaves the shock with a
higher pressure, lower speed, and with a subsonic Mach Number. In
the case of the moving shock, the gas over which the shock has passed
is a t a higher pressure than the undisturbed gas. The gas over which the
shock has passed travels leftward with the speed Vz- V,, and, since
this is less than the shock speed V,, a particle of high-pressure gas falls
farther and farther behind the shock front.
Let us define the Mach Number of the moving shock wave as the ratio
of the speed of the wave to the speed of sound in the stationary gas.
This ratio is identical with the value of M, for the stationary discon-
tinuity. Now, since for a wave of finite strength M, is greater than unity,
Stationory Shock Moving Shock
Stationory Gas
"Y '"x
(a) (b)
FIG.5.19. Transformation of coordinate system.
(a) Gas flowing through stationary shock.
(b) Shock moving into stationary gas.
it follows that such waves travel faster than the speed of sound in the
undisturbed medium. Only in the case of a wave of infinitesimal
strength does the wave propagate with the speed of sound.
Transformation Formulas. The equations of Art. 5.3 are, of course,
valid only in terms of Fig. 5.19a1i.e., only for quantities measured by an
observer who travels with the discontinuity. In order to obtain equa-
tions which are applicable to quantities seen by an observer a t rest with
respect to the low-pressure gas, all quantities containing a velocity term
must be modified in accord with the change in coordinate system.
Suppose we signify with primes those quantities measured by an
observer who is a t rest with respect to the gas toward which the dis-
continuity moves. Then we may write
Tz' -
- Tz; T,' = T, (5.41a, 5.41b)
M,' = 0 (5.42a, 5.4213)
Art. 5.10 CONVERGING-DIVERGING NOZZLE 139
Through the use of these transformations and the equations of Art

5.3, the shock relations may be worked out for the observer of Fig. 5.19b
It is worthy of note that the
change in stagnation temperature
is dependent on the observer's
motion, as indicated by the fol-
lowing expressions:
ATo = To, - To,
= T, - T, + vy2 - vz2
2%
ATO' = TO; - Toz'
= T, - T, + (Vz - V J 2
2%
and, since ATo = 0, we find that
VZ(V, - V,)
AT,' =
CP
FIG. 5.20. Metaphorical shock propa-
Similar transformation relations gating into a flowing medium which has
may be out for the case been suddenly brought to rest. (Reprinted
courtesy R. Courant and K. 0.Friedrichs,
where the flow in a pipe is sud- Supersonic Flow and Shock Waues, Inter-
denly brought to rest at one end, science Publishers, New York copyright
1948.)
thus initiating a shock wave which
propagates back through the pipe. The corresponding physical situa-
tion is pictured (with some poetic license) in Fig. 5.20.
5.10. Operating Characteristics of

Converging-Diverging Nozzle
We return now to the problem of the operating characteristics of
converging-diverging nozzles under varying pressure ratios, discussed
previously in Art. 4.7.
140 NORMAL SHOCK W A V m Gh.5
Simplified Theoretical Analysis. Fig. 5.21a shows the experimental
arrangement considered for purposes of discussion. Figs. 5.21b1c, and d
show the characteristics computed on the basis of frictionless flow, but
admitting the possibility of normal shocks. The measured charaeteris-
tics are in general agreement with those illustrated here and T i e r chiefly
because of boundary-layer effects.
-
Po ---
LOCUS of
*
P * / P ~
States Downstream
of Normal Shock
(61
(71
Distance Along Nozzle
(a)
IY
F I ~5.21.
. Performance of converging-diverging nozzle with various ratios of back
pressure to supply pressure.
(a) Curves of pressure versus distance along nozzle axis.
(b) Exit-plane pressure versus back pressure.
(c) Throat pressure versus back pressure.
(d) Mass flow parameter versus ratio of back pressure to supply pressure.
Four different regimes are possible. I n regime I the flow is entirely

subsonic, and the passage behaves like a conventional venturi tube. The
flow rate is sensitive to changes in back pressure.
At condition 2, which forms the dividing line between regimes I and
11, the Mach Number a t the throat is unity. As regime I1 is entered,
a normal shock appears downstream of the throat, and the process aft
of the shock comurises subsonic deceleration. As the back uressure is A
lowered, the shock moves down the nozzle until, at condition 4, it appears
Art. 5.10 CONVERGING-DIVERGING NOZZLE 141
in the exit plane of the nozzle. I n regime 11, as in regime I, the exit-
plane pressure p~ is virtually identical with the back pressure p ~ .On
the other hand, the flow rate in regime I1 is constant and is unaffected
by the back pressure. This is in accord with the fact that throughout
regime I1 all stream properties a t the throat section are constant.
In regime 111, as for condition 5, the flow within the entire nozzle is
supersonic, and the pressure in the exit plane 1s lower than the back
pressure. The compression which subsequently occurs outside the nozzl
involves oblique shock waves which cannot be treated on one-dimensional
grounds.
Condition 6 is termed the design condition for the nozzle under super-
sonic conditions, since the exit-plane pressure 1s then identical with the
back pressure. A reduction in the back pressure below that corre-
sponding to condition 6 has no effect whatsoever on the flow pattern
within the nozzle. In regime IV the expansion from the exit-plane
pressure to the back pressure occurs outside the nozzle in the form of
oblique expansion waves which also cannot be studied by a one-dimen-
sional analysis.
In both regimes I11 and IV the flow pattern within the nozzle is inde-
pendent of back pressure, and corresponds to the flow pattern for the
design condition. Adjustments to the back pressure are made outside
the nozzle.
For subsonic flow, there are an infinite number of possible pressure-
distance curves. For the supersonic region of flow, however, the pres-
sure-distance curve is unique. To put i t differently, in subsonic flow
the pressure ratio does not depend solely on the area ratio; in supersonic
flow the pressure ratio does depend solely on the area ratio.
Only over a narrow range of back-pressure ratios, namely, the range
covered by regime I, does the flow rate depend on the back pressure.
For regimes 11,111,and IV, the flow rate is independent of back pressure,
and, since M = 1 a t the throat, may be computed from Eq. 4.17.
Experimental Results. Fig. 5.22 shows schlieren photographs of the

flow through a converging-diverging nozzle. Figs. 5.22a, b, and c corre-
spond to regime I1 and show a normal shock advancing down the nozzle
as the back pressure is lowered. I n Figs. 5.22a and 5.2213 the shock is
weak and the boundary layer thin, and there is a single and substantially
normal shock. In Fig. 5 . 2 2 ~
the shock is fairly strong and there is bound-
ary-layer separation with a repeated shock pattern. With sufficiently
low back pressure (regimes I11 and IV) the shock moves out of the
nozzle, and the flow in the passage is entirely supersonic, as in Fig. 5.22d.
Measured pressure distributions on the nozzle axis are shown in Fig.
5.23. They are seen to agree generally with the simple model of Fig.
5.21a, the chief difference being a gradual rather than a steep rise in
FIG.5.22. Schlieren photographs of flow through convergingdiverging nozzle,

with flow from left to right (after Prandtl).
(a) Weak shock just downstream of throat.
(b) Shock further downstream than in (a).
(c) Strong shock accompanied by flow separation.
(d) Back pressure low enough to permit purely supersonic flow in passage. In-
clined lines are Mach wavelets generated by slight roughness on the walls.
0-05~ 1 2 3 4 5 6
Distance from Throol, In
Fro. 5.23. Experimentally determined curves of static pressure on axis versus

distance along axis for converging-diverging nozzle (after Stodola).
Art. 5.11 ONEDIMENSIONAL SUPERSONIC DIFFUSERS 143
pressure across the shock. This diierence is caused by boundary-layer

thickening and separation.
Flow pa.tterns in the exit jet for regimes I11 and IV are illustrated by
the schlieren photographs of Fig. 5.24. Analytical studies of the oblique
waves seen in these pictures are presented in Chapters 14, 15, and 16.
FIG. 5.24. Schlieren photographs of flow a t exit of converging-diverging nozzle,

with flow from left to right (after Prandtl).
(a) Pressure in exit plane greater than back pressure (regime IV).
($) Pressure in exit plane equal to back pressure (condition 6).
(c) Pressure in exit plane less than hack pressure (regime 111).
6.11. One-Dimensional Supersonic Diffusers

Diffusers, or passages which decelerate a stream to low velocity, are
important elements in such diierent devices as compressors, wind tun-
nels, and ram jets. The supersonic diffuser offers certain unusual prob-
lems not met with in the design of subsonic diffusers.
Special Problems of Supersonic Diffusers. At first thought it might
appear that a supersonic diffuser could be designed as though i t were the
reverse of a converging-diverging nozzle. Two difficulties arise, however.
First, if the diffuser is in a closed system and is preceded by a supersonic
nozzle, frictional effects between the nozzle and the diffuser require that
the diffuser throat be somewhat larger than the nozzle throat. If the
diffuser throat is made slightly too small, supersonic flow will not be
attained in the nozzle; and, if the diffuser throat is made slightly too
large, there will necessarily be a shock somewhere within the diffuser.
Indeed, even if the two throats did match perfectly, it appears that the
combined system would be unstable and that flow oscillations would
occur.
A second and more serious difficulty arises. Most flow systems start
from rest and accelerate to the operating velocity. During the starting
period normal shock waves pass through the system. Across a normal
shock there is no change in flow rate or in stagnation temperature.
According to Eq. 4.17, therefore, the product Az*poz must equal the
product A,*po,. As the stagnation pressure is reduced by a normal
shock, it follows that there is a corresponding increase in the minimum
area through which the flow can be made to pass. The design problems
raised by this aspect of the process will be discussed for two applications,
namely, wind tunnel diffusers and propulsion engine diffusers.
Supersonic Wind Tunnel Diffusers. I n order to focus attention on
the essential features of the problem, it will be assumed that the flow is
isentropic, except for entropy in-
creases which occur across normal
shocks. We shall assume also that
the flow is quasi-steady during the
period of starting the flow.
STARTING CONDITIONS. Consider
the supersonic wind tunnel system
of Fig. 5.25, comprising a converg-
ing-diverging nozzle, a test section,
a diffuser, and an exhauster. We
will suppose that the inlet stagna-
tion pressure and temperature are
constant. When the exhauster is
FIG.5.25. Starting of supersonic wind started up and thereby produces a
tunnel diffuser.
reduction in pressure a t the nozzle
(a) Most unfavorable starting con- exit, a shock moves down the
dition.
(b) Best operating condition. diverging portion of the nozzle
according to the discussion of Art.
5.10. During this period, the nozzle throat is passing the maximum
possible flow, and, since A*p, is constant across the shock, it follows
that the flow can pass through the diffuser throat only if the latter is
larger than the nozzle throat. The worst condition in this respect
occurs when the shock is moving through the test section (Fig. 5.25a)
inasmuch as the shock then occurs at the maximum possible Mach
Number and, consequently, produces the largest loss in stagnation pres-
sure. The minimum area of the diffuser throat is therefore given by
A m i n . diff. throat Ay* Poz

-- =- (5.48)
Anozzle t h r o a t Az* POV
where po,/pk is the stagnation pressure ratio for a shock a t the test-
Art. 5.11 ONE-DIMENSIONAL SUPERSONIC DIFFUSERS 145
section Mach Number. The limiting contraction ratio A,/A,* for the
diffuser is accordingly given by
The area ratio Ax/Ax*is given in terms of M , by the isentropic formula,

Eq. 4.19, and the ratio po,/p~, is given in terms of M , by the shock
formula, Eq. 5.26. Hence &/A,* depends only on the test-section
Mach Number. This area ratio is plotted versus M , in Fig. 5.26, tc-
gether with a curve showing the area contraction which would be pos-
sible if the stream a t M , could be diffused without a shock.
Test -Section Mach Number. M,

I
S
*
FIG. 5.26. Maximum area contraction FIG. 5.27. Starting operation of super-
for which supersonic wind tunnel diffuser sonic wind tunnel diffuser illustrated on
of fixed geometry will permit supersonic temperature-entropy diagram.
flow to be established in test section.
With the limiting diffuser throat area corresponding to Eq. 5.48, the
diffuser is barely able to "swallow" the flow during starting, and the flow
a t the diffuser throat is exactly sonic when the shock is in the test
section. If the diffuser throat is slightly smaller than that required by
Eq. 5.48, a normal shock will stand in the diverging portion of the nozzle,
and subsonic flow will exist in the test section. Should the diffuser
throat be considerably smaller than that given by Eq. 5.48 the flow will
be subsonic throughout the entire system, except possibly downstream
of the diffuser throat.
The worst starting shock is illustrated in the T-s diagram of Fig. 5.27.
I
In interpreting this diagram it is well to remember that all states on the
!
I
same Fanno line have a common stagnation temperature and flow per
I unit area. A change in cross-sectional area has associated with it a
shift from one Fanno line to another. It is clear from this diagram that
I during the "starting" condition there is a large loss in stagnation pressure
and a consequent increase in the area required to pass the flow. The
path of states during the most unfavorable starting condition (i.e.,
while the shock is in the test section) is from Ox to x to y to *y to Oy.
OPERATING CONDITIONS. Assuming now that the diffuser is lwge

enough for supersonic flow to be established in the test section, a sufficient
lowering of the exhaust pressure will cause the shock to move through
the test section and to be swallowed by the diffuser throat. The position
a t which the shock comes to rest depends on the manner in which the
operating characteristic of the exhauster is matched to the pressure-
flow characteristic of the remainder of the system. For best efficiency,
the system and exhauster should be matched so that the equilibrium
position of the shock is in the dif-
fuser throat, as the shock will then
occur a t the minimum possible
Mach Number in the diffuser. This
best operating condition is shown
in Fig. 5.2513. The corresponding
path of states in Fig. 5.27 is from
Ox to x to x' to y' to Oy'. The
stagnation pressure loss for best
operation is less than for starting
up because of the lower Mach
Number a t which the shock occurs.
~est-section M O C ~Number Hence the pressure ratio required
FIG.5.28. Stagnation pressure loss for of the exhauster is determined by
supersonic wind tunnel diffuser of fixed
geometry. Curve marked "least favor- the starting condition rather than
able starting condition" shows also the by the operating condition. A
stagnation pressure loss for a diffuser
comprising a normal shock followed by of the two limiting
isentropic subsonic diffusion. stagnation pressure ratios is shown
in Fig. 5.28.- The lower curve repre-
sents also the operating condition of a diiuser having no contraction
whatsoever, ie., a normal shock followed by isentropic subsonic diffusion.
For low test-section Mach Numbers it is evident that the normal shock
is so efficient as hardly to justify the use of a converging-diverging
diffuser.
In practice the shock is maintained slightly downstream of the diffuser
throat during operation. This is done because with a fixed exhaust.
pressure the shock is unstable in the converging portion of the diffuser.
For example, if the shock were maintained exactly a t the minimum
area, a slight disturbance might make it move temporarily into the
converging section. But this would augment the loss in stagnation
pressure, and, if the exhaust pressure were fixed, transient waves would
cause the shock to move further upstream. This would make the
situation still worse, and the shock would move upstream progressively
until it came to rest in the diverging part of the nozzle a t a point where
the stagnation pressure loss in the system matched the exhaust pressure
of the system. In order again to obtain supersonic flow in the test
Art. 5.11 ONE-DIMENSIONAL SUPERSONIC DIFFUSERS 147
section, it would be necessary to lower the exhaust pressure to the mini-

mum value required for starting, as determined from Fig. 5.28.
To insure that a supersonic diffuser of fixed geometry will start, the
throat area must be made slightly larger than the theoretical minimum
value to account for inaccurate estimates of the effects of friction, of the
departures from one-dimensionality, and so forth.
ADDITIONAL CONSIDERATIONS. Thus, because practical reasons re-
quire that the best design be compromised by an enlargement of the
throat and by an operating condition with the shock a t a Mach Number
greater than the minimum in the passage, the practically attainable
efficiencies of such diffusers fall somewhat short of the values which seem
possible in principle.
Experimental investigations confirm in a general way the theoretical
considerations outlined here, although there are modifications owing
to viscous effects. Wall friction of course produces slight additional
losses in stagnation pressure. The chief effect of viscosity, however, is
to make the shock occur in a length of several passage diameters, rather
-
than abruptly; unless this feature is taken

into account in the design, the diffuser
efficiency may be considerably lower than
anticipated.c4)
Some of the diiuser problems outlined
here may be avoided by such means as
(i) variable-geometry diffusers, (ii) variable-
geometry diffusers in conjunction with
variablegeometry nozzles, (iii) driving the
shock through the diffuser throat by means
of a large-amplitude pressure pulse, and
(iv) taking advantage of effects which are
not one-dimensional.
Supersonic Inlets. Let us now consider \oetached Shock
the inlets of air-breathing propulsion en- FIG.5.29. Supersonic inlet.
gines such as turbojets and ram jets when (a) Nomenclature.
these operate a t supersonic speeds. Such (b) No detached shock, with
engines always reach their operating speed all the free stream flow
corresponding to the area
by being accelerated from lower speeds, A1 entering the inlet.
(c) Detached shock, with spill-
and so we must investigate operation of over.
the supersonic inlet during starting. Here
again, for simplicity, we shall treat the flow as quasi-static and shall
ignore all losses except those occurring in normal shocks.
SWALLOWING OF SHOCK. Fig. 5.29a shows a converging-diverging
supersonic inlet having an inlet area A1 and minimum area A2. If no
shock stands ahead of the inlet when the latter travels a t supersonic
speeds, there is no possibility of the air flow being diverted before reach-
ing the inlet, and all the free-stream flow corresponding to the cross-
sectional area Al enters the engine (Fig. 5.29b). Should the inlet not
be able to pass this amount of flow, a detached shock stands ahead of
the inlet (Fig. 5 . 2 9 ~as
) ~then the flow behind the shock is subsonic and
may spill over the inlet lip of the diffuser. Fig. 5.30 is a schlieren photo-
graph illustrating this phenomenon.
F I ~ .5.30. Schlieren photograph of

supersonic inlet fed by supersonic free
jet. The streamlines were made visible
by means of a time exposure. The de-
tached normal shock permits some of the
flow to spill over the inlet lip (Depart-
ment of t,he Navy. U.S.A.).
The lower curve of Fig. 5.31 shows the maximum contraction possible
when there is no shock whatsoever, and comes directly from the isen-
tropic relations. The upper curve shows the maximum contraction
possible for the special case where a detached normal shock hangs on the
inlet lip, thus forcing all the air through the engine but with a reduced
stagnation pressure. The ordinates of these curves are the reciprocals
of the corresponding ordinates of Fig. 5.26.
CONSTANT-GEOMETRY INLET. Consider now a fixed-geometry inlet

designed to operate a t the speed M,, and having the area ratio (A2/A1),.
Fig. 5.32 shows various stages during starting. In Fig. 5.32a, the speed
is too low for the flow to be swallowed, and a detached shock is present
(point b of Fig. 5.31). When the speed is slightly less than the design
speed (Fig. 5.32b), the shock hangs on the inlet lip. A slight increase
in speed causes the shock to move into the inlet, provided of course that
the exhaust pressure is sufficiently low. The equilibrium position of the
shock depends on the way in which the performance characteristics of
for O p e r a t i o n ( N o S h o c k )
I Mm
FIG. 5.31. Limiting contraction ratio versus free-stream Mach Number.
FIG.5.32. Starting of fixed-geometry supersonic inlet, designed for free-stream

Mach Number M,~, and having contraction ratio ( A Z / A I ) ~
the inlet are matched to the performance characteristics of the remainder

of the engine. I n Fig. 5 . 3 2 ~the shock is too far downstream for good
efficiency, and in Fig. 5.32d it is in the farthest upstream position for
which the flow is stable. A slight disturbance in the flow pattern of
Fig. 5.32d would cause the shock to be disgorged.
Such a diffuser has an interesting hysteresis effect. Assuming that
the exhaust pressure is sufficiently low, the shock will not be swallowed
during acceleration until the speed M,, (Fig. 5.31) is reached. Once
swallowed, however, the shock will not be disgorged upon deceleration
until the speed Mmf (Fig. 5.31) is reached.
OVERSPEEDING O F CONSTANT-GEOMETRY INLET. It is possible to
eliminate altogether the operating shock of a supersonic fixed-geometry
inlet if the engine may be overspeeded. Fig. 5.33 shows various stages
- Mad
___)
MmiMm ,
yM =I
(Shock-Free)
(c) (dl
FIG.5.33. Overspeed starting of fixed-geometry supersonic inlet, designed for free-
stream Mach Number Ma, and having contraction ratio (A2/A&.
in the acceleration period of an inlet designed for the speed M,, and
having in Fig. 5.31 a contraction ratio (A2/AI),. By overspeeding the
engine to the speed Mmd (Fig. 5.31), the shock is swallowed, and then the
engine may be decelerated at most to M,, before the shock is dissorged.
VARIABLEGEOMETRY INLET. The operating shock may also be elim-
inated through the use of a variable-geometry diffuser. Suppose that
such a diiuser is operating a t the speed M,, of Fig. 5.31. If the throat
area is too small, corresponding perhaps to point c of Fig. 5.31, a de-
tached shock stands ahead of the inlet (Fig. 5.34a). By increasing the
throat area, the shock is brought closer to the inlet. When point a of
I
/'
4
-/
,
,
. ..
,. .'..:::!$:::::!;;
: ~,: ., ........ . .. . . .. . .
.!.;.$:!:.t.:*
..'.
( A t /A lo (AJA,)c
(a) (b) (c)
FIG.5.34. Starting of variable-geometry supersonic inlet, designed for free-stream
Mach Number M,.
Fig. 5.31 is reached, the shock reaches the inlet lip and may be swallowed
(Fig. 5.34b). Subsequently, the throat area may be reduced without
disgorging the shock until point c of Fig. 5.31 is reached, and, if the
exhaust pressure is properly ad-
justed, the diffuser is then free of
shocks (Fig. 5.34~).
Diffuser Efficiency. The most
common definition of diffuser effi-
ciency is parallel to the definition
employed for compressor efficiency.
Referring to Fig. 5.35, and assum-
ing that the velocity leaving the
diffuser is negligible, we define
where state 1 is the actual state f u ~ ~ e & ~ ~ A ~ ~ definition

! u s t r a of
~ sdif-
entering- the diffuser, 2 is the actual
state leaving the diffuser, and 3 is a fictitious state a t the actual leaving
pressure but a t the entering entropy.
For a perfect gas Eq. 5.49 becomes
VD =
7'3 - T1 -
T2 - Tl
T1 - - 1
(2
V12/%
)
and, since k-l
-
T3/7'1 = (p2/pi)
and
we get k- 1
The efficiency is uniquely related t o the stagnation pressure ratio
for a given inlet Mach Number. To demonstrate this, we write
-P2 = Po1
- .P- z Po1
= -Po2. -
Pl Pl Pol PI Pol
Employing the isentropic relation for p o l / p l , and substituting into Eq.
5.50, we get after rearrangement,
k-1
r l ~=
(I+? M12) ?):( 1
(5.51)
k-1
-M12
2
Corresponding to the two curves of Fig. 5.28, there are plotted in
Fig. 5.36, with the aid of Eq. 5.51, the maximum efficiencies for fixed-
geometry diffusers and normal shock diffusers.
Inlet Mach Number Mach Number at Tesl Section

FIG. 5.36. Maximum efficiency of one- FIG. 5.37. Summary of performance
dimensional supersonic diffusers. of high-speed wind tunnels. Triangles
represent measured ratio of inlet to exit
stagnation pressures for variable-geom-
etry wind-tunnel diffusers.
Wind Tunnel Performance. I n Fig. 5.37 are plotted the compressor

pressure ratios for a number of operating wind tunnels. This chart
shows that subsonic tunnels correspond roughly to a diffuser efficiency
of 75 per cent. It is evident that the principal loss in supersonic wind
tunnels is caused by the shock in the diffuser and that the maximum
gain in supersonic wind tunnel economy, a t least from the point of view
of machinery installation, is t o be had by improving the performance of
supersonic diffusers.
Art. 5.12 SUPERSONIC PITOT TUBE 153
5.12. Supersonic Pitot Tube

The pitot tube has long been an important instrument for measuring
the velocities of subsonic streams. When used with supersonic streams,
modifications must be made in the interpretation of the measured data.
FIG. 5.38. Supersonic pitot tube.
Rayleigh Supersonic Pitot-Tube Formula. Photographs reveal that

when an impact tube is placed in a supersonic stream, a curved shock
wave stands ahead of the mouth of the tube, as shown in Fig. 5.38.
The Mach Number of the undisturbed stream ahead of the shock wave
may be found from measured data
if the following assumptions are
made:
(a) The shock wave is locally
normal to the stagnation stream-
line a t the point where the latter
crosses it. This condition will, by
symmetry requirements, be true
when the tube is placed ~arallelto
a uniform stream.
(b) Particles following the stag-
nation streamline are brought to
rest isentropically in the subsonic
region aft of the shock.
These assumptions are illus-
1
I trated graphically on the T-s dia-
gram of Fig. 5.38. We write
Now, po,/p, may be expressed in

terms of M, by using Eqs. 4.14b FIG.5.39. Plot of Rayleigh supersonic
and 5.16b; and p,/p, is given in pitot-tube formula for k = 1.4.
terms of M, b y Eq. 5.18a. Introducing these expressions, we get the
Rayleigh pitot-tube formula:
Fig. 5.39 shows this relation for a gas with k = 1.4. T h e ratio pov/pz
is also tabulated in Table B.3.
Subsonic Pitot Tube. For subsonic flow, there is no shock (unless
M , is close ta unity), and Eq. 5.52 is not applicable. Instead, the cus-
tomary isentropic relation, Eq. 4.14b, is used for relating the Mach
Number, the static pressure, and the isentropic stagnation pressure.
Measurement of Static Pressure. I n using Eq. 5.52, t h e static pres-
sure p, must be measured upstream of t h e shock wave. If the stream
is in a wind tunnel, this might be done with a static pressure t a p in the
wall of the tunnel.
Pitot tubes often have a static t a p built into the side wall of the probe.
I n supersonic flow this t a p does not give a n easily interpretable measure-
ment. However, experimental data indicate t h a t if the static t a p is
placed some ten tube diameters aft of the nose, the static pressure at the
t a p is a close approximation t o the static pressure upstream of t h e shock.

1. SHAPIRO, A. H., and KLINE, S. J. On the Thickness of Normal Shock Waves in Air.
Istanbul, Turkey: Eighth International Congress on Theoretical and Applied
Mechanics, 1952.
2. NEWBERG, E. G., JR. Measurements of Diffuser Efficiency of Supersonic Pres-
sure Shocks, S.M. thesis, Dept. of Mech. Eng., Mass. Inst. of Tech. (1946).
3. KANTROWITZ, A., and DONALDSON, COLEMAN DUP. Preliminary Investigation of
Supersonic Diffusers, NACA, ARR No. L5D20 (1945).
4. NEUMANN, E. P., and LUSTWERS, F. High-Efficiency Supersonic Diffusers, Jour.
Aero. Sci., Vol. 51, No. 6 (1951), p. 369.
PROBLEMS
5.1. Air flows steadily through a pipe of constant cross-sectional area and
at a certain section has properties p, = 10 psia, T , = 124OoF,and V, = 3000
ft/sec.
(a) Plot the Fanno and Rayleigh lines corresponding to point x on the p,
T-v, and T-s diagrams. Find from these diagrams the pressure, temperature,
and velocity downstream of a normal shock occurring at condition x.
(b) Compare your results with the corresponding results found from the
shock tables.
5.2. Show that for a very weak shock in air,
6.3. Derive simple approximate formulas for very weak shocks, expressing
(1 - (P, - P=)/P=, (Tv - l1J/Tz, (V, - Vz)/Vz, and (PO*- P ~ / P O aU
*
in terms of (M, - 1).
5.4. (a) Find the limiting forms of the normal shock equations for a perfect
gas with k = 1. (b) Find simple asymptotic forms of the normal shock equa-
tions for a perfect gas when the Mach Number approaching the shock is verg
large compared with unity.
6.5. (a) Show that the equation of the Fanno line for a perfect gas is
where s, and T , denote respectively the entropy and temperature at the point
of maximum entropy on the Fanno line.
(b) Show that the equation of the Rayleigh line for a perfect gas is
where sa and Tt,are respectively the entropy and temperature a t the point of
maximum entropy on the Rayleigh line. Demonstrate that there is a point of
maximum temperature on the Rayleigh line and find a relation connecting the
entropy and temperature at this point with sb and Tb.
5.6. Consider a normal shock in a fluid (not a perfect gas) whose density
depends only on the pressure according to the relation
dp
P-=8
d~
where B is a ~ositiveconstant.
(a) show that the inlet and exit Mach Numbers are related by
(b) Find expressions for (p, - p,) andVy/V, in terms of M, and 8.

6.7. Show that the local Mach Number is unity at the point of maximum
entropy on the Rayleigh line.
5.8. Demonstrate with the help of the Rayleigh line that in the case of a
perfect gas the existence of an expansion shock wave would permit the con-
struction of a perpetual motion machine of the second kind.
6.9. Consider a pipe in which air a t 70°F and 10 psia flows uniformly with
a speed of 400 ft/sec. The end of the pipe is suddenly closed by a valve, and a
shock wave propagates back into the pipe. Compute the speed of the wave and
the pressure and temperature of the air which has been brought to rest.
5.10. A shock moves into stationary gas according to the sketch of Fig.
5.19b. Derive a formula for the speed of the high-pressure gas in terms of the
properties of the low-pressure gas and the strength of the wave. It is suggested
that this be put in the form of a relation between (V, - V,)/c, and M,.
5.11. Suppose that a blast wave, which might have been initiated by an
atomic bomb explosion, is traveling through air a t standard atmospheric condi-
tions with a speed of 200,000 ft/sec.
Estimate the changes in pressure (atm), temperature ( O F ) , stagnation pressure
(atm), stagnation temperature (OF), and velocity (ft/sec), ~roducedby the wave
with respect to an observer who is stationary with respect to the undisturbed air.
5.12. A particular convergingdiverging nozzle is designed for a Mach
Number of 2 on the basis of reversible, adiabatic flow. When the ratio of
exhaust-region pressure, p,, to supply-region pressure pa, is raised considerably
above the design value, a normal compression shock appears in the nozzle. The
pressures before and after the shock will be denoted by p, and p,, respectively.
Assuming reversible, adiabatic flow both upstream and downstream of the
shock, plot p,/po and pulpoagainst p,/po for the range in which the shock lies
between the throat and the exit plane. The shock may be considered as being
infinitesimal in thickness, and the fluid may be taken as a perfect gas with
k = 1.4.
Compare the curve of P,/P, with a curve representing the isentropic pressure
ratio a t the same location in the nozzle for the limiting case of subsonic operation,
that is, for the case where the Mach Number is unity a t the throat and the flow
in the diverging section is subsonic.
5.13. A rocket nozzle which is intended to operate a t an altitude of 100,000
ft has an area ratio of 5.16. At launching, however, the rocket exhausts to nor-
mal atmospheric pressure, 14.7 psia. The rocket gases are generated a t 400 psia
and have k = 1.3. Assuming frictionless, adiabatic flow in the nozzle, find the
ratio of the rocket thrust a t launching to the thrust a t 100,000 ft.
6.14. A fixed-geometry, convergent-divergent wind tunnel diffuser is to be
designed for Mach Number 2. Assuming no friction, compare the maximum
possible efficiencyand the minimum possible per cent loss in stagnation pressure
during operation for the following cases:
(a) The best possible design is employed.
(b) The design is conservative, with a throat area 5 per cent larger than
that required for starting, and with the shock located during operation
a t an area 5 per cent greater than the throat area.
(c) The converging portion is eliminated, and the process comprises a
normal shock followed by reversible subsonic compression.
5.15. Consider a fixed-geometry, converging-diverging wind tunnel diffuser
in which the minimum area is 15 per cent smaller than the entrance area, and
in which the efficiency for subsonic deceleration to zero velocity is 90 per cent.
Frictional effects upstream of the shock are negligible.
(a) Plot the maximum possible efficiency and the minimum possible per cent
loss in stagnation pressure versus the approach Mach Number. What is the
minimum value of the latter for starting purposes?
(b) Assume that during operation, because of conservative design, the shock
stands downstream of the throat a t a point where the area is 10 per cent larger
than the throat area. Plot the efficiency and the per cent loss in stagnation
pressure versus the approach Mach Number.
Ch. 5 PROBLEMS IS7
5.16. A ram jet aircraft is to fly a t 40,000 ft altitude with a speed of 2000
mah.
(a) Design the best fixed-geometry, convergent-divergent diffuser for this
aircraft, and compute for it the maximum efficiency and the least per cent loss
in stagnation pressure.
(b) Suppose that it were possible to overspeed the aircraft to 2400 mph.
Design the best convergent-divergent diffuser which could then be used, and find
for it the maximum efficiency and the least per cent loss in stagnation pressure
a t the operating speed of 2000 mph.
5.17. A supersonic wind tunnel is to he designed for a Mach Number of

2 with a test section one square foot in area. The general arrangement of the
tunnel will be as follows:
Air will be taken from the atmosphere (14.7 psia, 70°F) and will be accelerated
to a Mach Number of 2 in a convergingdiverging nozzle. From the test section
the air will be diffused to substantially zero velocity in a diffuser and will then
be discharged to the atmosphere by a compressor.
The design of the tunnel and compyessor will be based on the following assump
tions: (1) The nozzle is frictionless to the throat, while the over-all efficiency
(Ah/Ah,) of the nozzle, from entrance to exit, is 95 per cent. (2) Although an
attempt will be made to diffuse the supersonic stream through a throat, for pur-
poses of design the assumptions will be made that the stream with M = 2
passes through a normal shock and that the subsonic stream is then diffused
with an efficiency (Ah,/Ah) of 80 per cent. The conservative nature of this
assumption will tend to balance the fact that no account is taken of losses in the
test section. (3) The compressor has an efficiency, Ah,/Ah, based on the revers-
ible adiabatic work of compression, of 82 per cent.
(a) Make a sketch of this tunnel, showing the cross-sectional areas a t various
~ ~
sections.
(b) Indicate on the sketch the pressure (psia) and temperature ( O F ) a t the test
section, a t the entrance to the compressor, and a t the exit of the compressor.
(c) specify the diameter of the nozzle throat.
(d) Specify, for the compressor, the pressure ratio, the volume rate of flow at
inlet (cfm), and the horsepower required for operation.
5.18. A certain pitot tube is used for measuring the Mach Number M, of a
supersonic air stream, and the pressures shown in the sketch are recorded. It
is known that for the particular tube employed, the
pressure coefficient (referred to free-stream condi-
tions) a t the static holes is 0.1. Estimate the free-
stream Mach Number.
5.19. When an impact tube with a blunt nose is
placed in a supersonic stream, a curved shock front
4~~
shock Holes
f p - 2 0 psi0
PROB.5.18.
stands a t some distance in front of the tube. An
impact-tube traverse in a wind tunnel gives values of 2.34 psia and 10.02 psia
for the static pressure upstream of the shock and the pressure a t the mouth of
the tube, respectively. Estimate the Mach Number of the tunnel.
5.20. Explain why entropy decreases occur as a fluid particle nears the end
of its passage through a shock wave.
5.21. (a) Consider a rightward moving wave in a perfect gas. Show that,
one follows a part of the wave corresponding to a particular fluid velocity
rts
and pressure, the time rate of change of wave steepness is given by
(b) Show that the time required for a given part of the wave to acquire infinite
slope is given by
2
where aV/ax is the wave steepness at t = 0.

Consider a plane progressive sound wave in air at 70°F and 14.7 psia.
5.22.
The wave has a frequency of 10,000 cycles per second and an intensity level of
120 decibels. Neglecting damping of the wave, show that the wave tends to
become saw-toothed in form. Using the results of Problem 5.21, calculate the
time and distance required for the wave to develop into a series of shocks sepa-
rated by rarefactions.
6.23.Air flows through a frictionless, converging nozzle (see sketch) into a
sudden enlargement. At section 3 the flow is substantially uniform and parallel,
and between sections 2 and 3 wall friction is negligible.
Calculate the pressure acting on the annular surface of section 2 and the mass
rate of flow, assuming that &/A1 = 2.33.
HintS: (i) If the nozzleis choked, the pressure on the annular face of section 2
may be less than pl; but if the nozzle is unchoked, then p2 must equal pl.
(u) Apply the momentum, energy, and continuity equations between sections
2 and 3.
/ / / / / I I N / / / / / / ,/ //
/
Primary U
t !
-p0 = 4 0 Psia
~IOY
5.24.A jet pump (see sketch) uses a high-pressure stream of air to pump a
second stream of atmospheric air to a pressure of 18.25 psia. The mixing tube
of this particular pump has constant cross-sectional area. As a first approxima-
tion, all irreversibilities except those associated with viscous mixing are to be
ignored.
Calculate the mass rates of flow of primary air and secondary air.
Chapter 6
FLOW IN CONSTANT-AREA DUCTS
WITH FRICTION

The flow of compressible gases in ducts of constant area is of impor-
tance in many engineering fields, including stationary power plants,
aircraft propulsion engines, high-vacuum technology, transport of fluids
in chemical process plants, transport of natural gas in long pipe lines,
and various types of flow machinery. It is the purpose of this chapter to
discuss such flows from a simple one-dimensional point of view.
We shall here consider flows in which wall friction is the chief factor
bringing about changes in fluid properties, and we shall assume that no
special attempt is made t o transfer heat to or from the stream. When
the ducts are reasonably short, therefore, the flow is approximately
adiabatic. When the ducts are extremely long, however, as in the Case
of natural-gas pipe lines, there is sufficient area for heat transfer to
make the flow nonadiabatic and approximately isothermal. Both of
these limiting cases will be investigated in this chapter. More general
cases of compressible flow with heat transfer are treated in Chapters 7
and 8.
Additional material relevant to the subject matter of the present
chapter may be found in Volume 11, Chapters 24, 26, 27, and 28.
NOMENCLATURE
area L duct length

velocity of sound L, maximum duct length for con-
specific heat at constant pres- tinuous flow
sure M Mach Number
hydraulic diameter P pressure
coefficient of friction Q heat
length mean coefficient of fric- I R gas constant
, tion 1.9 entropy per unit mass
impulse function T absolute temperature
mass velocity, pV V velocity
enthalpy per unit mass w mass rate of flow
ratio of specific heats x Cartesian coordinate
159
160 now IN CONSTANT-AREA DUCTS WITH FRICTION Ch. 6
p
T,,,
( )*
density
wall shearing stress
signifies state at which M = 1
I ( )o
signifies state at which M
l/dkin isothermal flow
signifies stagnation state
=
for adiabatic. constant-area

flow with friction I
Preliminary Considerations of Adiabatic Flow. We now make the
assumptions that (i) the flow is one-dimensional, (ii) the flow is steady,
(iii) there is neither external heat exchange nor external shaft work, (iv)
differences in elevation produce negligible changes compared with fric-
tional effects, and (v) the duct is of constant area.
GOVERNING PHYSICAL LAWS. The energy equation of steady flow
may be written
v2
h+-=h 0 (6.1)
2
where h and V are respectively the corresponding values of the enthalpy
and velocity a t an arbitrary section of the duct, and ho (the stagnation
enthalpy) has a constant value for all sections of the duct. Physically,
ho is the enthalpy a t the section where the velocity is zero.
The equation of continuity is written
where p is the density a t the section where V and h are measured, and
G (the mass velocity) has a constant value for all sections of the duct.
Combining Eqs. 6.1 and 6.2, we get the equation of the Fanno line
in terms of the enthalpy and density:
Since ho and G are constants for a given flow, Eq. 6.3 defines a relation
between the local enthalpy and the local density. This relation is
shown graphically'in Fig. 6.1 for a single value of ho and for several
values of G. All possible states of the fluid for a given adiabatic, con-
stant-area flow lie on one of these lines.
THE FANNO LINE. For a pure substance, s = s(h,~),i.e., the entropy
is determined by the enthalpy and density., The curves of Fig. 6.1
may thus be transferred to the enthalpy-entropy diagram, giving the
Art. 6.1 INTRODUCTORY REMARKS 161
familiar Fanno curves of Fig. 6.2. For all substances thus far investi-
gated, the Fanno curves have the general shape shown in Fig. 6.2.
It was shown in Art. 5.2 that the upper branch of each Fanno curve
corresponds to subsonic flow, that the lower branch corresponds to super-
sonic flow, and that the Mach Number is unity a t the point of maxi-
mum entropy on each Fanno curve.
SECOND-LAW LIMITATIONS. Since the flow is adiabatic, the Second
Law of Thermodynamics tells us that the entropy may increase but
Lines o f Constont Entropy

A
'\ \i
FIG.6.1. Fanno lines on h-v diagram. FIG.6.2. Fanno lines on h-s diagram.
Solid lines have the same stagnation en- The three curves shown have the same
thalpy but different flows per unit area. stagnation enthalpy but different flows
Dashed lines are isentropes. per unit area.
may not decrease. Thus, the path of states along any one of the Fanno
curves must be toward the right.
Consequently, if the flow a t some point in the duct is subsonic (point a
of Fig. 6.2), the effects of friction will be to increase the velocity and
Mach Number and to decrease the enthalpy and pressure of the stream.
If, on the other hand, the flow is initially supersonic (point b of Fig. 6.2),
the effects of friction will be to decrease the velocity and Mach Number
and to increase the enthalpy and pressure of the stream. A subsonic
flow may therefore never become supersonic, and a supersonic flow
may never become subsonic, unless a discontinuity is present. Thus
we observe that, as in the case of isentropic flow, the qualitative char-
acter of the flow is markedly influenced by whether the flow is subsonic
or supersonic.
The limiting pressure, beyond which the entropy would suffer a d e
crease, occurs a t Mach Number unity and is denoted by p*. The asterisk
here denotes the state where M = 1 for the particular process under
consideration, namely, adiabatic flow a t constant area. Thus, referring
162 FLOW IN CONSTANT-AREA DUCTS WITH FRICTION Ch.6
to state a as an example, the value of pa* will be diierent for an isentropic
flow as compared with the value for an adiabatic, constant-area flow.
Fig. 8.2 indicates that the isentropic stagnation pressure is reduced
as a result of friction, irrespective of whether the flow is subsonic or
supersonic.
CHOKING DUE TO FRICTION. Consider a situation in which the stag-
nation enthalpy, flow per unit area, and length of duct are such that
Mach Number unity is reached a t the end of the duct. If the duct
length is increased, i t is evident from the foregoing considerations that
some sort of adjustment in the flow is necessary. When the flow is
subsonic, this adjustment is in the form of a reduction in the flow rate,
that is, the flow is choked. When the flow is supersonic, the adjustment
at first involves the appearance of shock waves, and, for sufficiently
large increases in duct length, involves ultimately a choking of the flow.
This will be discussed more fully in Arts. 6.2 and 6.3.
6.2. Adiabatic, Constant-Area Flow of a Perfect Gas

If it is assumed that the fluid is a perfect gas, the analytical treatment
is greatly simplified, and, moreover, it becomes possible to draw broad
conclusions which would otherwise not be apparent.
Conrrol S u r f a c e
FIG.6.3. Control surface for analysis of adiabatic, constant-area flow.
Our purpose is to find in analytical form the variations in all stream

properties along the length of a duct of constant area. Naturally, the
rate of change of properties depends upon the amount of friction, so that
the momentum equation must be introduced into the analysis. In
Chapter 4 the isentropic relations were derived by writing the various
physical relations for two sections a finite distance apart. To illustrate
another method of approach, we shall carry out the present analysis in
differential form, employing the infinitesimal control volume of Fig. 6.3.
Governing Physical Equations and Definitions. PERFECT-GAS EQUA-

TION. Starting with the perfect-gas relation,
p = pRT
we take logarithmic differentials, and thus obtain
DEFINITION OF MACH NUMBER. From the expression for the Mach

Number in a perfect gas, M2 = V2/k12T, we get by logarithmic diier-
entiation,
STEADY-FLOW E NERGY EQUATION. The energy equation of steady

flow is written for a perfect gas in the form
Dividing through by c,T, and using the definition of the Mach Number,
this becomes
dT k - 1 dV2
T
-+- 2
~2 --
V2
-0
CONSERVATION OF MASS. The continuity equation is

G = w/A = pV
or, since G is a constant,
d~ 1dv2
MOMENTUM EQUATION. Referring to Fig. 6.3, the momentum equa-

tion is written
-Adp-r,dA,=wdV
where A is the cross-sectional area, rw is the shearing stress exerted on
the stream by the walls, and dA, is the wetted wall area over which T,
acts.
DEFINITION OF FRICTION COEFFICIENT. The drag coefficient, or the
c o e p e n t of friction as it is usually called for flow in ducts, is defined as
the ratio of the wall shearing stress to the dynamic head of the stream.
Thus,
The hydraulic diameter is defined as four times the ratio of cross-sec-
tional area t o wetted perimeter,
We now introduce the latter two expressions and the continuity equa-
tion into the momentum equation, and thus obtain
Next, we divide through by p, and, upon noting that PV' = kpM2,

obtain
dp kM2 dx kM2 dV2
-+-4f-+-- = 0 (6.8)
P 2 D 2 V2
ISENTROPIC
STAGNATION PRESSURE. We retain our ~reviousdefini-
tion of the isentropic stagnation pressure corresponding to a given state
as that pressure which would be attained if the stream were isentropically
decelerated from the given state to the state of zero velocity. Then,
from Eq. 4.14b, L
or, in differential form,
IMPULSE FUNCTION. From the definition of the impuls~function,
we get, after noting that the area is constant,
Effects of Wall Friction on Fluid Properties. At this point we notice

that we have seven simultaneous, linear algebraic equations, namely,
Eqs. 6.4, 6.5, 6.6, 6.7, 6.8, 6.9, and 6.10, which relate eight differential
variables: dp/p, dp/p, dT/T, dM2/M2, dV2/V2, dpo/p,, dF/F, and
4 j dx/L). The physical phenomenon causing changes in state is viscous
friction. Hence we select the variable 4jdx/D as independent. The
remaining seven variables may consequently be found in terms of
4f dx/D with the aid of the seven equations. The usual methods for
solving simultaneous equations are used. For example, the term

d T / T may be eliminated from Eqs. 6.4 and 6.6, yielding
Next, d ~ is/ eliminated

~ from this expression with the help of Eq. 6.7,
thus giving
-dp = - 1 + ( k - 1)M 2 d V 2
-
P 2 v2
Finally, we use this relation for eliminating d V 2 / V 2 from Eq. 6.8, and
obtain, following algebraic rearrangement,
Using similar methods, the formulas which follow are obtained:
SECOND LAW OF THERMODYNAMICS. Since the flow is adiabatic, the

stagnation temperature is constant. From Eq. 5.23, therefore, the
entropy change is
ds k - 1 dpo
-=
CP k Po
or, uslng Eq. 6.16,
ds ( k - 1)M2 d x
-- 4f
CP 2
By convention, dx is positive in the direction of flow. The Second Law
of Thermodynamics states that the entropy may not decrease in an
adiabatic process, whence we conclude from Eq. 6.19 that the friction
coefficient j must always be a positive number. In formulating Eq. 6.8
it was assumed that the shearing stress acted on the stream in a direction
opposite to the direction of flow. Since j must always be positive, we
conclude that the shearing stress must always act in the direction
assumed.
It appears from Eqs. 6.16 and 6.17 that both the isentropic stagnation
pressure and the impulse function must decrease when friction is present.
This is true for either subsonic or supersonic fla~w. Wall friction there-
fore reduces the effectiveness of all types of flow machinery and also
reduces the thrust obtainable from jet propulsion devices.
SUMMARY OF FRICTIONAL EFFECTS. The directions of changes in the
remaining stream properties, according to Eqs. 6.11 through 6.15,
depend on whether the Mach Number is greater or less than unity,
since the term (1 - M2) appears in the denominator of each of these
equations. We may summarize the various changes as follows:
Subsonic Supersonic
Pressure, p decreases increases
Mach Number, M increases decreases
Velocity, V increases decreases
Temperature, T decreases increases
Density, p decreases increases
Stagnation Pressure, po decreases decreases
Impulse function, F decreases decreases
We see that the Mach Number always tends toward unity. Con-
tinuous transitions either from subsonic to supersonic flow, or from
supersonic to subsonic flow, are consequently impossible. For given
conditions a t an initial section of the duct, therefore, the maximum
possible duct length which can be employed without altering these given
initial conditions and without introducing discontinuities is that length
for which the exit Mach Number is exactly unity.
It is a t first surprising to see that friction has the net effect of accelerat-
ing a subsonic stream, and perhaps even astonishing t o observe that
friction causes a rise in pressure a t supersonic speeds.
Working Formulas. Our next step is to integrate the previously given
differential equations in order t o obtain formulas suitable for practical
computations. We shall use the Mach Number as the independent

variable for this purpose.
Eq. 6.12 is rearranged to read
where the limits of integration are taken as (i) the section where the
Mach Number is M, and where x is arbitrarily set equal to zero, and
(ii) the section where the Mach Number is unity, and x is the maximum
possible length of duct, Lmax.
Carrying out the integration, we obtain
wheref is the mean friction coefficient with respect to length, defined by
f =-i
1
Lmax
Lm,
jdz
Eq. 6.20 gives the maximum value of 4fL/D corresponding to any initial
Mach Number M.
Since 47LmaX/Dis a function only of M, the length of duct L required
for the flow to pass from a given initial Mach Number M1 to a given
final Mach Number Mz is found from the expression
To illustrate how the local stream properties are found in terms of

the local Mach Number, we shall take the pressure as an example.
First, we note that Eqs. 6.11 and 6.12 may be combined to give
168 FLOW IN CONSTANT-AREA DUCTS WITH FRICTION Ch. 6
Denoting the pressure a t M = 1 by the symbol p*, we integrate be-

tween the section where M = M, p = p, and the section where M = 1,
p = p*, and thus obtain
By similar methods we get the relations which follow:
r h e quantities marked with an asterisk in these expressions, such as

p*, V*, etc., represent the values of the stream properties a t the section
in the duct where M = 1. Since they are constants for a given adiabatic,
constant-area flow, they may be regarded as convenient reference values
for normalizing the equations. In order to find the change in some
Art. fjb2 ADIABATIC FLOW O F A PERFECT GAS 169
stream property, say t h e pressure, between the sections where the

Mach Numbers are MI and M2, respectively, we set
where (pip*),, is t h e value of t h e right-hand side of Eq. 6.22 corre-

sponding t o M2, and so forth.
Graphical Representation of Workiig Formulas. T h e formulas of
Eqs. 6.20 t o 6.27 are represented graphically in Figs. 6.4 and 6.5, from
which general orders of magnitude
and rates of change may b e found
with ease. On this chart the curve
of 4fLm,,/D is the key t o direction
of change, inasmuch as 4jL,,,/D
must always decrease.
Workiig Tables. For extensive
or accurate numerical calculations,
the important dimensionless ratios
are tabulated in Table B.4, with
Mach Number as the argument.
Mach Number Mach Number, M
FIQ.6.4. Variation of fluid properties FIG.6.5. Variation of fluid properties

on Fanno line, k = 1.4. on Fanno line, k = 1.4.
nlustrative Example.
P R o B ~ M With
. an experimental rig comprising a converging-diverging n o s
zle attached to a smooth, round tube, the following data were obtained with the
aim of measuring friction coefficients for the supersonic flow of air:
Pressure upstream of nozzle = 516 cm Hg abs
Temperature upstream of nozzle = 107.3"F
FLOW IN CONSTANT-AREA DUCTS WITH FRICTION Ch. 6
Throat diameter = 0.2416 inches

Diameter of nozzle exit and of tube = 0.5009 inches
Pressure of stream at a point 1.75 diameters from tube inlet
= 18.25 cm Hg abs
Pressure of stream a t a point 29.60 diameters from tube inlet
= 37.1 cm H g abs
It is desired to calculate the average friction coefficients between the sections

where x / D = 1.75 and x / D = 29.60. It will be assumed that the flow to the
throat of the nozzle is isentropic, and that the flow in the entire system is
adiabatic.
SOLUTION. Since the flow to the throat is isentropic and the entire flow is
adiabatic, the column labeled (p/po)(A/A*)in Table B.2 may be used, as
explained in Chapter 4, for finding the Mach Number at any section.
From Table B.2, then, M 1 = 2.524

Using this value of MI, we get, from Table B.4,
(4fLmar/D)1= 0.4371
Now, since
-
P2 = ( P / P * ) z
PI (P/P*)I
we obtain
37.1
( P / P * )=~ 18.25(0.2878) = 0.5850
Entering Table B.4 with this value of (p/p*)2, we find that
(4fLm,/D)? = 0.1512
Therefore,
AL
4f - = 0.4371 - 0.1512 = 0.2859
D
and, since
AL/D = 29.60 - 1.75 = 27.85
the average friction coefficient is
L i t i n g Duct Length for Supersonic Speeds. Frictional effects are

very serious at high Mach Numbers. T o illustrate, the following table
shows the maximum length-diameter ratio for several initial Mach

Numbers, assuming that j = 0.0025:
An interesting fact revealed by this table is that for purely supersonic

flow, no matter how great the initial Mach Number, the total length
of duct is limited to about 82 diameters for a friction coefficient of 0.0025.
Apart from this, the large losses in stagnation pressure which are incurred
a t supersonic speeds make i t desirable to avoid long runs of duct in which
the flow is supersonic. For example, with an initial Mach Number of
3.0 and a friction coefficient of 0.0025, the Mach Number is reduced to
2.0 in 21 duct diameters, and a t the same time the isentropic stagnation
pressure is reduced to about 40 per cent of the initial value.
Choking Due to Friction. If the value of 4 f L / D between two sections
is fixed, then for each initial Mach Number MI there will correspond a,
certain final Mach Number Mz
which may be easily found from
Fig. 6.5. Illustrative curves of Mz
versus M1 constructed in this way
are shown in Fig. 6.6 for values of
4jLID of 0 and 0.1. Me
For the limiting case where I
4jLID is equal to zero, there is of
course a trivial solution for which
MI = Mz and there are no changes
in any stream properties. On the
other hand, a shock might occur
in zero length. The corresponding o I 2
initial and final Mach Numbers M I
for a shock may be found from increase FIG. 6.6. Illustrates choking due to an
in 4fL,D.
Figs. 6.4 and 6.5 by connecting
the two branches of the F/F* curve with a horizontal line, since the
normal shock conditions of constant area, constant stagnation temper-
ature, and constant impulse function are then satisfied. I n Fig. 6.6 the
curved line labeled 4 j L / D = 0 represents, therefore, a normal shock.
The portion of this line for which M1is less than unity is imaginary,
since it corresponds to an expansion shock.
When there are no shocks present, the relative magnitudes of MI and
Mzfor a fixed length of duct are illustrated in Fig. 6.6 by the solid curves
labelled 4fL/D = 0.1. If the flow is initially supersonic, i.e., if MI > 1 ,
there is a possibility of shocks occurring in the given length of duct.
The relation between M1 and Mz will then depend on the location of the
172 FLOW I N CONSTANT-AREA DUCTS WITH FRICTION Ch. 6
shock. In Fig. 6.6 the curves labeled B and C represent the two extreme
locations: B referring to a normal shock a t the exit of the duct, and C
referring to a normal shock a t the inlet to the duct.
It is clear from Fig. 6.6 that for any initial Mach Number there is a
maximum value of 4jL/D for which a solution is possible, and that
M, = 1 a t this condition. From a different point of view, for a given
value of 4jL/D there is in subsonic flow a maximum initial Mach Number
for which a solution is possible, and in supersonic flow a minimum initial
Mach Number for which a solution may be found. It is pertinent to
ask, therefore, what happens when the duct length initially has its
maximum value for a given MI and the duct length is subsequently
increased. In answering this question we shall suppose that the back
pressure to which the duct exhausts is maintained as low as necessary,
so that any diminutions in the flow are due exclusively to limiting effects
produced by friction.
SUBSONIC FLOW. An increase in the value of 4fL/D over its maximum
value will act to decrease MI until a steady-state solution again becomes
possible with M2 = 1. This results in a reduction in the flow rate, i.e.,
the flow is "choked" by friction.
SUPERSONIC FLOW. An increase in the value of 4jL/D over its maxi-
-
mum value a t first produces a shock in the duct which moves upstream
as the duct length is increased. In
other words, for a fixed MI and a
variable 4jL/D, the value of Ma
remains a t unity, but the flow
pattern in the duct moves from a
C
condition corresponding to curve
,y
&
B in Fig. 6.6 toward a condition

corresponding t o curve C. After
- I the shock reaches the duct inlet,
P'
a further increase in duct length
' 0 0.1 0.2 0.3 0.4 0.5 a6 causes the shock to move into the
4 f L/D nozzle which feeds the duct. This
FIG.6.7. Mach Number and pressure causes the Mach Number a t the
distribution along a duct of varying
length which is fed by a converging- inlet to the duct t o become sub-
%verging nozzle. The exhaust pressure sonic. Further increases in duct,
tn the space to which the duct discharges
is assumed to be less than p*. length cause the shock to move
farther into the nozzle, which in
turn reduces the Mach Number a t the duct inlet. Finally, the shock
vanishes in the throat, and further increases in length produce reduc-
tions in flow rate.
Fig. 6.7 shows the course of states in a duct of varying length with an
initial Mach Number of 2. Length A represents the maximum length
Art. 6.3 PERFORMANCE OF LONG DUCTS 173
without a shock. When the length is increased to B, a shock stands in

the passage. Finally, when the length is increased to C, the shock
stands a t the inlet of the duct length, and the flow in the entire duct is
subsonic.
6.3. Performance of Long Ducts at Various
Pressure Ratios
To illustrate an important aspect of choking, let us investigate how
the adiabatic flow in a duct varies with the ratio of back pressure to
supply pressure. We must distinguish between two cases: (i) that in
which the duct is fed by a converging nozzle, and (ii) that in which it is
fed by a converging-diverging nozzle.
Duct Fed by Converging Nozzle. First we shall consider the case of a
duct fed by a frictionless, converging nozzle. Referring t o Fig. 6.8a,
suppose that the supply pressure, par, and the supply temperature, To,
are fixed, and that the back pressure, p ~is, variable. Conditions a t the
duct inlet are denoted by the subscript I, and a t the duct exit by the
subscript E. Four conditions of operation, namely, a, b, c, and d, are
- 6.8b through
shown in Figs. - 6.8f. The states a t the inlet for these con-
ditions are represented by the symbols a', b', c', and d', and the corre-
sponding states a t the exit by a", bff, c", and d".
In condition a the back pressure is slightly less than the supply pres-
sure. There are small pressure drops through the supply nozzle and
duct. The stream leaving- the duct is subsonic, and is therefore a t a
pressure equal to the back pressure.
-
A reduction in back messure to condition b acts to increase the flow
rate and the pressure drops in the nozzle and duct. The leaving stream
is a t a higher Mach Number than for condition a, but is still subsonic.
Thus, conditions a and b are qualitatively equivalent.
No qualitative changes are observed until the back pressure is reduced
to the value corresponding t o condition c, a t which point the exit Mach
Number is unity. Further reductions in back pressure cannot produce
further increases in flow rate, because ME cannot become greater than
unity. Thus the flow pattern within the duct for condition d is identical
with that for condition c, and the flow is choked. The expansion to the
back pressure for condition d involves oblique expansion waves in the
stream leaving the duct.
Based on these considerations, and using the isentropic flow equations
for the nozzle together with the relations derived in this chapter for
adiabatic flow in a duct, Fig. 6.9a shows the flow passed by ducts of
various length with various ratios of exhaust pressure to supply pressure.
When this ratio is sufficiently 'low, the flow is choked; then Fig. 6.91,
shows the ratio of the maximum flow which can be passed by the nozzle-
-
rt-..-:-=
"To I
L I I
0 Distance Along Duct
FIG.6.8. Performance of duct, fed by converging nozzle.

duct combination to the maximum

flow which can be passed by the
nozzle alone.
Duct F e d by Converging-
Diverging Nozzle. Let us consider
next the case of a long duct which
is fed by a frictionless converging-
diverging nozzle. Referring to
Fig. 6.10, we shall suppose the
supply pressure p o ~and supply
temperature To to be held con-
stant, and that the back pressure,
p ~ is
, variable. The pressure at
the inlet of the duct will be denoted
by pr, and a t the exit by p ~ .
The operation of the system of
Fig. 6.10 under a varying back
pressure is considerably more com-
plex than is that of Fig. 6.8, chiefly
Fro. 6.9. Flow passed by long adiabatic duct fed by isentropic converging nozzle
(k = 1.4).
(a) Dimensionless flow parameter versus ratio of exhaust pressure to supply
pressure.
(b) Ratio of choking flow to choking flow through nozzle alone, as function of
.4fL/D.
because the converging-diverging nozzle of fixed area ratio allows of

only one Mach Number greater than unity a t the duct inlet. We shall
confine our attention to the case where there are no shocks in the nozzle,
since the flow characteristics of the nozzle alone have been discussed in
Art. 5.10, and the characteristics of subsonic flow in a long duct have
just been examined.
Since the nozzle is without shocks, the pressure and Mach Number a t
section I are fixed by the area ratio of the nozzle. To the Mach Number
MI corresponds a particular value of 4jLm,,/D according to Eq. 6.20.
We therefore divide the types of flow into two main classes, depending
FIG.6.10. Duct fed by isentropic converging-diverging nozzle.
on whether the actual length of duct gives a value of 4jL/D greater or

less than the maximum allowable value. For convenience in classifica-
tion, the pressure a t the duct exit for shockless flow is denoted by p ~ , ,
regardless of whether shockless flow can in fact occur. The pressure
corresponding to Mach Number unity is denoted by p*. We may now
define the classes of flow as follows:
CLASS (1). The duct length is less than the maximum possible length,
so that < p* and L < L,, (Fig. 6.11).
FIG. 6.11. Operation of FIG. 6.12. Operation of

system of Fig. 6.10 when system of Fig. 6.10 when
p ~ '< p* and L < L
,
,. p ~>' p f and L > L,.
CLASS (2). The duct length is greater than the maximum possible
length, so that p p > p* and L > L,,, (Fig. 6.12).
The value of p* is a function only of MI and PI. The value of p ~ ,on
,
the other hand, depends on MI, p ~and, 4jL/D.
We shall now discuss the several types of flow in each class.

CLASS(la): p ~ < l p*, and PB < p ~ (Fig.f 6.11b). The duct length
is less than the maximum, and the back pressure is lower than the exit
pressure for shockless flow.
A shock cannot be present, for, if i t were, the flow a t the tube exit
would in general be subsonic. But, as the exit pressure would therefore
have to be identical with the back pressure, the entropy would decrease
adiabatically near the end of the tube.
Accordingly, we conclude that the flow in the tube is shockless. The
actual exit pressure, p ~ is, then determined completely by the magni-
tudes of PI, MI, and 4fL/D. I n other words, p~ = p ~ fand , the exit
flow is supersonic.
The stream leaves the tube a t a pressure higher than the back pressure,
and the adjustment to the back pressure occurs downstream of the duct
exit in the form of oblique expansion waves.
CLASS (lb): p ~ < t p*, and p~ > PB > p ~ (Fig.
l 6.11b). The duct
length is less than the maximum, and the back pressure is, within limits,
greater than the exit pressure for shockless flow.
If the back pressure, p ~ is, not too much greater than the pressure
p p , there will be no shocks in the duct, and the flow pattern in the duct
will be identical with that for Class (la). The adjustment of the stream
to the back pressure will then occur outside the duct in the form of
oblique shock waves.
After the back pressure is raised to the value corresponding to the
pressure pc downstream of a normal shock in the duct exit, further
increases in back pressure cause a normal shock to appear in the duct.
CLASS( 1 ~ )p: p < p*, and p~ > pc (Fig. 6.11~). The duct length is
less than the maximum, but the back pressure is so high that the stream
cannot leave the duct in supersonic flow (with the pressure pEt) and
adjust itself to the back pressure.
Instead, a normal shock stands in the duct, with subsonic flow in the
part of the duct downstream of the shock. As the flow a t the duct exit
is subsonic, the actual exit pressure p~ must be identical with the back
pressure p ~ .This condition, namely, that p~ = p ~ determines
, the
location of the shock.
An increase in back pressurc causcs the shock to move upstream.
Ultimately it moves into the diverging portion of the nozzle, and, after
the shock disappears into the throat, further increases in back pressure
produce a reduction in flow rate.
CLASS (28): p p > p*, and PB > p* (Fig. 6.l2b). Shockless flow to
the exit (shown by the dashed curve in Fig. 6.12b) would involve a
decrease of entropy. Consequently, a shock stands in the duct. Since
p~ > p*, the flow a t the duct exit must be subsonic, and, therefore, the
exit pressure p~ must be identical with the back pressure p ~ .This
condition, namely, that p~ = p ~ determines
, the location of the shock
in the tube.
Increases in the back pressure cause the shock to move upstream and
ultimately to vanish in the throat of the nozzle.
CLASS (2b): p p > p*, and p~ < p* (Fig. 6.12~). As in the case of
Class (2a), the duct length is so great that i t is impossible for the flow
to be supersonic in the entire length of duct. Accordingly, a shock is
present.
The flow a t the duct exit cannot be supersonic, for the flow downstream
of the shock is subsonic and can, therefore, not become supersonic
except through area chalige or heat transfer.
Furthermore, the flow a t the duct exit cannot be subsonic, with a
pressure greater than p*, for on leaving the duct it would diverge on
account of the exit pressure being greater than the back pressure, and
the pressure increase produced by this divergence would make it impos-
sible for the stream ever to adjust itself to the back pressure.
The sole remaining possibility is that the stream leaves the duct at
Mach Number unity, or, in other words, that p~ = p*. The adjustment
of the stream to the back pressure then occurs outside the duct in the
form of oblique expansion waves. As long as the back pressure is less
than p*, therefore, there is only one location in the duct a t which the
shock may stand.
Finally, it should be mentioned that the flow patterns of Figs. 6.10
and 6.11 are idealized in that the shocks are shown as discontinuities.
Because of boundary-layer effects, the transition from supersonic to
subsonic flow is more likely to be extended over a considerable distance,
as explained in Chapter 5 .
6.4. Isothermal Flow in Long Ducts

As pointed out earlier, isothermal flow with friction is of interest in
connection with pipe lines for transporting gas over long distances.
. Although the Mach Numbers for such flows are usually quite low, there
are substantial changes in pressure owing to the great lengths over which
friction acts, and so the flow may not be treated as incompressible.
Governing Physical Equations and Definitions. The analysis is paral-
lel to that for adiabatic flow, except that the energy equation now
includes changes in stagnation temperature. For a perfect gas the energy
equation may be written
Art. 6.4 ISOTHERMAL FLOW IN LONG DUCTS 179
where To is the local adiabatic stagnation temperature, i.e., the tem-

perature which the local stream would acquire if brought to rest adia-
batically. The change in To is a direct measure of the amount and
direction of heat transfer.
In Chapter 4 we derived the relation
Taking logarithmic differentials and noting that dT = 0,we get
For isothermal flow, the equation of state of a perfect gas becomes
~ P / P= ~ P / P
Similarly, Eq. 6.5 becomes
Eqs. 6.7, 6.8 and 6.9 are valid as they stand for isothermal flow.
Qualitative Effects of Friction Under Isothermal Conditions. Solving

the simultaneous equations, Eqs. 6.30, 6.31, 6.32, 6.7, 6.8, and 6.9, with
4j dx/D as the independent variable, we get
From these equations i t is seen that the direction of change depends

not on whether the flow is subsonic or supersonic, but principally on
whether kAA2 is greater or less than unity. Noting that 4jdx/D is always
positive, the directions of change are summarized for gases with k >1
in the table which follows.
M<I/& M> I/&
(Subsonic) (Subsonic or Supersonic)
Pressure decreases increases
Density decreases increases
Velocity increases decreases
Mach Number increases decreases
Stagnation temperature increases decreases
increases for M <
Stagnation pressure decreases
decreases for M >
It is seen that the Mach Number always tends towards I/%'% This
value therefore represents a limit for continuous isothermal flow, in the
same way that M = 1 represents a limit for continuous adiabatic flow.
When M is less than ljdi,heat is added to the stream; when M exceeds
I/&, heat is rejected from the stream.
Working Formulas. Eq. 6.33 may be rearranged t o give
where the lower limit of integration is taken a t x = 0, M = M, and the

upper limit is taken a t M = 1 / d k , beyond which the continuous iso-
thermal flow may not proceed. Integration yields
L,,, 1 - kM2
4fT -
-
kM2
+ In kM2
Now M2 = V2/kRT, and T is constant. Denoting properties a t
M = 1/2/7E by such symbols as V*t, P * ~ etc.,
, we may write, therefore,
from which
From the equation of continuity,
,~
The perfect-gas relation yields
Art. 6.4 ISOTHERMAL FLOW I N LONG DUCTS
From the formula for isentropic stagnation pressure, we get
Finally, the formula for stagnation temperature yields
Working; Charts. The formulas of Eqs. 6.36 to 6.41 are represented

graphically in Fig. 6.13. The procedure for employing the working for-
mulas and chart for computations is similar to that for adiabatic pipe
Bow.
M
FIG.6.13. Working chart for isothermal flow in ducts, k 3 1.4.
Relations for Low Mach Numbers. In long commercial pipe lines

the Mach Numbers employed are so low that the loss in stagnation
pressure is virtually identical with the loss in static pressure. For such
cases it is useful to have a direct relation connecting pz/pl, 4jL/D, and
MI, where subscripts 1 and 2 refer respectively to the inlet and exit
conditions for a pipe of length L. From Eq. 6.36,
1 - kM12 1 - kMZ2
-
m12
-
kMZ2
+ ln-M12
~2~
Also, from Eq. 6.39,
Substituting this valuc of M2 into the previous equation, and rearrang-

ing, we obtain
L
D
4f-=
1-T;(
~ , 2- ln (E)2
(6.42)
Fig. 6.14 is a convenient nomograph representing this relation. From

Eq. 6.39 and the fact that Ma cannot exceed l/d\/lict,
follows that
Hence, although for given values of MI and 4fL/D there are two solu-
tions for pz/pl, only the left-hand intersection may be used. The right-
hand intersection implies negative values of f and involves a violation of
the Second Law of Thermodynamics. A tangent-type intersection
indicates choking flow, and, when there is no intersection a t all, L
exceeds Lm,.
When the per cent pressure drop is fairly small, i t is convenient to
expand the right-hand side of Eq. 6.42 in a power series of the fractional
pressure drop, (pl - p2)/pl. Carrying out this expansion, and retaining
terms up to the second power of this variable, we obtain the following
useful approximation applicable to low pressure drops:
Art. 6.4 ISOTHERMAL FLOW I N LONG DUCTS 183
The conventional pressure-drop formula for incompressible flow is similar

to Eq. 6.43a except that the square bracket on the right-hand side has
the value unity. From this we see that the error incurred through
assuming the fluid incompressible is negligible only when both kM12
and ( p l - p 2 ) / p l are both negligible compared with unity.
use L e f t - H a n d
Intersection On
P,/P, Scale
FIG.6.14. Nomograph for determining static pressure drop for i ~ ~ t h e r mflow

a l in
duct of length L. Subscripts 1 and 2 refer respectively to inlet and exit conditions.
Example: If k = 1.4, MI = 0.313, and 4 f L / D = 3.43, then left-hand ordinate is
0.765 and right-hand ordinate is 0.137; using left-hand intersection, pz/pl = 0.638.
(After G . W. Thomson, I d . a d Eng. Chem., Vol. 34, No. 12, July, 1942. p. 1485.)
Eq. 6.43a is a quadratic in the fractional pressure drop, and may be

solved explicitly for the latter in the convenient form
In employing Eqs. 6.42 and 6.43 it is convenient to remember that
Choking Effects. Since, for a given value of MI, there is a maximum

length for continuous isothermal flow, i t follows that choking effects
may occur in similar fashion t o those for adiabatic pipe flow.
It should be kept in mind, however, that when a subsonic isothermal
flow approaches the limiting Mach Number, all fluid properties change
rapidly with distance. Unless heat is transferred purposefully, there-
fore, the flow process under these circumstances is likely to be more
nearly adiabatic rather than isothermal. At M = l/dfE,Eqs. 6.35 and
6.29 indicate the need for infinite heat transfer per unit length; thus this
limit is artificial and not physically real.
184 FLOW I N CONSTANT-AREA DUCTS WITH FRICTION Ch. 6
6.5. Experimental Friction Coefficients

Subsonic Flow. Keenan and Neumann (I) measured friction coeffi-
cients for turbulent flow in smooth pipes over a range of Mach Numbers
between zero and nearly unity (Volume 11, Chapter 28).
FIG. 6.15. Friction coefficient versus pipe Reynolds number for incompressible,
fully developed flow. Roughness of pipe is measured by e, and has typical values as
follows (after Moody):
P
- be c. f t
-
Drawn tubing 0.000005
Commercial steel 0.00015
Asphalted cast iron 0.0004
Galvanized iron 0.0005
Cast iron 0.00085
Concrete 0.0014.01
Riveted steel 0.0034.03
For a fully developed velocity profile, i.e., a t distances from the duct
inlet greater than about 50 diameters, no significant effect of Mach
Number was observed. In other words, the relation between friction
coefficient and pipe Reynolds Number for subsonic flow a t any Mach
Number was found to agree with the well-known Karman-Nikuradse
formula for incompressible flow:
where Rey is the pipe Reynolds Number based on diameter.

Art. 6.5 EXPERIMENTAL FRICTION COEFFICIENTS 185
For ready reference, the incompressible friction coefficient is plotted
versus Rey in Fig. 6.15 for both smooth pipes and for rough pipes, in a
form developed by Moody. (2)
SHORT PIPES. Near the duct inlet the one-dimensional analysis is
misleading, for the changes in velocity profile which occur there have
associated with them changes of momentum flux of appreciable magni-
tude. In the region of changing velocity profile, therefore, the symbol f
LL
in Eq. 6.8 represents an apparent friction coefficient,'' since it includes
momentum-flux effects as well as the effects of shearing stress.
The measurements of Reference 3 indicate that near the inlet the
flow is similar to that on a flat plate, inasmuch as the boundary-layer
thickness is a small fraction of the pipe diameter. The Reynolds Number
based on distance from the inlet is more significant than the pipe Reyn-
olds Number, and the boundary layer is laminar up to a length Reynolds
Number of about 5(10)5. For short pipes the average apparent friction
coefficient may be substantially in excess of that shown in Fig. 6.15.
Supersonic Flow. The foregoing remarks concerning flow in short
ducts apply with special force to supersonic flow, since the maximum
possible length of duct for supersonic flow is so small that a fully de-
veloped flow can never be established. Here it might be stated paren-
thetically that at high Mach Numbers, fully developed flow could not
be realized even in very long ducts, since all fluid properties change
rapidly along the length of duct.
The complexities thus introduced are discussed in some detail in
Chapter 28. Here we shall simply point out that the local apparent
friction coefficient is influenced by (i) Mach Number, (ii) whether the
local boundary layer is laminar or turbulent, (iii) changes in momentum:
flux associated with changes in velocity profile owing to the regular
development of either a laminar or turbulent boundary layer, (iv)
changes inmomentum flux shortly after transition because of a fairly
rapid change from a laminar velocity profile to a turbulent velocity
and~(v)changes in momentum flux occurring after the boundary
layer fills the pipe because of continuous alterations in the "fully de-
veloped velocity profile."
I t is evident from these considerations that the local apparent friction
coefficient is likely to be significantly dependent upon several factors,
including Mach Number, pipe Reynolds Number, length Reynolds
Number, initial thickness of boundary layer, and initial degree of turbu-
lence. Consequently the experimental data (If4)are most notable for
the scatter displayed, with no simple correlation seemingly possible;
under certain conditions the local apparent friction coefficient may even
be negative. To obtain a more coherent correlation of the data, it is
necessary to take account of two-dimensional effects because the one-
186 FLOW IN CONSTANT-AREA DUCTS WITH 'FRICTION Ch. 6
dimensional model of the flow omits certain essential features. Of
course this complicates considerably the otherwise simple calculations.
EXPERIMENTAL RESULTS. TO assist in making order-of-magnitude
estimates, the results of Keenan and Neumann ( I ) may be summarized
a s follows :
For tubes between 10 and 50 tube diameters in length, the average
apparent friction coefficient for the entire tube length was found t o vary
between 0.002 and 0.003. T h e Mach Number range was from about
1.2 t o 3, and the range of pipe Reynolds Numbers from 25,000 t o 700,000.
T h e corresponding range of friction coefficients for fully developed
incompressible flow is from about 0.003 t o 0.0065. T h e measured appar-
ent friction coefficients for supersonic flow were therefore only about
half a s great a s the coefficients commonly employed for incompressible
flow.
Additional details on friction coefficients for compressible flow in
ducts are presented in Volume 11, Chapter 28.

1. KEENAN,J. H., and NEUMANN, E. P. Measurements of Friction in a Pipe for
Subsonic and Supersonic Flow of Air, JOUT.A p p . Mech., Val. 13, No. 2 (1946),
p. A-91.
2. MOODY, L. F. Friction Factors for Pipe Flow, Trans A.S.M.E., Val. 66 (1948),
p. 671.
3. SHAPIRO, A. H., and SMITH, R. D. Friction Coefficients in the Inlet Length of
Smooth Round Tubes, N A C A Tech. Note, No. 1785 (1948).
4. KATE,J., KEENAN,J. H., and SHOULBERG, R. H. Measurement of Recovery
Factors and Friction Coefficients for Supersonic Flow of Air in a Tube, Pro-
ceedings of the General Discussion on Heat Transfer, London: The Institution
of Mechanical Engineers, September, 1951.
PROBLEMS
6.1: A stream of air flows in an insulated tube of constant area having a
cross-sectional area of 1 sq ft. At a section 1 the pressure is 10 psia, the temper-
ature is 40°F, and the mass velocity is 29.6 lb/sec ft 2 . The pressure in the space
to which the tube exhausts is so low that a "choking" condition prevails, i.e.,
alterations in the pressure of the exhaust region have no effect on the flow through
the tube.
(a) Calculate the Mach Number at section 1.
(b) Calcuhte the Mach Number, temperature, and pressure at the exit of
the tube.
(c) Calculate the total force in the axial direction which must be exerted to
hold statbnary the section of the duct between section 1 and the exit.
6.2. A perfect gas flows in an insulated pipe. Show that the pressure at'the
choking condition is given by
6.3. (a) Estimate the maximum flow rate of air (lbm/sec) which can flow
through the passage shown, assuming that the friction coefficient of the duct
is 0.005. (b) For what range of back pressures will
this maximum flow rate be achieved?
6.4. Consider a long, round insulated tube fed
with air by a frictionless nozzle. The critical pres-
sure ratio of this system is defined as the ratio of PROB. 6.3.
the minimum possible pressure a t the duct exit to
the nozzle supply pressure. Plot this critical pressure ratio versus the duct
4fL/D for values of the latter from 0 to 10.
6.5. Air enters an insulated tube of 1-inch diameter through a converging-
diverging nozzle with a throat diameter of 0.400 inches. The pressure and
temperature in the low-velocity region at the nozzle entrance are 100 psia and
140°F, respectively. The flow through the nozzle may be assumed frictionless,
and the average friction coefficientin the duct may be taken as 0.0025.
(a) Plot the pressure, temperature, Mach Number, and isentropic stagnation
pressure as a function of distance from the tube inlet, assuming shockless flow
to the exit. Determine the maximum length of duct for these conditions.
(b) Suppose that the tube is 40 inches long. Plot, versus the back pressure
in the space to which the duct discharges, the following quantities: mass rate
of flow, pressure and Mach Number at the duct exit, pressure and Mach Number
at the duct inlet, and location of shock.
(c) Repeat part (b) for a tube 90 inches long.
6.6. An isentropic nozzle having an area ratio of 2.00 discharges air into
an insulated pipe of length L and diameter D. The nozzle is supplied a t 100
psia and 70°F, and the duct discharges into a space where the pressure is 40
psia. Calculate the 4jL/D of the pipe and the mass flow per unit area in the
pipe.{&m/sec ft2) for the cases where a normal shock stands:
(a) In the 'nozzle throat
(b) In the nozzle exit plane
(c) In the duct exit plane
6.7. (a) Demonstrate that for incompressible flow of a perfect gas in a pipe
the fractional pressure drop is given by
(b) Derive the corresponding expression for the fractional loss in stagnation
pressure.
6.8. Natural gas (molecular weight 18, k = 1.3) is to be pumped through
a pipe of 36 inches I.D. connecting two compressor stations forty miles apart.
At the upstream station the pressure is not to exceed 90 psig and at%hedbwn-
stream station it is to be at least 10 psig. Calculate the maximum allowable
rate of flow (cubic feet per day at 70°F and 1 atm), assuming that there is
sufficient heat transfer through the pipe wall to maintain the gas a t 70°F.
6.9. A converging nozzle supplied with air at 100 psia and 70°F discharges
into a duct 10 inches in diameter and 1000 ft long. Assuming the flow in the
nozzle isentropic and the flow in the duct isothermal,
(a) Plot the mass rate of air flow (lb/sec) versus the exhaust pressure (psia)
in the space to which the duct discharges.
(b) What is the maximum mass rate of flow?
(c) What is the rate of heat transfer (Btu/hr) to the duct for the maximum
flow rate?
(d) What is the inlet Mach Number to the duct at the maximum flow rate?
6.10. Consider isothermal flow through a frictionless nozzle supplied with a
gas at low velocity, pressure pi, and temperature T,.
(a) Derive relations for p/pi, p/pi, V/V*t, To/Toi, po/poi, and A/A*" all in
terms of the local Mach Number M, where To and po are respectively the isen-
tropic stagnation temperature and pressure, and V*l and A*' are respectively
" and cross-sectional area at the section where w/A is a maximum.
the velocitv
(b) Show that at the minimum area the Mach Number is l/& and the
pressure ratio is p*l/p; = I/& = .-/I
(c) In what way are the results of parts (a) and (b) identical with what would
be found from the isentropic-flow relations by setting k = l ? Investigate the
reasons for the similarities and dissimilarities.
6.11. Derive a formula relating (T ds - dQ) to 4f dx/D for the isothermal
pipe flow of a perfect gas. From this show that the friction coefficient f must
always be positive.
6.12. The measurements shown on the sketch are made in a certain wind
tunnel where it is known that the boundary layers on the walls of the test
section are very thin.
In interpreting test data on the model, what would you consider the free-
stream Mach Number for the model? Defend your conclusion.
6.13. It is proposed to measure the chemical composition of the gases ap-

proaching the nozzle of a turbojet engine by withdrawing a sample of gas
through a capillary tube 0.500 inches in length and 0.010 inches in internal
diameter. The capillary will be attached to a much larger tube at A, as shown
in the sketch, and the gas will be pumped to the chemical apparatus by a vacuum
pump. One of the practical questions is whether or not the time required to
draw a sample of reasonable size is excessive, and it is therefore desired to esti-
mate the maximum rate at which gas can be drawn through the capillary.
The stream in which the probe is placed is a t 20 psia and 1200°F, and flows
at a speed of 1000 ft/sec. For purposes of estimating, it may be assumed that
Ch. 6 PROBLEMS IS9
the gas has properties k = 1.4 and R = 53.3 ft lbf/lbm F. The viscosity of the
gas may be taken as 2.6(10)-5 lbm/ft sec. Since the flow in the capillary is
likely to be laminar, the friction coefficient is given by the Poisseuille formula:
f = 16/Rey
(a) Estimate the maximum rate at which gas may be sampled, in pounds per
second.
(b) How much vacuum must the pump develop, in inches Hg, in order to
realize the maximum flow rate?
(c) What is the pressure a t the inlet of the capillary a t the condition of
maximum flow?
6.14. Consider the steady, laminar seepage of a perfect gas through a long,
fine capillary tube of diameter D and length L. Because of its very great length
and small diameter, the resistance to flow is very large, and consequently the
fluid velocities are so small that changes in fluid momentum and kinetic energy
are negligible. At the same time, the pressure drop is so large that there are
great variations in fluid density.
Show that the pressure drop is given by "d'ArcyJs formula,"
where p is the coefficient of viscosity.

Hint: Demonstrate first that the Hagen-Poisseuille law for laminar flow is
approximately valid for each increment of length.
Chapter 7
FLOW IN DUCTS WITH HEATING
OR COOLING

From a one-dimensional point of view, the three most common factors
tending to produce continuous changes in the state of a flowing stream
are (i) changes in cross-sectional area, (ii) wall friction, and (iii) energy
effects such as external heat exchange, combustion, or moisture con-
densation. I n Chapter 4 we studied the effects of area change with
frictional and energy effects absent; the corresponding process, which
we have called isentropic flow, might aptly be termed Simple Area
Change. In Chapter 6 we studied the effects of wall friction in the
absence of area change and energy effects; the corresponding process is
described by the Fanno curve and may aptly be termed Simple Friction.
In this chapter attention is focused on a process involving changes in
the stagnation enthalpy or stagnation temperature of a gas stream which
flows at constant area and without frictional effects. This process will
be called Simple To-Change.
Such a process is difficult to achieve in practice. If the stagnation
temperature is raised or lowered through external heat exchange, the ~ ~
connection between the mechanisms of friction and of heat transfer

assure that frictional effects will be present. Similarly, if the stagnation
temperature changes because of combustion, there will be a change in
chemical composition. To cite still a third example, a reduction in
stagnation temperature through evaporation of a liquid into the gas
stream means that both the mass rate of gas flow and the chemical com-
position will change.
Despite these objections, many important conclusions of practical
significance may be drawn by analyzing the process of Simple To-Change.
These conclusions can be expected to have a high degree of validity
when the departures from the assumptions of the model are small. For
example, when the heating or cooling is through external heat exchange.
frictional effects per unit length of duct will be relatively unimportant
compared to heating effects per unit length of duct in those cases where
the temperature differential between the duct wall and the moving
.cream is large. I n the combustion of hydrocarbons with air, the fuel-air
190
ratio is small; consequently, effects owing t o change in chemical com-

position are small compared with the effects owing t o changes in stag-
nation temperature.
Additional material relevant t o the subject matter of the present
chapter m a y be found in Volume 11, Chapters 25, 26, 27, and 28.
NOMENCLATURE
area (R recovery factor
speed of sound s entropy per unit mass
specific heat at constant pressure T absolute temperature
pipe diameter Taw adiabatic wall temperature
impulse function T, wall temperature
mass velocity V velocity
enthalpy per unit mass w mass rate of flow
latent heat per unit mass
heat of reaction at constant X coefficient of thermal conductiv-
temperature and pressure per ity
unit mass of mixture p coefficientof viscosity
heat transfer coefficient p density
Mach Number ( ) * signifies state at which M = 1
pressure for process of simple To-
rate of heat transfer per unit Change
wetted area ( )o signifies isentropic stagnation
heat state.
gas constant
Governing Physical Equatic ~ssumptionsof constant

area and no friction, the momentum equation may be written
p + pV2 = -FA = constant (7.1)
T h e equation of continuity is
w
pV = - = G = constant
A
Combining these, we have
The Rayleigh Lie. For nxed values ot,the now per unit area and t h e
impulse function per unit area, Eq. 7.3 defines a unique relation between
the pressure and the density, a relation which we have previously called
the Rayleigh line. Since both the enthalpy and entropy are functions of
192 FLOW I N DUCTS WITH HEATING OR COOLING Ch. 7
pressure and density, it follows that Eq. 7.3 may be used for representing
the Rayleigh line on the enthalpy-encropy diagram, as in Fig. 7.1.
Writing Eq. 7.3 in differential form and introducing Eq. 7.2, we get
dp G2
-=-- - V2
3
dp p2
or,
v = M
Now, eP
represents the local velocity of sound only for a special
circumstance, namely, when the
/PO, Ray'eigh, infinitesimal variation of pressure
h3L-+---
' dition
no
withchange
density
is fulfilled
of isentropy.
such
a t that
theThis
point
there
con-
of
is
maximum entropy on the Rayleigh

h
-q51 ~7 . <,.+
o/
P
/GOO
\\?
line (Fig. 7.1). This point is
marked with an asterisk to denote
7 /_-- +eo Raylelgh, that it is the state of Mach Num-
-- ber unity for the process ojSimple
M>I
/ To-Change. The words process of

Simple To-Change are italicized
here because, beginning with state
s 1 on Fig. 7.1, Mach Number unity
Fro. 7.1. Rayleigh curve for Simple might be reached in several ways
TO-Change. (isentropically, adiabatically at
constant area, etc.), and it is only
for simple heating that the asterisk point of Fig. 7.1 will correspond to
Mach Number unity.
Second Law of Thermodynamics. All fluids thus far investigated have
Rayleigh curves of the general form shown in Fig. 7.1.
The branch of the Rayleigh curve above the point of maximum
entropy generally corresponds to subsonic flow, whereas that below
corresponds to supersonic flow. Since the process of simple heating is
thermodynamically reversible, heat addition must correspond to an
entropy increase and heat rejection must correspond to an entropy
decrease. At subsonic speeds, therefore, the Mach Number is increased
by heating and decreased by cooling. At supersonic speeds, on the other
hand, the Mach Number is decreased by heating and increased by cool-
ing. Thus, heat addition is like friction in that i t always tends to make
the Mach Number approach unity. Cooling causes the Mach Number
to change always in the direction away from a Mach Number of unity.
For heat addition a t either subsonic or supersonic speeds, the amount
of heat input cannot be greater than that for which the leaving Mach
Art. 7.2 SIMPLEHEATING RELATIONS FOR A PERFECT GAS 193
Number is unity. If the heat addition is too great, the flow will be
choked, that is, the initial Mach Number will be reduced to a magnitude
which is consistent with the specified amount of heat input.
These phenomena will be investigated in detail in the articles which
follow.
7.2. Simple-Heating Relations for a Perfect Gas
As in the preceding chapters, we shall pay special attention to the
flow of a perfect gas. The inaccuracies which are thus introduced are
in most cases more than justified by the resulting simplicity.
Governing Physical Equations and Definitions. Consider the flow
through the control surface of Fig. 7.2. We shall write the physical
i T0 2
1
P
L -----------------J
FIG.7.2. Control surface for analysis of Simple TO-Change.
equations governing the flow in such a way as to obtain relations between

the ratios of corresponding stream properties a t sections 1 and 2 a finite
distance apart.
CONSERVATION OF MASS. The continuity equation for constant area is
MOMENTUM THEOREM. The momentum equation, in the absence of

frictional forces, is
w
P1 - P2 = - (V2 - V1)
A
Using the continuity equation, w / A = pV, and noting that for a perfect
gas pV2 = kpM2, this may be arranged to give
PERFECT-GAS LAW. From the pressure-density-temperature relation

of a perfect gas we get
P2 T2
P2 = --
-
PI pi TI
194 FLOW IN DUCTS WITH HEATING OR COOLING Ch. 7
Mam NUMBER. Using the definition of the Mach Number and the
formula for the sound velocity in a perfect gas, we find
IMPULSE FUNCTION. From the definition of the impulse function,
ISENTROPIC
STAGNATION PRESSURE. The definition of the isentropic
stagnation pressure, Eq. 4.14b, gives
ENTROPY CHANGE. The entropy change may be found from Eq. 2.32~.
Thus,
STAGNATION TEMPERATURE. The considerations thus far put forward

merely serve to establish the relations between different points on the
Rayleigh line. From a physical viewpoint the changes in stream proper-
ties are due primarily t o changes in stagnation temperature. Or, to put it
differently, the rate of change of stream properties along the Rayleigh
line is a function of the rate of change of stagnation temperature.
The stagnation temperature corresponding to a given state is that
temperature which the stream would assume if it were adiabatically
decelerated to zero velocity. From the energy equation, then,
When the process involves heat exchange, the change in stagnation

temperature is a direct measure of the amount of heat transfer. For,
from the energy equation,
When the process involves combustion or evaporation, it is usually

possible to devise an approximately equivalent process of Simple To-
Change. I n such cases the initial and final stagnation temperatures
would be made respectively identical for the real process and for the
equivalent process.
Eqs. 7.4 through 7.11 may be combined and rearranged algebraically
to give the ratios of stream properties a t sections 1 and 2 in terms of the
Mach Numbers a t these sections. For example, Eqs. 7.4 and 7.5 may be
inserted into Eq. 7.6, giving
hen using the value of V2/VI from Eq. 7.7, we obtain
Elimination of T2/T1 from Eqs. 7.11 and 7.13 next yields
Working Formulas. Similarly, expressions for pJP1, V2/V1, etc., all

in terms of MI and Mz, may be found. In the interest of constructing
simple and useful tables and charts, however, it is convenient to normal-
ize the equations by setting the Mach Number equal to unity a t one of
the sections. Let us suppose that the Mach Number is unity a t section 1,
and that section 2 is any other section of the duct. Section 1 will be
denoted by an asterisk, and section 2 will be without subscript. We
therefore write
Mi = 1, p~ = p*, Ti = T*, Tol = To*, etc.

and
Mz = M, PZ = p, T2 = T, To2 = To, etc.
Equation 7.13 then becomes

T
-=
(k + 1)2~2 (7.15)
T* (1 + k ~ " ) "
and Eq. 7.14 becomes
- -
-- (7.16)
To* (1 + kM2)2
Similarly, we obtain the following formulas:
(7.17)
(7.18)
(7.19)
(7.20)
Using these normalized expressions, the ratios of properties a t two

sections where the Mach Numbers
are M1 and M2 are given by ex-
pressions of the type
where (To/To*)M2is found from

Eq. 7.16, and so forth.
Working Charts and Tables.

T o facilitate calculations, i t is well
to have the working formulas given
above in graphical or tabular form.
Fig. 7.3 is a graphical form of
the Simple To-Change formulas,
with Mach Number as the inde-
M
.. pendent variable.
Fro. 7.3. Working chart for Simple To-
The corresponding functions are
Change, k = 1.4. also tabulated in Table B.5.
Art. 7.2 SIMPLE-HEATING RELATIONS FOR A PERFECT GAS 197
Illustrative Example.
PROBLEM. A gaseous mixture of air and fuel enters a ram-jet conlbustion
chamber with a velocity of 200 ft/sec, a t a temperature of 120°F, and a t a pres-
sure of 5 psia. The heat of reaction AH of the mixture for the particular fuel-
air ratio employed is 500 Btu per pound of mixture. It is desired to find the
stream properties a t the exit of the combustion chamber. I t will be assumed
that friction is negligible, that the cross-sectional area is constant, and that the
properties of both the reactants and the products are equivalent to air in respect
to molecular weight and specific heat.
SOLUTION. We begin by computing additional properties a t the inlet, which

will be called section 1. The sound velocity is
and
MI = Vl/cl = 200/1180.3 = 0.16945
The stagnation temperature and isentropic stagnation pressure a t scction 1 are

found for this value of MI from Table B.2. Thus,
5
= 0.9801; p a = -= 5.102 psia
0.9801
To analyze the changes in the combustion process, we first h d from Table

B.5 the ratios of stream properties to the corresponding properties at M = 1.
For MI = 0.16945, we find
The stagnation-temperature rise for the equivalent process of Simple To-

Change is given by
To2 - To1 = AH/cP
Now, AH = 500 Btu/lb, and
k 1.4 53.35
cp = -R - ---- = 0.23991 Btu/lb ° F
k-1 0.4 778.3
so that
500
To2 - To1 = = 2084.1°F
0.23991
The properties at section 2 are found with the help of the T o ~ / T oratio.
I We
&st compute
With this value of (To/To*)2,we again enter Table B.5 and find that
Finally, we obtain the remaining properties at section 2 with the relations
( P / P * ) z- 5-
1.9012 = 4.119 psia
Pz = Pl
( P / P * ) ~ 2.308
(PO/PO*)Z 1.1425
= 5.102 -= 4.688 psia
Po2 = Pol
(PO/PO*)~ 1.2435
It is also of interest to find the maximum heat of reaction for which flow with
the specified initial conditions may be maintained. Choking will occur when the
exit Mach Number is unity. Hence, denoting the choking condition by sub-
script 3,
(T0/T0*)3 1
= 583.0 -= 4550°R
= ( T O / T ~ * ) ~ 0.128
After the fuelair ratio has been increased to the value corresponding to this
maximum heat of reaction, further enrichment of the mixture will produce a
reduction in mass rate of flow and a consequent reduction in the initial Mach
Number.
Effects of Changes in To. T h e effects of changes in To on the
remaining stream properties may b e found from differential relations
of the type derived in t h e previous chapter. Such differential relations
are derived a s part of a more general analysis in the next chapter and so
are not presented here.
However, the desired information may also be found from Fig. 7.3,
a t least for the special value of k = 1.4. T h e key curve on this chart
is t h a t of To/To*,for it is directly connected with the magnitude and
direction of heat transfer. For example, if the speed is supersonic and

the stagnation temperature is increased, the curve of To/To*shows that
the Mach Number must decrease. From the curve of TIT* we then
see that the temperature will increase, and, from the curve of po/po*,
that the isentropic stagnation pressure will decrease. The directions of
these and other changes may be summarized as follows:
Heating Heating Cooling Cooling
M < 1 M > 1 M < 1 M >1
To increases increases decreases decreases
M increases decreases decreases increases
T Note (1) ' increases Note (2) decrease
p decreases increases increases decreases
po decreases decreases increases increases
V increases decreases decreases increases
Note (1): increases for M < I/&, and decreases for M >
Note (2): decreases for M < I/.&, and increases for M > I/&
Heating always acts to reduce the stagnation pressure, irrespective

of whether the speed is subsonic or supersonic. This has obvious prac-
tical implications for heat exchangers, combustion chambers, etc. In-
creases in stagnation pressure, on the other hand, may be obtained a t
either subsonic or supersonic speeds by a cooling process which reduces
the stagnation temperature; in practice this is difficult because other
effects are always present which tend to reduce the stagnation pressure.
The value of T/T* goes through a maximum a t M = .1 / corre-
sponding to point g on Fig. 7.1. In the case of air, therefore, for values
of M between 0.85 and 1, heat addition acts to reduce the stream temper-
ature and heat rejection acts to increase the stream temperature!
Choice of End State. A horizontal line connecting the two branches
of the To/To* curve of Fig. 7.3 defines a pair of end states having the
- unit area, the same value of the impulse function, and the
same flow uer
same stagnation temperature. Accordingly they are the corresponding
states on the two sides of a normal shock.
When the initial state is supersonic (point 1 on Fig. 7.4a), the h a 1
state may be either supersonic (state 2') or subsonic (state Y f ) , The
latter condition corresponds to the combination of simple heating and
a normal compression shock. Identical rcsults are obtained foz (i)
simple heating from 1 to Z f , followed by a normal shock to 2"; (ii) a
normal shock from 1 to 4, followed by simple heating a t subsonic speeds
from 4 to 2"; and (iii) simple heating a t supersonic speeds from 1 to a
state between 1 and 2', followed by a normal shock, and concluding with
simple heating a t subsonic speeds to state 2".
When the initial state is subsonic (point 1 on Fig. 7.4b), the final state
may be only subsonic (state 29, for supersonic speeds could only be
attained through a violation of the Second Law of Thermodynamics
in the form of a normal expansion shock from state 2' to state 2". This
conclusion is valid, of course, only when the heat transfer is in one
direction only. Otherwise a subsonic stream could be accelerated to
Mach Number unity by heating from states 1 to 3 in Fig. 7.4b, after
which the speed could be made supersonic by cooling from states 3 to 2".
To I
-
To
Violates
2nd Law
---Expansion Shock
0 +Compression Shock
0 I 2
0
0 I 2'
FIG.7.4.Illustrates possible end s t a h FIG.7.5. Curves of MZ versus MI for

for Simple To-Change. fixed values of Taz/Tol.
(a) Initial flow supersonic.
(b) Initial flow subsonic.
Fig. 7.5 shows illustrative curves of M2 versus MI for fixed values of

T o ~ / T o IWhen
. T02/T01 = I, one of the two curves is for the trivial
case in which no change of state occurs, and the other of the two curves
represents an adiabatic normal shock. For a value of Toz/Tolsuch as
1.1, corresponding to heat addition, there is a branch corresponding to
an initially subsonic flow and a branch corresponding to an initially
supersonic flow. On each such branch there are two values of Mz for
each value of M I . For initial states which are supersonic the entire
curve of Ma versus M I represents physically possible solutions. For
initially subsonic states, however, only that part of the curve is valid
in which the final Mach Number is less than unity. The remaining
part of the curve implies the existence of a rarefaction shock. Simi-
lar considerations apply to cooling curves, such as the curves labeled
ToZjTot = 0.9.
Art. 7.3 CHOKING EFFECTS I N SIMPLE TO-CHANGE
7.3. Choking Effects in Simple To-Change

Rise in Stagnation Temperature. It is evident from Figs. 7.3 and 7.5
that if the process involves a rise in stagnation temperature, there is
for each initial Mach Number a maximum possible ratio of stagnation
temperatures for which a solution is possible. In the limiting case where
the stagnation-temperature ratio of the process is a t its maximum allow-
able value, the leaving Mach Number is exactly unity. From another
viewpoint, if the initial stagnation temperature as well as the stagnation
temperature rise are fixed, there is in subsonic flow a maximum allowable
initial Mach Number and in supersonic flow a minimum allowable initial
Mach Number for which steady flow is possible.
Fall in Stagnation Temperature. When the process involves cooling,

no limitations on the stagnation-temperature ratio exist, provided that
the initial Mach Number is less than unity. For initially supersonic
speeds, however, and when no discontinuity in the form of a shock is
present, there is a maximum allowable reduction in stagnation-temper-
ature ratio corresponding to any given initial Mach Number. With
this minimum stagnation-temperature ratio the leaving Mach Number
is infinite; this condition corresponds to zero stream temperature at.
exit and to a leaving kinetic energy which is equal to the leaving value
of cpTo.
Problems involving a rise in stagnation temperature are more common
than those involving a reduction. Our attention will, therefore, be
focused on problems of this type.
Subsonic Flow. Considering subsonic flow first, let us suppose that

the stream flows with an initial Mach Number MI and that the imposed
increase of To yields a value of T02/Tol which is greater than the maxi-
mum value read from Fig. 7.3. It is clear that some sort of readjustment
in the flow pattern is necessary if the specified stagnation-temperature
rise is t o occur. This readjustment is brought about by a series of
transient effects characterized by pressure waves propagating in both
directions along the duct. When the flow finally reaches a steady con-
dition, we say that the flow is choked; that is, the initial Mach Number,
and hence the flow rate, is reduced sufficiently to permit the specified
amount of heating to occur. Assuming that the back pressure on the
duct is low enough, the choking will reduce the initial Mach Number
- to the point where the leaving velocity is exactly sonic.
The above considerations may best be illustrated by a practical
example. Suppose that a ram-jet combustion chamber of uniform
cross-sectional area and without friction is a t first operating cold
(that is, without combustion) with an air velocity of about 500 ft/sec
202 FLOW IiX DUCTS W I T H HEATING OR COOLING Ch.7
corresponding to M = 0.5 and To = 500°R. Suppose also that i t is
desired to burn enough fuel to raise the sta.gnation temperature to
2000°R, corresponding to a four-fold increase in To. From Fig. '7.3,
however, we see that, with an initial M of 0.5, the value of To (or of
To/To*) may be increased by only about 45 per cent. When combustion
begins, therefore, the flow is choked until in steady flow the leaving
Mach Number is unity. We therefore set M2 = 1, and write
Since Mz = 1, i t follows that
The corresponding value of MI is found from Fig. 7.3 as 0.246, which

means that the maximum possible air speed a t the inlet to the combus-
tion region is about 250 ft/sec. Fig. 7.3 also yields (po/pO*),= 1.219,
so that
The above results are significant because the assumed stagnation

temperatures are in the range of interest for both ram-jet propulsion
and for gas turbines. We see that the entering velocity is limited by
choking effects to about 250 ft/sec. This represents a restriction on the
capacity of high-output combustors over and above the limitation im-
posed by the maximum velocity for which a stable flame may be main-
tained. It is also noteworthy that with the maximum entering speed
a loss in stagnation pressure of a t least 18 per cent is inevitable, entirely
apart from the losses caused by friction.
Supersonic Flow. Let us now consider heat transfer to a supersonic
stream flowing in a duct which is fed by a converging-diverging nozzle.
If the stagnation-temperature rise is too great to allow of a solution with
the specified initial Mach Number, an adiabatic shock somewhere in
the heating region would not provide an adequate form of readjustment,
since there is no change in stagnation temperature across an adiabatic
shock. For example, if the state a t the duct inlet is represented by
point 1 in Fig. 7.4a, the maximum allowable ratio of final to initial
stagnation temperature is given by (To/ To*)a/(To/To*)l. The occur-
rence of a normal shock in the region of heating, say from 2' to 2", does
not alter the magnitude of the maximum stagnation temperature,
namely, that corresponding to state 3. The adjustment of the flow
must, therefore, occur in the nozzle which feeds the duct. A normal
shock in the diverging section of the nozzle will act to reduce the Mach
Art. 7 . 4 SHOCK WAVES WITH TwCHANGE 203
Number a t the duct inlet below the value corresponding to state 4;

this allows a greater stagnation-temperature ratio for the heating process
without any accompanying reduction in flow rate.
For very large increases in stagnation temperature, however, the flow
in the nozzle would have to be entirely subsonic, and the flow a t the duct
exit would a t most be exactly sonic. This would represent a choking
condition for the combined nozzle-duct system inasmuch as the flow
rate would, for a fixed stagnation-temperature ratio, be insensitive to
variations in back pressure as long as the exit velocity was sonic.
7.4. Shock Waves with Changes in Stagnation Temperature

The adiabatic normal shock wave discussed in Chapter 5 is only a
special type of quasi-discontinuity. Other types of discontinuities occur
in nature. These involve such effects as chemical reaction, evaporation
or condensation, as well as the associated effects of changes in the mass
rate of gas flow, changes in chemical composition, and changes in specific
heat and molecular weight.
We shall consider here two illustrative examples of discontinuities
involving changes in stagnation temperature, namely, (i) the moisture
condensation shock, and (ii) explosion waves in combustible mixtures.
We shall restrict the analysis to the assumptions of the process of Simple
To-Change. In other words, the changes in chemical composition,
specific heat, molecular weight, and mass flow of gas phase will be
ignored, and primary attention will be focused on the changes in stag-
nation temperature.
The analysis of Art. 7.2 may be used for these discontinuities be-
cause no restriction was made there as to whether the change from
section 1 to section 2 is continuous or discontinuous.
Moisture Condensation Shocks. The problem of moisture condensa-
tion shocks first received attention in connection with the flow through
the nozzles of steam turbines. In many cases the supply pressure, supply
temperature, ahd pressure ratio for such nozzles are such that at some
point in the nozzle the saturation condition is reached, and a wet mixture
of saturated vapor and saturated liquid might be expected upon further
expansion of the stream. Experiments show, however, that the precipi-
tation of moisture is delayed beyond the point where saturation is
reached. The phenomenon is called supersaturation, and, during the
period when the vapor pressure is greater than the saturation pressu-i
corresponding to the temperature, it involves the existence of states of
metastable equilibrium. The point of precipitation probably is con-
nected with the appearance of nuclei, perhaps comprising foreign
particles or accidentally formed groups of steam molecules. On the
Mollier (enthalpy-entropy) diagram the line of precipitation states,

called the Wilson line, has been found by experiment to lie about 60
Btu/lb below the line of saturation states.
The precipitation of moisture is quite sudden, as the state of the fluid
jumps quickly towards a state of stable equilibrium, and may in the
first instance be treated as a discontinuity (Fig. 7.6). In steam nozzles
it is observed to occur only in the diverging portion of the nozzle.
FIG. 7.6. Schlieren photographs (flow left to right) of moisture condensation shock
in air. The moisture shock is immediately downstream of the throat; as i t is not quite
normal near the wall, there is a reflected expansion wave a t each wall. Further down-
stream of the moisture shock is an adiabatic normal shock with attendant boundary-
layer separation (M.I.T. Gas Turbine Lab.).
(a) Schlieren knife-edge vertical.
(b) Schlieren knife-edge horizontal.
SUPERSONIC WIND TUNNELS. Moisture condensation shocks in super-

sonic wind tunnels present a serious problem because the sudden con-
densation of the supersaturated water vapor contained in the air stream
not only alters the anticipated Mach Number and pressure, but, and
of more importance, gives rise to nonuniformities in the direction and
,- ) velocity of flow a t the nozzle exit. These nonuniformities occur because
moisture condensation shocks are not strictly normal but rather assume
various shapes depending on the nozzle contour and on the degree of
supersaturation. The moisture condensation shock in wind tunnels
is observed only in the diverging portion of the nozzle. In experiments ('1
it was found that the water vapor does not precipitate until the temper-
ature of the stream is about 110°F below the temperature a t which
saturation is first reached. As the specific heat of steam a t low pressures
and temperatures is about 0.45 Btu/lb F, this means that the line of
condensation is about 50 Btu/lb below the saturation line, a figure which
is in moderately good agreement with the Wilson-line figure of about
60 Btu/lb for steam nozzles.
SIMPLIFIED ANALYSIS. Within the assumptions of the process of
Simple To-Change, the energy equation for the condensation shock may
be written
wcp(To2 - 7'01) = w ~ b g (7.22)
where w denotes the mass rate of flow of the main gas stream, c, the
specific heat a t constant pressure of the main gas stream, Tol and T02
the respective stagnation temperatures upstream and downstream of
Art. 7.4 SHOCK WAVES WITH T O-CHANGE 205
the shock, W L the mass rate of flow of condensing liquid, and h,, the
latent heat per unit mass of condensing liquid.
Eq. 7.22 may be rearranged to give
The simplified model of the moisture condensation shock is now taken

to be a process of Simple To-Change with the stagnation-temperature
ratio of Eq. 7.23. Referring to Fig. 7.4a, if the state before the shock
were represented by point 1, the state after the moisture shock might be
either 2' or 2", depending on the back pressure of the nozzle. For super-
sonic nozzles with low back pressures the downstream side of the shock
would be represented by state 2', and would correspond t o supersonic
flow. It is evident from Figs. 7.4a and 7.3 that moisture condensation
shocksltend t o reduce the Mach Number and increase the pressure a t
the nozzle exit.
It is t o be noted that the droplets formed during condensation are
exceedingly small. Because of the associated effects of surface tension,
the equilibrium relation between vapor pressure and temperature will
depend upon the radius of the droplet, and the term h,, in Eqs. 7.22
and 7.23 must be thought of as an equivalent latent heat which takes
into account the surface energy of the droplets. Little data are available
concerning the size of droplets initially formed, and virtually no data are
available concerning the rate of growth of the droplets.
CHOKING. Thermal choking may play a role in moisture condensation
shocks. For instance, suppose that the condensation begins a t state 1
in Fig. 7.4a. With an assumed droplet size, we may compute the
-
value of TO2/TOlrcorresponding to conditions after equilibrium is

established between vapor and liquid. If this ratio is greater than
the value of (To/To*)3/(To/To*)~ in Fig. 7.4a, it is clear that the assumed
solution is not possible. There seem to be two possible alternatives.
First, it may be that thermodynamic equilibrium is established after
the shock, but that thermal choking sets an upper limit to the size of
droplets which may be formed. Second, it may be that the size of drop-
lets formed is governed by an independent physical mechanism; in this
case thermodynamic equilibrium would not be established after the
shock, and the amount of water vapor condensed would be such as to
yield a Mach Number of unity downstream of the shock. The Mach
Number would remain a t unity as further expansion permitted additional
condensation until equilibrium was finally reached, whereupon the
iMach Number would once again increase. Further experimental data
are necessary to determine which, if either, of the above alternatives
best describes the course of events, or whether steady flow is impossible
and an oscillatory flow occurs.
Explosion Waves in Combustible Mixtures. A problem of practical
importance in connection with combustion chamber design, bomb
explosions, and detonation in internal combustion engines, concerns the
manner in which an explosion wave travels through a combustible mix-
ture. The significant quantities of interest include the speed of the wave
relative to the unburned gas, the motion of the burned gas through which
the wave has already passed, and the increases in pressure and temper-
ature through the wave.
We shall consider here only steady explosion waves, which might be
defined as waves in which the flow on both sides of the wave is steady
with respect to an observer who moves with the wave front. However,
the results are applicable also to unsteady waves provided that the
thickness of the discontinuity zone is extremely small. This is usually
the case in practice. The assumptions of simple heating will be adhered
to, so that no notice will be given the changes in chemical composition,
molecular weight, and specific heat.
Let us consider the case where the unburned gas is motionless with
respect to the earth, so that an ordinary observer sees the wave advanc-
'ing into the unburned gas, and the process appears to be unsteady.
However, steady-state conditions will prevail with respect to a second
observer who moves with the wave front. The equations of Section 7.2
would then be valid for this second observer. In order to obtain rela-
tions for the changes in stream properties for the observer a t rest with
respect to the unburned gas, transformation formulas like those of
Art. 5.9 must be employed.
SIMPLIFIED ANALYSIS. Writing the energy equation for the steady
flow viewed by the observer moving with the wave, we find that, within
the assumptions of the process of simple heating,
where AH is the heat of reaction at constant pressure and temperature

per unit mass of mixture, and c, is the specific heat at constant pressure.
In order to make use of the convenient tables and charts of Simple
To-Change, we put this expression in the form
The left-hand side is determined by the nature of the chemical reaction

Lrt. 7.4 SHOCK WAVES WITH To-CHANGE 207
and by the properties of the unburned gas. Assuming that these are
known in a given case, and remembering that To/To* depends only on
M and k, it follows that Eq. 7.26 is a relation between M1and M2. Fig. 7.7
shows illustrative curves of M2 versus M1for several values of AH/c,TI
in the range corresponding to hydrocarbon explosions; also shown are
curves of constant pz/pl.
AH To2 -To,
cp- T,
FIG.7.7. Curves of M2 versus MI for values of the parameter AH/c,TI in the usual
range of hydrocarbon explosions. Dotdash curves indicate values of pz/pl ( k = 1.4).
FLAME PROPAGATION SPEED. The Mach Number MI is the ratio of

the speed of the wave relative to the unburned gas to the speed of sound
in the unburned gas. Hence the leaving Mach Number M, seen by the
observer moving with the wave front depends, according to Eq. 7.26, on
the speed with which the flame front propagates into the unburned gas.
This flame propagation speed depends on such factors as the conduc-
tion of heat and the diffusion of mass through the combustion zone.
The detailed consideration of these factors is beyond the scope of this
book. We shall discuss here only the different classes of explosion waves
resulting from diierent flame propagation speeds; in this respect the most
distinctive criterion will be seen to be whether the speed of the flame
front is subsonic or supersonic relative to the unburned gas, i.e., whether
MI is less than or g r a t e r than unity.
It will be found most convenient to think in terms of the steady-state

process viewed by an observer traveling with the wave, but the results
will be expressed for a n observer at rest with respect to the unburned gas.
Fig. 7.3 may be employed for determining the velocity and pressure
changes in the steady-state frame of reference. The various configura-
-
tions discussed below are illustrated in Figs. 7.7 and 7.8.
Front Front
- -
(Stationary)
-4
'
VI
I
I V,
@
Vl
Unburned
( A t Rest) I
@ Burned
- -
I I -v2-v,
@ Unburned I @ Burned p~'P2 I
I Vl
Front Front
"1 Vl
a Unburned
( A t Rest)
I @
I
Burned @ Unburned
( A t Rest)
I @ Burned
I -v,-v, I- "1 -v2

Pl'P2 I Pl-=P, I
FIG. 7.8. Illustrates explosion waves (compare with Fig. 7.7).

(a) Process seen by observer in moving reference frame with respect t o which the
wave front is stationary. In (b), (c), and (d) the observer sees the unburned gas at
rest and the wave front moving leftward with speed V1.
(b) Slow burning, corresponding to process 1 + 2' of Fig. 7.4b; V1 < el; Vz < CZ;
12' > V1; burned gas flows away from unburned gas and pressure drops during - burn-
ing.
(c) Deflagration, corresponding to process 1 + 2" of Fig. 7.4b; V1 < cl;, VZ > cz;
Vz > V1; burned gas flows away from unburned gas and pressure drops durmg burn-
-
inn.
(d) Detonation, corresponding either to process 1 -+ 2' or p r o w s 1 -+ 2" of
Fig. 7.4a; VI > el; Vz < cz; Vz < VI; burned gas flows toward unburned gas and
>.
pressure rises during burnmg.
SLOW BURNING. This corresponds to a process from state 1 to state 2'

in Fig. 7.4b. The speed of the flame front is subsonic relative to the
unburned gas, and the burned gas flows away from the unburned gas.
The pressure drops as the front passes the observer.
DEFLAGRATION. This corresponds to a process from state 1 to state 2''
in Fig. 7.4b. The speed of the flame front is subsonic relative to the
unburned gas, and the burned gas flows away from the unburned gas.
The pressure drops as the flame front passes the observer. Although
these effects are qualitatively the same as in slow burning, the magni-
tudes of the changes in velocity and pressure are very much greater
than for slow burning. For the assumptions of simple heating, a station-
ary deflagration constitutes a violation of the Second Law of Thenno-
Art. 7.4 SHOCK WAVES WITH TO-CHANGE 209
dynamics, but it is not clear whether this is also the case for processes
involving chemical reaction.
DETONATION. This corresponds to a process from state 1 to either
state 2' or 2" in Fig. 7.4a. The speed of the flame front is supersonic
relative to the unburned gas, and the burned gas flows toward the
unburned gas. There is a great pressure rise as the flame front passes
the observer. Relative to the front, the flow of the burned gas may be
either subsonic or supersonic.
Assuming that an explosion occurs in a specified conlbustible mixture,
what will be the steady-state speed of the front relative to the unburned
gas? That is, what will be the value of MI, the ratio of the above-named
speed to the speed of sound in the unburned gas? As explained pre-
viously, the equations of continuity, momentum, and energy allow of
an infinite number of steady-state wave speeds MI (see Fig. 7.7). To
each of these values of M1 there correspond either two values of M2 or
none at all. The latter situation implies that a phenomenon similar
to thermal choking defines an upper limit to the value of MI for subsonic
wave speeds, and a lower limit to the value of M1 for supersonic wave
speeds. In both these limiting cases the value of Mz is unity.
The values of MI and M2 which will occur in a given instance depend
not only on the conservation principles of continuity, momentum, and
energy, but also on the kinetics of the chemical reaction and on the
boundary conditions in space and time, insofar as the latter influence
the formation of the steady-state wave front. Without going into detail,
the following facts may be stated.
(i) In slow combustion the speed of the steady-state flame front must
be identical with the speed of propagation of the flame relative to the
unburned gas. The speed of flame propagation depends on such factors
as the nature of the chemical reaction, the temperature of the unburned
gas, the degree of turbulence, and the size of the apparatus in which the
process is carried out. It is to be noted that steady-state slow combus-
tion is impossible if the flame speed is greater than the value correspond-
ing to the maximum value of M1 indicated by Eqs. 7.14 and 7.26. In
such instances it appears that the process, after passing through a tran-
sient phase, becomes one of detonation.
(ii) All detonation experiments thus far performed have yielded
steady-state wave speeds of magnitudes such as to make the speed of
the burned gas exactly sonic with respect to the wave front. Stated
differently, the observed values of MI correspond to the solution of
Eqs. 7.14 and 7.26 for Mz exactly equal to unity (point 3 of Fig. 7.4a),
and represent the minimum possible steady-state detonation speeds
in each experiment. This experimental result may be partially ration-
alized through considerat,ions of the stability of a detonation wave and
of the mechanism by which a wave starting from rest ultimately reaches

the stage of detonati0n.(~>3)The enormous flame propagation speeds
in a detonation wave are possible because the detonation is a combina-
tion of an adiabatic compression shock and a subsequent slow burning.
The shock precedes the flame zone and raises the temperature and pres-
sure of the unburned gas to such high values that the kinetics of the
chemical reaction are augmented to the point where the flame zone
propagates rapidly enough t o keep pace with the shock front.
pd2/pI (Dashed Llne)
FIG.7.9. Thermally-choked combustion waves (M2 = I), fork = 1.4 Solid curves
show maximum flame speed MI for slow combustion, and minimum flame speed M I
for detonation. The latter is the detonation Mach Number always observed in experi-
ments. Dashed curve shows, for detonation with M2 = 1, the stagnation pressure
ratio for an obsewer a t rest with respect to the unburned gas.
T HE CHAPMAN-JOUGUET RULE. The observed fact that M 2 = 1 for a

detonation wave is often called the Rule of Chapman and Jouguet. In
order to find the corresponding detonation speed M 1 , we set ( T O / T ~ * ) ~
equal to unity, and substitute for ( T O / T ~ in
* )Eq.
~ 7.26 the value given
by Eq. 7.16. Carrying out these operations, and simplifying, we obtain
The two values of M I which may be found from this equation corre-
spond to the maximum flame-front speed for slow combustion and the
Art. 7.5 THE RECOVERY FACTOR 21 1
minimum speed for detonation. These speeds are plotted in Fig. 7.9 as
a function of the energy parameter AH/c,Tl.
One measure of the potentialities for damage of a steady detonation
wave is the rise in stagnation pressure experienced by an obstacle over
which the wave passes. Denoting by pbzl the stagnation pressure of the
burned gas in a coordinate system where the unburned gas is a t rest,
it may be shown that
This ratio is also plotted in Fig. 7.9 for a Chapman-Jouguet detonation.

In the usual hydrocarbon detonation, TI is about 520°R, ATo/Tl is
about 6, and the speed of sound is about 1000 ft/sec. From Fig. 7.9 it
may be seen that the flame-front speed is then about 5500 ft/sec and the
stagnation pressure increases to about 25 atmospheres. The temper-
ature after an adiabatic shock traveling a t 5500 ft/sec is found from the
adiabatic normal shock tables to be about 3000°F: thus. as indicated
previously, there is considerable preheating effect available for increasing
the flame speed to values necessary for detonation.
The temperatures in an atomic explosion are said to be measured in
millions of degrees. If ATo/TI is assumed to equal lo4 for such an
explosion, Fig. 7.9 indicates a wave speed of the order of magnitude of
200,000 ft/sec and a stagnation pressure of the order of magnitude of
40,000 atmospheres!
7.5. The Recovery Factor

When a fluid flows through a pipe having an insulated wall, the tem-
perature a t the wall (called the adiabatic wall temperature, Taw) is
neither the average stagnation temperature To nor the average static
temperature T of the stream. In a gas flow the adiabatic wall temper-
ature usually lies between T and To. The factors influencing Tawand
theoretical methods for predicting its value are discussed in detail in
Volume 11, Chapters 27, 28, and 29. For the present we shall introduce
some elementary considerations and shall summarize some experimental
results of value for practical computations of pipe flow.
Physical Mechanism of Frictional Heating. The temperature as-
sumed by an insulated wall, past which flows a high-speed gas stream,
is controlled by complex phenomena in the boundary layer. Fluid
layers far from the wall deliver viscous shear work to layers near the wall.
The consequent temperature rise of the inner layer is necessarily ac-
companied by heat conduction through the fluid away from the wall,
212 FLOW I N 1)UCTS WITH HEATING OR COOLING Ch. 7
thus tending to limit the temperature rise. After these effects have
been brought into balance everywhere in the fluid, the wall is a t the
equilibrium temperature Taw.The
temperature distribution in the
boundary layer is then as shown
-E0,. I Oistrfbutlons
by the solid curve of Fig. 7.10.
0
I I Cold Woll Corresponding curves for a hot wall
a
*" .-. \
I Insulated W a l l
and a cold wall are also shown; it
0
-L"
c
XO'
.. Wall should be noted that whether heat
0 . .*
,' '. flows into or out of the wall depends
T-o-r
Temperature
on the direction of the temperature
FIG. 7.10. Illustrates definition of
gradient in the gas a t the gassolid
adiabatic wall temperature. boundary.
The Recovery Factor. The magnitude of Tawrelative to the one-
dimensional mean values of T and To is usually expressed by the so-
called recovery factor &, which is defined by
This may also be expressed as
T
-aw- Taw
- -
rn
1 1
-T
so that
In Volume 11, Chapter 28 it is pointed out that various experiments

have yielded values of the recovery factor for subsonic pipe flow between
0.87 and 0.91; and for supersonic pipe flow between 0.83 and 0.90.
7.6. The Coefficient of H e a t Transfer

Definition of the Film Coefficient. The film coeficient of heat trans-
fer X for incompressible pipe flow is ordinarily based on the difference
between the wall temperature and the mean stream temperature. For
high-speed compressible flow, however, i t is more appropriate to base X
on the difference between the actual wall temperature, T,, and the
adiabatic wall temperature, Taw.With any other method of defining X
it would be possible for X t o have both negative and positive values.
Moreover, i t is shown in Volume 11, Chapters 26,27, and 28 (cf. Fig. 28.4)
t h a t the value of X defined in terms of (Taw- T,) is the only one lead-
ing t o an experimental correlation which is independent of the rate of
heat transfer. Accordingly, we define X b y
where q is the rate of heat flow per unit wetted area. Substituting for
Tawwith the aid of Eq. 7.30, the heat transfer equation becomes
Summary of Experimental Results. Experimental results for sub-

sonic pipe flow are discussed in Volume 11, Chapter 28. They may be
summarized a s follows:
(i) Between M = 0 and M = 1 the curve of Nusselt Number X D / h
versus Reynolds Number V D P / 9is substantially independent of Mach
Number.
(ii) Between M = 0 and M = 1 the Reynolds analogy between friction
and heat transfer is substantially independent of Mach Number.
(iii) For turbulent flow u p t o Reynolds Numbers of 400,000 the
heat-transfer d a t a are correlated b y the equation
For theoretical considerations concerning the coefficient X in laminar

and turbulent flows the reader is referred t o Volume 11, Chapters 26
and 27.
1. SIMONS, W. W., and BOWEN,J. S. Investigation of the Condensation Shock in Air
~d -
of Tech., Cambridge, Mass. (1946).

.~ tof. Naval Arch.. Mass. Inst.
bv Use of the Schlieren Method. S.M. Thesis., D e - ~
2. SHAPIRO, A. H., and HAWTIIORNE, W. R. The Mechanics and Thermodynamics

of Steady. One-Dimensional Gas Flow. Jour. A .p.p . Mech.. Trans. A.S.M.E..
Vol. 14, N O . 4 (1947), p. 317.
3. COURANT, R., and FRIEDRICHS, K. 0. Supersonic Flow and Shock Waves. New
York: Interscience Publishers, Inc., 1948.
PROBLEMS
7 (a) Fuel is to be burned in a stream of air moving with a Mach Number
of 2.0, a temperature of - 160°F, and a pressure of 10 psia in a duct of constant
area. Estimate the maximum possible rise in stagnation temperature, and the
corresponding final values of the temperature, pressure, and stagnation pressure.
(b) If the supersonic stream of part (a) were to pass through a normal shock
before fuel was burned, estimate the maximum possible rise in the temperature,
214 FLOW I N DUCTS WITH HEATING OR COOLING Ch.7
and the corresponding final values of the temperature, pressure, and stagnation
pressure.
7.2. Make log-log plots of M2 versus MI, p2/pl versus MI, T2/T1 versus Mi,
and po,/po, versus Mi, for fixed values of To2/TO1
of 1.2 (heating) and 0.8 (cool-
ing). Note that there are two branches on each plot for each value of TodToi.
Use solid lines to indicate which portion of each branch is physically possible
and dashed lines to indicate which portions involve violations of the Second
Law of Thermodynamics.
7.3. Derive the asymptotic forms of the Simple To-Change formulas for:
(a) Very low Mach Numbers
(b) Very high Mach Numbers
7.4. Show that, in Simple To-Change a t Mach Numbers less than about 0.2,
changes in density are produced almost exclusively by changes in temperature.
Based on this observation, and considering two sections 1 and 2, demonstrate
that at very low Mach Numbers,
7.5. Fuel is injected and burned in air moving down a straight pipe of uniform
diameter. If the constant-pressure heat of reaction of the fuel is 19,000 Btu/lb,
the stagnation temperature of the air a t inlet is 90OoR, and the burned gas
stream discharges from the duct into a region where the pressure is 15 psia,
plot the isentropic stagnation pressure at the duct inlet section against the
Mach Number of the air at the inlet section for fuel-air ratios of 0.02 and 0.04.
To simplify the computations, it is agreed to neglect friction, the change in the
mass rate of flow, the change in chemical composition, and to assume that there
is no diffuser a t the exit of the duct.
7.6. Air is brought into a tube through a converging-diverging nozzle d e
signed for a Mach Number of 2, and has initially a stagnation temperature of
500°R and a stagnation pressure of 100 psia. The air is to be cooled by some
mechanism not involving friction.
(a) Assuming that there are no shocks, compute the maximum possible amount
of heat rejection per pound of air passing through the duct. Find, for this limit-
ing condition, the final temperature, pressure, and stagnation pressure.
(b) Investigate the types of flow patterns which might exist in the nozzle
duct system if the amount of cooling is greater than that computed for part (a).
7.7. A wind tunnel nozzle for a Mach Number of 2 is to be supplied with air
at 30 psia and 80°F. It is desired to estimate the maximum relative humidity
at the nozzle entrance for each of the following conditions:
(a) At no point in the nozzle is a condition of saturated water vapor reached.
(b) Precipitation of water vapor occurs a t no point in the nozzle. It will be
assumed for this purpose that condensation does not occur until the stream
temperature is 110°F below the temperature a t which saturation is first reached.
(c) The amount of water vapor is such that the Mach Number just down-
stream of the moisture condensation shock is exactly unity, assuming that
Ch.7 PROBLEMS 215
thermodynamic equilibrium with droplets of "infinite" size is attained. AB in

part (b), it will be assumed that condensation does not occur until the stream
temperature is 110°F below the temperature a t which saturation is first reached.
For this case, compute the Mach Number, pressure, temperature, stagnation
pressure, and stagnation temperature a t the nozzle exit and compare with the
corresponding values for flow without moisture condensation shocks.
7.8. The sketch shows a tube rocket motor, comprising a long tube which is
fed with liquid fuel and liquid oxidant at one end, with hot gases leaving the
open end.
In one test of such a motor, the measured thrust, P, was 676 lb; the pressure
on the back face of the motor, pt,, was
265 psia; and atmospheric pressure, p,,
was 15 psia.
With the fuel-oxidant combination
used, the gases are characterized by
k = 1.3 and R = 30 ft lbf/lbm F.
Because of uncertainties as to PROB. 7.8.
whether combustion is completed and
as to the degree of dissociation, the stagnation temperature of the leaving gases
is not known, but it is suspected to be between 3000°F and 5000°F.
(a) What is the pressure p, in the exit plane, in psia?
(b) What is the average shearing stress acting on the inside walls of the motor,
in psi?
7.9. It has been suggested that the combustion in rocket motors should be
carried out a t high speed in a tube of constant cross-section, as indicated in the
sketch. It is desired to compare the specific thrust (thrust force per unit mass
rate of flow, in lbf sec/slug) of such a rocket motor with the specific thrust of
the conventional motor, also shown in the sketch.
Proposed Design . Convent~onolDesign
Combusfion
Lomaustion
PROB. 7.9.
To simplify the calculations it will be assumed that friction is negligible, that

the jet leaves the nozzle at the pressure of the ambient atmosphere (say 5 psia),
that the fuel and oxidant are supplied as gases, and that both rockets are sup-
plied a t the same pressure (pol = pa, = 500 psia). Also, all processes will be
considered adiabatic, except that the combustion will be replaced by an equiva-
lent process of Simple To-Change, with Tol = To, = 500°R, and To3 = TO^
= 5000°R. As an approximation, take k = 1.4 and use a molecular weight of 20.
For the proposed design assume that the velocity is negligible a t section 1,
that all combustion occurs between 2 and 3, and that the Mach Number is
unity at section 3. For the conventional design assume negligible velocity a t a
and complete combustion between a and b.
Calculate for both motors the specific impulse (lbf sec/slug) and the throat
area per unit of thrust (sq ft/lbf) under the assumed conditions.
7.10. It has been proposed that a supersonic wind tunnel be operated by a
ram-jet device, as indicated by the accompanying sketch.
From Shock ,Fuel
Do you consider this scheme feasible? Justify your conclusion by suitable

arguments. Would your conclusion be altered if the fuel were burned in the
supersonic stream following the test section?
7.11. Consider a steady-state explosion wave traveling in a hydrocarbon-air
mixture for which the constant-pressure heat of reaction is 1200 Btu per lb of
mixture. The inflammable mixture is at rest and at a temperature of 70°F
and a pressure of 1 atmosphere. For the two limiting cases in which (i) the wave
speed is the maximum for slow combustion, and (ii) the wave speed is the mini-
mum for detonation, compute the following quantities of interest:
The speed of the wave front and of the burned gas
The pressure of the burned gas
The temperature of the burned gas
The isentropic stagnation pressure of the burned gas referred to an
observer a t rest with respect to the unburned gas
(e) The stagnation temperature of the burned gas referred to an observer
a t rest with respect to the unburned gas
(f) In the case of detonation waves, it is probable that the large flame
speeds necessary to support the wave are obtained with the help of the
preheathg effect caused by an approximately adiabatic shock which
precedes slightly the flame front proper. Compute, therefore, the
temperature behind an adiabatic shock which travels at the speed of
the detonation wave.
7.12. Verify Equation 7.28.
7.13. The sketch shows a slow combustion wave propagating spherically
from a point ignition source into a combustible mixture. The velocity of the
flame front is V,, and instantaneously it is at the radius T,. On the two sides
of this front the pressures are pl (unburned gas) and pz (burned gas), and the
Ch.7 PROBLEMS 217
corresponding densities are p l and pz. At any radius r the pressure is p and the
velocity is V'. At r = m, the pressure is pat,.
The speed of propagation V , is extremely low (i.e., a few feet per second).
It is, therefore, permissible to assume that the fractional variations in pressure
are so small that the density can be taken as constant except across the flame
front, where there is a large temperature change.
Burned Gas Unburned Gas
Show that the curve of p versus r is as shown in the sketch and that outside
the flame front it is represented by
Show also that
Investigate these results for endothermic and exothermic reactions.
7.14. Consider the flow of a combustible gas mixture through a sudden

expansion in a pipe (see sketch). Combustion occurs between section 1 a t the
exit of the small pipe and section 2 in the large pipe where the flow is uniform
and there is no longer any turbulence. Wall friction between sections 1 and 2 is
assumed to be negligible. The flow at section 1 is supplied from a converging
nozzle and has stagnation properties pol and TOI. If M1 is subsonic, the pressure
on the annular boundary AB is equal to pl. If MI = 1, however, the pressure
on this annular boundary may be less than PI. The heat of reaction is denoted
by AH, and the area ratio A2/A1 by a.
(a) When MI is subsonic, show that the leaving Mach Number is determined by
218 FLOW IN DUCTS WlTH HEATING OR COOLING Ch. 7
(b) When M, is subsonic, AH/c,Tol = 0, and a = a,show that
and comment on the significance of this.

(c) Derive appropriate relations for the case where Mq is exactly unity.
7.15. Air with a stagnation temperature of 70°F is supplied to an insulated
pipe at Mach Number 3 by means of a converging-diverging nozzle. Assuming
a recovery factor of 0.85, estimate the wall temperature of the pipe.
7.16. A ram jet flies at sea level with a speed of 1500 miles per hour. At
one point in the inlet duct the Mach Number is 0.5. Estimate the-adiabatic
wall temperature a t that section.
Chapter 8
GENERALIZED ONE-DIMENSIONAL
CONTINUOUS FLOW

Analyses were presented in the previous chapters for simple types of
flow, that is, for flow processes in which only a single independent
parameter was allowed to change. For example, in Chapter 4 the effects
of area change alone were considered; in Chapter 6 the effects of friction
alone were considered; and in Chapter 7 the effects of changes in stag-
nation temperature alone were considered.
In most practical problems these effects occur simultaneously, and,
in addition, there may be present such other phenomena as chemical
reaction, change of phase, injection or withdrawal of gases, and changes
in molecular weight and specific heat. As examples in which simul-
taneous effects are present, we may cite rocket nozzles, ram jets, com-
bustion chambers for gas turbines and turbojet engines, injectors and
ejectors, moisture condensation shocks, moving flame fronts, detonation
waves, heat exchangers, and gas-cooled nuclear reactors.
The phenomena which we shall take into account in our analysis
include:
(i) Area change
(ii) Wall friction
(iii) Drag of internal bodies
(iv) External heat exchange
(v) Chemical reaction
(vi) Change of phase, e.g., evaporation or condensation of water
or fuel
(vii) Mixing of gases which are injected into the main stream
(viii) Changes in molecular weight and specific heat occasioned b~
combustion, evaporation, gas injection, etc.
The assumptions made are as follows:
(i) The flow is one-dimensional and steady.
(ii) Changes in stream properties are continuous.
(iii) The gas is semiperfect, i.e., it obeys Boyle's and Charles'
laws and has a specific heat which varies only with temper-
ature andcomposition.
219
220 GENERALIZED ONE-DIMENSIONAL FLOW Ch. 8
T h e method of analysis follows closely t h a t of Shapiro and Haw-
thorne. (')
Additional material relevant t o the subject matter of the present
chapter may be found in Chapters 4, 5, 6, 7, and in Volume 11, Chapters
26, 27, and 28.
NOMENCLATURF,
area entropy per unit mass

wetted area of pipe wall (single phase)
speed of sound total entropy per unit
specific heat at con- mass of gas stream
stant pressure absolute temperature
mean hydraulic diam- adiabatic wall temper-
eter ature
base of natural log- temperature of pipe
arithms, 2.718 wall
coefficient of friction velocity
impulse function forward component of
influence coefficient de- velocity of injected
etc. fined by Eqs. 8.49 fluid
PT, mean influence coeffi- mass rate of flow of gas
cient evaluated a t stream
(Mi+ M2)/2 molecular weight
h enthalpy per unit mass work
h r constant-pressure heat drag force
of reaction per unit Cartesian coordinate
mass, positive when V1/V;also see Eq. 8 . 2 2 ~
exothermic
H energy term defined by density
Eq. 8.15 wall shearing stress
X heat transfer coefficient x/L
k ratio of specific heats signifies stagnation
L length state
M Mach Number signifies state at which
P pressure M = l
Q net heat per unit mass signifies injected gas
of gas dw,
R gas constant signifies evaporated
CR universal gas constant; liquid d w entering
~
recovery factor control volume
8.2. Physical Equations and Definitions

Consider the flow in a duct between two sections an infinitesimal
distance dx apart (Fig. 8.1). In this element of duct length gas is
injected into the stream at t h e mass rate of flow dw,, liquid evaporates
Art. 8.2 PHYSICAL EQUATIONS AND DEFINITIONS 221
into the stream a t the mass rate of flow d w ~heat

, in the amount dQ is
added to the stream from external sources, and shaft work, shearing
work, or electrical work is delivered by the stream to external bodies
in the amount dW,.
The various physical equations and definitions will be expressed in
logarithmic differential form. It will be seen that this procedure allows
easy separation of the physical variables.
A I 2oniro1 I P +dp
P I Surface I T +dT
T 1 I
F I ~8.1.
. Control surfaces defined for several methods of gas injection and liquid
evaporation.
Equation of State. The pressure-density-temperature relation is

p = pRT/W (8.1)
Taking logarithms, we get

log p = log p f log 6i + log T - log W
Then, taking the differential of each side, we have
Sound Velocity. The expression for the sound velocity in a semi-

perfect gas is
c2 = k@T/W (8.3)
or
222 GENERALIZED ONEDlMENSIONAL FLOW Ch. 8
Definition of Mach Number. From the definition of the Mach Num-

ber and Eq. 8.3, we find that
M 2 = V2/c2= V2W/k61T (8.5)
Equation of Continuity. Fig. 8.1 shows several methods by which the

injected gas and liquid may be brought into the main stream. Also
shown are the corresponding control surfaces employed for purposes of
analysis. Any liquid travelling through the duct, whether it be in the
form of droplets, filaments, or sheets, is considered to lie outside the
control volume. Accordingly, liquid which travels through the length
d x without evaporating neither enters nor leaves the control volume.
The infinitesimal flow dwL which evaporates in the length is considered
to be liquid which enters the control volume a t the liquid-gas interface
and which leaves as a gas with the main gas stream.
The injected gas flow dw, is assumed to be injected continuously
along the length of the duct, i.e., dw,/dx is assumed to be finite or zero.
This may be thought of as a simple model of real injection processes.
It is assumed further that the injected streams dwL and dw, are mixed
perfectly with the main stream a t the downstream boundary of the
control volume.
The mass flow of the main gas stream may be expressed as
In this expression dw denotes the total increase of mass flow in the

main stream and includes both injected gas and evaporated gas. The
equation of continuity is, therefore,
If vapor is condensed, d w is~ negative. It is important to note also that

d A refers to the change in cross-sectional area occupied by the main
gas stream and does not include the cross-sectional area of the injected
streams before mixing. Nor does it include that part of the cross section
of the pipe occupied by liquid, although usually the volume of liquid
may in fact be neglected in comparison with the volume of gas.
Energy Equation. Only the liquid crossing the control-surface bound-
ary and evaporating within the control surface is taken into account in
evaluating the flux of enthalpy. Changes in temperi~tureof the liquid
traveling along with the stream are taken to be the result of external
heat exchange to or from the main stream. As the flows d w and ~ dw,,
thoroughly mixed with the main stream, pass out of the control surface,
they are assumed to be a t the temperature of the latter.
The energy equation for the flow through the control surface of Fig.
8.1 may be written, assuming gravity effects to he negligible, as
w(d& - dWz) = [w(h + dh) + h , ~dw, + hv d w ~ l
In this expression dQ is the net heat added to the stream by conduction

or radiation from sources external to the main gaseous stream, per unit
mass of gas entering the control surface. Likewise, dWz is the net
external work delivered to outside bodies per unit mass of gas entering
the control boundary. The external work includes shaft work, elec-
trical work, and shear work on moving bodies adjacent to the control-
surface boundaries. The symbols hgT and h, denote, respectively, the
enthalpy of the injected gas dw, a t the temperature T, and the enthalpy
a t the temperature T, with which dw, enters the control volume. The
symbol hv denotes the enthalpy of the evaporated liquid d w a~t the tem-
perature T, and hL denotes the enthalpy of the liquid at out to evaporate
as it enters the control volume.
Eq. 8.9 may be rearranged in the form
The enthalpy change of the main gas stream, dh, is the sum of the
changes due t o chemical reaction and to temperature change. Thus
where dh,,, the enthalpy increase a t the temperature T and pressure p

for a change from products to reactants, is positive for exothermic reac-
tions. It is usually called the constant-pressure heat of reaction and, in
special cases, the constant-pressure heat of combustion. I n evaluating
dh,,, one considers of course only the chemical changes which actually
occur, but computes dh,, per unit mass of the main gas stream.
We may also write
where
- [h.
hg~ + T] = h . ~- hog = CpdT - To,), (8.12)
and hogand To, are, respectively, the stagnation enthalpy and stagnation
temperature of the injected gas stream.
Inserting Eqs. 8.11 and 8.12 into 8.10, and rearranging, the energy
equation is finally put into the convenient form
where dH is an energy term defined by

dH = dh,, - [CP,(T - Tog) + V2/2] -
dw,
W
Dividing Eq. 8.14 by c,T, we obtain
Momentum Equation. The net force acting on the material within

the control surface of Fig. 8.1 is equal to the increase of momentum flux
of the streams flowing through the control surface.
Consider forces in the direction of flow acting on the control surface
in Fig. 8.la. Assuming that the injected gas and liquid streams are
a t the control-surface pressure as they cross the boundary and that the
angle of divergence of the walls is small, the momentum equation may
be written *
I n the foregoing expression rw represents the frictional shearing stress

acting on the pipe wall area dA,; d X is the sum of (i) the drag of station-
ary bodies immersed in the stream within the control-surface boundaries,
(ii) the drag of liquid droplets and filaments traveling more slowly than
* If the injected gas stream is not a t the control-surface pressure as it crosses the
boundary, the term (p, - p)dA,' should be added to the left-hand side of Eq. 8.17,
where pt is the pressure of the injected stream a t the boundary and dA,' is the pro-
jected area of the part of the control surface occupied by the injected gas stream.
the main gas stream, and (iii) the component of body or gravity forces
acting on the material within the control surface in the direction opposite
to that of the velocity vector; V,' is the forward component of the
velocity V gwith which the injected gas dw, crosses the control surface,
and similarly for VL1.
The wall shearing stress is related to the coefficient of friction, f,
though the definition of the latter:
It is convenient to define the quantities

y, = V g l / V and y~ = VL'/V (8.19)
From the definition of the mean hydraulic diameter D, we obtain
Substituting Eqs. 8.18, 8.19, and 8.20 into Eq. 8.17, and noting that
PV - kP M', we obtain, following rearrangement,
dp k b ~ ' dv2 dw
-+-- + k ~ ~ -( y)-
1 =0
P 2 v2 2 W
(8.21)
where
dw
--
W
dw,
- -+-
W
d w ~
w
These equations apply also for the methods of injection shown in

Figs. 8.lb and 8.lc, provided that A is always taken as the cross-sec-
tional area of the main gas stream.
Definition of Impulse Function. The impulse function, which is de-
fined by
F = P A + ~ A V- - p A(l kbA2)+ (8.23)
is useful for evaluating the thrust of propulsion systems. The increase
in this function represents the total force exerted by the stream on the
internal walls of the duct and acting on the duct walls in the direction
opposite to the flow. In differential form,
226 GENERALIZED ONEDIMENSIONAL FLOW Ch. 8
Second Law of Thermodynamics. Application of the Second Law of

Thermodynamics is simplified through use of the entropy. For a semi-
perfect gas and no changes in chemical composition, the entropy change
of the main gas stream alone is
ds dT k - l d p
(8.25)
CP T k P
More important, however, is the total entropy change of the main
stream plus injected gas and evaporated liquid. It may be shown that
for unchanged chemical composition
where dS is the total entropy change per unit mass of main gas stream.
8.3. Working Equations and Tables of Influence Coefficients

Eight independent relations between the differential parameters have
been set forth, namely, Eqs. 8.2, 8.4, 8.6, 8.8, 8.16, 8.21, 8.24, and 8.25.
As there are fourteen differential variables, six may be chosen as inde-
pendent variables and eight as dependent variables. For the inde-
pendent variables we choose those most easily controlled in practice,
as indicated below.
Independent
The usual methods for solving a system of simultaneous, linear, alge-

braic equations may be employed for obtaining each dependent variable
in terms of the six independent, parameters. For example, by elimination
we obtain for dM2/M2the equation
Art. 8.3 TABLES OF INFLUENCE COEFFICIENTS 227
Influence Coefficients. We shall call the coefficients of the independ-

ent variables influence coeficients, since they indicate the influence of
each independent variable on each of the dependent parameters. Table
8.1 lists the influence coefficients found by solution of the simultaneous
equations. The examples presented later illustrate the usefulness of
these influence coefficients in treating a variety of practical problems.
Useful Integral Relations. In addition to the differential relations
summarized by Table 8.1, a number of useful integral relations are
obtainable from Eqs. 8.1, 8.5, 8.7, 8.16, and 8.23. They may be sum-
marized as follows:
These formulas are derived by writing the energy equation and con-
tinuity equation for two sections (1 and 2) of the flow, and by introducing
TABLE 8.1
I NFLUENCE COEFFICIENTS FOR VARIABLE SPECIFIC HEAT AND MOLECUWR WEIGHT
+
+
I + ~ M * ) ( I+ k7
-l
M (I M ) 2, M*)
1 - M= 1 - M'
WCP
(Note 2)
N OTES : (1) Each influence coefficient represents the partial derivative of the variable in the left-hand column with respect to the variable in the top
row: for example
(2) For unaltered chemical oom~ositiononly, and referring to entropy change of main stream. See Eq.8.26 for total entropy change.
Art. 8.4 FLOW WITH CONSTANT cp A N D W 229
the definitions of M and F together with the relations p = pRT and

c= m.
In deriving Eq. 8.28 it was assumed that for a moderate temperature
interval i t is valid to write
8.4. Flow with Constant Specific Heat and Molecular Weight

There are many instances where the changes in molecular weight
and specific heat are of secondary importance and may, therefore, be
neglected for the sake of simplicity.
The equations of Arts. 8.2 and 8.3 remain valid, of course, but are
simplified by noting that
dW = dc, = d k = 0
and
W1 = W p ; Cpl = cp2; k , = k2
Stagnation Temperature and Pressure. The stagnation temperature
and stagnation pressure are, as we have seen, useful parameters for
problems of one-dimensional flow.
Although these stagnation properties are always significant, they
have not been brought explicitly into the equations of Arts. 8.2 and 8.3
because there are no simple algebraic relations connecting the local
properties with the stagnation properties in the general case of variable
specific heat and molecular weight.
STAGNATION TEMPERATURE. With the assumptions of constant c,
and W, however, we may express the stagnation temperature as
(8.33)
or, in differential form, as
k-1
-M2
dTo - dT
To T
+ 2
k-1
dhP
M2
(8.34)
Eq. 8.14, when combined with Eq. 8.33, yields

dQ - d W , + dH = c, dT + d ( V 2 / 2 )= cp dTo (8.35:
230 GENERALIZED ONEDIMENSIONAL FLOW Cb.8
Eq. 8.35 shows that the change in To is a measure of the total energy
effects produced by heat transfer, shear and shaft work, chemical reac-
tion, change of phase, injection of gases, etc. We shall, therefore,
employ Eq. 8.34 rather than Eq. 8.16 as a representation of the energy
equation. The parameter (dQ - dW, + dH)/cpT is then replaced by
dTo/To in our system of equations. These parameters are related
through Eq. 8.36.
STAGNATION PRESSURE. The isentropic stagnation pressure is related
to the local pressure and Mach Number through the familiar formula
,, ,
or, in differential form,
The change of entropy, in terms of po and To, is then
Working Equations and Table of Influence Coefficients. As in Art.

8.3, the differential working equations are arranged in terms of inde-
pendent parameters, influence coefficients, and dependent parameters.
For example, Eq. 8.27 becomes
dM2
2 (1 + k--1
2 M a ) dA (1 + kM2) (1 + IC-I
2 M') dTo
The complete array of influence coefficients is given in Table 8.2.

TABLE 8.2
INFLUENCE COEFFICIENTS FOR CONSTANT SPECIFIC H EAT AN D MOLECULAR WEIGHT
I
-
NOTE: Each influence coefficient represents the partial derivative of the variable in the left-hand column with respect to thevariable in the top row; for
exam~le
USEFUL INTEGRAL RELATIONS. The several useful integral relations
between the properties a t two cross-sections may be summarized as
yM,j
k-1 k--l
I+- h2
Po2
- P2
=-
2
P OI Pl +
8.5. General Features of Flow Patterns

The influence coefficients are indicative of some inkeresting aspects of
compressible flow patterns.
Effects of Independent Parameters. The influence coefficients for
M, V, T, p, and p have, with the exception of those pertaining to the
independent parameter dk/k, the term (1 - M 2 ) in the denominator.
From this it is evident that the Mach Number, velocity, temperature,
pressure, and density, all change at an infinite rate in the neighborhood
of Mach Number unity. Moreover, wherever the signs of the influence
coefficients change a t M = 1, the effects of the independent variables
are of opposite sign for subsonic and supersonic flow. An exception
concerns the effect of heating on the temperature, for which it will be
noted that the influence coefficient changes sign a t M = I/& as well
as a t M = 1.
Art. 8.5 GENERAL FEATURES OF FLOW PATTERNS 233
The individual influence of each independent variable is readily found

from Table 8.1. From Eqs. 8.27 and 8.40 these influences are seen to
be independent of each other, i.e., the incremental effects of small changes
in the independent variables are linearly additive.
EFFECTS ON M. An increase in area tends to reduce M in subsonic flow
and to increase it in supersonic flow.
Heating or combustion tends to increase M a t subsonic speeds, and to
decrease M a t supersonic speeds.
The effect of friction, or drag of internal bodies, acts to increase M
a t subsonic speeds, and to decrease M at supersonic speeds. Note that
the friction term, jdx, and internal-drag term, dX, are always positive,
while the other independent variables may be either positive or negative.
An increase in mass flow (with y < 1 for the added flow) tends to
increase M a t subsonic speeds and to decrease M a t supersonic speeds.
When the molecular weight is increased, the Mach Number tends to
decrease for subsonic flow and to increase for supersonic flow.
An increase in the isentropic exponent k always tends to reduce M,
for either subsonic or supersonic speeds.
EFFECTS ON STAGNATION PRESSURE. Changes in area have no effect
on stagnation pressure, a t least when c, and W are constant.
An increase in stagnation temperature always tends toward a reduc-
tion in po. Thus the stagnation-pressure loss in combustion chambers
can never be entirely eliminated.
Friction and drag likewise always act to reduce po.
Gas injection tends to decrease po if y < 1, i.e., if the gas is injected
with a forward velocity less than that of the main stream; while the
stagnation pressure tends to be increased when gas is injected with a
forward velocity greater than that of the main stream (as in a jet pump).
The Second Law of Thermodynamics. Considering the simple case
where there is no combustion, no W,, and noinjection, we find from
Table 8.1 that
Application of the Second Law of Thermodynamics to the flow through

the control surface yields
from which it may be shown that both jdx and d X are always positive,
i.e., negative friction or negative drag would constitute violations of the
Second Law.
Considering next the case where there is no chemical change, no fric-
tion, no drag, and no W,, we find from Eq. 8.26 that
As in the preceding paragraph, the Second Law of Thermodynamics

requires that
Examination of this requirement for gas injection shows that i t is satis-

fied under all conditions.
If the foregoing equation for dS is to be valid for gas withdrawal, the
withdrawn gas must cross the boundary of the control surface with
exactly the same properties and direction of motion as the main stream.
In other words, the withdrawal must not include an unmixing process,
and, in addition, y must be unity. Obviously, gas may be withdrawn
a t other values of y, but in these instances the gas will cross the boundary
a t other than the control-surface pressure, and the equation for dS will
have to be modified to take into account the added term discussed in the
footnote to Eq. 8.17.
From this discussion i t is to be noted that for gas withdrawal the in-
fluence coefficients in Tables 8.1 and 8.2 are applicable only for the
condition y = 1.
Choking Effects. We have previously seen, for the simple flows of the
previous chapters, that under certain conditions it is impossible to
reduce the area, increase the amount of friction, or increase the amount
of heating without causing a complete readjustment of the steady-state
flow pattern. I n subsonic flow this readjustment consists of a reduction
in flow rate. In supersonic flow the readjustment usually involves, first,
the appearance of shocks and, ultimately, for large alterations in the
independent parameters, reductions in flow rate.
Similar effects occur for more complex types of flow patterns. To
illustrate the argument, let us first reconsider a simple case, namely, one
in which the area is the only independent parameter which changes.
From the first column and first line of Tables 8.1 and 8.2 we see that
(i) If the area increases, i.e., if dA > 0, we see that

Art. 8.5 GENERAL FEATURES OF FLOW PATTERNS
Therefore,
when M < 1, dM <0
and
when M > 1, dM >0
In other words, an increase in area causes the Mach Number t o decrease
in subsonic flow and to increase in supersonic flow, and there seem to
be no difficulties in allowing the area to increase indefinitely.
(ii) If the area decreases, i.e., if dA < 0, we see that
(1 - ~ 2 ) >0
Therefore,
when M < 1, dM >0
and
when M > 1, dM <0
In other words, a decrease in area causes the Mach Number to increase
in subsonic flow and to decrease in supersonic flow, and so the Mach
Number always approaches unity. If Mach Number unity is reached
a t some point in the duct, further reductions in area cannot be intro-
duced without altering the entire flow pattern. This alteration usually
is in the form of choking. I t is to be noted also that after reductions in
area have brought the Mach Number to unity, further changes in area
without alteration in flow rate can occur only if the area is increased.
Effects similar to these have been discussed previously for flow with
friction and for flow with increases in stagnation temperature.
RULES OF CHOKING. The foregoing arguments may be generalized as
follows (see also Art. 8-10>:
(i) If (1 - M') d~~ is always negative, as for an area increase, no
limiting effects are encountered in constructing flow patterns.
(ii) If (1 - kt2) dtvt2 is always positive, as for an area decrease, no
further changes in the stream parameters are possible after the Mach
Number has reached unity.
(iii) If, with (1 - M2) dM2 at first always positive, Mach Number
unity is reached, then further changes in stream properties can occur
only if (1 - d ~ becomes
' negative.
The three rules given above may be readily applied to complex flows
involving variations in two or more of the independent parameters.
For example, consider a process in which the area and stagnation tem-
perature change simultaneously, and in which there are also frictional
effects. Then, from the first line of T a b l ~
8.2,
236 GENERALIZED ONEDIMENSIONAL FLOW Ch.8
The term 4 j dx/D is always positive. Therefore, if the area continually
decreases and the stagnation temperature continually increases it ,is
clear that (1 - M2) dM2 will always be positive and that choking effects
must ensue. On the other hand, if the area were continually increased
and the stagnation temperature continually decreased, these effects
might outweigh frictional effects and act to maintain (1 - M ~ dM2 )
always negative, in which case there would be no difficulty in continuing
the flow pattern indefinitely.
CONTINUOUS TRANSITION THROUGH SPEED OF SOUND. In order to
obtain a continuous transition from subsonic to supersonic flow, or vice
versa, i t is evident that the rates of change of area, stagnation temper-
ature, and frictional length must be such as to produce a change of
sign in (1 - M2) dM2 a t Mach Number unity. That is, (I - M2) dM2
must be generally positive until Mach Number unity is reached, a t
which point (1 - /A2) dM2 must pass through zero and become negative,
after which the Mach Number will change in the direction away from
unity. This is a delicate process near M = 1, since dA/A, 4jdx/D
and dTo/To must be so controlled relative to each other as to make
(1 - M2) dM2 pass from positive values through zero to negative values
exactly at the moment when M = 1. When the flow is from subsonic
to supersonic speeds, this continuous transition is easy to carry out in
practice. The continuous transition from supersonic to subsonic speeds,
on the other hand, is hardly ever realized in practice, and is probably
unstagle under most conditions.
8.6. General Method of Solution

Integration of Differential Equations. A typical problem involving
simultaneous effects will usually involve known initial conditions
(MI, pl, TI, etc.) a t some point in the duct, together with certain pre-
scribed variations in area, stagnation temperature, and in the remaining
independent parameters of Tables 8.1 and 8.2. Integration of Eq.
8.27 then yields the value of the Mach Number a t each section of the
duct downstream of the initial section. Sometimes this integration may
be carried out analytically, but more often an approximate step-wise
integration is necessary.
In order to find the corresponding values of p, T, V, etc. at each section
of the duct, two procedures are open. First, the differential equations of
Table 8.1 for dp/p, dT/T, dV/V, etc. may be integrated analytically or
numerically. Or, secondly, since the value of M is known a t each section
from the previous integration, Eqs. 8.28 to 8.32 and 8.41 to 8.46 may be
used for finding the remaining stream properties.
Suppose that the values of MI, p,, TI, A,, etc. are known a t a section 1,
Art. 8.6 GENERAL METHOD O F SOLUTION 237
and that the values of Az, (Q - W , + AH)1-2,2 2 - XI, WZ, W2, k2,
etc. are known for a section 2, a small distance downstream of section 1.
Then, from Eq. 8.27, the value of M2 may be found by numerical or
graphical integration. Next, Eq. 8.28 is solved for T2, Eq. 8.29 for p2,
and Eqs. 8.30 and 8.31 for V2 and p2, respectively. The procedure is
then repeated for a step between sections 2 and 3, and so on.
Integration Near Mach Number Unity. Special precautions in numer-
ical integration must be taken near M = 1because many of the influence
coefficients there approach infinity. It is usually possible to avoid diffi-
culty by obtaining an approximate analytical formula for a step near
M = 1, making use of the approximation that 1 - M2 is negligible com-
pared with unity. For example, consider two of the terms of Eq. 8.40:
Define, for compactness, N = 1 - M2. Then, when M 1, N << 1,

and it may be shown by means of binomial expansions that
Retaining terms only up to the first order in N, and integrating, we finally

obtain To* - To
(1 - M2)2 , = (k + l)2
To
where To* is the stagnation temperature at Mach Number unity.
Tables of Influence Coefficients. For convenience in performing
numerical calculations the influence coefficients (multiplied by M2) are
tabulated in Table B.6 for k = 1.4, with M as the independent argument.
In order to explain the notation of these tables, we rewrite Eqs. 8.27
and 8.40 in the respective forms
dA
dM2 = FA - + FQ dQ - dW, + dH
A c P
where
The influence coefficients FA,FQ,etc., are seen to be greater than those

in Tables 8.1 and 8.2 by the factor M2.
Specific illustrations of both analytical and numerical procedures are
given in,the examples which follow.
8.7. Simple Types of Flow

A simple type of flow is defined here as one in which all but one of the
independent differentials in Table 8.2 are zero. Thus Simple Area-
Change refers to a process in which the area changes without frictional,
stagnation-temperature, or gas-injection effects. Similarly, Simple
To-Change refers to a process in which the stagnation temperature
changes (due to external heat exchange) without area change, friction,
or gas injection. I n each of the simple types of flow the specific heat
and molecular weight are constant.
Three of the simple types of flow have, of course, been discussed in
Chapters 4, 6, and 7.
The integration of Eq. 8.40 may be carried out analytically for each
of the simple types of flow. For example, consider Simple Area-Change.
Using the first influence coefficient in Table 8.2, we obtain
Integration yields
I C+l
A* MVL k+l J
0
I;:
240 GENERALIZED ONE-DIMENSION.4L FLOW Ch. 8
The integration is carried out between the sections where M = 1,

A = A*, and M = M, A = A . The asterisk then indicates the critical
value of A , or the value where the Mach Number is unity. Thus we
obtain the well-known formula giving the ratio of local area t o throat
area as a function of the local Mach Number for an isentropic process.
FIG.8.2. Comparison of process curves on temperature-entropy diagram for Simple

Area-Change, Simple Friction, and Simple To-Change. With d as the starting point,
the corresponding asterisk conditions (M = 1) are respectively a t points c, f, and h.
Referring again to Simple Area-Change, we obtain, from the first

column of Table 8.2,
Upon integration between M = I, V = V * and M = M, V = V, weget
where V * is the velocity where M = 1. This same equation might have

been obtained with the help of the integral relations of Eqs. 8.41, 8.43,
and 8.45.
Art. 8.8 COMBINED FRICTION AND AREA CHANGE 24 1
?
By continuing in this manner we obtain p / p * , T/T*, etc., for Simple

Area-Change. Similar formulas are obtainable by the same general
method for the other types of simple flow. These formulas are sum-
marized in Table 8.3. Note that the case of simple gas injection with
y = 1 is essentially identical with the case of isentropic area change;
that this must be so is evident when we remember that this type of gas
injection is thermodynamically reversible, involves no change in stagna-
tion temperature, and has as its
principal effect an alteration in the
mass flow per unit area.
Fig. 8.2 shows on a temperature-
entropydiagram the several courses
of states, beginning with state d,
corresponding to Simple Area-
Change, Simple To-Change, and
Simple Friction, respectively.
In interpreting the formulas of
Table 8.3, it must be remembered
that the conditions where M = 1
(asterisk conditions) are different
for the several types of simple flow.
For example, referring to Fig. 8.2,
points c, j, and h represent, re-
spectively, the asterisk condit,ions
for Simple Area-Change, Simple
Friction, and Simple To-Change.
The quantities
- marked with as-
terisks are best regarded as reference values which remain constant in
any one type of simple flow.
Charts and tables representing the formulas for Simple Area-Change,
Simple Friction, and Simple To-Change have been presented in previous
chapters. Fig. 8.3 shows curves for Simple Gas Injection with y = 0.
8.8 Example of Combined Friction and Area Change

As a simple instance where combined effects occur and a n analytical
solution is possible, let us examine the problem of designing a long insu-
lated duct for constant Mach Number along its length. Changes in
specific heat will be ignored.
From the first line of influence coefficients in Table 8.2, with dM,
dTo, d X , and dw set equal to zero, we get
242 GENERALIZED ONE-DIMENSIONAL LOW Ch. 8
Now, from the geometry of a duct of circular cross section
so that
where u is the half-angle of divergence of the duct. For a constant

friction coefficient, therefore, a must be constant, hence the duct must
be conical. As a specific example, let us take j = 0.0025, k = 1.4, and
M = 2; then a = 0.4 degree. Note that the duct must diverge for both
subsonic and supersonic flow. For typical values of f, the degree of
divergence in subsonic flow is very small.
From the second line of influence coefficients in Table 8.2 and Eq.
we get
-~ dV dA kM2 d x
Thus the velocity is constant. Similarly, the temperature and impulse

function are found to be constant.
For calculating the pressure change we find from Table 8.2 and Equa-
tion 8.50 that
dp - - - dA
- 7 whence plAl = p2A2
P A
In like manner, we obtain
kM2 M 2 j dD dD
-PO- - dx
- - 4 j - = - -- = -2-
Po 2 D tan u D D
or
Po2
- =A1
-
Pol A2
and, since the area must always increase, the stagnation pressure is
always diminished.
The foregoing results for V, T, F, p, and po might also have been found
from Eqs. 8.41 to 8.46.
8.9. Examples of Combined Friction and Heat Transfer

An important example of flow with combined effects is that of heating
of a gas flowing through a tube. The mechanisms of friction and heat
transfer are so similar that one cannot be had without the other. Hence,
to calculate the pressure drop and other properties of interest with
Art. 8.9 COMBINED FRICTION AND HEAT TRANSFER 243
reference to heat exchangers, it is necessary to take account of friction

as well as of heat transfer. We shall assume the area, specific heat, and
molecular weight all to be constant.
General Analysis. Considering an infinitesimal length of duct dx, the
rate of heat transfer may be expressed by means of the energy equation
in terms of the increase in stagnation enthalpy; or, by means of the
coefficient of heat transfer X, in terms of Tw - Taw. Thus,
When the flow is subsonic, the adiabatic wall temperature Tawdoes not
differ very much from the stagnation temperature To. Moreover, the
error in substituting To for Tawwill be small in any event if Twis con-
siderably in excess of Taw. We shall assume that this substitution
(which simplifies the subsequent calculations) is permissible; this is
equivalent to assuming that the recovery factor is unity. Then the
foregoing equation yields
dTn 4% dx
REYNOLDS ANALOGY. A further considerable simplification results

when the Reynolds Analogy between friction and heat transfer is as-
sumed to be valid. Experiments show that the analogy, which relates
the heat-transfer and friction coefficients according to
has an accuracy of a few per cent for fully developed turbulent gas flows.
Substituting this formula into the preceding equation, we get
With the help of Eq. 8.52, we may now write Eq. 8.48 as
The procedure beyond this point depends on the nature of the heat-
transfer process. To illustrate, we shall consider three specific problems:
(i) constant wall temperature, (ii) constant heat flux, and (iii) nuclear
reactors.
Case of Constant Wall Temperature. The wall temperature is ap-
proximately constant when the tube metal is a good c o : ~ J ~ ~ cand
t t , r is
surrounded by a constant-pressure bath of condensing vapor or boiling

liquid (e.g., condensing steam). Since T , is independent of x, Eq. 8.52
is easily integrated. Considering any pair of sections 1and 2, integration
yields
Tw
- - 1
Hence, for any initial value of Tol, and for known values off, D, and
T,, Eq. 8.54 yields Toz as a function of x2.
NUMERICAL INTEGRATION FOR MACH NUMBER VARIATION. TO com-
pute the variation of M with distance, we must integrate Eq. 8.53. In
the present example it is, unfortunately, not, possible to separate variables
and integrate in closed analytic form. Hence it is necessary to integrate
either numerically or graphically. We shall first illustrate numerical
integration.
Suppose that f = 0.005, M I = 0.5, and TWITol= 4. Let us specify
section 2 as that section downstream of section 1 where To2/TO1 = 1.05.
The location of section 2 is found by direct substitution into Eq. 8.54,
thus yielding
22 - 2 1 1 4- 1
-- - - -In = 1.68
D 0.01 4 - 1.05
In other words, section 2 is 1.68 duct diameters downstream of section 1.
Eq. 8.53 is now integrated approximately in finite-difference form
over the short interval between sections 1 and 2, the approximation
being that the coefficient of dTo is constant (at its mean value) during
the interval. Let P Tand~ Ff denote the values of the influence coeffi-
cients when evaluated a t &i = ( M I + M2)/2. Furthermore, let us set
+
To = (Tol T o 2 ) / 2for the interval. Then the approximate integration
Equation 8.55 must be solved by iteration, inasmuch as PToand FJ

are functions of M 2 , which is not known. We proceed as follows:
Trial I: We guess that M 2 = 0.60. From Table B.6, a t &i = 0.55, we
find that FTo= 0.655 and Ff = 0.195. Substituting into Eq. 8.55, we
get M 2 S 0.54.
Trial 11:We guess that M2 = 0.535. Then, a t &i = 0.518, we find that,
FTo = 0.5317 and Ff = 0.1453. From Eq. 8.55, then, M2 = 0.530.
Trial 111: After plotting for Trials I and I1 the computed values of M2
against the guessed values, we try Ma = 0.5293 and find that this value
makes the solution converge with adequate accuracy.
The corresponding values of all other properties a t section 2 may be
found either through numerical integration of the corresponding differ-
ential equations (as given in Table 8.2) or, more conveniently, through the
integral relations of Eqs. 8.41 to 8.46. The latter method is used here.
From Eq. 8.45,
b - 1
From Eqs. 8.42 and 8.45, furthermore,
Eqs. 8.43, 8.44, and 8.46 yield
Some of the numerical computations made immediately above may

be facilitated by noticing that several of the awkward functions of Mach
Number which occur are tabulated in one or another of the tables of
Appendix B. For example,
where (p/po);,,,, the pressure ratio for isentropic flow, is tabulated

versus M in Table B.2. The last equation of the present example might
therefore be written
~ 0 2 ~2 ! ( r ) / ~,&8i d
- = - -
Pr,, 111 I(I~//IO) ~s,.JMy
To illustrate still further the possibility of exploiting the tables of
functions in Appendix B, it is found by combining the foregoing expres-
, Tz/T~
sions for P O Z ~ P OPI ,~ / P Iand that
FIG.8.4. Illustrative results of calculations for combined friction and heat transfer
in a constant-area tube, with k = 1.4, (R = 1, and MI = 0.5. Solid curves are for
constant wall temperature and dashed curves are for constant heat flux per unit area.
The circled points of each curve indicate choking (M = 1).
(a) Stagnation temperature versus distance.
(b) Mach Number versus distance.
(c) Stagnation pressure versus distance.
(d) Temperature versus distance. Note that just prior to choking the tempera-
ture decreases as heat is added.
Art. 8.9 COMBINED FRICTION A N D HEAT TRANSFER 247
(e) Static pressure versus distance.

(f) Wall temperature versus distance.
(g) Duct length for choking versus initial wall temperature.
We now examine Table 8.3 to see whether any similar functions of M

and k appear and discover that
which again allows us to use the tabulated numerical values in Table B.2.
Returning now to our illustrative problem, the changes which occur
between section 2 and a section 3 further downstream are determined
in the same manner as the changes between sections 1 and 2. This
procedure is continued until the exit of the duct is reached, unless Mach
Number unity is reached first. The latter result indicates choking, and
then either the entrance Mach Number or the length of duct must be
reduced in order to obtain a steady-state solution.
The accuracy of the numerical integration naturally depends on the
size of interval chosen for each step and is improved by the selection of
small steps.
Fig. 8.4 shows the results of numerical calculations for the example
considered here. Also shown in this figure are corresponding results for
the case of constant heat flux per unit length of tube. It is seen that
many of the fluid properties change very rapidly as choking is a p
proached. Fig. 8.4g shows that high rates of heat transfer make the
choking problem very serious. For example, with TW1/Tol% 5 and
f 2 0.005, only about 10 diameters of duct length are allowable when
MI = 0.5. Apart from the large stagnation-pressure losses associated
with high Mach Numbers, therefore, considerations of choking often
prohibit the use of high values of M in heat exchangers.
GRAPHICAL INTEGRATION BY THE METHOD OF ISOCLINES. TO illustrate
this rapid method for obtaining solutions with the accuracy typical of
graphical constructions, we combine Eq. 8.53 with Eq. 8.52 and thus
obtain a differential equation of the form (assuming f is constant)
Now consider a chart with M~as ordinate and 4jx/D as abscissa (Fig.
8.5). Eq. 8.56 gives the slope on this chart of an integral curve satisfying
the differential equation. Consider-
ing the same illustrative problem
as before, we are given a constant
value of T,, and Tol is known.
The value of To corresponding to
any value of 4jx/D may, therefore,
be computed immediately from
Eq. 8.54. Hence T, and To are
known a t each value of 4fx/D.
If, for a given value of 4jx/D,
we choose various values of M~at
convenient intervals, the corre-
Oo
L sponding slopes may be found
from Eq. 8.56. These slopes are
4fx/D plotted in the chart as short
Fxa. 8.5. Illustrates graphical integra- streaks. By repeating this pro-
tion by method of isoclines.
cedure for a greatmany values of
4fx/D, the entire chart may be filled with such streaks indicative of
the local slopes.
The integration is carried out by starting with one of the streaks ai

4jxjD = 0 (corresponding to a certain MI2) and drawing with a french
curve a line which everywhere has the appropriate tangent. Such an
integral curve represents a solution for the chosen value of MI2, inas-
much as it satisfies the differential equation a t each point and also satis-
fies the initial condition.
It is evident that this method is especially adaptable to finding families
of solutions, for, once the slopes have been drawn in Fig. 8.5, i t is easy
to draw solution curves for many initial values of M2.
APPROXIMATE METHOD FOR LOW MACH NUMBERS. When the Mach
Numbers are low, which is the situation most often met with in practice,
an a -~-~ r o x i m amethod
te of calculation may be devised which has the
"
advantage of being in closed form and which, therefore, does not require
the tedious integrations described above.
Our interest lies primarily in the stagnation-pressure loss and the
increase in stagnation temperature occasioned by given values of T,,
pol, Tol, T,, D, and L. The changes in stagnation temperature are
easily calculable from Eq. 8.54 and need no further consideration.
For the variation in stagnation pressure we find from Table 8.2, for
the case under consideration, that
Eliminating dTo with the aid of Eq. 8.52, we get, in dimensionless form,
By eliminating p2/pl from Eqs. 8.42 and 8.46, and setting w2/w1 = 1
and A2/A1 = 1, we obtain k+l
k-1 k-1
(8.58)
Substitution of this relation for M2 into Eq. 8.57 then yields the differ-
ential equation
The left-hand side is immediately integrable, but the rightrhand side

is not because of the term in M? However, we find by a binomial ex-
pansion that
This suggests that, if M is always fairly small, this function of M is nearly

constant. For example, if M never exceeds 0.3, then this function of M
cannot be less than 1.0 or more than 1.113. Hence, if the function is
assumed to have a constant mean value during the integration, the
consequent error in Ape can a t most be a few per cent.
We therefore integrate as indicated above between the following
limits: (i) po/pol = 1 when x/L = 0, and (ii) po/pol = poz/pol when
x / L = 1, the subscript 2 now denoting the exit of the duct. Thus we
obtain
where
and
The quantity (13 depends on MI and M2, and may be set equal to unity
as a first approximation. The definite integral D depends on the varia-
tion of T , and of To with x; this integral, accordingly, depends in part on
the method of heating.
As shown previously, and in accord with Eq. 8.58,
I n order to evaluate we take the mean of the two values of the

quantity on the right-hand side of Eq. 8.59b; one of these values is found
by setting M = MI, and the second is found by setting M = M2. When
we set M = MI, the right-hand side of Eq. 8.59b is unity; when we set
M = Mz, we find with the help of Eq. 8.58 that
k+l
-
k-1 k-l
Hence it follows that
Assuming that T,, Tol, L, D, and pol are known t o begin with, and
that we wish to calculate p02 and M2 corresponding to a certain inlet
Mach Number MI, we proceed as follows: The value of a)is found from
Eq. 8.62, which is derived below. A first approximation for po2is found
from Eq. 8.59a by setting @ = 1 (or perhaps a slightly higher value,
depending on one's experience with prior calculations). The corre-
sponding value of M2 may then be found from Eq. 8.60, inasmuch as T02
is easily calculated from Eq. 8.54.
If it is desired to improve on this approximation (often i t is not neces-
sary), we employ the first approximations to po2 and M2 for calculating
a more accurate estimate of 03 from Eq. 8.61. The calculation procedure
is then repeated with this new value of 03. The recalculation of and
repetition of the calculation procedure are continued until convergence
is obtained; in practice convergence is usually adequate on the second
approximation.
In order to evaluate a), we first write Eq. 8.54 aa
which we may solve for To in the form

To = T, - (T, - Tol)e2f; (;)
Substituting this into Eq. 8.59c, 9 may be worked out in terms of ele-
mentary integrals. The final result is
Case of Constant Heat Flux. We now take up briefly the case where
the heat flux per unit area is the same for all values of x. Such a situation
results, for example, when the tube is heated electrically by passing
current either through the wall of the tube itself or through resistance
wires wrapped uniformly around the tube.
Assuming that X is constant, it follows from the first equation of this

article that T, - To is independent of x. Hence we write
T, - To = Twl- Tol (8.63)
Substituting this into Eq. 8.52, we have
and integration then yields
which allows us to compute To a t any value of x.

The variation of M with x may, as in the case of constant T,, be
found either by (i) an iterative integration using finite-difference meth-
ods, or (ii) a graphical integration using the method of isoclines. For
the former, the finite-difference equation is written as
The corresponding values of p, T, etc., are found from the integral

relations, Eqs. 8.41 to 8.46.
The approximate analytical method for low Mach Numbers, as em-
bodied in Eqs. 8.59 to 8.61, is also applicable here. It should be noted,
however, that a) is no longer given by Eq. 8.62 but rather has t o be
recalculated for the specific temperature distributions of the present
example.
Case of Heat Transfer in a Nuclear Reactor. One possible arrange-
ment of a gas-cooled nuclear reactor is to have the gas flow through tubes
in parallel, with the radioactive material incorporated in the tube walls.
Apart from the problems of choking and of maintaining reasonable
pressure drops, i t is desirable to have large rates of heat flux and yet not
exceed wall temperatures injurious to the material. Hence we are
interested in T,,-,, the maximum wall temperature.
One interesting feature of a nuclear reactor is that, as a result of
neutron leakage, the heat flux is usually a maximum a t the center of the
reactor and tends to become small a t the outer periphery. We shall not
concern ourselves with the actual heat-flux distribution, but shall,
rather, use for illustrative purposes a simple distribution which embodies
the heat-flux characteristics described above.
A simple heat-flux distribution which lends itself to calculation is
parabolic, with a maximum value a t the center of the tube and a zero
value at the ends. Assuming that the tube has a length L, and employ-
ing the dimensionless length variable t = x / L , the equation representing
such a parabolic distribution of heat input F
dt max
where the term c, dTo is found from the energy equation.

Separating variables and integrating, w? find
where To, is the stagnation temperature a t the tube exit, corresponding

to ( = 1. Hence
To - To1
= 3t2 - 2E3 (8.67~)
To2 - To1
Furthermore, from Eqs. 8.66 and 8.67b,
Eq. 8.52 yields, moreover,
We now form a dimensionless expression for the wall temperature

by means of the identity
The three ratios on the right-hand side are now expressed in terms of E
by means of Eqs. 8.67c, 8.67d, and 8.68, thus yielding
In order to find T,,,,, we set
Carrying out the indicated differentiation and setting the derivative

Fra. 8.6. Some general performance features of gas-cooled nuclear reactor tube
with parabolic heat-flux distribution.
(a) Heat-flux distribution.
(b) Stagnation-temperature distribution.
(c) Wall-temperature distribution.
(d) Location of maximum wall temperature.
(e) Maximum wall temperature.
254
Art. 8.10 SPECIAL CONDITIONS AT THE SONIC POINT 255
equal to zero, we find the following value of &, for which the wall tempera-
ture is a maximum:
The maximum wall temperature corresponding to selected values of

Tol, To2, and 4jL/D is then easily found by substitution in Eq. 8.69.
It is of interest to note that TWmax depends only on the three variables
mentioned and is independent of MI, of pressure level, of pressure drop,
of the gas constant, etc. Indeed, the ratio T,/Tol depends only on
To2/Toland on 4jL/D.
The temperature characteristics of a gas-cooled reactor tube with
parabolic heat-flux distribution are indicated in dimensionless form
in Fig. 8.6.
The calculation of the Mach Number distribution and of the dis-
tribution of fluid properties is carried out in substantially the same
way as for the case of constant T,, and the remarks made in the latter
case are applicable here.
Eqs. 8.59 to 8.61, comprising a rapid, approximate method for deter-
mining the loss in stagnation pressure, are also valid for the present
case. However, the definite integral a)of Eq. 8 . 5 9 ~
must be recalculated,
as Eq. 8.62 is no longer applicable.
8.10. Special Conditions at the Sonic P o i n t

The discussion on pp. 234-36 is here amplified and generalized to in-
dicate clearly the special circumstances of choking and of the continu-
ous transition through Mach Number unity.
Form of Differential Equation. To illustrate the thinking, let us con-
sider the case of constant specific heat and molecular weight. he;
Eq. 8.40 may be written as
dM2 G(x)
wilci~d(ln A ) / d x , d(ln To)/dx, and 4f/D are assumed to be functions

cd .I:. :I 11: I \\-I we dx is of course always positive.
,I : ,111 1 . S.7 1 n it is clear that whether the local Mach Number in-
; .. : . .(.,,:-rS:lsrs tlepends on whether the local Mach Number is
greater or less than unity and also on whether G is positive or negative,

rlccording to the following table:
VALUE OF d~'/dx
M < 1 M = l M > 1
G<O
G=0
-
0
m
indeterminate
+0
G>O + m -
Therefore the curve of M versus x may have many different forms, de-
pending on whether the initial Mach Number (MI) is less than or greater
than unity, and also depending on whether G is always positive, always
negative, or changes sign. We shall now explore the various possibilities.
CASE I: G IS ALWAYS NEGATIVE. (Fig. 8.7a.) If the flow is initially
subsonic, the Mach Number continually decreases. If it is initially
supersonic, on the other hand, the Mach Number continually increases.
(a) G < 0
I
X
- x X
(d) G Changes from Negative to Positive (e) G Changes from Positive to Negative
FIG. 8.7. Illustrates possible variation s of Mach Number versus distance.
In brief, the Mach Number always proceeds away from unity. There
are no limits on the flow.
Examples: (i) Isentropic flow in a diverging passage; (ii) Flow with
cooling only; (iii) Adiabatic flow with friction in a diverging passage,
with d(1n A)/dx > 2 k ~ ~ f / D .
Art. 8.10 SPECIAL CONDITIONS A T THE SONIC POINT 257
CASE11: G IS ALWAYS ZERO. (Fig. 8.7b.) The Mach Number al-

ways remains constant, irrespective of whether the flow is subsonic or
supersonic. There are no limits on the flow.
Example: Adiabatic flow with friction in a diverging passage, with
d(ln A)/dx = 2kM2f/D.
CASE111:G IS ALWAYS POSITIVE. (Fig. 8.7c.) If the flov is initially
subsonic, the Mach Number tends continually to increase. If it is ini-
tially supersonic, on the other hand, the Mach Number tends continu-
ally to decrease. In brief, the Mach Number always proceeds toward
unity. Consequently, the flow reaches a limiting condition when
M = 1, for it may then proceed no further. This phenomenon, called
choking, is discussed more fully below.
Examples:
(i) Isentropic flow in a converging passage
(ii) Flow with heating only
(iii) Adiabatic flow with friction in a straight or in a converging
passage
(iv) Adiabatic flow with friction in a diverging passage with
d(ln A)/dx < 2kM2f/D
Choking. When G is always positive, i t is impossible for the flow to
pass continuously through M = 1. To prove this, assume that the flow
is initially subsonic, and that Mach Number unity is reached a t some
point where x = so. Suppose that the flow subsequently became super-
sonic. Then, at a value of x = xo + ax, where 6x is positive, dtvt2/dx
would, according to Eq. 8.71a, be negative. This, however, contradicts
+
the supposition, for if M~a t xo 8 s is greater than unity, dM2/dx must
be positive in the interval between xo and s o + 62. Hence it is impos-
sible for the flow to become supersonic if it is initially subsonic. A
similar argument may be used to demonstrate that the flow cannot
pass continuously from supersonic to subsonic speeds.
At the choking point, dM2/dx is infinite, and in fact most of the fluid
ropert ties change a t an infinite rate.
If choking occurs, and there is an increase in x while G remains posi-
tive, it will not be possible for a steady-state flow to exist. After a
transient period of readjustment, a new steady-state condition will be
established for which choking does not occur (this usually involves a
reduction in the initial Mach Number, the appearance of shock waves,
or both).
CASE IV: G CHANGES FROM NEGATIVE TO POSITIVE. (Fig. 8.7d.) If
the flow is initially subsonic, the Mach Number first decreases, passes
through a minimum where G = 0, and then increases until finally chok-
ing occurs. If the flow is initially supersonic, however, the Mach Num-
ber first increases, passes through a maximum where G = 0, and then

decreases until choking occurs.
Example: Isentropic flow through a diverging-converging passage.
CASE V: G CHANGES FROM POSITIVE TO NEGATIVE. (Fig. 8.7e.)
There are several interesting possibilities for either initially subsonic
or initially supersonic flow. We shall consider in detail only flows which
are initially subsonic, but we note that similar circumstances prevail
for initially supersonic flows.
(i) If MI in the initially subsonic flow is quite low, the Mach Number
a t first increases, passes through a maximum where G = 0, and then
continually decreases.
(ii) If MI is sufficiently high, the flow chokes before G = 0, and the
solution may not be continued.
(iii) If, for a given variation of G with x, a particular value of MI is
chosen, the flow will reach Mach Number unity at the exact point
where G also passes through zero. Under these conditions, as described
in detail below, the flow may pass continuously into the supersonic
range or may pass continuously back into the subsonic zone.
Continuous Passage Through Sonic Point. Examining the right-hand
side of Eq. 8.71a, we see (i) that in general there is a maximum or mini-
mum in the curve of M versus x a t the point where G = 0 ; (ii) that in
general choking occurs when M = I ; (iii) that dbA2/dx is of the indeter-
minate form 0/0 when G passes through zero simultaneously with M
becoming unity. The latter singular condition is the one we shall now
examine, for it permits a continuous passage from subsonic to supersonic
speeds (or vice versa).
RELATION BETWEEN INDEPENDENT VARIABLES AT CRITICAL POINT.
Using an asterisk to denote conditions at the critical point where G = 0
and M = 1 simultaneously, Eq. 8.?lb yields the following relation
among the independent variables which must be fulfilled at this point:
VALUE OF (dht2/dx)* A~ CRITICAL POINT. The limiting value of

(dM2/dx)* at the critical point is found by applying L'Hospital's rule
to the right-hand side of Eq. 8.71a. Thus,
where ,Gr(x) = dG/dx, from which we obtain

Art. 8.10 SPECIAL CONDITIONS AT THE SONIC POINT 259
as the value of the slope of MZ versus x at the critical point. Thus, at

the critical point, the flow may be continued through the speed of sound
provided that [G1(x)]*is negative and is not zero. Under these condi-
tions it is clear from Eq. 8.74 that the solution may be continued from
the critical point in either of two directions: into the supersonic range,
or back into the subsonic range.
The critical point is a singular point in the solution and is in fact a
saddle point, (see Fig. 8.7e). The critical curve passing through the
critical point is a singular curve of the family of solutions.
Since G(x) depends on hi2 as well as on the independent variables
d(ln A)/dx, d(ln To)/dx, etc., i t follows that G1(x) in Eq. 8.74 contains
terms in dM2/dx. However, if, as is usually the case, Eq. 8.71 is being
integrated numerically or graphically, the terms in dM2/dx implicitly
contained in G1(x) will be automatically accounted for provided that
one plots G versus x and then determines the value of the slope G1(x)
as M2 -+ 1 and G 4 0.
SPECIAL RELATION FOR (dM2/dx)* AT CRITICAL POINT. When the
only independent variables are d(ln A)/dx, d(ln To)/dx, and 4f/D, and
when these are functions of x only, then one may proceed as follows:
The value of [G'(x)]* is found from Eq. 8.71b by differentiating with
respect to x and then setting M2 = 1. Thus we obtain
(G1)*= q* + @*(dM2/dx)* (8.75a)
Using Eq. 8.72, Eq. 8.75b may also be written as
Inserting Eq. 8.75a into Eq. 8.74, and solving the resulting quadratic
for (dM2/dx)*, we finally obtain
In most practical situations, however, such independent variables as

d(ln To)/dx will depend on M2 as well as on x. I n such cases it is best
to proceed from Eq. 8.74.
260 GENERALIZED ONE- DIMENSIONAL FLOW Ch. b
Examples: (i) The most familiar example of continuous passage

through the speed of sound is a t the throat of a nozzle. Assuming fric-
tion and heat transfer t o be negligible, Eq. 8.72 shows that dA/dx must
be zero a t the sonic point, i.e., the area must have an extremum. I n
addition, Eqs. 8.76 and 8.75 show that d2(ln A)/dx2 must be positive
a t the sonic point, i.e., the area must be decreasing upstream of x* and
must be increasing downstream of x*, and the sonic point must there-
fore have a minimum area.
(ii) The author and his associates have succeeded experimentally in
producing the critical curve of Fig. 8.7e by injecting water a t low speed
into a high-speed subsonic hot gas stream. At first drag is controlling,
and G is positive. After the water droplets have been accelerated, evap-
orative cooling is controlling, and G is negative. In this way all the
solid curves of Fig. 8.7e have been realized, and an acceleration from
subsonic to supersonic speeds has been achieved in a constant-area pipe.
Stability of Passage Through Sonic Point. The foregoing considera-
tions refer t o steady flows, and provide no information as to whether
the steady-state solutions are in fact stable to small time-dependent
disturbances in the boundary conditions. This question has hardly be-
gun to be explored, but what has been done theoretically suggests that
the transition from subsonic to supersonic flow is stable whereas the
transition from supersonic flow to subsonic flow is unstable. Experi-
mental results thus far seem to confirm this tentative finding.

1. SHAPIRO, A. H., and HAWTHORNE, W. R. The Mechanics and Thermodynamics
of Steady, One-Dimensional Gas Flow, Jour. App. Mach., Vol. 14, No. 4
(1947), p. A317.
2. T URNER, L. R., ADDIE, A. N., and Z IMMERMAN, R. H. Charts for the Analysis of
One-Dimensional Steady Compressible Flow, NACA Tech. Note, No. 1419
(1948).
3. HICKS, B. L., MONTGOMERY, D. J., and WASSERMAN, R. H. The One-Dimen-
sional Theory of Steady Compressible Fluid Flow in Ducts with Friction and
Heat Addition, NACA Tech. Note, No. 1336 (1947).
PROBLEMS
8.1. Plot on a temperature-entropy diagram the curves corresponding to
Simple Friction, Simple To-Change, and Simple Gas Injection with y = 0.
Choose the starting point of each curve such that in all the processes the cor-
responding stream properties are identical at M = 2.
8.2. A stream flowing in an insulated duct with friction is to be maintained
at constant Mach Number through suitable changes in duct area.
Assume that a one-dimensional treatment is acceptable and that at section
1 of the duct the properties are MI,pl, Al, etc.
(a) Starting with fundamental principles, show that the product of area and
pressure is the same for all cross sections.
(b) Find an expression for the area a t a point downstream of section 1 in
terms of the area a t 1, the Mach Number, the friction coefficient, anti the
number of length-diameter ratios (based on Dl) between section 1 and the
downstream section.
(c) If MI = 0.5, pl = 1 atm, and f = 0.005, compute the ratios A2/Al, p2/pl,
and poz/pol if section 2 is 50 diameters downstream of section 1.
8.3. Air flows adiabatically in a tube of circular cross section with an initial
Mach Number of 0.5, temperature of 100O0R, and pressure of 100 psia, all a t
section 1.
The tube is to be changed in cross-sectional area so that, taking friction into
account, there is no change in the temperature of the stream.
Assuming the friction coefficient to be 0.005 and that the exit section (2) is
downstream of the inlet by a distance 100 times as large as the initial diameter,
find :
(a) The final Mach Number, Mz
(b) The ratio of diameters, Dz/Dl
(c) The find stagnation pressure, p02, in psia
8.4. A hydrocarbon gas is to be injected step-wise and burned in a high-
speed stream of air. It is desired to carry out the combustion a t a constant
Mach Number of 0.5, and changes in cross-sectional area will therefore be
necessary.
To simplify the analysis, a one-dimensional point of view will be adopted,
and the following assumptions will be made: (a) variations in specific heat and
molecular weight are negligible, (b) wall friction is negligible, (c) the hydrocarbon
is injected a t low speed, and (d) enthalpy effects associated with the difference
between the gas temperature and the stream temperature are negligible com-
pared with the enthalpy effects due to combustion.
Assuming that the constant-pressure heat of reaction is 15,000 Btu per pound
of hydrocarbon gas, Tol = 1000°R, and Toz = 250OoR, estimate the area ratio
A2/A1 and the pressure ratio pz/p~.
8.5. A gas flows in a conical duct of circular cross section, the included angle
between the walls being 8. If j is the coefficient of friction, find a relation be-
tween 0, DID*,M, and f, where D is the diameter a t Mach Number M and D* is
the diameter a t Mach Number unity. Find a similar relation between pO/po*,
M, 8, and f.
8.6. Consider the compressible flow of air in an electrically heated tube of

constant cross-sectional area. It is agrced to assume a recovery factor of unity
and to employ the Reynolds analogy between heat transfer and friction.
The electrical resistance of the tube wall is such that the rate of heat transfer
varies linearly from a maximum a t the inlet end to zero a t the outlet end, i.e.,
262 GENERALIZED ONELDIMENSIONAL FLOW Ch. 8
(a) Plot 4fL/D versus Tw,, - To1

-.
To2 - To1
(b) Plot M and po versus x/L, if MI = 0.25, To, = 500°R, TOP= 1500°R,
and TwmSx= 2000°R.
8.7. Consider the compressible flow of air in a heated tube with friction.
A constant Mach Number of 0.5 is to be maintained through suitable changes
in area. It is agreed to assume a recovery factor of unity, a constant wall
temperature T,, and to employ the Reynolds analogy between friction and
heat transfer. The stream properties a t the inlet section will be denoted by
flow is turbulent, f varies as D to the -0.2 power, i.e., f -

Al, Tol, TI, etc., and a t a distance x from the inlet by A , To, T, etc. Since the
D-0.2.
Plot A/A1, TITI, To/Tol, p/pl, and po/pol, all versus 4flx/D1, for a value of
T,/Tol of 2.0. Carry the computations up to a value of 1 for 4flx/D1.
8.8. Investigate the possibility of
raising the stagnation pressure of a
high-speed gas stream by cooling the
stream as it flows through a duct.
I P, Show with the help of Reynolds'
+"a analogy that there is always a net
T3
loss in stagnation pressure if the
cooling is accomplished by conduc-
tion to the walls.
Ill
P ROB. 8.9. 8.9. Consider the constant-area
mixing between two streams, as
shown in the sketch. It is agreed to assume
(i) That just before mixing the individual streams have equal cross-
sectional areas
(ii) That wall friction in the mixing section is negligible and that the
walls of the mixing section are insulated against heat transfer
(iii) That the streams are completely mixed a t section 3
(iv) That air behaves as a perfect gas
Calculate
(a) The isentropic stagnation pressure (psia) for stream 1
(b) The mass flow per unit area (Ibm/sec ft2) for stream 1
(c) The stagnation temperature a t section 3 (degrees R)
(d) The isentropic stagnation pressure a t section 3 (psia.)
8.10. The throat of a circular passage has a wall of the shape r/ro =
+
1 f f ( i 7 / ~ ~where
) ~ , ro is the radius a t the minimum section, r is the radius a t
the longitudinal distance x, x is the distance from the minimum section, and
a is a dimensionless
*
constant. Assuming adiabatic flow with a constant fric-
tion factor f, determine
(a) The location x*/ro of the critical point
(b) The value of dM/dx a t the geometric throat and also a t the critical
point
P A R T I11
INTRODUCTION TO FLOW IN TWO
AND THREE DIMENSIONS
Chapter 9
THE EQUATIONS OF MOTION FOR STEADY,

IRROTATIONAL FLOW

The analyses previously given for one-dimensional flow are exactly
correct only for the flow through an infinitesimal stream tube. For all
real problems the assumption of one-dimensionality for the entire flow
is at best an approximation. In many instances, especially those relating
to the flow in ducts, the one-dimensional treatment is adequate. In
other cases, however, the one-dimensional methods are not only inade-
quate but often can provide no information whatsoever about important
aspects of the flow. As examples where the flow field must be thought-
of as two-dimensional or three-dimensional in order to obtain results of

value, we may mention the flow past the wings, fuselage, and control
surfaces of aircraft; the flow through the blade passages of turbines and
compressors; and the flow through ducts of rapidly varying cross-sec-
tional area or through ducts the axes of which undergo rapid changes in
direction.
'
To treat the most general case of three-dimensional motion-including
friction, heat transfer, shocks, and a fluid with a complex equation of
state-involves mathematical difficulties so -nreat that the task is well-
nigh hopeless using present-day methods of analysis. Hence, as in most
engineering problems, i t is necessary to conceive simple models of the
flow which lend themselves to analytical treatment, but which a t the
same time furnish information of value concerning the real, and more
complex, flow patterns.
Using Prandtl's concept of the boundary layer-a concept well verified
by experiment-it is possible to ignore friction and heat transfer for
the region of potential $ow outside of the boundary layer. According
to this concept shearing stresses and heat transfer may be ignored
compared to other effects, except in a thin film near solid boundaries.
In this thin film, usually called the boundary layer, the velocity and
temperature gradients are so large that the shearing stresses and heat
transfer cannot be neglected. Thus, the flow is assumed to be adiabatic
and frictionless outside a fictitious body whose outline is displaced
265
266 EQUATIONS OF STEADY, IRROTATIONAL FLOW Ch. 9
from the outline of the actual body by an amount equal to the displace-
ment thickness of the boundary layer a t each point. The flow within the
boundary layer is treated separately by methods specially adapted to
the characteristics of flow in a boundary layer. An important premise
in this line of attack is that to first-order effects the potential flow is
independent of the flow in the boundary layer.
Shock waves cannot occur in regions where the flow is entirely sub-
sonic. In transonic and supersonic flow, on the other hand, it is difficult
if not impossible to avoid shocks. Even with shocks present, however,
the flow patterns upstream and downstream of the shock can each be
treated as shock-free, and the shock is then thought of as a discontinuity
dividing the two continuous domains of flow. It will be shown in this
chapter that a flow which is initially uniform and parallel has the special
property of being irrotational outside the boundary layer as long as shocks
are absent, and that the assumption of irrotationality leads to important
simplifications in the analysis.
When shocks appear in an initially irrotational flow, however, the flow
downstream of the shock is in general rotational, except in the special
circumstance of a plane or conical shock. The amount of rotation intro-
duced by a curved shock is often so small, however, that there is little
error in treating the downstream flow as irrotational.
In connection with the boundary-layer concept, i t should be noted
that when shocks are present, the flow outside the boundary layer may
be substantially influenced by the boundary layer itself. Experiments
indicate a marked interaction between shock waves and boundary layers,
in which the extraordinary pressure gradient of the shock wave signifi-
cantly alters the flow in the boundary layer, and this alteration in turn
influences the external flow to such an extent that the shock pattern
becomes itself dependent on the effects it produces on the boundary
layer.
For most problems in compressible flow the assumption that the fluid
is a perfect gas permits great mathematical simplifications with scarcely
any sacrifice in accuracy. This assumption would have to be abandoned,
however, for themiddynamic states near the critical state or near the
condensation state.
In this chapter the differential equations of motion will be derived for
the irrotational, frictionless, adiabatic, shock-free motion of a perfect
gas. The fluid will be treated as a continuum and all stream properties
will be assumed to vary in a continuous manner in all directions. Grav-
ity effects will be ignored since they are seldom of importance in the flow
of gases. For convenience in later reference, several coordinate systenis
will be employed.
~ ~~
~
Art. 9.2 SIGNIFICANCE OF IRROTATIONAL MOTION
NOMENCLATURE
acceleration absolute temperature
area velocity component in xdirec-
area vector tion
speed of sound velocity component in ydirec-
force tion
acceleration of gravity velocity
enthalpy per unit mass velocity vector
unit vectors in the x-, y-, and velocity component in zdirec-
zdirections respectively tion; mass rate of flow
ratio of specific heats Cartesian coordinates
length
mass angle
curvilinear coordinate meas- circulation
ured normal to streamline angle
pressure density
rotation vector velocity potential
radius; cylindrical coordinate stream function
radius vector rotation; cylindrical coordi-
radius of curvature of stream- nate
line angular velocity
entropy per unit mass
time signifies stagnation state
9.2. The Physical Significance of Irrotational Motion

Before proceeding t o a development of t h e differential equations of
motion, we shall first discuss the meaning of rotation in a fluid and the
implications of irrotational motion as t o the nature of the flow pattern.
Circulation. The circulation is defined as the line integral of the veloc-
ity around any closed curve. Mathematically,
the line integral of the velocity is the sum of
the products of the element dl of the curve and
the corresponding component of velocity tan-
gent t o the curve. Referring t o the closed
curve C of Fig. 9.1, and using the symbol r
to denote circulation, we may write
(9.la) FIG. 9.1. Tllustrates the

concept of circulation.
In vector notation this may be written in terms of the scalar product ars
268 EQUATIONS O F STEADY, IRROTATIONAL FLOW Ch. 9
where V denotes the vector velocity - and r the vector radius from any-
fixed origin. If u, v, and w denote respectively the velocity components
in the x-, y-, and z-directions, the Cartesian
form of Eqs. 9.la and 9.lb becomes
dx
I
I
dx
I - '"
- CIRCULATION PER UNIT AREA. Let us now
consider the circulation d r , around a small
FIG. 9.2. Circulation square element in the x,y-plane, as shown in
f o r elementary
curve.
Fig. 9.2. Employing Eq. 9.lc with the velocity
components shown in Fig. 9.2, and proceeding
counter-clockwise from the origin, we obtain
dr,=udx+
or, simplifying,
The circulation per unit area in the x,y-plane is therefore given by
Similar expressions for the other directions may be found b y the usual
rotation of indices.
GREEN'STHEOREM. Considering now the circulation around an area
of finite size, we note from Fig. 9.3 that the line integral around the
bounding curve C is the algebraic sum of the line integrals around the
elementary square elements comprising the area
within C, since each interior line of each ele-
ment is traversed an equal number of times in
opposite directions, and hence only the exterior
lines of the elements (which comprise curve C)
contribute to the circulation. Therefore, em-
ploying Eq. 9.2, we find that the circulation
around any finite closed curve C in the x,y- FIG. 9.3. The circuia-
plane may in the limit be expressed in terms tion around any closed
of a surface integral, namely, curve is the sum of the cir-
culations around all the
closed curves bounding
rz =g:( - $)axdy (9.4)

the elementary areas com-
prising the region within
C.
Art. 9.2 SIGNIFICANCE OF IRROTATIONAL MOTION 269
Likewise, Eq. 9.lc gives
r, =
f(u~x+u~Y)
A comparison between Eqs. 9.4 and 9.5 shows that taken together they
represent Green's Theorem for the transformation from line to surface
integrals in two-dimensional space.
STOKES' THEOREM. The general form of Eq. 9.4, expressed in vector
notation, is,
This equation taken together with Eq. 9.lb is an expression of Stokes'

Theorem for converting from a line integral to a surface integral in three-
dimensional space.
Fluid Rotation at a Point. The angular velocity of a rigid body is defined
as the time rate of change of the angle between any fixed line on the
object and a reference line fixed in space. With deformable bodies such
as fluids or elastic solids the angular velocity thus defined is ambiguous
and must be modified. The difficulty here is that an initially straight
line of the body becomes curved owing t o deformations, and the angle
referred to above can have many values. This suggests that we consider
a point of the body and measure the angle with respect to the tangent
to a curve fixed in the body and passing through the point. However,
a new difficulty is thus raised, inasmuch as the angular velocities of
tangents to different curves passing through the same point will not be
identical. Obviously what is required is some generalization of the
concept of angular velocity which defines the average angular velocity
a t a point in a manner which is invariant with orientation.
Such a generalization is given by the following definition: The fluid
rotation at a point is the mean angular velocity of two infinitesimal and
mutually perpendicular fluid curves instantaneously passing through
the point. A Jluid curve is defined as a curve passing through a set of
fluid particles of fixed identity.
~ h rotation
k a t a point, i t should be noted, reduces to the familiar
angular velocity for rigid-body motion. Hence the rotation as defined
above is an extension of the concept of angular velocity to bodies which
are deformable.
ROTATION I N TWO DIMENSIONS. In order to find an analytical expres-
sion for the fluid rotation a t a point in a two-dimensional flow, let us
consider the infinitesimal and mutually perpendicular fluid lines OA
and OB of Fig. 9.4. The motion of each of these lines may be resolved
into a translation plus a rotation. Considering the fluid line OA, a
rotation will take place only if the y-component of velocity a t A is differ-
ent from that a t 0 . Since only the difference in velocity is important,
we imagine that we move with point 0 ; the upward velocity a t A relative
to that at point 0 is, then, (au/ax) dx. During an infinitesimal time
interval dt, OA rotates to the posi-
'1
tion OA', and the relative vertical
gdydt displacement AA' is
av
AA' = - dx dl
ax
Hence the angle AOA', taken posi-
tive when measured counter-clock-
=dl gdxdl
U
X
wise, is given by
dx
v
au
LAOA' = -dt
F IG . 9.4. Illustrates fluid rotation at
ax
a point. OA and OB are mutually per-
pendicular fluid lines. and the time rate of change of this
-
angle is avlax.
Similarly, it may be shown that the angular velocity of the fluid line
OB is - au/ay.
From the definition of rotation, therefore,
Referring to Eq. 9.3, we see that the circulation per unit area is twice
the average rotation of a fluid particle:
ROTATION IN THREE DIMENSIONS. The rotation of a fluid at a point,

like the angular velocity of a rigid body, is a vector quantity. Eq. 9.7
can be taken to represent the z-component of rotation, and the x- and
y-components may then be found from Eq. 9.7 by rotation of indices.
In terms of the unit vectors i, j, k, the vector rotation q is then given by
which is more compactly expressed as the curl of the velocity vector:
RATE OF SHEAR DEFORMATION. Before leaving Fig. 9.4, it is of inter-

est to note that the rate of shear deformation in a fluid is defined as the
Art. 9.2 SIGNIFICANCE OF IRROTATIONAL MOTION 271
rate of change of the angle between two mutually perpendicular fluid

lines. Referring to the angle between the lines OA and OB, we get
Rate of shear (av/ax) dt + (du/ay) dt = -

a v au
+- (9.9)
deformation dt ax 8y
The shear deformation is connected with the shearing stresses in the

fluid through the coefficient of viscosity.
I t may be of interest here to point out also the analogue between the
expressions given above and corresponding expressions encountered in
the mechanics of deformable solids. I n the case of a fluid particle we
deal with velocities, whereas for a deformable solid particle we deal with
displacements. The average rotation of a fluid
particle corresponds to the average rotation of v=nr
a solid particle, and the rate of shear deforma-
tion of a fluid particle corresponds to the shear-
ing strain of an elastic solid particle.
@a
EXAMPLE OF ROTATIONAL MOTION. A sim-
ple example of rotational motion in a fluid is
the solid-body type (or wheel type) of angular
motion about an axis. Referring to Fig. 9.5,
the streamlines are concentric circles, and, since FIG. 9.5. Forced vortex
the individual fluid particles maintain their
relative positions as in the case of a rigid solid, we may write for the
velocity along a streamline
V = fir
where fi is the angular velocity in the rigid-body sense. Then, evaluating
the circulation along any circular streamline of radius r, we get
and, referring to Eq. 9.8, we conclude that the angular velocity fi of the
entire mass is, as we expect, identical with the average rotation of each
particle.
EXAMPLE OF IRROTATIONAL MOTION. A simple but important exam-
ple of irrotational motion is the potential vortex, which describes approxi-
mately the motion in tornadoes and whirlpools, and which plays an
important role in the theory of flight and the theory of turbines and
compressors. In its simplest form a potential vortex is a two-dimensional
motion in which the streamlines form concentric circles, and the tan-
gential velocity along any streamline is inversely proportional to the
radius of the streamline. Thus,
where K is a constant. Consider the circulation around the element

ABCD of Fig. 9.6. Noting that the line integral is zero along the sides
AB and CD, we obtain
~ ~ A B C=D (V + dV)(r + dr) dB - Vr d0 = (V dr + r dV) dB
Then, using Eq. 9.10, we see that
Thus the circulation of any element not enclosing the origin is zero, and,
employing the theorem illustrated by Fig. 9.3, we conclude that the
circulation of any area not enclos-
ing the origin of the vortex is zero.
Considering next the circulation
around the streamline of radius r
in Fig. 9.6, we obtain
2r
r =lZ1VrdB = K l d0 = 2rK
so that Eq. 9.10 becomes

'I
Vr =- (9.11)
2lr
where r is the circulation for any
closed circuit enclosing the origin.
FIG.9.6. Free, or potential, vortex. Each particle of fluid in the
vortex undergoes a shearing defor-
mation but the average rotation of each particle is zero. We conclude
that a potent,ial vortex is in general irrotational, except for a singularity
a t the origin where the rotation is infinite.
To illustrate in another way the fact that the motion in a potential
vortex is irrotational even though the fluid particles follow circular
paths, consider a cork floating in a pan of water in which the water moves
as in a potential vortex. The cork will travel in circles, but if a straight
line is painted on the top of the cork it will be observed that this line
has a purely rectilinear motion, that is, the line always remains parallel
to its original direction and consequently has zero angular velocity
Art,. 9.3 EULER'S EQUATIONS OF MOTION 273
The line on the cork will in fact have the same type of rectilinear motion
as the magnetic needle of a compass when the compass travels through
a circular path in a horizontal plane.
9.3. Euler's Equations of Motion

In this article we shall consider the application of Newton's second
law of motion to a fluid flowing without friction and without body forces
(gravity, electromagnetic, etc.) in three-dimensional space. Since we
are interested in the details of the flow pattern, we shall write the equa-
tions in differential form. It will be assumed that all stream properties
vary continuously in all directions; this implies that shocks are excluded.
Euler's Equations in Cartesian Coordinates. We begin by considering
the forces acting on and the accelerations experienced by a particular
FIG.9.7. Pressure forces acting on isolated body.

(a) At time t.
+
(b) At time t dl.
particle of fluid of fixed mass and identity. This system is taken to be

an elementary cube of dimensions 62, 6y, and 62. The respective loca-
tions of the system a t the times t and t +dt are indicated in Figs. 9.7a
and 9.7b, respectively.
Newton's second law can be written for each of the three Cartesian
directions as
Du
Fz z =m- (9.12a)
Dt
Dw
F , = ma, = m-
Dt
where F,, F,, and F, denote the respective forces acting on the system
in the x-, y-, and z-directions, and u, v, and w are the corresponding
components of velocity. The symbol DjDt represents the substantial
derivative with respect to time and denotes that the differentiation is to
be carried out while following a particular fluid particle.
Since shearing forces and body forces are excluded from this analysis,
the only forces acting on the system are pressure-area forces. The
magnitudes of these forces acting on the various faces in the x,z- and
+
y,z-planes are shown in Figs. 9.7a and 9.7b a t the times t and t dt,
respectively.
Considering first the x-direction, the average force during the time
interval dt is
or, after simplifying,
For the y- and z-directions we obtain, similarly,
The mass of the system is the product of the mass density and the
volume:
m = p 6% 6y 62
The acceleration in the x-direction, a,, may be written
Art. 9.3 EULER'S EQUATIONS OF MOTION 275
Now, since dx is the x-displacement of the fluid particle during the time
dt, it follows that u = dxjdt. Likewise, v = dy/dt and w = dz/dt
Hence
For the y- and z-directions we obtain similar expressions:
Inserting the foregoing expressions for m, for F,, F,, and F,, and for
a,, a,, and a , into Eqs. 9.12, and simplifying, we finally obtain Euler's
equations of motion in Cartesian coordinates:
There are two special sets of conditions for which this system of equa-
tions may be easily integrated: first, along a streamline; and second,
throughout the entire flow field, provided that the flow is irrotational.
Integration of Euler Equation Along a Streamline in Steady Flow.
Since a streamline is defined by
Y
the condition that a t each instant
the velocity vector is tangent t o
the streamline (Fig. 9.8), we may
write the equation
- of the stream-
line in the form
dy -d =z- . w d- x =u-
_ = -v.
dx u' dy v ' dz w
FIG.9.8. Illustrates the equation of a

We now multiply Eq. 9.13a by dx, streamline.
Eq. 9.13b by dy, and Eq. 9 . 1 3 ~by
dz, and we introduce the assumption of steady flow by setting time de-
rivatives equal to zero. Furthermore, by introducing Eq. 9.14 we restrict
276 EQUATIONS OF STEADY. IRROTATIONAL FLOW Ch. 9
ourselves to considering only changes along a particular streamline:
Adding these equations, and noting that
au au au
du = - dx
ax
-t - d y
ay
+-
az
dz; etc.
we obtain
where V is the resultant velocity. Eq. 9.15 may be put in the form
or i t may be integrated to give
J v 2
+ = constant along streamline (9.16b)
The constant of integration is usually called the Bernoulli constant

of the streamline. It should be especially noted that Eq. 9.16 is, within
Art. 9.3 EULER'S EQUATIONS OF hIOTION 277
the assumptions given here, valid only from point to point of a given
streamline. Or, to put it differently, the Bernoulli constant may in
general vary from one streamline to another. Of course, if all the stream-
lines originated in a region of uniform flow, the same Bernoulli constant
would prevail throughout the entire field of flow.
Integration of Euler Equation for Steady, Irrotational Flow. We now
abandon the restriction that only points along the same streamline are
to be considered and introduce instead the assumption that the motion
is irrotational and steady. According to Eq. 9.7, then,
the last two equations being obtained from the first through rotation of
indices.
We now multiply Eq. 9.13a by dx, Eq. 9.13b by dy, and 9 . 1 3 ~by dz,
and introduce the assumption of irrotationality through Eq. 9.17, thus
getting
Adding these equat,ions, and noting that
that
a + v2 + w 2
4 )= t ( ~
etc. ) *
u2
ax 2
278 EQUATIONS OF STEADY, IRROTATIONAL FLOW . Ch.9
and that
which is identical with Eq. 9.15 as derived along a streamline in the

preceding section. Integration yields
dp V2
constant throughout flow field (9.18)
P
Thus, for steady, continuous, frictionless, and irrotational flow the

Bernoulli constant is the same for all streamlines, and Eq. 9.18 may be
used for relating the flow properties between any two points in the flow
field. The asumption of irrotational flow is therefore an important
simplification, as will be seen more fully in subsequent sections.
BERNOULLI'S EQUATION FOR INCOMPRESSIBLE FLOW. Finally, we
note that for frictionless, steady, incompressible flow the Euler equation
reduces to the familiar Bernoulli equation,
p + *pV2 = constant (9.19)
where the constant has a particular value for each streamline, except for
irrotational flow in which case the constant is the same for all streamlines.
9.4. Kelvin's Theorem

Several significant conclusions follow from the theorem first demon-
strated by Thomson (Lord Kelvin), which states in part that i n a
frictionless, homogeneous JEuid without body forces the circulation along a
closedJluid line remains constant with
respect to time. A homogeneous, or
barotropic, fluid is one whose den-
sity depends only on the pressure.
To demonstrate this theorem, we
first evaluate the substantial deriva-
tive of the line integral of the
velocity along a fluid line (Fig. 9.9),
namely
t
X
D
Dtc ( u d x + u d y + w & )
FIG.9.9. Illustrates Kelvin's theorem.
Art. 9.4 KELVIN'S THEOREM 279
Considering the first term, we may write
inasmuch as a consideration of the meaning of the integration and

differentiation operations indicates that the two operations may be
performed in any order. We may then differentiate by parts, thus
getting
Dt
Now, from Eq. 9.13a, we note that in the absence of friction,
and, since we are following a particular fluid particle,

Dx D
- - u ; and -(dx) = du
Dt Dt
Hence
Carrying out similar operations on the other two terms of the line
integral, we obtain
Now, if the density is a function only of the pressure, it is clear that the
line integral depends only on the fluid properties at points A and B, and
not on the path of integration. Points A and B coincide if the fluid
line is closed, and then the line integral is zero. The proof of the theorem
is completed by noting that the line integral of the velocity around a
closed curve is by definition the circulation.
280 EQUATIONS OF STEADY, IRROTATIONAL FLOW Ch.Y
When body forces, such as gravity, are present, Kelvin's Theorem re-
mains valid provided that the body force may be written as the gradient
of a potential. such body forces are usually said to be conservative.
IRROTATI~NAL MOTION. If the fluid starts from rest, or if in some
region the flow is uniform and parallel, the rotation in these regions is
zero. Kelvin's theorem then leads to the important conclusion that,
the entire flow is irrotational, subject to the assumptions of no friction,
a homogeneous fluid, the existence of potentials for the body forces, and
the absence of discontinuities.
Since in many practical problems the flow may as a first approxima-
tion be thought of as frictionless and as having started as a uniform,
parallel flow, Kelvin's theorem and the concept of irrotational motion
play an important role in the understanding of many flow phenomena.
They indicate, for example, that while in most flow patterns observed
in nature the fluid particles follow curved paths, the flow is nevertheless
likely to be irrotational except in those regions where viscous stresses are
an important factor.
9.5. The Connection Between the Rotation and the

Thermodynamic Properties of the Flow
It was shown in Art. 9.4 that a steady flow which starts from a region
of uniform, parallel motion remains irrotational if the fluid is frictionless
and homogeneous and if there are no discontinuities.
We shall now demonstrate the connection between the rotation,
which is a purely kinematic property of the motion, and some of the
thermodynamic properties of the fluid. To bring out the salient points,
we shall make the simplifying as-
sumptions that the flow is steady,
that the fluid is frictionless, and
that all fluid properties vary in a
continuous manner from point to
- point. Such would be the case, for
example, downstream of a curved
shock. We shall also assume the
flow to be two-dimensional.
Rotation in Streamline Coordi-
nates. It will be convenient to
FIG. 9.10. Pressure forces and veloc- work in a system of curvilinear co-
ities for an elementary fluid particle
in a streamline coordinate system. ordinates comprising the stream-
lines and the system of lines nor-
mal to the streamlines. Referring to Fig. 9.10, let us first calculate the
rotation of the element ABCD. Since the line integral of the velocity is
Art. 9.5 ROTATION AND THERMODYNAMIC PROPERTIES 281
zero along sides BC and DA, we get
dr = V R d0 - (R + dn) dB = -
Dynamic Equation Normal to Streamlines. Next we apply Newton's

second law of motion in the direction n normal to the streamlines. The
mass of the element is pR dB dn, and its acceleration toward the center
of curvature of the streamline is the familiar expression for centrifugal
acceleration, V 2 /R. In the absence of friction and body forces, the only
forces acting are those owing to pressure. From Fig. 9.10 we see that
the net force acting toward the center of curvature is given by

aP R dn dB
-
an
Finally, after equating the force to the product of mass and accelera-
tion, and simplifying, we get the steady-state form of Euler's equation
in a direction normal to the streamline direction:
Combining Eqs. 9.20 and 9.21, we find that
Crocco's Theorem. Now the stagnation enthalpy, ho, is related to

the velocity and enthalpy through the steady-flow energy relation:
or, differentiating in the n-direction,

282 EQUATIONS OF STEADY, IRROTATIONAL FLOW Ch.9
The thermodynamic relation for the entropy,
1
T ds = dh - - dp
P
may be expressed in derivative form as
ah as 1 ap
--
- T-+--
an an pan
Combination of the two thermodynamic equations then yields
Substituting this into the expression for the rotation, we obtain finally
a special form of Crocco's Theorem for two-dimensional flows:
The rotation is therefore seen to depend on the rates of change of

entropy and stagnation enthalpy normal to the streamlines.
If the fluid is originally in parallel motion with uniform properties,
and if the fluid along each streamline undergoes adiabatic, reversible
changes, it follows from Eq. 9.22 that the flow is everywhere irrotational.
But the stipulation of reversible flow is equivalent to the statement
that there is no friction, and hence the results obtained from this
partially thermodynamic treatment coincide with the results of Kelvin's
Theorem.
It is conceivable that all streamlines might have the same stagnation
enthalpy, but that the entropy would vary from streamline to stream-
line. Such would be the case in the region downstream of a curved shock,
because the entropy change depends on the angle of the shock. Under
these circumstances the flow downstream of a shock would be rotational
even though the upstream flow were irrotational.
In a boundary layer near a solid wall viscous stresses and heat con-
duction give rise to variations in entropy and stagnation enthalpy be-
tween neighboring streamlines. Hence, according to Eq. 9.22, the
boundary layer flow is rotational. From a purely kinematic point of
view, the rotational nature of the flow may be seen by examining the
angular velocities of the two arms of a fluid cross or by computing the
circulation around a closed line in the boundary layer.
Art. 9.6 THE EQUATION OF CONTINUITY 283
9.6. The Equation of Continuity

To express the principle of conservation of mass, we consider a control
surface of dimensions dx, dy, dz a t a fixed location in space, and equate
the net incoming mass rate of flow to the time rate of change of mass
within the control surface.
Referring to Fig. 9.11, which shows the mass rates of flow entering
and leaving the dy dz and the dx dz
{ypzld t
areas of the control surface, i t is pvdxdzt
seen that the net flow into the con- d p~d~dz)dy
trol surface in the x-direction only p:dydZ +

face
is theminus
flow through
the flow the
through
left-hand
the pudydz
right-hand face, or, con rol

surloce
a
-- pvdxdz
(pu dy dz) dx
ax *
=
a
-- (pu) dx dy dz
x
FIG.9.11. M a s flows through elemen-
ax tary control surface.
Likewise, the net flows into the control surface in the y- and z-directions
only are, respectively,
and
- -a (pw) dz dx dy
az
so that the total net flow into the control surface is
The time rate of change of mass within the control surface is expressed by
a ap
- (p dx dy dz) = - dx dy dz
at at
Hence the principle of conservation of mass takes the general form
This expression is usually called the equation of continuity in Cartesian

coordinates.
For steady jlow the equation of continuity becomes
When the jlow is both incompressible and steady, we have
9.7. The Laws of Thermodynamics

In the treatment of irrotational, frictionless motion, we are dealing
with a conservative system which can be almost completely described
by Newton's laws of motion and by the principle of conservation of
mass, the sole contribution of thermodynamics being information con-
cerning the relation between pressure and density for the isentropic
process under consideration. For a perfect gas, as an example, we have
the familiar isentropic equation,
p/p" constant (9.26)
For adiabatic or isentropic flows of a perfect gas an important rela-
tion previously derived (see Eq. 4.8) is
Although, in the previous derivation, use was made of the First Law of
Thermodynamics in the form of the steady-flow energy equation, Eq.
9.27 may be obtained for isentropic flows with no recourse to thermo-
dynamics except the p-p relation of Eq. 9.26. The reader may verify
that Eq. 9.27 may be derived from the Euler equation (Eq. 3.18), the
definition of the sound velocity (Eq. 3.3), and the isentropic pressure-
density relation for a perfect gas (Eq. 9.26). The steady-flow energy
equation in reality embodies Newton's second law in the form of the
kinetic-energy term, and it is only the latter term which entered into
the previous derivation of Eq. 9.27.
Thermodynamics plays a more important role, it will be seen later,
when shock waves, friction, heat transfer, or work effects are important
factors in the flow pattern.
9.8. Differential Equations in Terms of the Velocity Potential

The Velocity Potential. Restricting our remarks now to irrotational
motion, the definition of circulation shows that for irrotational motion
the line integral of the velocity around a closed curve is zero. ~ r o this
h
Art. 9.8 DIFFERENTIAL E Q U A T I O N S V E L O C I T Y POTENTlAL 285
it follows that the line integral of the velocity between any two points
A and B depends only on the location of the points and not on the path
of integration. For, referring to Fig. 9.12 (see also Fig. 9.1), we may
write for the closed curves AlB3A and A2B3A :
from which we see that FIG. 9.12. Line integral

of velocity between two
0
lBv cos a! dl = lBv
0
00s a! dl
points in a n irrotational
motion is independent of
path of integration.
Since paths 1 and 2 were chosen arbitrarily, it follows that the line
integral of the velocity is independent of the path.
This means that V cos a! dl is an exact differential, and may, therefore,
be considered as the differential of a point function whose value depends
only on x, y, and z. This function is called the velocity potential, p, and
is defined through the relation
dp = V cos a! dl
d~ =
- V cos a!
dl
Thus, the derivative of the velocity potential in a given direction repre-
sents the component of velocity in the same direction. In Cartesian
coordinates we have
u = ap/az = p, (9.29a)
where the symbol p, is shorthand for ap/ax, pzz is correspondingly

aZp/ax2, and cp,, is a2p/ax ay.
Using the Cartesian notation, it is easy t o see that the condition of
irrotationality is the necessary and sufficient condition for the existence
of a scalar potential function whose gradient is the velocity. For,
using Eqs. 9.29, and noting that the order of differentiation of a point
function with continuous derivatives is immaterial, we see that
which show that the conditions of irrotationality specified by Eq. 9.17

are automatically satisfied by the assumption that a velocity potential
exists.
Eqs. 9.29 may be summarized in vector notation by
V = grad p = Vp
That this definition of q automatically satisfies the condition of irro-

tationality is seen by writing
where the last term is set equal to zero because the cross product of two
parallel vectors is zero.
The concept of the velocity potential derives historically from the
concept in mechanics of the potential of a force. It may be recalled
that for conservative systems, the work, or the line integral of the force,
is zero around a closed curve, and hence a force potential exists whose
derivative in any direction is the force in that direction.
Because a velocity potential always exists for irrotational motion,
the terms potential motion and irrotational motion are often used inter-
changeably.
The Differential Equation of the Velocity Potential. We shall now

derive a single differential equation representing the steady, irrotational,
frictionless, isentropic flow of a perfect gas.
The condition of irrotationality is automatically satisfied through the
introduction of the velocity potential, (p, which is assumed to be a point
function with continuous derivatives, and which is defined through
Eq. 9.29.
The principle of conservation of mass is represented by the continuity
equation, Eq. 9.24, which, in terms of the velocity potential, takes the
form
or, expanding,
Art. 9.8 DIFFERENTIAL EQUATIONS-VELOCITY POTENTIAL 287
Newton's second law under the conditions assumed here may be

written in the form of the Euler equation:
The sound velocity is given by c2 = (ap/ap),, but, since we are in

fact dealing with isentropic flow, this may be written in terms of the
,.
actual pressure and density changes. Thus, c2 = dp/dp. Introducing
Eq. 9.31, we obtain
dp = -dp = - - dP
2
(a2 + + .:)
c cZ
from which, by differentiation, we get
Substituting these expressions into Eq. 9.30 and rearrang

finally obtain
The local sound velocity is, of course, not constant, but varies with p
according to Eq. 9.27:
Substitution of Eq. 9.33 into Eq. 9.32 would yield a single differential
equation for p in terms of x, y, and z, an equation which simultaneously
satisfies the law of conservation of mass, Newton's second principle of
motion, and the laws of thermodynamics.
DIRECT AND INVERSE METHODS. Having obtained Eqs. 9.32 and 9.33,
how do we employ them for the solution of problems? Two approaches
are possible:
(i) Most often, we are interested in the flow through a duct or around
a body of known shape, and with certain additional boundary conditions
as to velocity and pressure. We must, therefore, seek a function p(x, y, Z)
which satisfies Eq. 9.32 and which also satisfies all the boundary condi-
tions. This is often difficult, if not impossible, and experience and
intuition rather than a prescribed procedure are likely to lead to success-
ful results. Thus, although for practical reasons the direct approach
is preferable, it is often necessary to use the second approach.
(ii) I n this case we find, more or less a t random, functions P(X, y, Z )
which satisfy the differential equation, Eq. 9.32, and then proceed to
determine the corresponding streamline patterns. If one or more of the
streamlines coincides approximately with the shape of a solid body or
duct which is of interest, then the solution is useful; otherwise, it is not.
Evidently, this procedure is at best haphazard.
In either case, if we have a solution of Eq. 9.32 for p in terms of x, y,
and z , the entire flow pattern may then be determined. From Eq. 9.29
it is possible to calculate the three velocity components and the Mach
Number a t each point and thus to construct the streamline pattern.
Furthermore, from the equations for isentropic flow, Eqs. 4.14a, 4.14b,
and 4.14c, all the fluid properties may be found a t each point. The
integration of the pressure-area forces on the solid boundaries then
gives the resultant force acting on the duct or on the body immersed
in the stream.
MATHEMATICAL REMARKS. Mathematically speaking, Eq. 9.32 is of
the second order. A differential equation is said to be linear when the
dependent variable and its derivatives appear only in linear form. In
Eq. 9.32, however, we find quadratic terms of the form e2, and cubic
terms of the forms pZ2p,, and rp,p,cp,,, which means that Eq. 9.32 is
nonlinear. An important feature of linear equations is that solutions
may be superposed; that is, if +y(x, y, z) and p2(x, y, Z) are respective
solutions of a linear differential equation, then Alpl + A2pz is also a
solution, where A1 and Az are arbitrary constants. Many of the mathe-
matical difficulties encountered in compressible fluid mechanics are
traceable to the fact that the equations are nonlinear, for i t is not
possible to build up complex solutions by superposition of a few simple,
known solutions. Later, it will be seen that, a t the expense of other com-
plications, the hodograph method removes this difficulty, for the differen-
tial equations prove to be linear when the velocity components rather
than the physical coordinates are used as the independent variables.
Afi. 9.8 DIFFERENTIAL EQUATIONS-VELOCITY POTENTIAL 289
PLANE, TWO-DIMENSIONAL FLOW. For plane, two-dimensional flow,

Eqs. 9.32 and 9.33 are simplified to give
INCOMPRESSIBLE FLOW. When the flow is incompressible, the sound

velocity is infinite compared to the stream velocity, so that Eq. 9.32
becomes
For plane, two-dimensional flow Eq. 9.34 takes the form
These may be recognized as two forms of Laplace's equation, an equa-

tion which plays an important role in such other fields as heat conduction
through a solid and flow of electricity through a conductor. Many of
the solutions which have been found in these well-explored fields may
therefore be borrowed to give solutions in the field of incompressible
fluid mechanics; furthermore, analogue techniques may be employed so
that experimental results in one field may be directly applied to one of
the other fields.
In two-dimensional, incompressible flow, the mathematical techniques
are especially convenient, for it may be shown that Eq. 9.37 is satisfied
by both the real and imaginary parts of any analytic function of a com-
plex variable. Thus, the formal mathematical methods in the theory
of functions of a complex variable may be applied to plane, incompres-
sible flow.
Since Eqs. 9.36 and 9.37 are linear, solutions may be superposed.
Consequently, even the second method described previously has been
fruitful in incompressible flow, for after some experience it becomes
possible to select a few simple solutions which when superposed will
give a flow pattern approximating the pattern desired.
The Differential Equation in Cylindrical Coordinates. For certain
types of problems, such as flow over bodies of revolution, it is convenient'
to use cylindrical coordinates. We shall adopt the notation shown in
Fig. 9.13, where u, v: and zu are the velocity components in the r-, U-,
and z-directions, respectively.
290 EQUATIONS O F STEADY. IRROTATIONAL FLOW Ch. 9
The various equations in this

system of coordinates may be
found either by a direct mathe-
matical transformation of the pre-
vious equations using the relations
x = r cos w (9.38a)
y = r sin w (9.38b)
0 z = z (9.3%)
FIG.9.13. Nomenclature for cylindri-
cal coordinate qstem. or by carrying out the analysis
from the beginning in cylindrical
coordinates. By either method we obtain the following relations:
Irrotationulity.
Definition oj (o.
Equation oj continuity.
a a a
- bur)
ar
+-
am (PV)+ r - (pw)
a2
= 0
Sound velocity in isentropic process.
The combination of Eqs. 9.40 to 9.43 in the same manner as that used
previously for Cartesian coordinates yields finally
where I
Art. 9.9 DIFFERENTIAL EQUATIONS-STREAM FUNCTION 291
AXIALLY SYMMETRIC FLOW. When a body of revolution is placed in a

uniform flow a t zero angle of attack, the flow pattern will possess axial
symmetry. The flow pattern is then the same in all meridian planes
passing through the axis of symmetry, and the flow is two-dimensional.
Placing the z-axis along the direction of flow, and noting that for axial
symmetry all the derivatives with respect to w vanish, we find that
Eqs. 9.44and 9.45 reduce to
PLANE, TWO-DIMENSIONAL FLOW. When the flow pattern is identical

in all planes normal to the z-axis, we obtain in polar coordinates the
following equations corresponding to Eqs. 9.34 and 9.35:
9.9. Differential Equations in Terms of the Stream Function

The Stream Function. Just as the condition of irrotationality is the
necessary and sufficient condition for the existence of a velocity potential,
so the equation of continuity for steady, twodimensional $ow is the
necessary and sufficient condition that another point function, the
stream junction, exist. The flow need be two-dimensional in the sense
that only two space coordinates are necessary to describe the motion,
so that both plane flow and axially symmetric flow fall in this category.
Considering two-dimensional, plane, steady flow, and using Cartesian
coordinates, the continuity equation may be written
We now define the stream function through the relations

292 EQUATIONS O F STEADY, IRROTATIONAL FLOW Ch. 9
where po is the isentropic stagnation density (or any other convenient

+
constant) and is brought in so that will have the same dimensions as
the velocity potential.
Next, we form the expression
+
If is a point function, the order of differentiation is immaterial. By
comparison of the foregoing equation with Eq. 9.50 we see that the
+
statement that there exists a point function defined through Eq. 9.51
is equivalent to the statement of the continuity equation. Conversely,
the continuity equation is the necessary and sufficient condition that +
be a point function.
The magnitude of the velocity vector a t any point is given in terms of
the stream function by the relation
PHYSICAL INTERPRETATION OF STREAM FUNCTION. What is the phys-

+
ical significance, if any, of the stream function? Since is a point func-
tion we may write
Now, referring to Fig. 9.14, the mass rate of flow dw across any surface
AB (of unit width normal to the paper) connecting two neighboring
surfaces of constant +
is given by
dw = pu dy - pv dx
Y and hence we see that the change
+
in the magnitude of between two
+
lines of constant is a measure of
the mass rate of flow between the
+
two lines, according to the rela-
tions
d w = po d+ (9.54a)
FIG.9.14. Illustrates physical signif-
icance of stream function. w = PO(+B - # A ) (9.54b)
Since, for steady flow, the mass rate of flow between streamlines is
+
constant, it follows that lines of constant are streamlines, hence the
term "stream function."
The Differential Equation of the Stream Function. We shall now
derive a single differential equation representing the steady, two-dimen-
sional, irrotational, frictionless, isentropic flow of a perfect gas.
Art. 9.9 DIFFERENTIAL EQUATIONSSTREAM FUNCTION 293
The equation of continuity is automatically satisfied through the

+
introduction of the point function as defined by Eq. 9.51.
In terms of the stream function, the condition of irrotationality
(Eq. 9.174 takes the form
or, after differentiation and rearrangement, the equation of irrotation-

alitv becomes
From the Euler equation and the expression for the sound velocity
in isentropic flow, we obtain
The partial derivatives d p / a x . and a p / a y may now be formed from

Eq. 9.56. When these derivatives are substituted into Eq. 9.55, and
the resulting expression simplified, there is obtained
where, from Eqs. 9.27 and 9.52,
and, from Eqs. 4 . 1 4 ~and 9.52,

1
Eqs. 9.57, 9.58, and 9.59 on the one hand are completely equivalent
to Eqs. 9.34 and 9.35 on the other. However, the differential equations
in terms of q are simpler in form than those in terms of J., and hence it is
customary to work with Eq. 9.34.
INCOMPRESSIBLE FLOW. For incompressible flow, the sound velocity
is infinite relative to the velocity of the gas, and Eq. 9.57 reduces to
This is the Laplace equation in two dimensions, and because of its
similarity in form with Eq. 9.37, a solution for the potential lines in plane,
incompressible flow represents also a solution for the streamlines, and
vice versa.
Axially Symmetric Flow. Since in steady, axially symmetric flow
there are only two space coordinates, the statement of the continuity
equation is the necessary and sufficient condition for the existence of a
stream function. Bearing in mind that partial derivatives with respect
to w are zero, Eq. 9.41 takes the form
a a
-
ar
(pur) +r-
az
(pw) = 0
But, since r is independent of z, this may be written
I t may be verified by direct substitution that this relation is auto-

+
matically satisfied by a point function defined by
The equation of irrotationality is now
Using the Euler equation and the expression for the sound velocity
in isentropic flow, we get
Combination of Eqs. 9.63 and 9.64 in a manner similar to that de-

.scribed previously finally yields
Art. 9.10 VELOCITY POTENTIAL A N D STREAM FUNCTION 295
9.10. Relations Between the Velocity Potential and the

Stream Function
For the velocity potential to exist i t is only necessary that the flow be
irrotational; hence we may employ the velocity potential for unsteady
flow or for three-dimensional flow. For the stream function to exist i t
is only necessary that the flow be steady and describable by two space
coordinates; hence the stream function may be used when the flow is
rotational. However, it should be noted that the differential equations
for p and which have been developed in this chapter refer to flow pat-
terns which are both steady and irrotational. If the flow were unsteady,
Eq. 9.32 would be different in form. Likewise, if the flow were rotational,
Eq. 9.57 would be different in form.
Orthogonality of Streamlines and Equipotential Lines. When the
flow is steady, irrotational, and two-dimensional, the stream function
and velocity potential are related through Eqs. 9.29 and 9.51:
That the lines of constant J. are also streamlines may be shown by

noting that, since $ is a point function,
Setting d$ equal to zero and using Eqs. 9.68 and 9.69, we find the slopo
of a line of constant J, to be given by
But, by definition, the slope of a streamline is given by
and hence the $ lines and streamlines are identical.

Turning next to a consideration of the equipotential lines, we have,
similarly,
drp = (az d x +
rpy 41
Using Eqs. 9.68 and 9.69, we get for the slope of an equipotential line,
($)+ >, - 'Pz

- - - u-
'Par v
so that
from which we conclude that a line of constant p is normal to a line of

constant +.I n other words, the equipotential lines and streamlines form
an orthogmal network.
Spacing of Net Lies. To investigate the relative spacing of the net-
work, we note that the gradient of
the potential in any direction is
the velocity component in that
JI+ AJI direction. Referring to Fig. 9.15,
the gradient of p along the stream-
line gives the magnitude of the
++ J A + velocity vector,
V= A'P/Z
Similarly, the gradient of J, in a

FIG.9.15. Illustrates relation between given direction is related to the
streamlines and equipotential lines velocity component normal to that
direction. Taking the gradient of
+ along the equipotential line, we have, for the resultant velocity
We may arbitrarily select equal numerical magnitudes for the values

of A+ and Acp between neighboring lines in the grid. In that case we find
that
Since p/po cannot be greater than unity, i t follows that the potential
lines are more closely spaced than the streamlines. And, since p/po
is near unity a t low Mach Numbers and far less than unity a t high Mach
Numbers, the difference between the spacings will be most marked in
Art. 9.10 VELOCITY POTENTIAL AND STREAM FUNCTION 297
regions of high Mach Number. These results may be verified by in-

spection of the potential and streamline nets of Fig. 9.16 for the two-
dimensional flow past a circular cylinder in the x direction.
INCOMPRESSIBLE FLOW. For incompressible flow each small element
of the net is, in the limit, a square. Steady, two-dimensional, incom-
pressible, irrotational motion might even be said to be a problem in pure
geometry, for any orthogonal network of lines, each small element of
FIG.9.16. Streamlines and equipotentiar lines for two-dimensional, irrotational,

isentropic, frictionless flow past circular cylinder (after Busemann).
(a) Incompressible (M = 0).
(b) Free-stream Mach Number of 0.4, giving nearly sonic velocity a t shoulders
of cylinder.
which is a square, represents a possible solution to the equations of

motion. In this respect we note that the conditions of irrotationality
and steady flow are purely kinematic for an incompressible fluid, and
only these conditions enter into the derivation of Laplace's equation
for the velocity potential or stream function. The dynamic conditions,
represented by Bernoulli's equation, are not a t all involved in the deter-
mination of the streamline pattern. For these reasons, incompressible
flow patterns may be a t least rudely constructed by drawing square nets
by trial with the help of paper, pencil, and eraser. Or a more organized
procedure, such as the relaxation method, may be used. In any case,
for incompressible flow the 9 lines and J, lines are interchangeable, so
that any single flow pattern embodies two possible solutions.
COMPRESSIBLE FLOW. The situation is quite different for compressible
flow. Although the condition of irrotationality remains kinematic, the
' equation of continuity contains the density, which is in turn related to
the pressure and, therefore, to the dynamic requirements. In com-
pressible flow the streamline and potential network depends on both
+
kinematic and dynamic considerations, and the q and lines when inter-
changed do not represent a solution.

1. PRANDTL, L., and TIETJENS, 0.G. Fundamentals of Hydro- and Aeromechanies.
2. PRANDTL, L., in The Physics of Solids and Fluids, by EWALD, POSCHL, and PRANDTL.
2d Ed. London: Blackie & Sons, 1936.
3. LAMB,H. Hydrodynamics. 6th Ed. New York: Dover Publications, 1945.
4. SAUER,R. Theoretical Gas Dynamics. Translated by F. K. HILL and R. A. ALPHER.
Ann Arbor: J. W. Edwards, 1947.
5. MILNE-THOMPSON, L. M. Theoretical Hydromechanics. 2d Ed. London: Mac-
millan & Co., Ltd. 1949.
PROBLEMS
9.1. Show that Euler's equation for unsteady, irrotational motion can, with
the introduction of the velocity potential q,be integrated to give
where the "constant" of integration f ( t ) is a function of time only.
9.2. I n Art. 9.3 Euler's equations in Cartesian coordinates (Eq. 9.13) were
obtained by applying Newton's second law of motion to a fluid particle of
fixed mass and identity, and with sides 82, 6 y , 82.
Show that the same results may be derived by applying the momentum theo-
rem to the flow through a cubical control surface fixed in space and with sides
ax,d y , dz.
9.3. I n Art. 9.6 the equation of continuity (Eq. 9.23) in Cartesian coordinates
was derived by applying the principle of conservation of mass to conditions
within and a t the boundaries of a control surface fixed in space and with sides
ax, d y , dz.
Show that the same result may be obtained by applying the principle of con-
servation of mass to the motion of a particle of fixed mass and identity and with
sides 62, 6y, 62.
9.4. Starting with first principles, verify Eqs. 9.39 through 9.45, expressing
the steady, irrotational, frictionless, isentropic motion of a perfect gas in terms
of the velocity potential and in terms of cylindrical coordinates.
9.5. Verify the derivation of Eq. 9.57.
9.6.Derive a single differential equation for steady, plane, two-dimensional,

irrotational, frictionless, isentropic motion, expressing the stream function in
terms of the polar coordinates r and w.
9.7. Show that the stream function for axially symmetric flow, as defined
by Eq. 9.62, defines surfaces which are generated by streamlines. Derive the
+
relation between the magnitudes of between two streamline surfaces and the
corresponding mass rate of flow between these surfaces.
Ch.9 PROBLEMS 299
9.8. Consider axially symmetric flow, with the streamlines lying in planes
containing the axis of symmetry. Use the following notation for a spherical
aystem of coordinates:
R radius vector, measured from the origin 0
cr angle of the radius vector with respect to the axis of symmetry
u velocity component along radius vector
u velocity component perpendicular to radius vector
(a) Show that the condition of irrotationality may be written
/b) Show that the equation of continuity may be written
9.9. Show that the stream function
represents in polar coordinates a plane source motion for which w is the maas
flow issuing from the line source per unit length of source.
Assuming the flow is frictionless and adiabatic, calculate the corresponding
expression for the velocity potential for
(a) Incompressible flow
(b) Compressible flow with a stagnation sound velocity co
9.10. Show that the potential function
satisfies Eq. 9.48 and that it represents a line vortex of circulation I

'.
Assuming the motion to be steady and frictionless, and that all streamlines
have the same entropy and stagnation temperature, determine the corresponding
+
expression for the stream function for
(a) Incompressible flow
(b) Compressible flow with a stagnation sound velocity co
PART IV
SUBSONIC FLOW
Chapter 10
TWO-DIMENSIONAL, SUBSONIC FLOW WITH SMALL

PERTURBATIONS

Although the differential equations in terms of the velocity potential
for steady, irrotational, isentropic motion were derived rather easily in
the preceding chapter, exact solutions of these equations for particular
flow problems often involve such complex or tedious mathematical
procedures as to be impractical. Two courses then seem to be open-
either (i) to find exact solutions for a relatively few problems in the hope
of obtaining a qualitative understanding of the nature of other flow
patterns for which solutions are not obtainable; or (ii) to find simple,
though approximate, solutions suitable for practical computations. Both
methods of approach yield useful information and, in a sense, comple-
ment each other, as the few exact solutions serve as a check on the ac-
curacy and reliability of the approximate methods. In this chapter we
shall show how the second method may be applied to some important
problems of two-dimensional, subsonic flow.
The assumption of two-dimensional flow serves as a first approxima-
tion to the flow past the wings of airplanes and to the flow through the
blade systems of propellers and of axial-flow compressors and turbines.
I n many such applications the velocity perturbations produced by the
body or bodies immersed in the flowing stream are small, because the
bodies are very thin. Here we have the essence of the linearized method
-that the flow pattern may be thought of as the combination of a uni-
form, parallel velocity on which is superposed small perturbation veloc-
ities.
The advantage of making such an assumption, we shall see, lies in the
fact that the equation of motion is greatly simplified and also becomes
linear. From this method, which is called the linearized theory or the
method of small perturbations, we shall draw useful approximate informa-
tion as to the effects of Mach Number for subsonic flow. The linearized
theory also makes evident an approximate similarity law for subsonic
flow; that is, a rule connecting the pressure field of a given compressible
flow with that of a related incompressible flow. Analogous similarity
laws for transonic, supersonic, and hypersonic flow are derived in later
chapters.
303
TWO-DIMENSIONAL, SUBSONIC FLOW Ch.10
NOMENCLATURE
angle of attack V velocity
area x, y Cartesian coordinates
sonic speed
specific heat a t constant pressure dl - h L 2
coefficient of lift camber ratio
moment coefficient thickness ratio
pressure coefficient interference pressure coefficient
amplitude of wave-shaped wall affine transformation factor
ratio of specific heats mass density
chord length of air foil; also perturbation velocity potential
wave length of wave-shaped velocity potential
wall
a characteristic dimension; also ( )o signifies stagnation state
lift force ( ) signifies free-stream condition
Mach Number ( ) signifies state at solid boundaries
pressure ( ) signifies incompressible flow
gas constant (M, = 0)
absolute temperature
perturbation velocity compo- [0] signifies "of the order of mag-
nent in xdirection nitude of"
perturbation velocity compo-
nent in 21-direction
10.2. Linearization of the Potential Equation

Equation of Motion. For steady, two-dimensional, irrotational, isen-
tropic motion, the differential equation for t h e velocity potential is
given by Eqs. 9.34 and 9.35:
where the symbol represents the complete velocity potential a s defined

in Chapter 9, and is t o be distinguished from the perturbation velocity
potential denoted b y rp in this chapter.
T h e Perturbation Components. We now assume the velocity at any

point t o be given b y the vector sum of a constant velocity V , along the
x-axis, together with perturbation velocities u and v in the x- and y-
directions, respectively. As a n example, for a thin profile moving
through a n infinite medium (Fig. 10.1) a t the speed - V,, the velocitv
A r t . 10.2 LINEARIZATION OF THE POTENTIAL EQUATION 30 :
of the gas relative to the profile would be T.', a t points very far from tll:
profile; and near the profile the x- and y-components would be (Vm 2:; +
and v, respectively. The perturba-
tion velocities, u and v, are con-
sidered to be small compared with
the jree-stream velocity, 77,.
Following this line of thought,
we express the velocity potential a
as the sum of the potential due to
the main velocity V, and of a per-
turbation potential q associated with
the perturbation velocities u and v:
a = V,x +v (10.3) FIG. 10.1. Nomenclature.
Taking derivatives, and noting that V, is a constant, we obtain
a,, = a,, = aulay = av/ax (10.8)

Comparisons of Orders of Magnitude. We wish to simplify Eq. 10.1
by dropping those terms which, according to the assumption of small
perturbations, may be regarded as negligible compared with the remain-
ing terms. The second derivatives in Eq. 10.1, according to Eqs. 10.5,
10.7, and 10.8, are given by
au --av , Qjz,=--=-au av
az2= - ; a,, - .
ax ay ay ax
Now the perturbation velocities u and v are of the same order of magni-
tude. Also, since Eq. 10.1 is to be applied throughout the entire flow
field, the lengths in the denominators of &/ax, &lay, and av/ax must
all be regarded as being of the same order of magnitude. Hence, the
derivatives a,,, a,,, and a,, are all of the same order of magnitude.
We next examine the relative magnitudes of the coefficients of a,,,
a,,, and QZ,. Using Eqs. 10.2, 10.4, and 10.6, we obtain, upon noting
that c, is constant,
where cmis the sound velocity a t points where the stream velocity is Vm.
306 TWO-DIMENSIONAL, SUBSONIC FLOW Ch. 10
Simplifying and rearranging this expression, we obtain
where Mm is the free-stream Mach Number, defined by

M, = Vm/cm (10.10)
Expanding Eq. 10.9 with the help of the binomial theorem, we obtain
where the symbol [0]means "of the order of magnitude of."

For the coefficient of a=, we get, using Eq. 10.4,
a21 -
1--= (Vm + uI2 = 1-
+ 2
Vm2 cw2
(Vm uI2 --
c2 c2 Vm cw2 c2
Introducing Eqs. 10.10 and 10.11, this expression becomes
@z2 U U
1--=1-~,2- 2MW2
- - (k - l)Mm4-
c2 v* V.3,
+ 101(Mm2 $)
Vw
+...
together with terms of higher order. We now assume that the perturba-
tion velocities are so small compared with Vw that we may write
"2($)<< 1 2
(10.12)
2
k2(&) << 1 (10.13)
Mm2 (s) << 1 (10.14)
Consequently the equation preceding Eq. 10.12 may be approximated by
I---
*m2
c2
2 1 - Mm2 - [2Mm2 + (k - 1)Mm4]-
v,
U
(10.15)
Turning now to the second term of Eq. 10.1, we have

Art. 10.2 LINEARIZATION OF THE POTENTIAL EQUATION 307
Introducing Eqs. 10.10 and 10.11 as before, we ohtair!
However, with the assumption of Eq. 10.13, this becomes
For the coefficient of the third term of Eq. 10.1, we have

@zip, (VW + u)v = 2
vwv+ uu --
2
Vm cm2
2c,
= c2 vw2 cm2 c2
Again introducing Eqs. 10.10 and 10.11, we obtain
Using the assumption of Eq. 10.14, however, this becomes
Substituting Eqs. 10.5, 10.7, 10.8, 10.15, 10.16, and 10.17 into Eq. 10.1,
we get the approximate differential equation
The Linearized Equation of the Velocity Potential. Unfortunately,

despite the assumptions already made, this equation is not linear, since
there remain the quadratic terms u au/ax and v au/ax. We therefore
introduce the further assumptions that
and thus we get the following linear differential equation:
where
8==- (10.21b)
Examination of Assumptions Underlying Linearized Equations. The
assumptions underlying the linearized differential equation are embodied
in Eqs. 10.12, 10.13, 10.14, 10.19, and 10.20. Examination of these
equations indicates that the latter two include the former three, and are,
therefore, the most stringent.
As M, approaches unity, Eq. 10.19 becomes invalid. Hence, except
for extraordinarily small perturbations, the conclusions set out in the
remainder of the chapter are not valid for transonic flow, i.e., for values
of M, greater than about 0.8.
The perturbation ratios u/Vw and v / V , depend chiefly on the thick-
ness ratio of the body. The accuracy of Eqs. 10.19 and 10.20 will, for
thin profiles, be good up to high subsonic (but not transonic) Mach
Numbers. For thick profiles, on the other hand, the assumptions are
reasonably valid only for very low Mach Numbers.
10.3. Linearization of the Pressure Coefficient

The pressure distribution in a given flow pattern is of primary prac-
tical importance. A knowledge of the pressure distributior~is necessary
for the calculation of the forces and moments produced by the fluid on
solid boundaries. Moreover, the pressure gradient a t a solid boundary
influences the development of the boundary layer, and the nature of the
pressure gradient must be known before the behavior of the boundary
layer may be predicted.
Dimensional Analysis. Since we are dealing with an inviscid fluid,
the motion of a fluid particle is determined by a balance between pres-
sure forces and inertia forces. Let us assume that we are dealing with
the flow of a particular gas past a series of geometrically similar bodies,
so that in any particular experiment the specification of one character-
istic dimension L (perhaps the length of the body) suffices to determine
the body geometry.
The pressure forces on a given fluid particle are due to the differences
in pressure acting on the various faces of the particle. They may, there-
fore, be represented by some characteristic pressure difference, p - pw,
and by the characteristic length L. Here p is the pressure a t the desired
point in the flow pattern and p, is the free-stream static pressure.
The inertia force of the particle depends upon the local density and
the local acceleration as well as upon the particle size. The local density
is characterized by the free-stream density p, together with some index
of compressibility which is conveniently taken as the free-stream speed
of sound cw. The local acceleration may be characterized by the char-
acteristic velocity Vw and the characteristic length L. The particle
size is, of course, characterized by L.
Art. 10.3 LINEARIZATION O F THE PRESSURE COEFFICIENT 309
From the foregoing considerations, we may write a physical equation

in the implicit form
P - Pm = f W m , PWI cm1 L)
which may be rearranged to give, implicitly,
Since the left-hand side has zero dimensions, the right-hand side must
also be dimensionless. But p,, c,, and L can in no way be formed into
a dimensionless group. From this we conclude that, when the physical
equation is rearranged as indicated, these three terms disappear from
the equation except insofar as pm and c, appear in the dimensionless
groups.
PRESSURE COEFFICIENT. The group containing the local pressure is

called the pressure coeficient, and may be variously written as
From the dimensional analysis we conclude that the local pressure

coefficient a t a given point depends only on the free-stream Mach Num-
ber M,. The size of the body is seen to enter not a t all. For the same
free-stream Mach Number the pressures a t corresponding points in two
flows are, therefore, related through the condition that the value of C,
is the same a t both points.
The Pressure Coefficient for Small Perturbations. We shall now
develop a special form of Eq. 10.22 which is consistent with the assump-
tions of the linear theory. From the energy equation and the velocity
vector diagram of Fig. 10.1, we may write
We now solve this expression for T/T,, and, aftcr cmploying the relations
cm2= kRT,
Mm = VJc,
we obtain
Since the flow is isentropic, the pressure ratio may be written as

k
(2- + v+m 2 ")lA

P T Z i k-l
Pm
- = =[ 1-----~,~
2 m; u2
We now note that the second-term in the bracket is less than unity, and
so we develop a series expansion for p/pm with the help of the binomial
theorem:
Substituting this expression into Eq. 10.22, and simplifying, we get
We now introduce the assumptions of Eqs. 10.12, 10.14, and 10.19, and,
since mathematical simplicity is the main purpose of the iinearized
theory, we make the additional assumption that
The final expression for the pressure coefficient of the linearized theory
is then
u
Cp 2 -2- (10.26)
V,
Eq. 10.25 is a more severe restriction than Eq. 10.19. However, in
many applications, Eq. 10.26 for C, is used only when comparing cor-
responding values of C;, for two flows wherein the two values of C, do
not differ by much. Such is the case, for example, in the formulation of
the Gothert and Prandtl-Glauert similarity rules to be discussed later
in this chapter. In cases of this sort, i t may be shown by detailed con-
sideration that the use of Eq. 10.26 in comput,ing the ratio of two values
of Cp involves only the assumption that ( u / V ~ < )<~ I, rather than the
assumption of Eq. 10.25.
10.4. Flow Past a Wave-Shaped Wall
5
We shall now investigate a particular solution (I) o the linearized
equations, the practical significance of which rests on the fact t,hat in a
relatively simple way we may demonstrate the quantitative and yualita-
tive effects of compressibility on the streamline pattern and on the pres-
sure distribution.
Art. 10.4 FLOW PAST A WAVE-SHAPED WALL 311
Velocity Potential. Let us consider the following expression for the

perturbation velocity potential:
where h and I are length constants whose physical meanings will be evi-
dent later and 13 is defined by Eq. 10.21b. Whether or not this assumed
function satisfies the conditions of continuity, momentum, energy, and
irrotationality depends on whether it satisfies Eq. 10.21a. To determine
this, we first take derivatives as follows:
au a2, - sin - (10.29)

ax ax2 13
Substitution of Eqs. 10.29 and 10.31 into Eq. 10.21a shows that the
function cp(x,y)of Eq. 10.27 satisfies the differential Eq. 10.21a identi-
cally. Hence we conclude that Eq. 10.27 represents a flow pattern
satisfying the conditions of continuity, momentum, energy, and irro-
tationality.
Streamlines. To determine the form of the streamlines, we make use
of the definition that the slope of the streamline is identical with the
slope of the velocity vector, so that, referring to Fig. 10.1, we write
The values of v/V, and u/V, are known in terms of x and y through Eqs.
10.28 and 10.30, but on substituting these values into Eq. 10.32 there
seems to be no simple way of separating variables and thus integrating
to find the equations of the streamlines. However, this difficulty may be
avoided by introducing approximations in keeping with the assumptions
of small perturbations. We write the identity
Now, according to the assumption of Eq. 10.19, the value of MW2u/Vw

is negligible compared with unity, hence
Now, substituting u / V , and v/VW from Eqs. 10.28 and 10.30 into
Ea. 10.33, and rearranging, we obtain
hI
dy = - 2 r (sin ?)
2rx
e - 2 T h l z dz - 277~
This may now be integrated directly to give the equation of the stream-
line,
y = constant +h (10.34)
where each streamline has associated with i t a particular value of the

constant of integration.
Suppose that the streamline for which the constant of integration is
zero is taken to be a solid wall, defined by the coordinates y, and x,. If
we also stipulate that the solid wall lies very close to the x-axis (i.e., y,
is very small), the value of the exponential term of Eq. 10.34 is nearly
unity, and hence the shape of the solid wall may be expressed by the
relation
27rx,
Y, S h cos --
I
At very large distances from the wall (i.e., very large values of y), it is
evident from Eq. 10.34 that the streamlines have constant values of y
and are therefore parallel to the x-axis.
The physical nature of the flow pattern defined by Eq. 10.27 is now
clear. We are dealing with the flow past a wave-shaped wall which has
the shape of a cosine curve. Referring t o Eq. 10.35 and to Fig. 10.2, we
see that I and h are the wave length and lateral amplitude, respectively,
of the wave-shaped wall.
From Eqs. 10.28 and 10.30 we see that the perturbation velocities u
and v are zero a t very large values of y. This means that a t an infinite
distance from the wall the flow is parallel to the x-axis and has the uni-
form velocity V-.
rbt. 10.4 FLOW PAST A WAVE-SHAPED WALL 313
Streamlines f
\
FIG. 10.2. Flow past wave-shaped wall, showing streamlines and pressure distri-
bution a t surface. Amplitude of wave in wall is esaggerated.
(a) Incompressible flow (M, = 0).
(b) Subsonic flow (M, < 1).
Pressure Distribution. The perturbation velocities a t the wall, ac-

cording t o Eqs. 10.28 and 10.30, are
Using the approximate expression for the pressure coefficient given

by Eq. 10.26, we obtain
cam= 0 (10.39)
From Eq. 10.38 i t is evident that on the solid wall the pressure is a
maximum in the lowest points of the troughs and is a minimum a t the
highest points of the crests. The variation in pressure along the contom
is one-half wave length out of phase with the shape of the wall (Fig. 10.2).
By tracing the cross-sectional area of a stream tube formed by two
neighboring streamlines, i t will be observed that the pressure is a maxi-
mum at the largest cross-sectional area and a minimum a t the smallest
cross-sectional area, a result which fits in with one-dimensional con-

siderations.
The streamlines a t any distance from the wall are also approximately
cosine in shape and are in phase with the wave-shaped wall, but the
amplitude of tlhe wave in the streamline shape dies off with increasing
distance from the wall, according to the negative exponential term of
Eq. 10.34. Likewise, the disturbance produced by the wall, as measured
by the perturbation velocity u and by the pressure coefficient, decays
with increasing distance from the wall according to the factor
Effect of Mach Number. How does compressibility influence the flow

pattern? Suppose that we have a wall whose contour is given by Eq.
10.35, with fixed values of h and 1. Now, for high Mach Numbers, 0 is
small, while for low Mach Xumbers, 0 approaches unity. Hence we see
first of all from Eq. 10.38 that the difference between the free-stream
pressure and wall pressure increases with Mach Number according to the
factor .-/I Examination of Eqs. 10.34 and 10.37 shows also
that with increasing Mach Number the disturbance produced by the
wall on the flowing stream decays at a slower rate. Thus, increasing
the Mach Number acts not only to augment the pressure difference acting
on the solid body, but the effects of the solid body are felt a t greater
lateral distances from the body.
The conclusions reached in the preceding paragraphs are illustrated
in Figs. 10.2 and 10.3.
FIG.10.3. Effect of Mach Number on pressure disturbance for flow past wave-
shaped wall.
(a) Peak pressure coefficient at wall.
(b) Lateral decay of peak pressure coefficient.
Although the flow past a wave-shaped surface is by no means the

same as the flow past a thin profile, it nevertheless seems plausible that
the effects of Mach Number would be much the same and that we may
use the example of the wave-shaped wall as a guide in estimating com-
Art. 10.5 GOTHERT'S RULE 315
pressibility effects. Thus, since the lift and moment acting on a profile
are found by suitable integration of the pressures acting over the sur-
face of the profile, we might expect from Eq. 10.38 that as a first approxi-
mation the lift and moment coefficients would be inversely proportional
to 8. i.e..
Still another important conclusion to be drawn from this simple

example is that in wind tunnel experiments the interference effects due
to the walls of the tunnel are likely to become increasingly important at
high Mach Numbers, because the lateral influence of a solid profile
immersed in a stream increases with increasing Mach Number. For
similar reasons, the mutual interference effects between blades in an
axial flow turbine or compressor may be expected to grow larger as the
Mach Number is increased.
A further interesting conc1usion:is that for small perturbations the
effect of Mach Number on the pressure coefficient is independent of the
specific-heat ratio. Of course this is valid only to the order of approxi-
mation of the linearized method.
10.5. Gothert's Rule

General Remarks Concerning Similarity Laws. I n this article we
shall derive an expression which relates the subsonic compressible flow
past a certain profile to the incompressible flow past a second profile
derived from the first through an affine transformation. Such an expres-
sion is called a similarity law. Since we shall subsequently derive other
similarity laws for transonic, supersonic, and hypersonic flow, it is
appropriate to discuss the philosophy underlying such similarity laws.
If the equations of motion could be solved easily, the solutions them-
selves would indicate quite clearly the nature of any similarities which
might exist among members of a family of flow patterns. A separate
derivation of similarity laws would, therefore, be superfluous.
We are in fact unable, in the majority of instances, to solve the equa-
tions of motion. However, even though solutions are lacking, we may
exploit our knowledge of the forms of the differential equations (and the
associated boundary conditions), and thus derive the similarity laws
which of necessity would appear in the solutions if the latter could be
found.
The derivation of similarity laws is a form of dimepsional analysis
involving distorted models rather than geometrically similar models.
Similarity laws may be useful in a number of practical ways. For
example, Gothert's rule for subsonic flow, which is derived in this article,
316 T\VO-DIMENSION.%L, SUBSONIC l%O\V Ch. 10
allows us to predict the details of the subsonic flow past a body a t sub-
sonic speeds if we know the details of an incompressible flow past an
affinely related body. Similarly, the transonic and hypersonic similar-
ity rules show how experimental data for a certain body a t a certain M,
can be made applicable to a related body at a different M,.
Transformation of Variables Leading to Laplace's Equation. Suppose
that a thin, two-dimensional airfoil is moving a t the constant speed
- Vw through an infinite gas medium which is at rest a t great distances
from the airfoil. To an observer traveling with the uniform velocity of
the airfoil, the airfoil is stationary, and the gas a t great distances from
the airfoil flows past the airfoil with the uniform, parallel velocity Vm
(Fig. 10.1).
The flow pattern is in first approximation governed by the linearized
differential equation of the perturbation velocity potential,
This equation, however, is not a complete description of the problem,

since it is necessary also to specify the boundary conditions. At an
infinite distance in any direction from the airfoil the velocity is Vw,and
the perturbation velocity is zero. Hence we may write
u = av/ax = o
aty = fa:
= av/ay =o
In addition, the velocity vector a t the surface of the profile must be

tangent to the profile itself. Denoting the coordinates of the surface of
the profile by x,, y,, we express this boundary condition (Fig. 10.1) by
the relation
-
~ Y S
at xs, y8: - -
vs
- vs/Vm
s Vw + us +
1 us/Vw
where the subscript s denotes conditions at the surface of the profile.
According to Eq. 10.25, however, we have agreed in the linearized theory
to ignore u/Vm as compared with unity. Hence the boundary condition
a t the profile becomes
For a profile of given shape, that is, with a contour given by a certain
function f(x,, y,) = 0, the solution of Eq. 10.21a, subject to the boundary
conditions given by Eqs. 10.40 and'l0.41, is a t best not easy. However,
it will be noticed that Eq. 10.21a is not very different in form from the ex-
act equation for incompressible flow, the latter being found merely by set-
ting p equal to unity. Also, it may be recalled that one of the devices
used for solving unfamiliar differential equations is to seek a transforma-
tion of variables which reduces the differential equation to a familiar
form for which a solution is available. These remarks suggest that by a
suitable transformation it might be possible to reduce Eq. 10.21a to
the Laplace equation and thus open the possibility of employing the
well-developed store of theory and experiment for incompressible flow.
A generalized transformation which leads to the desired result is
found by defining new variables x', y' and cp1(x', y') as follows:
where A,, Xu, and X, are constants whose values are to be determined.
We now transform Eq. 10.21a into the x', y', cp' system of variables
through the following relations:
Substituting these into Eq. 10.21a1we obtain

2
p--xZ2 a 2 cp1 hV2aZcp'
+--= 0 (10.46)
A, axf2 A, aYl2
Now, if we set
P2Xz 2 --X 2 . , or XV/h2 = p (10.47)
we obtain
which means that the x', y', p' system of variables describes the motion
of an incompressible fluid if A, and A, are so chosen that X,/X, = p.
Transformed Boundary Conditions. To discover the complete rela-

tion between the compressible flow (o(x, y) and the incompressible flow
~'(x',y'), we must see what form the transformed boundary conditions
take.
At infinite distance from the airfoil, by employing Eq. 10.40 with
Eq. 10.45, we see that
From these relations we conclude that in the incompressible flow there

are no velocity perturbations at great distances from the airfoil and
hence the velocity is uniform and parallel a t infinity.
Turning now to the boundary condition a t the airfoil, we 6rst note
that the profile slopes are related by
where x,' and y,' represent the coordinates of the transformed profile
as given by
x,' = X,x, (10.51a)
In addition, we note that the transformation formulas give
Substituting Eqs. 10.52 and 10.50 into the boundary condition of Eq.
10.41, we get after rearrangement
dys' 1
- = -- (10.53)
dx' hzX, V,
If we now set
X; = XzXp (10.54)
and if we also define V,' to be equal to V , (which may be done without
loss of generality), we obtain
Comparison of Eq. 10.55 with Eq. 10.41 shows that the boundary con-
ditions for the incompressible flow past a new profile defined by x,', y,'
are satisfied when V,' = V, and = XZX,.
To recapitulate what has been accomplished thus far, we have found

that the function p'(xf, y') describes the uniform, parallel incompressible
flow past a new profile x,: y,', provided that Ax, A,, and A, are chosen so
as to satisfy Eqs. 10.47 and 10.54 and that the velocity a t infinity is the
same for both flows.
Relation Between Geometries of Corresponding Profiles. How does
the newly found incompressible flow compare with the compressible
flow to which it is related? We see from Eqs. 10.50 and 10.47 that
which means that the slope of the profile in the compressible flow pat-
tern is larger by the factor 1/0 than the slope of the corresponding pro-
file in the related incompressible flow pattern. But if the slope of the
profile a t each point is greater by the factor I/& it is also true that the
camber ratio, angle of attack, and thickness ratio must all be greater for
the compressible airfoil by the factor I/& Thus, denoting the camber
ratio by y , the angle of attack by a and the thickness ratio by 6, we have
To demonstrate this relation for the thickness ratio, for example, we

note that
where t, is the difference in profile ordinates a t the point of maximum

thickness of the profile, and 1 is the chord length (which lies in the x-
direction). But, from Eq. 10.51,
t,, = tf,,/A,; I = If/A,
and hence
which was to be proved.

The geometrical results discussed above are illustrated in exaggerated
form in Fig. 10.4. It will be noted that the individual contraction ratios
A. and &, are unimportant, since i t is the ratio of the two which deter-
mines changes in the shape of the profile. Thus, we can set A, = 1 and
A, = 8, thereby keeping the same chord but changing the ordinates; or
we can get equivalent results by setting h i = 1 and A, = 1/8, thereby
keeping the same ordinates but changing the chord. The type of geo-
metrical transformation involved here is termed afine, meaning that all
coordinates in a given direction are changed by a uniform ratio. If a
profile were printed on a thin, uniform sheet of rubber, and the rubber
were then stretched with uniform strain in both the x- and y-directions,
the profile would change shapc in an afine manner.
"I Incompressible
1 I /
Camber Line
Fro. 10.4. Coordinate transformation for Gothert's rule.

2' = x,z; stg= XZZ8
Relation Between Forces on Corresponding Profiles. The lift on the

profile is determined by the pressure distribution on the surface of the
airfoil, which in turn may be expressed in terms of the pressure coeffi-
cient a t the surface. From Eq. 10.26 we have
Similarly, noting that we have already set V,' = Vm,
But
and hence, wit;h the aid of Eqs. 10.47 and 10.54, we find that
The lift is found by integrating the vertical component of pressure

force over the surface of the profile. Hence the lift coefficient is found
by integrating the pressure coefficient over the surface, as indicated by
the following operations:
Art. 10.6 T H E PRANDTL-GLAUERT RULE 321
At equal values of x,/l, Cp, and Cp,' are related by Eq. 10.58. Therefore
the ratio CI/C{ must be the same as the ratio Cp,/Cp,'. A similar
argument leads to the same conclusion for the moment coefficient, Cn.
Thus we may write
which means that if the two profiles are related affinely as described
previously, the lift and moment coefficients for the compressible flow
pattern will be greater than for the related incompressible flow pattern
by the factor 1/b2.
We might inquire whether these results have any practical significance.
There has been accumulated fairly extensive theoretical and experi-
mental information on the properties of classes of affinely related pro-
files in incompressible flow, with systematic variations in camber,
thickness ratio, and angle of attack. If it is desired t o find the lift coeffi-
cient of one of these profiles a t a finite Mach Number M,, one first finds
(either theoretically or experimentally) the lift coefficient in incom-
pressible flow of an affinely related profile whose camber, thickness
ratio, and angle of attack are all smaller than the corresponding values
for the original profile by the ratio P. Then, by multiplying this lift
coefficient by l/p2, one finds the desired lift coefficient for the compress-
ible flow.
This method of projecting experimental data for incompressible flow
is sometimes awkward, since it requires incompressible data for a large
range of thickness ratios. It would be more convenient in many cases
to know how Mach Number affects the performance of a profile of fixed
shape. The Prandtl-Glauert rule, which is discussed in the next article,
yields information of this type.
10.6. The Prandtl-Glauert Rule

Similarity Law for Incompressible Flow. For the present let us con-
sider the incompressible, two-dimensional flow past a family of affinely
related thin profiles. It is known that for such affinely related profiles
the pressure coefficients a t corresponding points are approximately
proportional t o the thickness ratios (and, of course, simultaneously t o
the camber ratios and angles of attack). For example, if in the case of
the wave-shaped wall we consider h/l to be the appropriate "thickness
ratio," i t is evident from Eqs. 10.37 and 10.38 that the pressure co-
efficient is proportional to the thickness ratio. As a second example,
consider incompressible flow past an inclined flat plate: classical aero-
dynamic theory shows that the lift is proportional to the angle of attack.
Following this line of thought, consider two affine profiles in incom-

pressible flow which are related as follows:
Then it follows that, a t least approximately,
Combination of Gothert's Rule with Incompressible Similarity Law.

The foregoing rule for incompressible flow gives us an opportunity to
derive further similarity criteria between incompressible and compress-
ible flow. For if we combine the relations immediately above with
Eqs. 10.57, 10.58, and 10.59, we find that if the compressible profile
(unprimed) is affinely related t o the incompressible profile (double
primed) in such a way that
then the forces acting in the two flows are related through
These equations mean that for the same profile at the same angle of
attack the pressure coeficient, lift coeficient and moment coeficient are all
aflected by Mach Number in proportion to the factor 1 / 4 1 - b2.
From Eqs. 10.60 and 10.61 it also follows that for a given profile the
lift-curve slope is proportional to I/@:
It is profitable t o carry this line of attack one step further. Let us

now consider two affinely related profiles in incompressible flow for
which
Y' a' 6'
a"' 6"' - B2
Y"'
so that, a t least approximately,
Art. 10.6 THE PRANDTL-GLAUERT RULE 323
Now let us combine these relations with Eqs. 10.57, 10.58, and 10.59.
We find that if the compressible profile (unprimed) is affinely related to
the incompressible profile (triple primed) in such a way that
then
These equations mean that the dimensionless pressure distribution, lift

coefiient, and moment coeficient will be the same for compressible.and
incompressible flow if the profiles are afinely related in such a way that the
compressible profile i s smaller in camber ratio, thickness ratio, and angle
of attack by the factorm.
This form of the similarity rule is sometimes useful for making pre-- -
dictions concerning boundary-layer separation, for the latter is largely

dependent upon the dimensionless pressure gradient a t the surface of
the profile. If the effect of Mach Number on boundary-layer separation
is assumed to be small, for example, then Eqs. 10.63 and 10.64 show
that, in order to avoid stalling the profile in a compressible flow, the
profile must be thinner, have a smaller value of 7 , and be at a lesser
angle of attack than would be allowable with an incompressible flow.
Summary of Similarity Laws. The three laws of similarity may be
conveniently summarized by the following symbolic statements:
Rule I :
Rule 11:
Identical statements apply to CL and Cm.

Eq. 10.65 corresponds to Eqs. 10.57 and 10.58, and is to be read:
"The pressure coefficient at a certain point for compressible flow with
Mach Number M, past a profile having geometric parameters of magni-
tude a, 6, and 7 is greater by the factor than the pressure coefficient
st the corresponding point for incompressible flow (M, = 0 ) past an
affinely related profile having geometric parameters of magnitude
Ba, @ti,and By."
The three rules are illustrated in Fig. 10.5 Rule I is Gothert's rule,
and is the most exact of the three. Rules I1 and 111 are known as the
Prandtl-Glauert rule, and were originally derived incorrectly, inasmuch
as the transformed boundary conditions were not taken into account.
For two-dimensional flows this oversight was not serious, or a t least in-
troduced errors no worse than are already inherent in the linearized
equations. The oversight was not appreciated for some twenty years,
until anomalous results were obtained in comparing Rule I1 with ex-
perimental and exact theoretical results for flow past thin bodies of
Compressrble ( M), lncompress~bleI

Cp = Press. Coeff. C& = l/P2
Incompy~ssibleI[ Incompr=2~sibleIII
Cp/Cp = I/P Cp/Cp = I
FIG.10.5. Similarity Rules I, 11, and I11 for subsonic flow.
revolution. For such axi-symmetric flows the incompressible pressure

coefficient is not proportional to the thickness ratio, and hence Rules
I1 and I11 are not applicable. Gothert (2) formulated the more general
Rule I and showed the correct way of extending it to three-dimensional
flows (Chapter 13).
Effect of Mach Number on Flow Far from Prafile. Thus far we have
considered the influence of M, on the pressure distribution a t the surface
of the profile. At great distances from a given profile it may be shown
that, a t a given point (see Problem 10.6)
Laitone's Modification of Prandtl-Glauert Rule. An improvemmt

on ltulc I1 has been suggested by Lait~ne.'~'
Art. 10.6 THE PRANDTL-GL.4UERT RULE 325
The linearized theory is evidently not applicable for large perturba-

tions. Or, stated differently, it is only applicable for regions where Cp
is very small and for low values of M,. A better approximation is
obtained by writing Rule I1 in terms of the local Mach Number ML
rather than in terms of the free-stream Mach Number M,. This pro-
cedure implies that the linearized equations are applied not to the entire
field of flow but instead to the local flow field which is considered to be
approximately uniform in a small region.
Let Cp be the pressure coefficient a t a certain point for compressible
flow past a profile, and let C,, be the pressure coefficient a t the same
point for incompressible flow past the same profile. Then, by Laitone's
hypothesis, n
Eq. 10.22 for Cp is in this case written with p ~ / p , on the right-hand side.
The latter ratio is given by the isentropic relations of Chapter 4 as
Substituting this into Eq. 10.22, and solving for tAL2, we obtain
However, since the linearized theory ignores terms of greater than lirst
order in Cp, the foregoing expression may be expanded by the binomia!
theorem to give
k-1
M L2~ M , ~- Mm2)MmZcP - . +
Substituting this into Eq. 10.68, and solving for Cp, we finally obtain
." fin,
For very small values of C,, this is seen to be substantially identical

with the Prandtl-Glauert rule. Usually we are interested in regions
where the pressure coefficient is negative. For such regions Eq. 10.69
326 TWO-DIMENSIONA4L,SUBSONIC FLOW Ch. 10
indicates a greater influence of Mach Number on C , than does the

Prandtl-Glauert rule. Moreover, Eq. 10.69 is in agreement with experi-
mental results in predicting that the effects of compressibility are propor-
tionately greater for regions of high negative pressure coefficient than
for regions of low negative pressure coefficient.
The pressure correction rule of Eq. 10.69 cannot be extended to the
prediction of forces, because the effect of Mach Number on C, is d i e r e n t
a t each point of the profile. If the pressure distribution for incompress-
ible flow is known, however, the pressure distribution for compressible
flow may be found from Eq. 10.69. Then, by integration, the lift and
moment coefficients for compressible flow may be computed.
10.7. Experimental Results for Thin Profiles

In comparing the similarity rules with experimental data, it must be
remembered that they represent primitive a.ttempts to obtain useful
information in the simplest possible manner. It is evident that the
FIG. 10.6. Comparison of measured FIG. 10.7. Comparison of measured

lift-curve slope for NACA 4412 profile lift-curve slope for propeller section of
(12% thickness) with Prandtl-Glauert 6% thickness with Prandtl-Glauert rule
rule (after Stack, Lindsey, and Littell). (after Ferri).
assumptions on which the rule is based are in most practical cases

satisfied only in a very approximate way. For example, although the
perturbation velocity is taken t o be negligible compared to the free-
stream velocity, a t the stagnation point the perturbation velocity is
exactly equal to the free-stream velocity.
Fig. 10.6 shows in a typical way the experimental effect of M, on the
lift-curve slope for a 12%-thick profile. Similar data for a propeller
section of 6'% thickness ratio is shown in Fig. 10.7. Up to the Mach
Number where the lift begins to drop sharply because of "compressi-
bility burble," the Prandtl-Glauert rule predicts the general effect of
M, but underestimates the magnitude of the effect. As might be ex-
pected from the nature of the linear theory, the Prandtl-Glauert rule
is more accurate for the thinner profile. The "compressibility burble"
.4rt. 10.8 WIND TUNNEL CORRECTIONS 327
is associated with the appearance of shock waves near the profile which,
by virtue of adverse pressure gradients, produce flow separation. This
effect is very similar t o stalling a t high angles of attack.
Fig. 10.8 shows the effect of M, on the pressure coefficient a t 30%
chord on the upper surface of NACA 4412 operating with an angle of
attack of -2". Here again it is evident that the Prandtl-Glauert rule is
correct as regards order of mag-
nitude, but underestimates the
effects of compressibility.
From these typical results we
may conclude that the Prandtl-
Glauert rule predicts the effects of
M, with good accuracy for very
thin profiles a t low Mach Num-
bers. For thick profiles and high
Mach Numbers, the rule provides
a guide to orders of magnitude
but is not accurate enough for FIG.10.8. Pressure coefficient versus M,
design purposes. at 30% chord on upper surface of NACA
4412, with a = -2". Comparison of
For additional experimental measured data with Eqs. 10.66 and 10.69.
dataillustrating thevalidity of the The curve marked C,* denotes the local
sonic condition (after Stack, Lindsey, and
Prandtl-Glauert rule, the reader is Littell).
referred to Chapters 11 and 12.
Laitone's modification of the Prandtl-Glauert rule is seen from Fig.
10.8 to predict the variation of Cp with M, somewhat better than does
the Prandtl-Glauert rule. It should be noted here that no single simi-
larity rule can be correct for all profiles, and hence the different similar-
ity laws compare more or less favorably with experiment, depending
upon the profile shape, angle of attack, position of measurement, etc,
It is not amiss here to point out that the Karman-Tsien pressure
correction formula (Chapter 11) is generally in better agreement with
experiment than is the Prandtl-Glauert rule.
10.8 Wind Tunnel Corrections

Definition of Problem. When a wing profile is tested in a wind tunnel,
corrections must be made to the measured lift and moment coefficients
in order to obtain the corresponding coefficients for the same profile in
an infinite air stream. These corrections are necessary because the
boundary conditions of the flow in a tunnel are different from those in a
large air stream. For two-dimensional tests, with the wing spanning
a rectangular test section, the tunnel correction depends chiefly on the
ratio l/h, where 1 is the chord of the profile and h is the half-height of
the tunnel (Fig. 10.9a). In incompressible flow the methods of making
328 TWO-DIMENSIOTAL. SUBSONIC FLOW Ch. 10
FIG. 10.9. Similarity Rules I, 11, and I11 for thin profile in wind tunnel.
such corrections are well developed. We shall now derive similarity

rules which show the effect of compressibility on the tunnel correction
factors.
Similarity Transformations. The differential equation describing the

flow is, as before, Eq. 10.21a. Introducing the transformation formulas,
Eqs. 10.42, 10.43, and 10.44, we get Eq. 10.46 and find as before that
' p l ( x ' , y') represents an incompressible flow pattern if XJX, = 0.
Far upstream and far downstream of the profile, the boundary con-
dition is
Using Eqs. 10.42 and 10.45, we see that the corresponding boundary
condition in the incompressible flow is
which means that in the incompressible flow the stream is undisturbed

far upstream and far downstream of the profile.
The boundary condition a t the tunnel wall is
Art. 10.8 WIND TUNNEL CORRECTIONS 329
For the incompressible case, the corresponding boundary condition,

according to the transformation formulas, is
at y' = f h h : -
A,
[
hy av1(x', Y')
ayl ]y'= +Ayh
= 0
from which i t follows that v' is zero a t y' = f h', where

h' = X,h
In the incompressible flow the tunnel-wall boundary condition is there-
fore satisfied when the half-height is h'.
Hence it follows that
h/l - h 1' A, 1
- =-
hl/ h' 1 X, B
Regarding the boundary conditions a t the profile, Eqs. 10.50 to
10.55 are applicable here, and yield the requirement that Xu2 = X,Xv.
Similarly, Eq. 10.58, showing the connection between the correspond-
ing pressure coefficients, is also valid for the present case.
Similarity Rules. Combining the foregoing considerations, we obtain
the following similarity rule:
Rule I: 1
h
[cpslk,,
a, 6.7. = [c~slo,oar 66, 07, (10.70)
If h/l is considerably larger than unity, then, as in the case of a single
profile in an infinite stream, there is a similarity law for incompressible
flow which states that
1
[CpJo, a, a,.,:
=-n [CpJo, ,a, 6 ,n, -hI
where n is a constant.
By employing this approximate rule for incompressible flow, there
may be derived from Eq. 10.70 the following rules:
None of these, it may be noted, shows the effect of M, on a test arrange-

ment of fixed geometry.
The three rules are illustrated in Fig. 10.9.
Identical rules apply t o the lift and moment coefficients.
The interference pressure coeficient ACp is defined as the value of Cp
with the tunnel walls present minus the value of C, with the tunnel walls
330 TWO-DIMENSIONAL. SUBSONIC FLOW Ch. 10
absent. I t may be shown (see Problem 10.10) thzt the effect of com-
pressibility on the interference pressure coefficient a t a fixed point on
a profile of fixed shape in a tunnel of fixed size is given approximately by
10.9 Flow Inside Two-Dimensional Passages

Consider the two-dimensional flow inside a passage such as is shown
in Fig. 10.10. We shall suppose that the wall contours are such that
FIG.10.10. Flow in two-dimensional duct.
the velocity perturbations are small enough to permit application of the

linear theory.
Application of Gothert's Rule. I n applying the similarity analysis,
only Rule I (Gothert's rule) is permissible, since the approximations in-
volved in the boundary conditions near thin profiles for Rules I1 and I11
are not applicable when applied to the boundary conditions at the duct
walls of Fig. 10.10.
Using the previous results, Gothert's rule for the present case may be
easily applied. We consider an affine transformation of the duct such
that I
xs = Xd,; Y,' = X,Y, (10.73)
and, in order that the primed quantities shall correspond to incompress-
ible flow, we stipulate that
At corresponding points in the incompressible and compressible flow,

the following relations then prevail:
k t . 10.9 now INSIDE TWO-DIMENSIONAL PASSAGES 33 1
Thus, using one of the many methods for obtaining incompressible flow
patterns, it is possible with this rule to construct approximately a series
of corresponding patterns for compressible flow.
One-Dimensional Mow. The method given here refers t o flows which
are essentially two-dimensional and serves no useful purpose for one-
dimensional flow because in the latter case there are simple and more
accurate methods a t hand. However, we may employ one-dimensional
considerations to verify for a limiting case the validity of the rule as
given above. Suppose that we have a long duct of uniform width fol-
lowed by a slight change in area, after which the duct is again constant
in width (Fig. 10.11). Far upstream and far downstream of the change,
IA
IV
I
lM
- A+ d ~ i
V + dV;
M+dMl
FIG.10.11. Illustrates verification of Gothert's rule for one-dimensional flow.

the flow is nearly one-dimensional, and therefore we may use one-
dimensional formulas for calculating changes between the two sections.
What does Gothert's rule tell us about this problem? First of all,
since the transformation is affine, Eq. 10.73 shows that the fractional
change in area, dA/A, taken in a one-dimensional sense, is the same for
the compressible and incompressible patterns which are to be compared.
Eq. 10.75 then indicates that, for the same fractional change in area,
Using- the Euler equation, the pressure coefficient for small changes
may be written
dp p V dV dV
C p = c = - - -2-
2p'V iPV2 v
Now, employing the one-dimensional considerations previously referred
to, we find from Table 8.1 that for isentropic flow,
and, accordingly,
which shows that the pressure coefficient is indeed proportional to

1/(1 - kt2) for a given area ratio. Consequently we have verified

Gothert's rule for the limiting case considered here.

1. ACKERET, J. tfber Luftkrafte bei sehr grossen Geschwindigkeiten insbesondere bei
ebenen StrBmungen, Helvelica Physica Acta, Val. 1 (1928), pp. 301-22.
2. GOTHERT, B. Plane and Three-Dimensional Flow a t High Subsonic Speeds,
NACA Tech. Memo., No. 1105 (1946). (Translation from original German
paper in Lilienthal Gesellschaft 127.)
3. STACK, J., LINDSEY, W. F., and LITTELL,R. E. The Compressibility Burble and
the Effect of Compressibility on the Pressure and Forces Acting on an Airfoil,
NACA Tech. Rep., No. 640 (1938).
4. FERRI, A. Investigations and Experiments in the Guidonia Supersonic Wind
Tunnel, NACA Tech. Memo., No. 946 (1940).
5. TSIEN, H. S., and LEES, L. The Glauert-Prandtl Approximation for Subsonic
Flows of a Compressible Fluid, Jour. Aero. Sci., Val. 12, No. 2 (1945), p. 173.
6. HUNSAKER, J. C., and RIGHTMIRE, B. G. Engineering Applications of Fluid
Mechanics, New York: McGraw-Hill Book Co., Inc., 1947.
7. GLAUERT, H. The Effect of Compressibility on the Lift of Airfoils, PTOC.Roy.
Soc. (A), Vol. 118 (1927), p. 113.
8. GOLDSTEIN, S., and YOUNG, A. D. The Application of the Linear Perturbation
Theory of Compressible Fluid to Wind Tunnel Interference, British A.R.C.,
R & M No. 1909 (1943).
9. LAITONE,E. V. New Compressibility Correction for Two-Dimensional Subsonic
Flow, Jour. Aero. Sci., Vol. 18, No. 5 (1951), p. 350.
PROBLEMS
10.1. Consider subsonic, two-dimensional flow with small perturbations, and
assume that the Prandtl-Glauert rule is valid.
If t h e free-stream Mach Number a t which the local speed on a profile first
becomes sonic is 0.80, what is t h e maximum negative pressure coefficient on the
profile at very low Mach Numbers?
10.2. The local pressure coefficient is defined a s
Show that the value of the pressure coefficient corresponding to the critical
velocity on the surface of a n airfoil is given by
Plot Cp*versus (M,),,it between values of 0 and 1 for k = 1.4.

10.3. The wing of a high-speed airplane has, owing to inaccuracies of manu-
facture, a bump in a region where the local Mach Number is 0.70. The bump
extends 0.1 inch above the surrounding surface and is about two inches long
in the direction of flow.
(a) Estimate the peak Mach Number on the bump, neglecting the boundary
layer.
(b) Suppose that a static pressure tap were a t the highest point of the bump.
What would be the per cent error in static pressure owing to the presence of the
bump?
(c) What effect would the presence of a boundary layer have on the foregoing
results?
10.4. Show that the potential function
is a solution of the linearized equation of motion and that it represents the flow
in a two-dimensional passage formed by one wavy wall (amplitude h and wave
length I ) and one straight wall, the two walls being a t a lateral distance H from
each other.
Explore the flow pattern corresponding to this solution.
Compare the streamlines and wall-pressure distributions with the corre-
sponding quantities when H is infinite.
Compare the influence of the straight wall at a fixed value of H in incompress-
ible flow with that in subsonic flow.
Determine how the "interference velocity perturbation" varies with &.
10.5. I n deriving the similarity rules of this chapter it was assumed arbi-
trarily that V,' = V,. Demonstrate that this is not a restriction on the
results by starting anew with the more general form, V,' = XvV , and showing
that the similarity rules thus obtained are the same as those previously obtained.
10.6. I n estimating the effect of distance from a profile on the disturbance
produced by the profile in incompressible flow, the shape of the profile is not
important a t very large distances from the profile. For large distances, then,
we might consider the disturbance of a doublet to be representative of that of
any profile.
The perturbation potential for a doublet a t the origin which is superposed
on a parallel flow V , is
v,x
= x 2 + Y2
and from this it may be shown that, for incompressible flow,

Y2 - x 2
U v.
+ y2Y
(x2
(a) Using this result show that a t a fixed point a t a large lateral distance
( y / x -+ &co) from a fixed profile a t a fixed angle of attack, the effect of com-
pressibility on pressure coefficient is given by
334 TWO-DIMENSIONAL, SUBSONIC FLOW Ch.10
(b) Show that at a fixed large longitudinal distance (z/y -+ +co) from a fixed
profile a t a fixed angle of attack,
10.7. The NACA profiles in the 4-digit series are defined in the following
way. The first digit represents the maximum camber, in per cent of the chord;
the second digit the location of the point of maximum camber, in tenths of the
chord; and the last two digits the maximum thickness, in per cent of chord.
Arbitrary camber equations and thickness functions, together with the number
of the airfoil, serve to specify the shape of the profile.
It is proposed to w d i c t the performance of the NACA 4412 profile at a Mach
Number of 0.6 and a t a2 angle of attack of 4", starting with data taken in a
low-speed wind tunnel.
For each of the three similarity rules, stipulate the number of the profile
which should be tested at zero Mach Number, specify in each case the angle
of attack for the low-speed test, and explain how you would extrapolate the
data to Mach Number 0.6.
10.8. NACA Report No. 646 gives the following data on lift-curve slopes
for the NACA 4412 airfoil, where M is the Mach Number and a is the angle of
attack in degrees:
(a) Plot d C ~ / d aversus M, using these data.

(b) Starting with the value of d C ~ / d aat M = 0.2, calculate the value of
d C ~ / d aat high Mach Numbers with the aid of the Prandtl-Glauert rule. Plot
these results on the chart of part (a).
(c) Plot on the same chart the theoretical lift-curve slope for a flat plate.
The lit coefficient of the latter, for zero Mach Number, is CL = h a , , where
a, is in radians.
10.9. Wind tunnel tests of a series of affinely related profiles at low speeds
and at zero angle of attack yielded the attached curve of maximum negative
pressure coefficient versus thickness ratio.
Ch. 10 PROBLEMS 335
Using the method of small perturbations, estimate the free-stream Mach

Number at which sonic velocity would first be reached a t zero angle of attack
a t the surface of a profile in this series having a 10% thickness ratio.
10.10. In wind tunnel testing the "interference" velocity is defined as the
perturbation velocity at a certain point (usually on the profile) minus the per-
turbation velocity which would exist a t the same point if there were no tunnel
walls.
For incompressibk, two-dimensional flow it may be shown that. a t least
approximately,
where 1 is the profile chord, 2h is the height of the tunnel, and Au is the "inter-
ference" velocity.
From this and the similarity rules demonstrate that the effect of compressi-
bility on the interference velocity for a profile of fixed shape in a tunnel of fixed
10.11. Test Gothert's rule for flow inside two-dimensional ducts against
results obtained from a one-dimensional analysis according to the type of
argument presented in Art. 10.9. Assume that two long sections of straight
duct having diierent cross-sectional areas are joined together. Defining
~ 1.05. Compare the results

plot Cp/C,' versus M I for a fixed ratio A I / A =
thus obtained with the corresponding results predicted by Gothert's rule.
10.12. Consider the two-dimensional flow past an elliptical cylinder a t zero
angle of attack.
For incompressible flow the maximum velocity at the surface of the ellipse is
+
related to the free-stream velocity by Vm,/Vm = 1 6, where 6 is the thick-
ness ratio.
From this information, compute the lower critical Mach Number (the value
of M, for which sonic velocity is first reached at the profile surface) for an ellipse
with 6 = 0.15, using Gothert's rule, the Prandtl-Glauert rule, and Laitone's
modification of the latter.
Compare your result with the nearly exact result of &,,it = 0.771 found by
the Rayleigh-Janzen method (cf. Chapter 12).
Chapter I1
HODOGRAPH METHOD FOR TWO-DIMENSIONAL,

SUBSONIC FLOW

It has already been observed that the nonlinearity and complexity of
the partial differential relation (Eq. 9.34) for the velocity potential cp as
a function of x and y make it very difficult to obtain solutions for subsonic
flow. One of the few simple approaches is the approximate linearized
method of the preceding chapter, leading to the Gothert and Prandtl-
Glauert rules. Other techniques are discussed in Chapter 12, including
the Rayleigh-Janzen method, numerical relaxation methods, etc., but
these are all so cumbersome that they have been used thus far only t o
solve in a nearly exact manner a few simple problems which are useful
for indicating typical results and as a basis of comp?rison for approxi-
mate methods.
The hodograph method is a powerful mathematical approach t o
problems of two-dimensional flows which are either subsonic or mixed
subsonic-supersonic. The term hodograph stems from a choice of the
velocity coordinates (i.e., u and v in Cartesian coordinates, or V and e in
polar coordinates) as the independent variables, in place of the physical
coordinates x and y which have hitherto been employed. One great
advantage of this change in variables is that the differential relation
for p or J. as a function of V and 0 is linear, thus allowing complex
solutions to be formed by linear superposition of elementary solutions.
If, in addition, an approximate form of the isentropic pressurdensity
relation for the gas (the tangent-gas relation) is adopted, the hodograph
equations are reducible to the Laplace equation, thus making them
accessible to the powerful tool of analytic functions of a complex variable.
Relation Between Physical and Hodograph Streamlines. The nature
of the transformation between the physical and hodograph planes and
some of the difficulties of the hodograph method may be illustrated by
considering the flow past the symmetrical profile of Fig. 1l.la. Two
streamlines are identified-the boundary streamline 1-2-3-4-5, and an
external streamline 6-7-8. The images of these streamlines in the hodo-
graph plane, Fig. l l . l b , are sketched in by noting that V is the resultant
magnitude of the velocity a t a point and that 0 is the slope of a stream-
336
line. Along 1-2, for example, 8, for reasons of symmetry, is zero, but
the velocity decreases from its free-stream value to zero at the stagna-
tion point. Along 2-3 the speed increases, while the angle 8 first increases
A Streamline
2,
4
Streamline
..
(a) (b)
FIG.11.1. Corresponding streamlines in:
(a) Physical plane.
(b) Hodograph plane.
and then decreases until it is zero a t 3. A continuation of this reasoning

leads to the streamline shapes of Fig. 1l.lb.
Several difficulties are now evident, namely:
(i) The entire physical flow a t infinity (points 1, 5, 6,8, etc.) is mapped
into the singular point 1 of the hodograph plane. Furthermore, all
stagnation points (2, 4, etc.) are mapped as a single point in the hodo-
graph plane.
(ii) The entire upper half of the physical plane maps into a finite
region of the hodograph plane, and this latter region represents also the
lower half of the physical plane.
(iii) By far the most serious problem is that the streamline patterns
in the two planes are totally dissimilar. Therefore, given a complete
solution in the hodograph plane, the physical streamlines can be found
only by a laborious transformation. For practical problems, given a
body shape in the physical plane, it is very difficult to obtain a solution
for the given boundaries by the hodograph method.
Scope of This Chapter. Although the hodograph method is almost
unique in providing analytical flow patterns involving both subsonic and
supersonic flow, we shall reserve the discussion of mixed flow to Volume
11, Chapter 20, and consider now only subsonic flow.
Furthermore, we shall in this chapter discuss in detail only the
Karman-Tsien pressure-correction formula for subsonic flow. Other
procedures for subsonic flow using the hodograph variables are too
lengthy to be treated here, and will be mentioned only briefly.
We shall assume throughout the chapter that the flow is two-dimen-
sional, steady, frictionless, and irrotational.
Additional material relevant to the subject matter of this chapter may
be found in Chapter 12 and in Volume 11, Chapter 20.
HODOGRAPH METHOD Ch. 11
NOMENCLATURE
see Fig. 11.10 Cartesian coordinates

speed of sound complex coordinate vari-
pressure coefficient +
able, x iy
critical pressure coefficient, conjugate complex coor-
corresponding to M = 1 dinate, z - i y
peak negative pressure co-
efficient thickness ratio
drag coefficient complex variable (see Fig.
6, imaginary operator 11.10)
for complex variable direction of velocity vector
ratio of specific heats see Eq. 11.27
Mach Number mass density
free-stream Mach Number see Fig. 11.10
for which local sonic velocity potential
velocity is first attained stream function
pressure
Cartesian components of denotes stagnation state
velocity vector denotes free-stream condi-
velocity tions
complex velocity, u iv+ denotes incompressibleflow
conjugate complex veloc-
ity, u - iv
11.2. Derivation of the Hodograph Equations

Our first task is t o derive the differential equations representing the
governing physical laws in terms of V and 0 as independent variables.
Introduction of Potential a n d Stream Functions. The condition of

irrotationality is satisfied through the introduction of a velocity potential
function (see Chapter 9) :
Similarly, t h e equation of continuity is satisfied through the intro-

"4 duction of a stream function (see Chapter 9):
v = - ( P o / P ) ( ~ ~ x=) -(p0/p)fi2 (11.2)

Polar Coordinates in Velocity Plane. It is
convenient t o use polar coordinates for the ve-
locity, as illustrated b y Fig. 11.2. T h u s
FIG. 11.2. Velocity
coordinates. u = V cos 8 : v = V sin 0 (11.3)
Art. 11.2 DERIVATION OF THE HODOGRAPH J3QUATIONS 339
Now, noting that p and 1C. are functions of x and y, and making use of
Eqs. 11.1 and 11.3, we may write
dp = cpz d x+ cp, d y = V(cos e d x + sin 0 d y ) (1 1.4a)
d# = +, d x + $C.yd y = ( p / p 0 ) V ( - sin e d x + cos e d y ) (11.4b)
Elimination of Physical Coordinates. Considering these relations as
two simultaneous equations for d x and d y in terms of d p and d*, we may
by elimination solve for d x and d y , thus obtaining
cos e po sin 0
dx = -d p - --d* (11.5a)
v P v
sin 0
dy = -drp
v + --d*cosv B
Po
P
Now, V and e are functions of x and y. Since p and $ are also func-
tions of x and y, it follows that p and J. may be considered to be func-
tions of V and 0. Thus, we may write
Next, we substitute d p and d$ from Eqs. 11.6 into Eqs. 11.5:
dx = (6
cos 0 po
vv - --$v)
P
sin
v
e
d v + (-cosV e PO -
P
Sme
v
h )dB (ll.7a)
sin 0 Po cos e sin 0

dy = ( y ~ Y + - - tPv ) v
d~+ (Tpe+- P v
*+e) d~ (1i.m)
Since V and e are independent of each other,
Therefore from Eqs. 11.7 we may form

ax cos e po sin e
- = xv = -- CPV- --rLv (11.8a)
av v P v
ax cos e po sin e
- = xg = -w - - - $ 8 (11.8b)
ae v P v
a~
-=
sin 0 Po cos e
Y v =-pv+--*v
av v P v
340 HODOGRAPH METHOD Ch. 11
We may now eliminate x and y from these equations by noting that
=a2x_ _a2x . a2y --a2y

--
avae a e a v ' aeav avae
The derivative of Eq. 11.8a with respect to 0 is therefore set equal to
the derivative of Eq. 11.8b with respect to V, and similar operations are
performed on the y-derivatives. Thus we obtain after simplification
(noting that p depends only on Tr),
cos 0 po sin 0 sin 0 po cos 0

-pv - ----$v = - -- ve - --$6
V P v v2 P v2
Now we multiply Eq. 11.9a by cos 0 and Eq. 11.9b by sin 8, and add,
thus getting
ve = ( P O / P ) ~ + V (11.10)
Similarly, we multiply Eq. 11.9a by sin 0 and Eq. 11.9b by cos 8, a.nd
subtract, thus getting
Introduction of Dynamic Equation. Since the flow is isentropic, the

pressure is uniquely related to the density through the speed of sound,
i.e., dp/dp = c2. Also, Euler's equation for frictionless, irrotational
motion is dp = -pV dV. With these relations we now form the expres-
sion
d PO dp PO dp dp
_ - _ - = _ - _ _ = PO v
P dp dV P P c2
(11.12)
Substituting this into Eq. 11.11, we have, after rearrangement,
Art. 11.3 THE TANGENT-GAS APPROXIMATION 341
To eliminate cp from Eqs. 11.10 and 11.13, we set 9ev = cpvb Thus,
noting that p and M depend only on V , and using Eq. 11.12,
Differential Equation of the Stream Function in the Hodograph Plane.

Equating the two preceding equations, and simplifying, we finally get
+
which is the desired relation giving as a function of V and 8, since c is a
function of V only. This differential equation is linear in the dependent
variable +.
Procedure for Obtaining Physical Flow Pattern from Hodograph
+
Solution. Assuming that a function = +(V, 6) has been found which
satisfies Eq. 11.14, the procedure for determining the flow pattern is as
follows:
(i) The derivatives +v and +o as functions of V and 8 are found by
+
direct differentiation of = +(V, 8).
(ii) Using Eqs. 11.10 and 11.13, cpe and cpv are found as functions
of V and 8.
(iii) Eqs. 11.7 are integrated to give x and y as functions of V and 8.
Since we started with a solution for $ in terms of V and 0, we thus have
+ in terms of x and y. That is, we have the physical streamlines. If
these physical streamlines represent a practical problem of interest, the
solution is useful-otherwise not. The great disadvantage of the hodo-
graph method may now clearly be seen, namely, that the problem of
flow past given boundaries is not susceptible to direct treatment by the
hodograph method.
11.3. The Tangent-Gas Approximation

The principal application of Eq. 11.14 which we shall discuss in this
chapter is the Karman-Tsien pressure correction formula. This is
arrived a t through the use of an approximate pressure-density relation,
the use of which allows us to construct compressible flows from known
incompressible flows.
Definition of Tangent Gas. Referring to Fig. 11.3, the p-p relation

which we shall call the tangent gas approximation is defined by drawing a
tangent to the true isentropic curve of p versus a t the point repre-
senting free-stream conditions, pm and p,.
FIG.11.3. Illustrates definition of tan- FIG. 11.4. Limiting velocity for tan-
gent gas. gent gas.
The straight-line relation between pressure and specific volume has a

constant slope which may be written (remembering that dp/dp = c 2 ) as
- dp 2 dp 2 2 -
= p - = p c - pm2cm2 = pO2co2 = constant (11.15)
d(llp) dp
Separation of variables and integration of Eq. 11.15 now yields the

tangent-gas pressure-density relation in terms of free-stream properties:
Dynamic Equation for Tangent Gas. Combining Euler's equation of

motion with Eq. 11.15, we get
Integrating, we get
from which i t follows that

V2 - c2 = va2- cm2 = constant (11.17a)
Art. 11.3 THE TANGENT-GAS APPROXIMATION 343
Furthermore, since V is zero a t the stagnation state,
Now, combining Eq. 11.17 with Eq. '11.15 in various ways, we find
Equations 11.15, 11.16, 11.17, and 11.18 give the relations between the
various fluid properties for the tangent gas, and replace the customary
isentropic relations. The accuracy of these formulas of course depends
on how close the fluid properties in the flow field are to the free-stream
properties.
Let us see what happens during an acceleration of this fictitious gas.
As V increases, Eq. 11.18b shows that p decreases. But Eq. 11.17 shows
that c increases! From Eq. 11.18a, on the other hand, M increases, but
can never exceed unity. Hence the approximate methods outlined
here are applicable only to subsonic flow.
Limitations of Tangent-Gas Approximation. Another way of examin-
ing the limitations of the tangent gas is to note that for sufficiently
large accelerations the pressure may become negative. We proceed to
determine when the latter condition is reached. Setting p = 0, Eq.
11.16 becomes
Pm - Pm
2
1
Cm Pm P
But, a t the point of tangency, the true isentropic relations may be

employed. Thus, cm2 = k p m l p m , and the preceding relation become?
simply
Pm
- -1 1 P,
. - -- -k+l
P k' P k
Combining this expression with Eqs. 11.18~and 11.18a1we have
Then, forming V,,/Vm, we obtain

which shows, for any given values of M, and V,,, the maximum local
velocity which can be reached without the attainment of negative pres-
sures. The relation is shown graphically in Fig. 11.4. For thin profiles,
it is known that the maximum local velocity seldom exceeds twice the
free-stream velocity. Fig. 11.4 shows, therefore, that the possibility
of the pressure becoming negative is not often an important Iimitation.
11.4. The Karrnan-Tsien Pressure Correction Formula

Differential Equation for Case of Tangent Gas. Multiplication of
Eq. 11.14a by VP/po(l - M2) gives the differential relabion for J. in
terms of V and 6 in the form
For the tangent gas, however, Eq. 11.18a shows that
We introduce this relation into Eq. 11.20, and thus obtain, subject to
the tangent-gas approximation,
Introduction of Transformed Incompressible Velocity. By means of

a transformation, this equation may be reduced to that of an incompress-
ible flow. We define a new variable, Vi,which is a function only of the
velocity V, through the relation
Since Vi and V are uniquely related through Eq. 11.22, Eq. 11.21 may
now be expressed in terms of Vi rather than V. Noting that, according
to Eq. 11.22,
7
d1 - M - d
dV dVi
and replacing V-derivatives in Eq. 11.21 by Vt-derivatives, we get

a
Art. 11.4 KARMAN-TSIEN PRESSURF, CORRECTION FORMULA 345
Incompressible Flows as a Basis for Constructing Compressible Flows.

Eq. 11.23a is evidently identical in form with Eq. 11.21 if we set M = 0
in the latter relation. Hence Eq. 11.23a is the differential equation, in
hodograph coordinates, for an incompressible flow with the velocity
(V;, B), and its solution, J.i = J.i(V;, 8), therefore represents the flow
pattern of an incompressible fluid.
The significance of this result is that i t now allows us to obtain results,
a t least approximately, for given boundary conditions (e.g., a, specific
profile) in the physical plane. Suppose we have the solution for incom-
pressible flow past the given profile, J.i = J.i(xi, yi). From this we may
find the incompressible velocity components in terms of xi and yi, and
so we have the incompressible solution in hodograph coordinates,
+i = Gi(Vi, 8). But, if the latter is a solution of Eq. 11.23a, a function
+(V, 8) of the same form must also be a solution of Eq. 11.21 because
Eqs. 11.23a and 11.21 are transformations of each other through the
medium of Eq. 11.22. Note that the compressible solution is found
from the incompressible solution merely by replacing Vi by V, ie., we set
From the compressible solution, J. = $(V, 8), the profile shape in.the
physical plane may be found by integration of Eq. 11.7. The compress-
ible-flow profile thus found will, in general, be diierent from the original
incompressible-flow profile. However, we can a t least make them
similar by insuring that the compressible solution merges continuously
into the incompressible one as M, approaches zero. This is done by
integrating Eq. 11.22 with such limits that V + V i as M, 4 0.
Two distinct steps are involved in this method: first, finding the re-
lation between the corresponding velocities in the compressible and
incompressible flows; second, finding the relation between the corre-
sponding profile shapes. Let us proceed now to the determination of
the relation between the corresponding velocities.
Relation Between Corresponding Velocities. Combining Eq. 11.21
with Eqs. 11.18a and 11.18b, we have,
dV; p dV dV cn dV
Integrating, and letting K be the constant of integration, we have
log vi = - log co +7
co + v 2
+ log K
v
Now, for reasons given previously, we let V = Vi when V / Q = 0.

Thus we get K = 2co, and, therefore,
Solving explicitly for V in terms of V; we obtain
We would like to obtain the relation between V and Vi in terms of Mm

rather than Q . To do this, we proceed as follows. In terms of free-
stream conditions, Eq. 11.24b may be written
Dividing Eq. 11.24b by Eq. 11.25, we obtain

v;
- (1 - A)
Vi
- Vim
2
Vim
Vim
(11.26)
where, for brevity, we have defined
Now, combining the latter relation with Eq. 11.24a written for free-
stream conditions, we have
But, from Eq. 11.18a,

) ~ 1 - Mm2
( C ~ / C ,=
and thus we obtain after simplification,
Eqs. 11.26 and 11.27 taken together comprise the relation between the
corresponding compressible and incompressible velocities. We now
come to the final and most important step, namely, finding a relation
between the corresponding pressure coefficients.
Art. 11.4 KARMAN-TSIEN PRESSURE CORRECTION FORMULA 347
Pressure-Coefficient Correction Formula. Using Eq. 11.16, the

pressure coefficient becomes
Dividing Eq. 11.17a by cm2,and introducing Eq. 11.15, we form
Solving for p,/p, and substituting in the formula for C,, we get
Now, for incompressible flow, Bernoulli's equation yields
cpi= 1 - (V;/Vim)2
Therefore, Eq. 11.26 may be written as
Substituting this expression for V/Vm into the previous relation for
C,, and simplifying, we finally obtain the Karman-Tsien pressure-cor-
rection formula :
17
Application of Karman-Tsien Rule. As mentioned previously, the

two pressure coefficients in this formula correspond to slightly diierent
profiles, and so do not strictly indicate the influence of Mach Number for
a profile of given shape. However, the distortion of the profile during
the transformation is usually very small, and Eq. 11.28 is commonly
employed to estimate the effect of Mm on the pressure coefficient for a
profile of fixed shape. It is the most widely accepted pressure-correction
formula for two-dimensional, subsonic flow.
Fig. 11.5a is a graphical version of Eq. 11.28, showing how Cp varies
with M, for various values of CPi. Also indicated are lines of constant
Fro. 11.5. (a) Variation of C with M,
according to Karman-Tsien rug.
(b) Corresponding curve of lower crit-
ical Mach Number versus maximum neg-
ative pressure coefficient for incompres-
sible flow.
Peok Suction Pressure Coefficient

For !ncompressible Flow.
Art. 11.4 KARMAN-TSIEN PRESSURE CORRECTION FORMULA 349
local Mach Number, based on the isentropic relation between Mach

Number and pressure ratio, from which may be found
Lower Critical Mach Number. Where the line for a particular value
of Cpi intersects the line marked M = 1 (sonic line) in Fig. 11.5a1there
may be read on the horizontal scale the value of M, for which the sonic
velocity is reached locally a t the point having the particular low-speed
coefficient C,,. It is known from experience that soon after supersonic
speeds are reached locally, there are usually rapid increases in the drag
coefficient, and the corresponding value of M, is called the lower critical
Mach Number, M, ,. Fig. 11.5b, which is based on the Karman-Tsien
pressure-correction curves of Fig. 11.5a, shows M, ,,as a function of C,.
Given the peak negative pressure coefficient for a profile a t low speeds,
Fig. 11.5b shows approximately the free-stream Mach Number a t which
compressibility effects begin to have a serious adverse effect on drag.
Comparison with Prandtl-Glauert Rule. For very small values of the
pressure coefficient, that is, for very small perturbations from free-
stream conditions, Eq. 11.28 reduces to the Prandtl-Glauert rule,
c, = cp,/l/l--M,2.
When C, is negative, corresponding to regions of velocity higher than
free-stream velocity, the Karman-Tsien (K-T) formula indicates a
greater influence of compressibility than does the Prandtl-Glauert (P-G)
formula. As a result, the K-T rule leads to a value of the lower critical
Mach Number smaller than that computed by the P-G rule.
For positive values of C,, on the other hand, the I<-T rule predicts
a compressibility effect which is smaller than that predicted by the P-G
rule. This is not of great practical importance, however.
Since the effects of compressibility depend on the local pressure coeffi-
cient, the K-T rule does not allow the derivation of a simple relation
showing the effects of compressibility on the force coefficients of a given
profile. If low-speed pressure distributions for the profile are available,
however, these may be corrected to higher Mach Numbers by the K-T
formula, and then, by integration, the high-speed force coefficients may
be found.
Accuracy of Kannan-Tsien Pressure-Correction Formula. The
reliability of the Karman-Tsien pressure-correction formula may be
tested by (i) comparing it with experimental data, or (ii) comparing i t
350 HODOGRAFH METHOD Ch.11
with more accurate analytical results for simple cases where this is
possible.
Fig. 11.6, which is a typical comparison (2) between experimental data,
the K-T rule, and the P-G rule, indicates good agreement between the
measurements and the K-T rule, but shows the P-T rule t o underesti-
mate the effect of compressibility.
FIG. 11.6. Comparison of Karman- FIG. 11.7. Comparison of Karman-

Tsien rule with experiment. Tsien rule with exact theoretical result
(Kaplan) for flow over a bump.
Fig. 11.7 shows excellent agreement between (i) exact analytical

results for the maximum negative pressure coefficient as a function of
& for flow over a bump (4) and (ii) the predictions of the Karman-Tsien
formula as to the shape of the curve of (-C,),,, versus M,.
An examination of the assumptions underlying both the Prandtl-
Glauert and Karman-Tsien rules shows that both rules are limited bv"
the restriction of small perturbations, and hence can reasonably be
expected to be reliable only for thin profiles a t small angles of attack.
However, the P-G rule is also limited to small Mach Numbers, whereas
the K-T rule is not so limited. Hence, as verified by experimental and
analytical results, the K-T rule may be expected t o be more accurate
than the P-G rule a t high subsonic speeds.
Fig. 11.3 shows that in the range of negative pressure coefficients, i.e.,
the range of greatest practical significance, the tangent gas produces a
lesser density change for a given pressure change than does the true
isentropic. It would seem, therefore, that the Karman-Tsien rule should
underestimate the effects of compressibility. However, as discussed
below, the coordinate correction in the corresponding flows is such as to
make the compressible profile thicker than the corresponding incom-
pressible profile. Thus the compressibility effect would be overestimated
if we took no account of the coordinate correction in comparing the
compressible flow with its corresponding incompressible flow. By apply-
Art. 11.5 CALCULATION OF PROFILE SHAPE CORRECTION 351
ing the K-T formula to profiles of fixed shape we thus produce two errors
in opposite directions, and the experimental evidence seems to indicate
that they nearly cancel each other.
11.5. Calculation of Profile Shape Correction

Incompressible flow patterns may, for the tangent-gas approximation,
be transformed to corresponding compressible flow patterns as indicated
by Eq. 11.23b. The correspond-
ing velocities in the two flows are Compressible
related by Eq. 11.22, from which ?;;Ff
we derived Eqs. 11.26 and 11.28. w=u+iv =Ve
It now remains to find the corre-
sponding streamline patterns in z=x+~)-
the physical plane, or, in other X u
V
words, to find the relation between
the incompressible and compressi- Hodogroph
ble profiles. The relation between
the various planes in the corre- .#.=". +iv.=Veie
v,
sponding flows is shown by
Fig. 11.8.
Xi Ui
Introduction of Complex Vmi- FIG.11.8. Complex physical and veloc-

ables, The calculations are greatly ity planes for incompressible and corn-
pressible flow.
abbreviated through the use of
complex variables. Let the locations of corresponding points in the
compressible and incompressible physical planes be represented respec-
tively by the complex quantities
z=x + iy; 2; = xi + iyi
where i = 6. Also, after noting that the flow direction e is the
same for the corresponding flows, let the corresponding velocities be rep-
resented by the complex quantities
w =u+iv=Ve a
w; = u; + iv; = Vie i0
Now, for incompressible, steady, irrotational, two-dimensional flow,

the velocity components may be written in terms of the velocity poten-
tial and stream function as
ui= arpi/azi= alc.i/ayi
These may be recognized as the Cauchy-Riemann equations. We know

from the theory of functions that the incompressible complex potential,

pi+ i+;,may accordingly be expressed as an analytic function of zi, i.e.,
It is also well known (5) that the conjugate complex velocity is then
given by
Since this relation gives fii as an analytic function oi(zi) of zi, it follows
that z; is an analytic function of fii. Then, from Eq. 11.29a, pi i+i +
must be expressible as an analytic function of fii:
By changing the signs of the imaginary parts of Eqs. 11.29 and 11.30,
we get -
qi - iGi = P(&) = G(wi) (11.31)
and
w; = ui +
ivi = Vieie = dF/di$ (11.32)
Now Eq. 11.23b tells us that an incompressible solution 4i(Vi, 8) = 0
yields a corresponding compressible solution +(V, 8) = 0 if we merely
replaced Vi by V in the former equation, provided that Vi and V are
connected by Eq. 11.24. Noting that 8 is constant in the transformation,
it follows that
9 + i+= G(fi); p - i+ = G(w)
It seams reasonable to suppose that the functions G(o) and G(w) may be
respectively approximated with good accuracy by the functions G(fii)
and G(wi), inasmuch as w and wi are nearly alike. Hence we write
In Eqs. 11.5 there appears the term Po/P, for which we now write with
the help of Eqs. ll.18b and 11.24b,
Coordinate Correction Formula. Having set down these preliminaries,

we now rewrite Eqs. 11.5 in the following form, using Eq. 11.24b to
substitute for V, and Eq. 11.35 to substitute for pojp:
4cO2- Vi2 4c02 f vi2
dx = cos I9 dv - sin 19d#
4 ~ 0 ~ ~ ; 4c02vi
4cO2- 4c02 + Vi2
dy =
k o 2 Vi
+
sin I9 drp
4c02Vi
cos I9 d#
With these relations we now form, noting that (cos I9 + i sin 8) = e",
Now, making use of Eqs. 11.29 to 11.34, this may be brought into the
fo m
Finally, noting that X = (Vim/2c,J2, we get
where X is given in terms of M, by Eq. 11.27.

Eq. 11.36 is the desired relation connecting corresponding points in
the two physical planes. Upon integration, and noting that the con-
stant of integration is immaterial to the shape
.
of the flow pattern, we have
i i
Thus, the compressible physical coordinates

are equal to the corresponding incompressible
physical coordinates plus a correction term
proportional to A. As shown by Fig. 11.9, Mm
is small, especially a t moderate Mach Num- FIG. 11.9. The factor x
in the Karman-Tsien co-
bers, and goes to zero as M, approaches zero, ordinate correctionformula.
thus proving that the compressible flow pat-
-
tern merges gradually into the incompressible flow pattern. In specific
examples the correction term usually proves to be very small, and, con-
sequent!~,there is little difference in profile shape between the corre-
sponding flows.
The complete Karman-Tsien method may now be summarized thus:
If the pressure distribution in incompressible flow for a given profile is
known, then a corresponding compressible flow a t any other Mach

Number, &, may be constructed, using Eq. 11.37 as a coordinate
correction formula and Eq.11.28 as a pressure correction formula to
relate the location and pressure coefficient a t corresponding points in
the two flows.
Application of Shape-Correction Formula to Flow Past Elliptical
Cylinder. To illustrate the procedure for profile correction, let us con-
sider the incompressible flow past the ellipse of Fig. 11.10a. The incom-
pressible flow past an ellipse in the zi-plane may be found from the in-
FIG.11.10. Transformation of a circle to an ellipse (incompressible flow).
compressible flow past the circle of radius b in the {%-planeof Fig. 11.10b
through the conformal transformation
To prove this, note that, a t the surface of the circle of radius b, the l,-
coordinate is 3.2 = bei", and hence
Separating real and imaginary parts, we have
z i= ()+ k) oos u; y;=

(b 3
- - sin o
and, eliminating u through the relation, sin2 a + cos 2

u= 1, we get
This is the equation of an ellipse in the zi-plane, with thickness ratio
The complex potential for the flow past the circle in the r,-plane is
doublet:
F(li) = vim (ri + z)
well known c5) and is given by the superposition of a uniform flow and a
Changing the sign of the imaginary part gives the conjugate complex
(11.4Oa)
potential,
p(fi) = Vim (11.40b)
Now, since z; and Ci are uniquely related, we may write, after taking
the derivatives of Eqs. 11.38 and 11.40b,
Taking the differential of Eq. 11.38, we have also
Inserting the foregoing relations into Eq. 11.37, we find
1 - -2
Ti
Carrying out the integration, we get
Since ri is related to z;, i t is evident that any point of the incompress-

ible (zi) flow yields a corresponding point (2).of the compressible flow.
We are of course particularly interested in the distortion of the elliptical
profile. Accordingly, inserting into Eq. 11.41 the coordinates of the

profile in the <a-plane, as listed helow,
ci' = be" = b(cos a + i sin a)
1 . 1
= - e-'c = - (cos a - i sin a)
b b
fi' = be-i" = b(cos a - i sin a)
1 . 1
l/f[ = - em = - (cos u
b b
+ i sin a)
and taking the real and imaginary parts of the resulting expression, we
obtain the x- and y-coordinates of the profile in the compressible flow:
x' =
(b + 3cos- u
(b2 - 1)' (b2 - 1)2 + 4b 2

sin2 a
b(l + b ) cos a +
2
4
In
b2 + 2b cos u + 1 1
y1 = (6 - i) sin r
(b2 -
2
b ( l - b ) sina + 2
tan-'
2b sin a
b2 - 1 I
-
The compressible profile is therefore approximately elliptical if X is
small. To find the half-chord, x',,,,
thickness, y',,,,
we set u = 0; and to find the half-
we set a = ~ / 2 . By taking the ratio of the two, we
find, after rearrangement, the thickness ratio of the compressible profile:
"I
b(b2 - 1)
tan-'
b2 - 1
&
6 =
'- Lb2 +
b(b2-1)2
2(b2 1) +ln
b- 1
b 1
-1+ (11.42)
Analysis of Eq. 11.42 shows that 6 > ail which means that the trans-
formed compressible profile is slightly thicker than the original incom-
pressible profile. For example, a t M, with ai = 0.095, Eq. 11.42 yields
6/6i = 1.02. This seems to be generally the case for other examples
treated in the literature. It appears, therefore, that if profile distortion
is neglected in applying the K-T method, the effects of compressibility
are exaggerated, which is exactly opposite to the sense of the error owing
to the linear pressure-density relation.
Art. 11.7 MISCELLANEOUS EXAMPLES 357
11.6. Extension of Karman-Tsien Method

Flows with Circulation. One of the weaknesses of the Karman-Tsien
method as given here is that it is restricted to flows past closed bodies
without circulation. In the case of a profile with circulation, the coordi-
nate transformation changes a closed profile into one which is not
closed.
This difficulty was resolved by Lin,(6)among others, who devised a
logical extension of the Karman-Tsien method which not only may be
used for flows with circulation but has the further advantage of leading
to the calculation of the compressible flow past a given profile with about
the same degree of difficulty as for incompressible flow past a given
profile.
For details of the method, the reader is referred to References 6 and 3.
It is worth mentioning here, however, that Lin's extension shows that
the Karman-Tsien pressure-correction formula (Eq. 11.28) is equally
applicable to flows with circulation as to flows without circulation.
A det,ailed application of the method for determining the airfoil shape
having a prescribed velocity distribution is given in Reference 7.
Polygonal Approximation Method. A natural extension of the

Karman-Tsien method involves using for the tangent-gas approxima-
tion not one but several straight-line segments in the pressure-volume
plane. The method was suggested and developed by Poritsky.@)
11.7. Miscellaneous Examples

The mathematical details involved in the application of the hodograph
method with the tangent-gas approximation to various examples are too
lengthy to be given here. For these details the reader is referred t o the
original papers and to Reference 3. However, it seems desirable to
summarize briefly those results of examples worked out in the literature
which are of greatest practical interest.
Flow Past Circular Cylinder in Two-Dimensional Wind TunneL

Suzuki (9) has worked out the calculations for the physical arrangement
of Fig. 1 l . l l a . Perhaps the most significant effect to be looked for here
is the influence of wind tunnel interference on the lower critical Mach
Number, M, .,,that is, the value of M, a t which sonic velocity is first
reached a t the surface of the profile. This effect, based on the Karman-
Tsien pressure-correction formula, is shown in Fig. ll.llb, wherein it
may be seen that a 10 per cent blockage of the channel reduces the lower
critical Mach Number from 0.40 to 0.393. Also shown is the difference
in the calculated results produced by ignoring the coordinate correction
in comparing the compressible and incompressible flows. For reasons
given previously, i t may be desirable to ignore this correction.
FIG. 11.11. Theoretical effect of wind tunnel walls on lower critical Mach Number
for circle (after Suzuki).
Elliptic Cylinder at Angle of Attack with Circulation. This case has

been worked out by the Karman-Tsien method in Reference 10, wherein
it was found that as a result of the coordinate
I
! correction the compressible profile has a
Mmcr 5 / i
-6 -
slightly greater thickness ratio but a slightly
smaller angle of attack. The most interest-
e I
ing results are shown in Fig. 11.12, where
the lower critical Mach Number is shown as
a function of thickness ratio and angle of
0
o I attack. For thin profiles, e.g., for thickness
Thickness Ratio ratios of the order of 0.10, M, ,,is very sensi-
G 11.12. Theofetical tive to angle of attack. For thick profiles, e.g.,
effect of thickness ratio and
angle of attack on lower for thickness ratios of the order of 0.5, the
critical Mach Number for critical Mach Number is insensitive to thick-
ellipse (after Tamatiko and
Tamada). ness ratio and to angle of attack.
Contraction Ratio for Jet from Sharp-Edged Two-Dimensional Ori-

fice. A jet escaping from a two-dimensional slit, Fig. 11.13a, was
studied by B u ~ e m a n n , ( ~ ~ ) the tangent-gas approximation. At
using
infinite distance from the slit, the streamlines become uniform and paral-
lel, the Mach Number is Mz, and the jet width is bz. The entire free
surface of the jet is at uniform pressure and a t the speed Ma. For
practical purposes, the asymptotic conditions are substantially arrived
at within a few jet widths of the slit.
Art. 11.7 MISCELLANEOUS EXAMPLES 359
7fs ....,
.,.... .
,..
Me
(a) (b)
FIG.11.13. Theoretical coefficient of contraction for jet escaping from sharp-edged
slit (after Busemann).
The contraction ratio of the jet was found t o be
If we evaluate p21Po for the tangent gas, we have
whereas if we use the exact isentropic relation for p 2 / p o , we get
Both these relations are plotted in Fig. 11.13b, where there is plotted
also the results of typical experiments ('4) on round sharp-edged orifices.
The theoretical results follow the
experimental results in a qualita-
tive way. Both indicate a sub-
stantial rise in discharge coefficient
owing t o compressibility effects.
Elbows for Accelerated Mows.
A method for designing passages
-
which simultaneously turn and ac- 11.14, turning channel
celerate the flow, Fig. 11.14, has (after Carrier).
been developed by based on the tangent-gas approximation.

The passages have the following desirable properties:
(i) The velocity increases monotonically along each streamline,
thus reducing the possibility of boundary-layer separation and
transition to turbulence.
(ii) Each streamline turns monotonically.
(iii) The ratio of total length of bend to exit width is a minimum,
thus reducing frictional effects and space requirements.
Drag of Flat Plate Placed Normal to Gas Stream. I n a pioneer paper
on the hodograph method, Chaplygin (I3) solved the problem of a two-
dimensional flat plate immersed normal to the flow in an infinite gas
stream, assuming as in Fig. 11.15a a surface of discontinuity extending
downstream from the edges.
2.0, I I I I 1
(a) Mm
(b)
FIG.11.15. Theoretical drag of flat plate placed normal to stream (after Chaplygin).
The calculated drag coefficient of the plate, Fig. 11.15b, begins a t

0.88 for incompressible flow, and increases to 1.16 a t Mach Number
unity. Although the calculated coefficient is less t,han half the measured
coefficient for flat plates, the rise in drag coefficient with Mach Number
is similar to the observed increase in drag for blunt bodies. The dashed
portion of the curve is an approximate extrapolation to supersonic
speeds, and also agrees well with the general shape of the experimentally
determined curves of CD versus M, for blunt projectiles.

1. TSIEN, H. S. Two-Dimensional Subsonic Flow of Compressible Fluids, Jour.
Amo. Sci., Vol. 6, No. 10 (Aug., 1939), p. 399.
2. VON K ARMAN, TH. Compressibility Effects in Aerodynamics, Jour. Aero. Sci.,
Val. 8, No. 9 (July, 1941), p. 337.
3. EHLERS, F. E., and CARRIER, G. F. Monograph IV-Methods of Linearization
in Compressible Flow. Part 11: Hodograph Method, Air Material Command
Report, NO. F-TR-1180B-ND (Feb.. 1948).
4. KAPLAN, C. The Flow of a Compressible Fluid Past a Curved Surface, N A C A
Rep., No. 768 (1943).
5. PRANDTL, L., and TIETJENS, 0. G. Fundamentals of Hydro- and Aeromechanics,
6. LIN, C. C. On an Extension of the van Karman-Tsien Method to Two-Dimen-
sional Subsonic Flows with Circulation Around Closed Profiles, Quarterly App.
Math., Vol. 4, No. 3 (Oct., 1946), p. 291.
7. COSTELLO, G. R. Method of Designing Airfoils with Prescribed Velocity D i s
tributions in Compressible Potential Flows, N A C A Tech. Note, No. 1913
(Aurr..
. - , 1949).
8. PORITSKY,H. 'Polygonal Approximation Method in the Hodograph Plane,
Jour. A p p . Mech., Vol. 16, No. 1 (June, 1949), p. 123.
9. SUZUKI,H. On the Subsonic Flow of a Compressible Fluid Past a Circular
Cylinder Between Parallel Walls, Proc. Phys. Math. Soc. Japan, Val. 23
(1941), p. 337.
10. TAMATIKO, S., and T AMADA, K. Application of the Hodograph Method to the
Flow of a Compressible Fluid Past an Elliptic Cylinder, Proc. Phys. Math.
Soc. Japan, Val. 23 (19411, p. 958.
11. BUSEMANN, A. Hodographenmethode der Gasdynamik, Z.a.M.M., Vol. 17
(1937), p. 73.
12. CARRIER, G. F. Elbows for Accelerated Flows, Jour. App. Mech., Vol. 14 (1947),
p. 108.
13. CHAPLYGIN, S. A. On Gas Jets, NACA Tech. Memo., No. 1063 (1944).
14. PERRY, J. A,, Jr. Critical Flow Through Sharp-Edged Orifices, Trans. A.S.M.E.,
Vol. 71, No. 7, p. 757.
PROBLEMS
11.1. Derive the partial differential equation for two-dimensional, steady,
irrotational, isentropic motion with V and 0 as independent variables and the
velocity potential cp as the dependent variable.
11.2. (a) Demonstrate by series expansions t h a t the exact isentropic relation
between density ratio and Mach Number may be written
whereas the corresponding expression for the tangent-gas approximation is
Determine the per cent error for M = 0.1, 0.5, and 0.8. Note from these results
that the Karman-Tsien pressure correction formula (ignoring coordinate correc-
tion) is correct u p to terms of the order of M2.
(b) Determine corresponding expressions for p/p, in terms of M and &.
11.3. It is often said that the tangent gas is equivalent t o a perfect gas
with k = - 1.
(a) Compare the various tangent-gas relations with the corresponding isen-
tropic relations for a perfect gas with k = -1.
(b) Compare the pressure-specific volume curve of the tangent gas with that
corresponding to isentropic expansion of a perfect gas having k = -1.
(c) Compare the various tangent-gas relations with approximate formulas,
found by developing the exact isentropic relations for small changes from f r e e
stream conditions.
(d) What conclusions do you reach?
362 HODOGRAPH METHOD Ch.11
11.4. (a) Demonstrate, for the tangent-gas approximation, that the stag-
nation pressure is related to the free-stream properties by the formula
For a gas like air, should this formula be used with a value for k of 1.4 or of
(- 1 Explain.
(b) Show that if a perfect gas changes state along the line of states defined
by the tangent gas, the entropy change is given by
Verify the argument that if the processes are without friction, the tangent-gas
formulas may be thought of as referring to the nonadiabatic flow of a perfect
gas. Show that if the velocity changes are such that the pressure is changing
in a direction away from the free-stream pressure, heat must be rejected from
the gas; whereas if the pressure is changing toward the free-stream pressure,
heat must be added to the gas.
11.5. Determine the isentropic relations between density, pressure, sound
velocity, speed, and Mach Number, for a tangent gas having the point of tan-
gency a t the stagnation point.
11.6. Demonstrate that in the Karman-Tsien method,
11.7. The complex potential

pi + i+i = V&(z, + b ln 3)
b
represents the superposition of a uniform, parallel flow and a source flow, and
hence yields the incompressible flow past an infinite half-body.
Using the Karman-Tsien method, determine the corresponding profile shapes
for incompressible flow and for flow a t M, = 0.5.
11.8. The incompressible flow past a circular cylinder of radius R has the
complex potential
pi +
i*i = V,,
Using the Karman-Tsien method, determine the shape of the transformed

cylinder as a function of M,.
11.9. Consider incompressible flow past an ellipse of thickness ratio 0.5.
profile for M,-

Plot on the same sheet this elliptical profile together with the corresponding
0.5 according to the Karman-Tsien method.
11.10. The incompressible flow without circulation past a circle of radius b
at an angle of attack a is given by the complex potential
Ch. 11 PROBLEMS 363
Using the mapping function between a circle and an ellipse, namely,
find an expression for the coordinates in compressible flow corresponding to the

incompressible flow past an elliptical profile, according to the Karman-Tsien
method.
If the incompressible flow is past an elliptical profile of 25% thickness ratio
at an angle of attack of loo, and if the compressible flow is at M, = 0.5, plot
on the same sheet the corresponding incompressible and compressible profiles.
Note the important result that the compressible profile is slightly thicker
but is a t a smaller angle of attack.
11.11. Compare the blockage effects shown in Fig. 11.11 with a simple one-
dimensional approach in which the profile is assumed to be in an infinite channel
at a free-stream Mach Number corresponding to the flow in the minimum section
of the actual channel.
What conclusions do you draw?
Chapter 12
MISCELLANEOUS METHODS AND RESULTS FOR TWO-

DIMENSIONAL, SUBSONIC FLOW

The purpose of this chapter is to summarize briefly a variety of im-
portant methads and results for subsonic flow, both analytical and exper-
imental, falling into categories not classified elsewhere in this book.
On the analytical side, we shall discuss certain methods which are
too lengthy for general use, but which provide nearly exact answers to
a limited number of simple problems. These exact solutions are useful
for testing the validity and range of application of such quick but ap-
proximate methods as the Prandtl-Glauert rule and the Karman-Tsien
rule. Equally important, the exact solutions are exemplary: that is,
they indicate correctly the general nature of the flow pattern for certain
simple cases, and these results are a useful guide to the thinking of the
designer for more complex problems.
The principal lines of attack underlying the theoretical methods will
be outlined briefly, without going into details. Significant results found
with each method will be summarized, again omitting laborious details.
On the experimental side, we shall present some important results
connected with drag and lift for subsonic flow. These are important for
practical purposes but are not easily connected with analytical studies.
chapter may be found in Volume 11, Chapters 21, 22, and 28.
NOMENCLATURE
c speed of sound tional, steady flow is not
CD drag coefficient possible
CL lift coefficient peak negative pressure co-
CP pressure coefficient efficient
CP.w pressure coefficient corre- ratio of specific heats
sponding to M = 1 Mach Number
C ) limiting peak negative free-stream Mach Number
pressure coefficient, for which sonic velocity
above which irrota- is first reached on profiie
3
Art. 12.2 THE RAYLEIGH-JANZEN METHOD
n empirical exponent defined 6 thickness ratio

by Eq. 12.14; also dis- P mass density
tance along equipoten- 9 velocity potential
tial line, measured away
from center of streamline ( )m signifies free-stream condi-
curvature tions
P pressure ( 18 signifies conditions at pro-
R radius of curvature of file surface
streamline ( )o signifies stagnation condi-
t characteristic thickness tions
parameter ( )i signifies incompressible
um, v m free-stream velocity flow
v velocity
2,Y Cartesian coordinates
12.2. The Rayleigh-Janzen Method of Expansion in Series of

the Mach Number
The Rayleigh-Janzen method is a method of successive approxima-
tions, applicable to either two-dimensional or three-dimensional sub-
sonic potential flows.
Outline of Method. To illustrate the procedure, let us consider two-
dimensional potential flow, for which the differential equation of the
velocity potential may be written (Chapter 9) as
where
It is well known that subsonic compressible flows about a given body

are not different qualitatively from incompressible flow about the same
body. This suggests that the exact expression for the velocity potential,
,p(x, y), of a compressible flow may be expressed as the velocity potential,
po(x, y), for incompressible flow past the same body, plus a series of
correction terms whose magnitudes would depend on the Mach Number.
Accordingly, the zeroth approximation, po(x, y), is found by setting
the right-hand side of Eq. 12.1 equal t o zero (c = m for incompressible
flow), and using classical potential theory or function theory for solving
the resultant Laplace equation for the given boundary conditions.
Having thus determined po(x, y), the right-hand side of Eq. 12.1 is
evaluated approximately on the basis of this zeroth approximation,
but retaining terms only up to the order of MW2. Let u$ call the right-
hand side, as so evaluated, Fo(x, y). The next step is to obtain the first
approximation to compressible flow, pl(x, y), by solving the linear
Poisson equation,
+
c ~ z z P,, = Fo(x1Y) (12.3)
This process is continued to get higher approximations, but becomes

prohibitively laborious, even for simple cases, after the second or third
approximation.
When the mathematical details are carried out, i t is found that the
exact solution for the velocity potential may be expressed as
where the function po is found by solving the Laplace equation,
after which the function pl is found by solving the Poisson equation,
and so on, inserting the appropriate boundary conditions:
avo
-=
avo
Uw and - - 0 a t x = y = a ,
ax
avo
- -- 0 a t solid boundary (n is normal to boundary)
an
av1
- -- 0 a t solid boundary
an
Note that the method gives the compressible and incompressible flows
about the same body. The iteration method converges, however, only
if the flow is everywhere subsonic. It is not, however, restricted to
thin bodies nor to flows without stagnation points.
Important improvements in the details of carrying out the method
were developed by Kaplan and Imai.(6) For complete details as to
Art. 12.2 THE RAYLEIGH-JANZEN METHOD 367
the techniques employed by various investigators for numerous ex-
amples, the reader is referred to Reference 1 and the complete bibliog-
raphy contained therein.
We now turn to some of the significant results which have been found
with the Rayleigh-Janzen method.
Flow Past Circular Cylinder. Fig. 12.1 shows the velocity distribu-
tion for flow past a circular cylinder a t M, = 0.4. Besides indicating a
comparison with incompressible flow, the various curves show the differ-
ences between the first, second, and
third approximations and indicate 2 5
the degree of convergence of the
method. For 0 < 45", the increase
in local surface velocity is greater
15
for incompressible flow than for x
compressible flow; whereas, for Urn ,
90" > 6 > 45O, the opposite effect
holds. o5
The lower critical Mach Number
I I I I I
M, ,, i.e., the free-stream Mach 0 20 40 60 30 100 120
Number a t which sonic velocity is" 8.. Degrees
-
first reached local'~in the flow, is F I ~ 12.1.
. Velocity distribution on cir-
found by the third approximation cular cylinder (after Imai).
to be 0.404. The Prandtl-Glauert
rule together with the incompressible solution yields M, ,,= 0.418,
and the Karman-Tsien rule correspondingly yields M, ,,= 0.40.
The maximum local velocity, which occurs a t the point of maximum
thickness, is related to the free-stream velocity and Mach Number as
follows:
This illustrates that up to terms in Mm2the specific-heat ratio has no

effect on the flow pattern, a typical result not confined to circular
cylinders, and demonstrated also by the various linearized theories.
When the above expression is evaluated for k = 1.40, one gets
which illustrates that the convergence of the method becomes poorer as

M, increases. This is the main reason why the method is so clumsy for
high subsonic Mach Numbers.
Circular Cylinder in Wind Tunnel. Fig. 12.2 0)shows how the pres-
ence of wind tunnel walls decreases the lower critical Mach Number for
FIG. 12.2. Lower critical Mach Number for circular cylinder in wind tunnel (after
Hasimoto).
flow past a circular cylinder, and indicates that caution must be used
in interpreting wind tunnel results.
Circular Cylinder in Two-Dimensional Free Jet. For a circular cyl-
inder in a free jet of limited lateral extent, as in an open wind tunnel,
Lamla gives the following results (see Fig. 12.3):
FIG. 12.3. Circular cylinder in free wind tunnel jet.
The maximum increase in local velocity is seen to be less when the free
stream is limited than when the latter is infinite, and hence the lower
critical Mach Number is increased as the jet becomes more limited in
width. Notice that this effect is exactly opposite to that in a closed
wind tunnel.
The maximum widening of the jet owing to the presence of the profile
is increased as the result of compressibility effects.
Mow Past Elliptic Cylinder. An elliptic cylinder may be thought of
as approximating a wing of large span or a compressor or turbine blade,
Art. 12.2 THE RAYLEIGH-JANZEN METHOD 369
and, because of its relatively simple geometry, has received much

theoretical a t t e n t i ~ n .and
' ~')
Fig. 12.4 is a comparison of the pressure distributions over an ellipse
for incompressible and compressible flow. The effect of compressibility
1.2
1 .o
0.8
0.6
%
0.4
0.2
-0.2
0 20 40 60 80 100
m e g Thickness Ratio, 8
(4 (b)
FIQ.12.4. Flow past elliptic cylinder (after Kaplan).
(a) Effect of Mach Number on pressure distribution.
(b) h w e r critical Mach Number as function of thickness ratio and angle of
incidence.
is to steepen pressure gradients and to increase considerably the peak

negative pressure coefficient.
On Fig. 12.4b is shown the lower critical Mach Number as a function
of thickness ratio and angle of attack. For moderate angles of attack,
there is a particular thickness ratio having the highest critical Mach
Number. Below this particular thickness ratio, the nose of the profile
is too sharp, whereas above this ratio the profile is too thick.
Symmetrical Joukowski Profile. The Joukowski class of profiles,

although not used in practice, do resemble some real profiles both
geometrically and in respect to flow pattern. Also, since their aero-
dynamic properties may be found relatively easily, they are suitable
for analytic studies, the results of which are sometimes a useful guide to
the characteristics of r e d profile=
The most significant results which have been worked out for this type
of profile (lo) are summarized in Fig. 12.5. The lower critical Mach
Number for symmetrical Joukowski profiles is shown in dependence on
thickness ratio and angle of attack in Fig. 12.5b. The curves are like
thoee of Fig. 12.4b1for much the same physical reasons.
1.0
0.8
3
0.6
z
8
2 0.4
I 0.2
0
0 0.1 0.2 0.3 0.4 0.5 O 0.2 Q4 0.6 0.8 1
.0
Thickness Rotio t/!. Thickness Rotio. t/P
(4 (b)
FIG.12.5. Flow past symmetrical Joukowski profile (after Tamatiko and Umemoto).
(a) S h p e of profile.
(b) Lower critical Mach Number as function of thickness ratio and angle of
incidence.
(c) Effect of M, on lift coefficient.
In practice, the Mach Number a t which large drag increases appear

for modem profiles is significantly greater than that indicated by Fig.
12.5b. This is explained by the fact that modern high-speed profiles
are designed to give much flatter pressure distributions than do Jou-
kowski profiles, and hence have lower peak negative pressures.
Fig. 1 2 . 5 ~shows the theoretical effect of Mach Number on the lift
of symmetrical Joukowski profiles. The correction to the incompressible
lift is seen to depend on thickness ratio and angle of attack as well
as on M,. In this connection, it may be remembered that the Prandtl-
Glauert rule takes no account of these, whereas the Karman-Tsien rule
does, at least approximately. For small thickness ratios, small angles of
attack, and low Mach Numbers, Fig. 1 2 . 5 ~yields the same compressi-
bility correction as the Prandtl-Glauert rule.
12.3. The Prandtl-Glauert Method of Expansion in Series

of a Shape Parameter
Whereas in the Rayleigh-Janzen method we expand the velocity
potential in a series of the Mach Number, in the Prandtl-Glauert methoci
Art. 12.3 THE PRANDTL-GLAUERT METHOD 371
we expand the velocity potential (or stream function) in a series of some

shape parameter, such as the thickness ratio of a class of related profiles.
Although the difficulty of working out the Rayleigh-Janzen method is
scarcely influenced by the thickness of the body under study, the method
converges very slowly a t high Mach Numbers and becomes very clumsy.
The Prandtl-Glauert method, on the other hand, works well at high
Mach Numbers, but converges slowly for bodies of large thickness
ratio or camber ratio. Thus, the Rayleigh-Janzen method is especially
suitable for finding the flow past thick bodies a t low Mach Numbers,
whereas the Prandtl-Glauert method is especially suitable for the flow
past thin bodies a t high Mach Numbers. The latter method, inciden-
tally, is not limited to flows where the velocity is everywhere subsonic.
It is, however, limited to two-dimensional flows.
Outline of Method. The basic differential relation describing the flow
is given by Eqs. 12.1 and 12.2. We now assume that we are dealing with
an initially uniform, parallel flow past a thin body. Then it is physically
plausible to assume that the velocity potential is the sum of the velocity
potential for the undisturbed flow plus a series of terms whose magni-
tudes depend on a shape parameter. This is expressed mathematically
by
$9 = U,X + ~(PI(x, +
Y) f t2?9zpz(x,Y) . . a-
2
(12.7)
where t is a characteristic parameter of the shape, such as the thickness

ratio or camber ratio. The problem now is to find the functions q1, q2,
etc.
The derivatives qz,q,, pZz, p ,,, and (p,, are taken in Eq. 12.7 and
these, together with Eq. 12.2, are introduced into Eq. 12.1. As the equa-
tion is to be valid for all values of t, the coefficient of each power of t in
the resulting equation is equated to zero. In this manner the following
differential equations for ql, q2, etc., are found:
The corresponding expression for q3 is given in Reference 11.

Eq. 12.8 is recognized as the linearized equation for two-dimensional
Bow, and its solution for subsonic flow leads to the familiar Prandtl-
Glauert rule, which states that for a given profile the pressure coefficient
a t any point varies in proportion to 1/41 - M,~. The solutions of
372 TWO-DIMENSIONAL. SUBSONIC FLOW Ch. 12
Eq. 12.9 and of the equations for a,a,etc., provide higher approxima-
tions to the flow of a compressible fluid and may be expected to hold
good for larger departures from a uniform, parallel flow.
To apply the method to the flow past boundaries of specified shape,
Eq. 12.8 is first solved for the function pl. In finding p l it is only
necessary to determine the incompressible solution for the chosen
boundaries and then to apply the Prandtl-Glauert rule. Now, having
determined pl(x, y), the right-hand side of Eq. 12.9 may be expressed
as a function of x and y. In order to determine the function pz it is
then necessary to solve the resulting linear Poisson equation. The
process is repeated for higher approximations. Naturally the boundary
conditions for each approximation need be satisfied for each approxima-
tion only to the same order of t which is involved in the velocity poten-
tial.
The detailed calculations for specific problems are too lengthy to be
given here, but we shall summarize some of the results of practical
significance which have been found with this method.
Flow Past a Bump and Past a Symmetrical Profile Without Stagnation
Points. Kaplan (11) has developed the method for uniform flow past a
symmetrical shape with cusps at both the leading and trailing edges
(thus insuring against stagnation points). The results are applicable
to the flow past the cusped profile of Fig. 12.6a, and also to the flow
(a) (b)
F I ~12.6.
. Flow past Kaplan bump (after Kaplan).
(a) Application to profile without stagnation points.
(b) Applicat,ion to bump on wall.
<
near the bump on the plane wall of Fig. 12.6b. The latter interpreta-
tion, for example, would be useful in estimating the effects of roughness
or waviness on a surface in subsonic flow.
Fig. 12.7a shows the velocity distribution a t the surface of a profile
of 10% thickness ratio, for values of M, from zero to 0.83. As found
previously, compressibility tends generally to augment the local de-
partures from free-stream velocity. For the 10yo profile, the value of
M, ,is 0.742, i.e., a t this free-stream Mach Number the velocity is
exactly sonic a t the point of maximum thickness. It is found that the
solution is regular up to M, of 0.83, beyond which the series diverges.
Beyond this limiting Mach Number, denoted by M, Ii,, and sometimes
called the upper critical Mach Number, no irrotational potential flow
for the given boundary is possible. Presumably, for M, > M,li, shock
Art. 12.3 THE PRANDTL-GLAUERT METHOD 373
waves would appear in the flow field. With the limiting value M, li,
for the 10% profile, the zone of supcrsonic flow is shown in Fig. 12.7b,
1.5
14
I. 3
-..
;I 2
_
3
m
1. 1
0
8
3
, 1.0
>
0 9
O.e 0
0.7 m 0.2 0.4
(a)
zx/e
0.6 0.8 I
Subsonic / Subsonic
/ \
/ I \
Region of \
1
Supersonic
F l o w Without
Shock f o r M,=0.83
\
\
,
I
I
-
\
urn .- -- -..
Mm
(b)
FIG. 12.7. Flow past cusped profile with 1070 thickness ratio (after Kaplan).
(a) Velocity distribution a t surface.
(b) Region of supersonic flow for M, = 0.83
and is seen to extend to a distance on each side of about seven times the
half-width. This example is significant in that it demonstrates that
there is no theoretical reason to expect shock waves in a potential flow
merely because the flow becomes supersonic locally.
-7 1 TWO-DIMENSIONAL, SUBSONIC FLOW Ch. 12
I 12.8a shows, for various thickness ratios, the variation of peak

negative pressure coefficient with M,. Also shown is the curve denoting
the local attainment of the sonic speed, marked (-C,),,, and the curve
beyond which irrotational flow is impossible, marked (-Cp)li,. The
Thickness R a t i o
(b)
FIG.12.8. Flow past cusped profile (after Kaplan).
(a) Maximum negative pressure coefficient. Also shown are the sonic line and
the limiting line beyond which steady potential flow is impossible.
(b) U ~ p e and
r lower critical Mach Numbers.
Karman-Tsien rule proves t o be in good accord with the curves of

(- Cp)max up to M, .,,but of course deviates considerably beyond this
value, because the tangent-gas approximation of the K-T rule limits the
latter to purely subsonic flow.
From the intersections on Fig. 12.8a it is possible to construct Fig.
12.8b, showing the upper and lower critical Mach Numbers as a function
of thickness ratio. The difference between the two curves is of the order
of 0.10 for conventional thickness ratios.
Art. 12.3 T H E PRANDTL-GLAUERT METHOD 375
Flow Past a Circular Arc Profile. The results given above refer to an
uncambered profile. To investigate the effects of camber, Kaplan (I2)
studied the flow past a circular arc profile of zero thickness (Fig. 12.9a)
a t zero angle of attack, with the circulation adjusted to satisfy the Kutta
condition of a stagnation point a t the trailing edge.
Fra. 12.9. Flow past circular arc profile (after Kaplan).

(a) Streamlines.
(b) Effect of M, on lift coefficient.
(c) Lower critical M, as function of camber.
The increase of lift coefficient with M, for a typical camber ratio is

seen from Fig. 12.9b to bc slightly more rapid than the increase pre-
dicted by the Karman-Tsien and Prandtl-Glauert rules, although the
former is a good approximation.
Fig. 1 2 . 9 ~shows the relation between lower critical Mach Number
and camber ratio, and indicates that the K-T and P-G rules both over-
estimate M, .,
376 TWO-DIMEXSIONAL, SUBSONIC FLOW Ch. 12
Lift of Elliptic Cylinders at Small Angles of Attack. To investigate

further the effects of compressibility on the lift of profiles of finite thick-
ness, Kaplan (I3) studied the compressible flow past an elliptic cylinder
a t an angle of attack (Fig. 12.10a), with the circulation adjusted to
FIG. 12.10. Flow past elliptic cylinder

a t small angle of attack (after Kaplan).
(a) Streamlines.
(b) Effect of M, on lift coefficient.
give the Kutta condition a t the trailing edge of the major axis. The
results of this study are summarized by the following formula relating
the lift coefficients about the same elliptical profile in compressible and
incompressible flow:
where 6 is the thickness ratio. It is shown in Reference 13 that this

formula is approximately true also for any slender symmetrical profile.
For zero thickness ratio this relation reduces to the Prandtl-Glauert
rule. The second term on the right-hand side represents the first-order
thickness correction t o the P-G rule. The order of magnitude of the
thickness effect is illustrated by Fig. 12.10b, which is a plot of Eq. 12.10.
For profiles of finite thickness, this analysis shows that the P-G rule
underestimates the compressibility effect, a result which is in accord
Art. 12.4 RELAXATION METHOD 377
with experiment. The curves of Fig. 12.10b are in fairly good agreement
with experimenta'l results on profiles having the same thickness coeffi-
cients as the elliptical profiles.
12.4. Relaxation Method

The so-called relaxation method of obtaining solutions to differential
equations is a numerical technique which occupies a unique place in that
it can usually be made to give answers where other methods fail. In
the field of high-speed flow, for example, it is the only available method
of calculation for flow patterns containing regions of subsonic and super-
sonic flow together with shock waves.
Some General Remarks. The procedure for solving Eq. 12.1 with
specified boundary conditions begins with guesses of the values of p a t
the net points of a grid in the x, y plane. Then, by writing Eq. 12.1 in
finite difference form, it is possible t o calculate a t each point by how
much the differential equation is violated. The values of p are then
adjusted systematically from point to point until ultimately the finite-
difference equations are satisfied everywhere. This indicates that the
correct value of p a t each point has been found. When this is done,
the velocity and pressure distributions are easily computed.
For further details of the method and hints as t o how the method may
be made to converge most rapidly, the reader is referred to Volume 11,
Chapter 21 and to References 14 and 15.
When the flow is incompressible, i.e., when- Laplace's equation is
valid, the relaxation technique gives final answers to practical problems
in a time measured in hours. For compressible flow, on the other hand,
the time is measured in days, weeks, or months, so that for this class
of flows the method is useful chiefly in an exemplary way, i.e., to provide
theoretical solutions to a few typical problems which are not accessible
to other methods of analysis.
Illustrative Results. As an illustration of the type of results obtained with
this method, consider the flow past an NACA 0012 profile at zero angle of attack
in a wind tunnel (Fig. 12.11a). The theoretical flow patterns were calculated in
Reference 16 and corresponding experimental results were obtained in Reference
17.
A typical theoretical result is shown in Fig. 12.11b for a free-stream Mach
Number of 0.75. This example is of special interest because it shows a shock
attached to the airfoil at about 32% chord, followed immediately by a small
expansion.
Up to a free-stream Mach Number of 0.60, the theoretical and experimental
pressure distributions, Fig. 12.11c,are in good accord. At higher Mach Numbers
there is serious disagreement. For example, at M, = 0.75, Fig. 12.11d shows
that the measured peak negative pressure coefficient is considerably less than
t/P=O.12
H/4=3.6
W 1 : 30
NACA 0014
- 1.0
-0.8
-0.6
CD
-0.4
.O 2
0
0.2
0 10 20 30 40 5 0 6 0 7 0 8 0 9 0 IOU
Percent Chord oercent Chord
FIG. 12.11. (a) Flow past NACA 0012 profile in wind tunnel.
(b) Streamlines and lines of constant Mach Number, as worked out by relaxation
method for M, = 0.75 (after Emmons).
(c) Theoretical and ex~erimentalDressure distributions on profile for M, = 0.60
(after Emmons; Amick). '
(d) Theoretical and experimental pressure distributions on profile for M, = 0.75
(after Emmons; Amick).
Art. 12.5 SOME MEASURED EFFECTS OF COMPRESSIBILITY 379
the theory indicates. Furthermore, in practice there seem to occur a series of

weak shocks near the surface rather than a single large shock. The probable
explanation of the discrepancy lies in the change in the effective boundaries of
the profile owing to a shock-boundary layer interaction.
12.5. Some Measured Effects of Compressibility in Subsonic

Flow
General Considerations. From theoretical and experimental work
on compressible boundary layers (Volume 11, Chapters 26 and 27), i t
appears that for subsonic flow the local skin-friction coefficient is almost
independent of Mach Number, all other things remaining constant.
Nevertheless, one might expect the drag coefficient of airfoil profiles to
increase with free-stream Mach Number, for three reasons:
(i) As M, increases, the local velocity outside the boundary layer a t a
given point on the profile generally increases more rapidly than V,.
But the local shear stress is a function primarily of the local V, whereas
the drag coefficient CD is defined in terms of V,, whence it follows that
CD increases even though the local skin-friction remains approximately
constant.
(ii) An increase in M, generally increases the local pressure coeffi-
cients for the profile, and hence increases pressure gradients. This is
particularly significant on the rearward portions of the profile where
pronounced thickening of the boundary layer usually occurs. When the
adverse pressure gradient near the trailing edge is steepened by com-
pressibility, this thickening becomes greater and there is consequently
a rise in drag coefficient.
(iii) As M, increases, the maximum local Mach Number increases
even more rapidly, until, after a zone of supersonic flow has appeared,
shock waves are found t o occur. These shock waves are serious not
only because of the losses they occasion directly, but even more so
because of the way in which they produce a very rapid thickening and
sometimes separation of the boundary-layer flow.
PRESSURE DISTRIBUTION. Fig. 12.12, showing the measured pressure

distributions ( l a ) a t various Mach Numbers for the NACA 4412 profile,
illustrates the steepening adverse gradient of point (ii) above. At
M, = 0.66, there is also evident a t about 55% of chord a sudden rise
in pressure on the upper (convex) surface, which, by the simultaneous
evidence of schlieren photographs, is associated with a shock wave
near the surface a t that point. As M, increases further, the shock moves
rearwards, and the pressure distribution aft of the shock is considerably
altered as the result of boundary-layer separation. In Fig. 12.12, the
short horizontal dashes marked "M = 1" indicate the point where the
-1.2
-0.8
-0.4
0
0.4
0.8
1.2
0 20 40 60 80 100
Chord, Percent C h o r d , Percenl
(b)
- 1.6 -1.6
- 1.2 -1.2
-0.8 -0.8
-0.4 -0.4
Cp 0 CP 0
0.4 0.4
0.8 0.8
1.2 1.2
0 20 40 60 80 I00 0 " 20 40 60 80 100
Chord , Percenr C h o r d , Percent
(c) (d)
F I ~12.12.
. Pressure distribution on NACA 4412 profile a t angle of attack of l052'
(after Stack, Lindsey, and Littell).
(a) M, = 0.19.
(b) M, = 0.60.
(c) M, = 0.66 (note shock a t 55% chord).
(d) M, = 0.73 (note shock at 67% chord).
Woke
.
Low M a .,
Wake
-
---__ D
H i g h Mm
F I ~12.13.
. (a) Flow past profile a t low Mach Number.
(b) Flow at high Mach Number, with boundary-layer separation owing to shock
wave.
Art. 12.5 SOME MEASURED EFFECTS O F COMPRESSIBILITY 381
lock1 sonic velocity is first reached on the profile. It is of particular

interest that shock waves do not necessarily appear immediately upon
the appearance of supersonic speeds somewhere in the flow.
FIG. 12.14. Schlieren photographs of flow past NACA 23015 profile a t angle of
attack of 3" (after Stack). Flow from left to right.
(a) M, = 0.40. Flow entirely subsonic.
(b) M, = 0.60. Small zone of supersonic flow a t 25% chord, with weak repeated
shocks.
(c) M, = 0.70. Large zone of supersonic flow, with lambda shock and boundary-
layer separation.
(d) M, = 0.80. Extensive shock pattern with so much boundary-layer separa-
tion that profile is stalled.
SHOCK-BOUNDARY LAYER INTERACTION. The effect of the interaction

between the shock wave and the boundary layer is shown in Figs. 12.13,
12.14, and 12.15. As a result of the very large adverse pressure
gradient imposed by the shock wave, the boundary layer thickens
rapidly, generally becomes turbulent, and often separates from the

profile, thus producing a large increase in the thickness of the wake.
The change in effective body shape usually acts to decrease the lift and
increase the drag.
FIG. 12.15. Schlieren photograph of flow past Llustsng wing at M, = 0.78, with
angle of attack of 2" (after Hilton). Flow from right to left.
Fig. 12.16 shows the results of wake-survey measurements ('9' behind

a symmetrical profile. At M, = 0.708, the wake is narrow, the maximum
loss in total pressure is small, and t>hedrag coefficient is small. At
M, = 0.910, however, for which schlieren photographs show strong shock
waves, the wake is wide, with a large maximum loss in total pressure.
The middle portion of this diagram represents viscous and separation
losses in the boundary layer, whereas the "wings" of the diagram repre-
sent direct losses in stagnation pressure across the shocks.
FORCE COEFFICIENTS. Fig. 12.17 shows typical effects of compres-
sibility on the force coefficients of airfoil^.^^) For Mach Numbers up
to and somewhat b e y o ~ dthe lower critical Mach Number (correspond-
Art. 12.5 SOME MEASURED EFFECTS OF COMPRESSIBILITY 383
Boundary -Layer
Losses
\
Shock
.-
c "7
g 30 Losses Losses
"7 L Mw=0.910
oa 20
ap 0 10
, n
Distance Across Woke, Percent Chord Mm

FIG. 12.16. Loss of total pressure in FIG. 12.17. Effect of Mach Number on
wake of symmetrical NACA 0012 profile force coefficients of NACA 23015 profile
(12% thick) at zero incidence (after at zero incidence (after Stack).
Stack).
ing to the initial appearance locally of supersonic flow), the drag and
moment coefficients remain nearly
constant, while the lift coefficient
increases approximately according
to the Prandtl-Glauert rule. At a
Mach Number of about 0.7, as
compared with M, ,,of about 0.58,
there is a "force break,," i.e., the
drag coefficient suddenly increases,
the lift coefficient suddenly falls
+
off, and the moment coefficient
C, Measurements
decreases. An extensive summary -
of such results is given in Refer- o C , Measurements
ence 20.
Application of t h e P r a n d t l -
Glauert Rule to Drag. Karman (21)
has suggested a method, based on
the Prandtl-Glauert rule, for pre-
dicting the effect of M, on drag
coefficient in the range of Mach w
Numbers where shock waves are
not present. It is assumed that if Calculated DI I.Curv
the surface distributions of pres-
sure coefficient are the same in an
incompressible and in a compressi-
ble flow, then the local skin-fric-
tion coefficient will be alike a t FIG. 12.18. CL and CD versus M, for
NACA 4412 prome a t a = lo 52', show-
corresponding points, and hence the ing comparison of calculated drag with
drag coefficient will be the same measured drag (after von Karman).
for both profiles. The drag coefficient of a given profile a t M, and with
a certain value of CLM,is, therefore, taken to be equal to the drag coef-
ficient in incompressible flow of an affinely related profile operating a t
an angle of attack to give the same lift coefficient CLM-;the incompressi-
ble profile must have its thickness and camber larger by the factor
1/- than the corresponding quantities for the original profile.
A comparison of this method with experimental results for the NACA
4412 profile a t a = 1.88' (Fig. 12.18) shows, for this example a t least,
that the method works well up to the value of M, a t which the "force
break" or "compressibility burble" takes place.
Effects of Thickness, Camber, and Angle of Attack on Relation Be-

tween Mach Number and Force Coefficients. Fig. 12.19 shows the
effects of thickness, camber, and angle of att,ack on the way in which the
lift coefficient is altered by increasing M,.'22)
Considering first the uncambered sections, the effect of thickness is
seen to be quite marked. Whereas the 6% profile experiences a t constant
incidence a significant increase in CL followed by a moderate drop, the
CL of the 12% profile a t first remains nearly constant, then suddenly
drops severely, and finally recovers slightly. Whereas the M, for peak
CL a t constant incidence is about 0.85 for the 6% section, the correspond-
ing figure for the 12% profile is only about 0.75. A comparison of the
two cambered profiles, Figs. 1 2 . 1 9 ~and 12.19d, shows similar effects
of thickness.
By comparing Fig. 12.19~with Fig. 12.19a, and Fig. 12.19d with Fig.
12.19b, i t is evident that an increase in camber reduces the Mach Num-
ber for lift-force break a t a prescribed CL.
A similar set of results showing variations in drag coefficient (22) is
shown in Fig. 12.20. Here it is seen that the typical sudden increase in
CD a t high speeds occurs a t lower and lower Mach Numbers as the inci-
dence is increased. By comparing the thin with the t>hicksections, it is
evident that the Mach Number for sudden drag rise is more favorable
for the thin sections.
Force-Divergence Mach Numbers. The NACA 4-digit series pro-

files of Figs. 12.19 and 12.20 do not have as favorable high-speed char-
acteristics as the NACA 6-series profiles. Fig. 12.21 summarizes the
high-speed characteristics of the NACA 66-210 profile in terms of
the Mach Numbers for drag divergence and lift divergence a t selected
values of CL. These special Mach Numbers are defined as the values
of M, where (d2C~/dhtm2), = 0 (for lift), and (dCD/dM,), = 0.1 (for
drag). In this example the force-divergent M, for lift is about 0.01
higher than that for drag; and the latter, over a range of lift coefficients
near the design lift coefficient of 0.2, is about 0.02 higher than the theo-
FIG.12.19. Effect of Mach Number on lift coefficient (after Ferri).
(a) NACA 0006-34, having 6% t,hickness and 0% camber.
(b) NACA 0012-34, having 12% thickness and 0% camber.
(c) NACA 2306, having 60jo thickness and 2% camber.
(d) NACA 2312, having 12% t.hicknessand 20jo camber.
385
.I3
.I2
I I
.I 0
.09
.08
.07
CL
.06
.05
.o 4
.03
.o2
O
-=
'00.4 0.5 0.6 0.7 0.8 0.9 1.0
Mm
Mm
(b) Fro. 12.20. (Explana.tion on page 387.) (4
Art. 12.6 THE STREAMLINE CURVATURE METHOD 387
retical lower critical Mach Number a t each CL. It is especially significant

that the range of CL over which the force-divergent Mach Numbers
b& .,
remain high is about three times as wide as the range of CL for which
is high.
GL
FIG.12.21. Force-divergence Mach Numbers for NACA 66-210 profile designed
for CL of 0.2 (NACA).
12.6. The Streamline Curvature Method

A simple, approximate method of analyzing two-dimensional friction-
less flows has been developed by Per1.(23)
The method is a so-called integral method, inasmuch as i t satisfies the
equations of motion in the large, but does not satisfy them in the small.
Governing Physical Equations. Consider the potential flow past the
thin profile of Fig. 12.22. Point P is any point in the flow field and has
the local velocity V . Point s is the point on the airfoil surface lying
on the same equipotential line as P, and has the local velocity V,. The
streamlines and equipotential lines may be regarded as curvilinear
coordinates, with n representing distance along the equipotential line
(in the direction away from the center of curvature of the streamlines).
IRROTATIONALITP. In Chapter 9 it was shown that the condition of
irrotationality a t point P is
-av+ - = vo
an R
where R is the radius of curvature of the streamline at P.
FIG.12.20. Effect of Mach Number on drag coefficient (after Ferri).

(a) NACA 0006-34, having 6% thickness and 0% camber.
(b) NACA 0012-34, having 12% thickness and 0% camber.
(c) NACA 2306, having 6% thickness and 2% camber.
(d) NACA 2312, having 12% thickness and 2% camber.
TWO-DIMENSIONAL, SUBSONIC FLOW Ch. 12
FIG. 12.22. Nomenclature for streamline curvature method.
If the profile is thin the equipotential lines are nearly parallel to the
y-axis. Hence, we may, with little error, write dn Z dy, and the condi-
tion of irrotationality accordingly becomes
where it is to be understood that the differentials are to be taken a t

constant x.
CONTINUITY. The mass flow between the stagnation streamline and
a streamline far from the profile must be the same for all values of x.
Hence, we write, in accordance with the previous assumption that
dn 2 d ~ ,
where ym is so large that the corresponding streamline is parallel to the

x-axis and the corresponding velocity is Vm. The integral is evaluated
a t constant x on a vertical line passing through point s. Rearrange-
ment of the foregoing equation yields
and introduction of Eq. 12.12 gives
where the variable of integration has been changed to V/Vm.

Art. 12.6 THE STREAMLINE CURVATURE METHOD 389
Approximation of Streamline Curvature. If R were known as a func-

tion either of y or of V / V , , then Eq. 12.13 could be integrated to yield
the value of V s / V , for each value of y,. This is tantamount to deter-
mining the pressure distribution on the profile, and constitutes the
desired solution. In order to determine R exactly, however, i t would be
necessary to solve the complete equations of motion, and this is just
what we are trying to avoid.
It is known that R is infinite a t y = m, and has the radius of curvature
of the profile, R,, a t y = y,. Hence we postulate that R may be approxi-
mately represented in terms of V / V mby mea.ns of the empirical relation
where n is a constant the value of which is to be determined later. Eq.

12.14 shows a monotonic variation of R with V , which is as it should be
for stations near the station of maximum thickness.
From the isentropic relations of Chapter 4, it may be shown that
Substituting this equation for p/p,, together with Eq. 12.14 for R, into
Eq. 12.13, we obtain
Determination of Parameter n. The pressure distribution for in-

compressible flow past the given profile can be found relatively easily.
Assuming that this is known, the value of V , / V w is computed a t each
value of x for incompressible flow. Then, integrating Eq. 12.15 with
M, = 0, and inserting the incompressible value of VJV,, we are able
to compute n for each value of x.
By assuming the values of n as found above to be independent of M,,
Eq. 12.15 may be solved for the velocity distribution on the airfoil for
any value of M , less than unity.
Details of the method, together with useful tables and charts for
numerical computations, are given in Reference 24. The method may
also be extended to supersonic flows (23) and to flow through turning

channels, where the equipotential lines may not be assumed parallel.(24)
Application to Flow with Small Perturbations. As an example of

the method, consider flow with small perturbations. Then it is appro-
priate to expand the integrand of Eq. 12.15 in powers of (V/V, - 1).
Proceeding in this way and assuming this variable to be small compared
with unity, we obtain by detailed expansions
This may be readily integrated and solved for (V,/Vm - 1) in the form
The value of n is now found by setting M, = 0 . Thus
where Vaiis the velocity a t point s for incompressible flow. Combining

Eqs. 12.16 and 12.17, we obtain finally
which may be recognized as the Prandtl-Glauert rule. This result

gives some assurance that the curvature function, Eq. 12.14, is reason-
ably accurate for flows involving large perturbations.
Accuracy of Streamline Curnature Method. To appraise the accuracy

of the streamline curvature method, i t has been compared a3)with the
substantially exact results in the example of Fig. 12.7. As may be seen
Art. 12.6 REFERENCES AND SELECTED BlBLIOGRAPHY 391
from Fig. 12.23, the streamline curvature method gives an excellent
approximation to the exact results a t M, = 0.75; a similar comparison
(not shown here) a t M, = 0.50 shows almost perfect agreement.
FIG.12.23. Velocity distribution (at M, = 0.75) for flow past 10%-thick profile of
Fig. 12.7b. Curve marked "exact" is taken from Fig. 12.7a (after Perl).

1. EHLERS, F. E., and C ARRIER, G. F. Monograph IV, Methods of Linearization
in Compressible Flow. Part I: Janzen-Rayleigh Method, Air Materiel Com-
mand Report, No. F-TR-ll8OA-ND (Feb., 1948).
2. SAUER, R . Theoretical Gas Dynamics. Ann Arbgr: J. W. Edwards, 1947, p. 135.
3. RAYLEIGH, LORD. On the Flow of a Compressible Fluid Past an Obstacle, Phil.
Mag., Vol. 32 (1916), p. 1.
4. KAPLAN, C. On the Use of Residue Theory for Treating the Subsonic Flow of a
Compressible Fluid, N A C A Tech. Rep., No. 728 (1942).
5. IMAI, I. A New Method of Successive Approximation for Dealing with the Two-
Dimensional Flow of a Compressible Fluid, Proc. Phys. Math. Sac. Japan,
Vol. 24 (1942).
6. IMAI, I. On the Flow of a Compressible Fluid Past a Circular Cylinder, 11,
Proc. Phys. Math. Soc. Japan, Vol. 23 (1941), p. 181.
7. HASIMOTO, Z. On the Subsonic Flow of a Compressible Fluid Past a Circular
Cylinder Between Two Parallel Walls, Proc. Phys. Math. Soc. Japan. Vol. 25
(1943), pp. 563-74.
8. LAMLA, E. Die symmetrische Potentialstromung eines kompressiblen Gases um
einen Kreiszylinder in Freistrahl im unterkritischen Gebiet, Luftfahrtjorsch-
urn. Vol. 19 (1943)... DD. 358-62.
A A
9. KAPLAN, C. A Theoretical Study of the Moment on a Body in a Compressible

Fluid, NACA Tech. Rep., No. 671 (1939).
LO. TAMATIRO. S.. and UMEMOTO. H. On the Subsonic Flow of a Compressible Fluid
Past a ~ & n e t r i c a l~oukowskiAerofoil, Rcport 201 of ~ e r o .Res. Inst.,
Tokyo Imp. Univ. (1941).
11. K APLAN , C. The Flow of 3, Compressible Fluid Past a Curved Surface, N A C A
Rep., No. 768 (1943).
12. K APLAN , C. The Flow of a Compressible Fluid Past a Circular Arc Profile,
NACA Rep., No. 994 (1944).
392 TWO-DIMENSIONAL, SUBSONlC FLOW Ch. 12
13. KAPLAN, C. Effect of Compressibility a t High Subsonic Velocities on the Lift-
ing Force Acting on an Elliptic Cylinder, N A C A Rep., No. 834 (1946).
14. EMMONS, H. W. The Numerical Solution of Compressible Fluid Flow Problems,
N A C A Tech. Note, No. 932 (1944).
15. EMMONS, H. W. The Theoretical Flow of a Frictionless, Adiabatic, Perfect Gas
Inside of a Two-Dimensional Hyperbolic Nozzle, NACA Tech. Note, No. 1003
(1946).
16. EMMONS, H. W. Flow of a Compressible Fluid Past a Symmetrical Airfoil in a
Wind Tunnel and in Free Air, NACA Tech. Note, No. 1746 (1948).
17. AMICK, J. L. Comparison of the Experimental Pressure Distribution on an
NACA 0012 Profile a t High Speeds with that Calculated by the Relaxation
Method, N A C A Tech. Note, No. 2174 (1950).
18. STACK, J. L., LINDSEY, W. F., and LITTELL,R. E. The Compressibility Burble
and the Effects of Compressibility on Pressures and Forces Acting on an
Airfoil, NACA Rep., No. 646 (1838).
19. STACK, J. Compressible Flows in Aeronautics, Jour. Aero. Sci., Val. 12, No. 2
(1945).
20. ABBOW,I. H., and v o D~OENHOFF, A. E. Theory of Wing Sections. New York:
McGraw-Hill Book Co., Inc., 1949.
21. VON K ARMAN, TH. Compressibility Effects in Aerodynamics, Jour. Aero. Sci..
Val. 8, No. 9 (1941), p. 337.
22. FERRI, A. Completed Tabulation in the United States of Tests of 24 Airfoils at
High Mach Numbers, N A C A ACR, No. L5E21 (1945). (Wartime Report,
No. L-143.)
23. PERL,W. Calculation of Compressible Flows Past Aerodynamic Shapes by Use
of the Streamline Curvature, NACA Tech. Note, No. 1328 (1947).
24. ALPERT,S. Design Method for Two-Dimensional Channels for Compressible
Flow with Application to High-Solidity Cascades, N A C A Tech. Note, 1931
(1949).
Chapter 13
THREE-DIMENSIONAL, SUBSONIC FLOW

Most practical problems are three-dimensional in nature but their
mathematical complexity usually makes the exact analysis of such
problems prohibitively difficult. An idealized approach is usually
adopted, therefore, the results of which are taken as a general guide
to what might take place in more complex situations. This simplified
approach might consist of the exact analysis of certain mathematically
simple forms, such as ellipsoids; or i t might involve the determination
of approximate rules of comparison between three-dimensional flows.
and plane two-dimensional flows, since analysis of the latter is less
difficult; or it might depend on a linearized theory, based on the assiimp-
tion of small perturbations, and leading to approximate rules showing
the effect of Mach Number.
All these techniques, alone and in combination, will be discussed in
this chapter, the purpose mainly being t o arrive a t a general understand-
ing of subsonic flow about three-dimensional bodies.
Our analytical considerations will be limited t o potential flows, and
will, therefore, give information concerning the pressure distribution
and streamline pattern in the flow outside of the boundary layer. How-
ever, the pressure distribution of the potential flow controls to a large
extent the behavior of the boundary layer. We shall therefore inject
boundary-layer considerations into the discussion of experimental drag
results.
NOMENCLATURE
aspect ratio etry normal to leading

angle of attack (radians) edge
span of wing moment coefficient
speed of sound chord
pressure coefficient Mach Number
peak negative pressure co- lower critical free-stream
efficient Mach Number
drag coefficient
lift coefficient radius in cylindrical coor-
section l i f t coefficient. dinates
based on flow and geom: 1t thickness of profile
393
394 THREE-DIMENSIONAL, SUBSONIC FLOW Ch. 13
x-component of perturba- u sweepback angle

tion velocity 'P perturbation velocity po-
y-component of perturba- tential
tion ve1ocit.y
resultant velocity ( 1' signifies transformed coor-
z-component of perturba- dinates
tion velocity ( 18
signifies conditions at pro-
Cartesian coordinates file surface
( )e signifies compressible flow
( )i signifies incompressible
thickness ratio flow
thickness ratio in plane ( la signifies free-stream con-
normal to leading edge ditions
13.2. Gothert's Rule for Uniform Flow with Small Perturbations

We shall consider in this article the irrotational, isentropic, steady,
subsonic flow of a perfect gas under the assumption that there are only
small deviations from a uniform, parallel flow. Let the uniform flow be
along the x-axis with a speed V,. The body producing the disturbances
in the flow must then be very thin in a t least one direction normal to
X, say the y-direction; that is, the thickness ratio of the body in the
x, y-plane must be small compared to unity, but may be anything what-
soever in the x, z-plane (Fig. 13.1).
Linearization of the Differential Equation. The exact differential
equation of the velocity potential, without the assumption of small
perturbations, is Eq. 9.32. I n Chapter 10, the linearization of this
equation for two-dimensional flow was undertaken very carefully,
with special consideration of the significance of the various assumptions,
and Eq. 10.21a for the perturbation velocity potential was arrived at.
Employing similar assumptions, Eq. 9.32 for three-dimensional flow
may be linearized to give
where B = m, M, = VJc,, and p is the perturbation velocity

potential, the derivatives of which are the perturbation velocity com-
ponents u, v, and w i n the x-, y-, and z-directions, respectively:
At any point in the flow the complete velocity components in the x-, y-,
+
z-directions are, respectively, (V, u), v, and w.
A review of Art. 10.2 will show that the most stringent assumptions
underlying Eq. 13.1 are that
( M m ) 2 ( ~ / V<<m1); Mm2(v/Vw)<< 1 ; M ~ ' ( W / V<<~1) (13.3)
Mm2(u/Vm)/(1- Mm2) << 1 (13.4)
Equation 13.4 bars from the present analysis the transonic region, that
is to say, free-stream Mach Num-
bers near unity. In the subsonic
region Eq. 13.3 shows that the
value of M , up to which the analy-
sis is accurate depends on the rela-
tive magnitudes of the perturba-
tion velocities produced by the
body.
Boundary Conditions. Let us
suppose that the fluid streams past
a thin body B (Fig. 13.1) whose
surface is defined implicitly by the
relation S ( x s , y,, z,) = 0. At in-
finity the boundary conditions are
FIG. 13.1. Nomenclature for flow with

small perturbations past body B.
At the surface of the body the
boundary condition is that the streamline be tangent to the surface, or,
mathematically,
+
(Vm us):us:W , = dxs:dys:dza
This may also be written
where u,, v,, and w, are the perturbation velocities a t a point on the
surface S. On the surface S we may also write, however,
dS
as
= - dx,
as
+- as
dy, + -dz, = 0
axs ayS az,
Elimination of dx,,,dy,, and dz, with the help of the preceding relations
then gives us the boundary condition in the form
or, within the approximations of the linear theory,

vmas/ax. + v, a s l a g , + W, as/az, = o (13.6)
396 T ~ ~ SUBSONIC~ FLOW ~ Ch.13 - ~
M n e Transformation to Corresponding Incom~ressibleFlow. Now,

using the detailed analysis of Art. 10.5 as a guide, we may effect a trans-
formation xhich permits a comparison of the original compressible flow
with an incompressible flow over a modified body B' having a surface
defined by Sf (x,', y,', zsl) = 0. The results of Art. 10.5 suggest that we
define new coordinates x', y', z' and a new perturbation potential func-
tion cpl as follows:
- x;
xI = y1 = By; z' = Pz; rp'(xf, y', z') = P2q(x, y, z) (13.7)
These same relations define the new body B' with surface coordinates
-
x,I = 2,; y,' -- py,; z,' = Pz,. Hence, we may write S(xs, y,, z,) =
S1(xs, Ys , zs') = 0.
I I
Substituting these expressions for z, y, z, and rp into Eq. 13.1, and

simplifying, we get
aZcpf azP1 aZpf
, . -
which is evidently the differential equation for the perturbation potential

of an incompressible flow.
From Eqs. 13.2 and 13.7, we may find
where u', Y', W' are the velocity components in the incompressible flow.
By applying the transformation formulas of Eqs. 13.7 and 13.9 to
the boundary condition of Eq. 13.5, we get
which means that in the incompressible flow, the flow is uniform and
parallel a t infinity.
Next, applying Eqs. 13.7 and 13.9 to the boundary condition a t the
body as expressed by Eq. 13.6, we obtain, after simplification,
Without loss of generality, we may set V,' = V,. Then Eq. 13.11 (by
comparison with the form of Eq. 13.6) is exactly the condition that the
incompressible flow follow the boundaries of B'. The latter, according
to Eq. 13.7, is thinner than B in all lateral dimensions by the factor 0.
Linearized Pressure Coefficient. In Art. 10.3, it was shown that the

linearized pressure coefficient is C, = -2u/V,. This expression is
based on the assumption that u / V , << 1. However, i t is important to
note that the ratio of two linearized pressure coefficients when based on.
this expression has an accuracy dependent on the assumption ( u / V , ) ~<< 1,
and hence the comparison of pressure coefficients by the linear theory is
considerably more accurate than the calculation of a single pressure
coefficient by linearized methods.
Now, using Eqs. 13.9, we get
Gothert's Rule. The bodies B' and B are affinely related, so that in
going from B to B', the thickness ratio in the x,y-plane, the aspect
ratio seen in the x,z-plane, the angle of attack seen in the x,y-plane, etc.,
are all altered in the same proportion, namely, they are reduced by the
factor p. Letting 6 denote the thickness ratio, and using the convention
that in affine transformations a change in 6 produces proportionate
changes in angle of attack, a , aspect ratio, B , etc., we may summarize
the entire analysis by the symbolic relation
or, in words, the pressure coe$cient at a given point for the compressible
flow at Mach Number M, past a body of thickness 6 i s 1/b2 times a s large
a s the pressure coeficient at the correspondzng point for incompressible
flow past a thinner afine body o j thickness P6.
This will be recognized as Rule I of Chapter 10. But, whereas for
two-dimensional incompressible flow the pressure coefficient for affine
thin profiles is approximately proportional to the thickness ratio, no
such simple rule applies to three-dimensional incompressible flow. For
three-dimensional flows, therefore, it is not possible to arrive a t the
simple Prandtl-Glauert rule showing the effect of Mach Number on the
pressure coefficient for a body of fixed shape.
In applying Eq. 13.13 to a compressible flow a t Mach Number t&,
it is necessary t o consider the incompressible flow past a different body
all of whose lateral dimensions are reduced by the ratio 4 1 - Mm2.
The pressure coefficient for the incompressible flow is found either from
systematic experiments or by means of analytical methods.
THREE-DIMENSIONAL, SUBSONIC FLOW Ch. 13
Aspect Ratio
Locol Sonic Speed I
of Revolution
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
1.o
Mm
(4
FIG.13.2. Flow past ellipsoids (after Hess and Gardner). (See explanation on page 399..
Art. 13.3 FLOW PAST ELLIPSOIDS 399
The lift and moment coefficients are found by integrating the pressure
forces acting on the surface of the body. Within the linear theory
Eq. 13.13 holds a t all corresponding points of the two flows. It follows,
therefore, that the rule expressed by Eq. 13.13 applies also to the lift
and moment coefficients.
13.3. Flow Past Ellipsoids

Method of Calculation. Since in three-dimensional flow there is no
simple rule showing the effect of Mach Number on the pressure distribu-
tion around a given body, a knowledge of this effect may be gained only
by working out typical examples.
A particularly convenient example is the family of ellipsoids (Fig.
13.2a) with axes 1, t, and b. As b/l -+ m, we get the case of an elliptic
cylinder, whereas as b/l + t/l we get the case of an ellipsoid of revolu-
tion. For intermediate values of b/1 the ellipsoid is similar to a wing of
elliptic planform with thickness ratio t/l.
In applying Gothert 7s rule to an ellipsoid in compressible flow, with
thickness ratio 6 = t / l , and aspect ratio
span bZ 4b
A = - = --
planform area/span ?r(b/2)(1/2) ?r I
we consider the incompressible flow about an ellipsoid with 6' = B(t/l)
and B' = /3(4b/?rl). Using the classical solution for incompressible
flow past an ellipsoid (e.g., Reference 8), the incremental velocity for
the compressible flow a t any point is found from the incremental velocity
of the incompressible flow through Eq. 13.9a:
Typical results of calculations of this type (2) are shown in Fig. 13.2.
Three-Dimensional Relief Effect. The maximum local velocity on the

surface of the profile occurs a t the half-chord line, i.e., a t . %= 0, and is
the same a t all points on the half-chord line. Fig. 13.2b shows the
effect of Mach Number on u,,/V, for various aspect ratios, the thick-
(Explanation for Fig. 13.2):

(a) Nomenclature.
(b) Maximum perturbation velocity on ellipsoid of 15% thickness ratio, show-
ing effects of M, and aspect ratio.
(c) Ratio of maximum perturbation velocity for compressible flow to that for
incompressible flow.
(d) Effect of aspect ratio and thickness ratio on lower critical Mach Number.
ness ratio remaining fixed. It is clear that the pressure coefficient for a
given thickness ratio is very much less in three-dimensional flow than
in two-dimensional flow. This is the result of a three-dimensional relief
effect which arises because, in a three-dimensional flow, the stream may
deviate from the original direction in both the y- and z-directions,
whereas onljr the y-direction is allowable for a two-dimensional flow.
Thus a three-dimensional body produces a lesser disturbance in the
uniform, parallel flow, and, consequently, it has a smaller peak negative
pressure coefficient.
One of the very important conclusions to be drawn from this example
is that for practical problems the linearized theory is more reliable for
three-dimensional flow than for two-dimensional flow because the relief
effect makes the assumptions of Eq. 13.3 more likely to be met.
Influence of Aspect Ratio on Compressibility Effect. To bring out
more clearly the effect of M, a t various aspect ratios, the results of Fig.
13.2b are replotted in Fig. 1 3 . 2 ~to show the ratio of the peak negative
- pressure coefficients for compressible and incompressible flow past ellip-
soids of 15% thickness ratio. For infinite aspect ratio, corresponding to
two-dimensional flow past an elliptic cylinder, the curve of Fig. 1 3 . 2 ~
corresponds to the Prandtl-Glauert rule, namely, C, .-/I
For finite aspect ratios the effect of Mach Number is considerably less,
until for ellipsoids of revolution with small thickness ratios the pressure
coefficient is almost independent of Mach Number. For a thickness
ratio of 0.15 and an aspect ratio of 2 the effect of Mach Number is ap-
proximately midway between that of the ellipsoid of revolution and that
of the elliptical cylinder.
Lower Critical Mach Number. On Fig. 13.2b is plotted the curve of
u,,,/V, versus M, for which the value of u,, is exactly sonic. The
intersection of this curve with each of the aspect ratio curves gives the
value of M, , the lower critical Mach Number, at which sonic velocity
is first reached a t the surface of the ellipsoid. By plotting M, ,,versus
aspect ratio for constant thickness ratio, the curves of Fig. 13.2d are
obtained. This chart shows that the lower critical Mach Number is
markedly increased as the aspect ratio decreases to values less than
about 2.0. The increase in lower critical Mach Number is the result
of the three-dimensional relief effect, which reduces both the magnitude
of the incremental velocity and also the effect of Mach Number on this
magnitude.
Effect of Aspect Ratio on "Compressibility Burble." The lower critical
Mach Number, M, ,, is of practical significance because experience
has shown that, as the result of a shock wave-boundary layer inter-
action, the drag of a body rises when velocities on the surface substan-
Art. 13.4 BODIES OF REVOLUTION 401
tially exceed the sonic velocity. From the results on ellipsoids, there-
fore, it would be expected that the &TachNumber a t which the compressi-
bilitg burble appears would be increased as the aspect ratio is decreased.
No direct data for ellipsoids are available, but Fig. 13.3,(9)showing the
results of measurements on a wing having a rectangular planform and r
symmetrical pr'ofile of 12y0 thick-
ness ratio, gives evidence of the
beneficial influence of three-di-
mensional effects in delaying the
sharp drag rise associated with
compressibility. Corresponding
schlieren photographs are given
in Fig. 13.16.
It is also well known that shock
waves first appear on the wings
of high-speed subsonic aircraft
rather than on the fuselage or on
the tail surfaces. This is further
confirmation of the three-dimen-
sional relief effect.
13.4. Bodies of Revolution

A special but important case of
three-dimensional flow is that of
flow with axial symmetry past FIG. 13.3. Experimental curves showing
bodies of revolution. Strictly, influence of wing aspect ratio on curve of
drag coefficient versus Mach Number (af-
only two length dimensions are ter Stack and Lindsey).
involved here, which are in cylin-
+
drical coordinates x and r = dy2 z2, but the flow possesses the
"relief ' effects of three-dimensional space flow, and is therefore included
J
in the latter category.
Method of Sources and Sinks. Incompressible flows with axial

symmetry may be analyzed in the classical manner by distributing
sources and sinks along the x-axis, the distribution being determined
by the requirement that one of the streamlines have the form of the
desired body of revolution.(") By applying Gothert's rule to the formulas
of this method, a wt of similar formulas for compressible flow with small
perturbations may 'be derived.
Using this procedure, Laitone derived the following approximate
formula for the pressure coefficient a t the surface of a body of revolution :
THREE-DIMENSIONAL, SUBSONIC FLOW Ch. 13
.I
(13.14)
where L is the length of the body; x is the distance aft of the nose; r, is
the radius from the axis to the body surface; A = T T ~A'; = dA/dx;
A" = d2A/dx2;and so forth.
The results given by this approximate formula agree well with the
exact results for an ellipsoid of revolution, discussed later.
Disturbance Produced by Body of Revolution. To determine the
effects of wind tunnel wall interference, or to estimate the effects of a
body of revolution on the pressure coefficient a t large distances from the
body, we may assume that in incompressible flow the influence of the
body at remote points may be approximately represented by a three-
dimensional doublet. The incompressible velocity potential of the latter
is
x
(P-
(x2 + r2)%
so that, for a given body of revolution in incompressible flow,
Furthermore, exact solutions for incompressible flow past very slender

ellipsoids of revolution show the effect of thickness ratio on the per-
u -
turbation velocity far from the body to be given approximately by
6'.
Now, referring to Fig. 13.4, let points 1 and 2 be corresponding points
in compressible and incompressible flow about corresponding bodies,
S and Sf, in the sense demanded by Gothert's rule; let point 3 refer to
incompressible flow past Sf, but a t the same physical coordinates as
point 1; and let point 4 refer to incompressible flow, a t the coordinates
of point 1, past the original body S. Then to find the effect of Mach
Number on the pressure coefficient for a $xed body a t a fixed point, we
first write the identity
FIG.13.4. Illustrates application of Gothert's rule.

By using Gothert's rule and the approximate rules for incompressible
flow mentioned above, we may evaluate the three ratios on the right-
hand side, and thus we get
By the definitions of terms, however,
Making these substitutions, and simplifying, we get
At great lateral distances from the body, i.e., as xl/rl -+ 0, this

formula becomes u1/u42 1/p3, which means that the effect of com-
pressibility is given by
-
Cp l / ( l - Mm2 1$5
This is the same result as was obtained for two-dimensional flow. It
shows that if Au represents the interference velocity produced at the body
the interference velocity is given by Au -

in a test because of wind tunnel walls, the effect of compressibility on
1/p3.
It is especially curious that a t the surfaces of two-dimensional and
axially symmetric bodies the effects of compressibility are markedly
different; whereas a t great lateral distances the effects are the same.
At great longitudinal distances from a body of revolution, i.e., rl/xl-+O
the results of this analysis show the pressure coefficient to be independent
404 THREEDIMENSIONAL, SUBSONIC FLOW Ch. 13
of Mach Number. I n Chapter 10 it was shown that for two-dimensional

bodies the pressure coefficient a t great longitudinal distances varies
as 118.
Comparison of Gothert's Rule with Experiment. Fig. 13.5 shows the
measured surface pressure coefficients 0 ) a t various sections of a semi-
infinite half-body of revolution with an ellipsoidal nose. Also shown
A B c D
F I ~13.5.
. Flow past semi-infinite half-body of revolution (after van Driest).
(a) Nomenclature.
(b) Comparison of Gothert's rule with experimental data.
are solid lines indicating the effect of Mach Number on pressure co-
efficient as found by Gothert's rule. To apply the latter, the theoretical
solution for incompressible flow was first worked out as a function of
thickness ratio by the method of sources and sinks, and then Gothert's
rule applied in the usual manner.
The agreement between theory and experiment is on the whole fairly
satisfactory, even up to Mach Numbers of 0.9. Thanks to the several
effects mentioned earlier in reference to the flow past ellipsoids, the
method of small perturbations is seen to be considerably more reliable
at high Mach Numbers for axi-symmetric flow than for two-dimensional

flow.
Thin Ellipsoid of Re3olution. The ellipsoid of revolution remains such
during an affine transformation. Since the exact incompressible solu-
tion is known, we may derive an analytical formula in closed form relat-
ing the peak negative pressure coefficient to the Mach Number and
thickness ratio 6.
PRESSURE CORRECTION F ORMULA. According to Lamb,(lQ)the maxi-
mum disturbance velocity for incompressible flow occurs a t the point
of maximum thickness, and for potential flow is given exactly by
By restricting our considerations to small thickness ratios (less than

about 207&), i t may be shown by expanding this expression and ignoring
terms of higher order that
Applying Gothert's rule to this expression, we obtain for the maximum

incremental velocity in compressible flow past a thin ellipsoid of revolu-
tion of thickness 6,
To find the effect of compressibility for a fixed value of 6, we combine

Eqs. 13.17 and 13.16 to give
(-Cpi) max ( ~ i n e o m p max

)
In the absence of better information, Eqs; 13.17 and 13.18 may also
be used as a guide for estimating the peak negative pressure coefficient
on slender bodies of revolution of shapes other than ellipsoidal. For this
purpose an equivalent 6 should be used, so chosen as to produce equal

peak negative pressure coefficients in incompressible flow for the ellipsoid
and for the body under investigation.
A theoretical investigation of the flow past a prolate spheroid eO)gives
exactly the same result as Eq. 13.18a, thus lending strength to the
FIG.13.6. Substantially exact theoretical results for flow past ellipsoid of revolu-
tion (after Schmieden and Kawalki).
(a) and (b) Maximum perturbation velocity.
(c) Ratio of maximum perturbation velocity for compressible flow to that for
incompressible flow.
hypothesis that Eq. 13.18a may be used as a general rule for thin bodies
of revolution. I n Reference 20 it is also shown that, within the approxi-
mations of the perturbation theory,
- (-Cp,)mx = 2 1 (13.18b)
RESULTS OF EXACT THEORETICAL ANALYSIS. Substantially exact
results for potential, compressible flow past ellipsoids of revolution (3)
are shown in Fig. 13.6. For thickness ratios below 0.20, the approximate
formulas of Eqs. 13.17 and 13.18 are found to fit these results to within
a few per cent.
Figs. 13.6a, b show, as was mentioned previously, that the maximum
dis'turbance velocity produced by a thin body of revolution is small
compared with that of a two-
dimensional body. Fig. 1 3 . 6 ~
,
shows further the relatively small

effect of Mach Number on the
pressure coefficient of slender
bodies of revolution, as compared
with the factor 1 / d 1 - M> for
two-dimensional flow. Indeed. for
the limiting case of zero thickness
ratio, Eq. 13.18 shows the pres-
sure coefficient to be independent
of Mach Number.
By plotting on Figs. 13.6a, b the
relation between (V*- Tr,)/Vm
and M, for isentropic flow, corre-
sponding to the local appearance 0.7
of sonic velocity, the value of the o 0.1 0.2 0.3 0.4
lower critical Mach Number for Thickness Rotio, 8
each thickness ratio may be found FIG. 13.7. Lower critical Mach Number
for ellipsoid of revolution and for elliptic
a t the appropriate intersection cylinder.
point. Fig. 13.7 accordingly.shows
the lower critical Mach Number as a function of thickness ratio. For
comparison, the corresponding relation for the elliptic cylinder is also
shown. For 6 = 0.20, as an example, the lower critical Mach Number
of the cylinder is 0.74, as com-
pared with 0.90 for the body of
revolution, thus showing again
that compressibility effects be-
come serious a t an earlier stage
for wing-type bodies than for
fuselage-type bodies.
COMPARISON WITH EXPERI-

MENT. Fig. 13.8 shows the ex-
FIG. 13.8. Experimental results for flow perimental effect of Mm on the
past prolate spheroid and ogival body of peak negative pressure coefficient
revolution (after Matthews). Theoretical
curves represent Eq. 13.18b. of two prolate spheroids and of
an ogival body of revolution. Also
shown is the pressure correction formula of Eq. 13.l8b for the relevant
thickness ratios of 1/6 and 1/10.
408 THREE-DIMENSIONAL. SUBSONIC FLOW Ch. 13
Up to M, = 0.9 the agreement of theory with experiment is quite

good. It is also of practical interest that the two shapes with 6 = 116
show substantially the same effect of Mach Kumber.
Drag. Fig. 13.9 shows the results of drag measurements (12) made on
a bomb.
This body of revolution has a comparatively large thickness ratio,
namely, 6 2 0.25. Hence i t is to be expected that the drag will be pre-
dominantly form drag, i.e., pressure drag. In the range of M, where
Mm
FIG.13.9. Drag of bomb a t subsonic speed (after Gothert).
the flow is entirely subsonic, the pressure drag should, therefore, exhibit
a variation with M, somewhat similar to the variation of pressure coef-
ficient for inviscid flow. A comparison of the bomb drag in the range of
low Mach Numbers with the pressure-coefficient curve for 6 2 0.25 in
Fig. 13.6 shows that the rise of drag coefficient is about twice as large as
this argument would indicate, most likely because the shape of the bomb
produces higher incremental velocities than does an ellipsoid of like
thickness ratio.
According to Fig. 13.6, for 6 0.25, the lower critical Mach Number
is about 0.86. At about this point Fig. 13.9 indicates a very rapid in-
crease in drag coefficient, owing probably to the appearance of shock
waves.
.4rt. 13.5 SPHERES 409
13.5. Spheres
Gothert's Rule. The sphere is of some interest because of its simple
geometry and also because it represents the limiting case of a body of
revolution. Since a sphere is far from slender, Eq. 13.15 must be used
for the exact incompressible solution. Then, noting that 6 = 1 for a
sphere, Gothert's rule applied to Eq. 13.15 yields the prediction that for
compressible flow
Rayleigh-Janzen Method. The accuracy of Gothert's rule may be

tested by comparing Eq. 13.18~
with the substantially exact solution for
0 30 60 90
0 , Deg.
FIG.13.10. Maximum perturbation ve- FIG. 13.11. Pressure dis-
locity for flow past sphere as function of tribution on sphere for com-
Mach Number. Comparison of exact so- pressible and incompressible
lution (after Kaplan) with Gothert's rule. flow (after Kaplan).
flow past a sphere obtained by the Janzen-Rayleigh method.(l31 This

comparison appears in Fig. 13.10. Considering the severity of this test
of the method of small perturbations when applied to a body as thick as
a sphere, this comparison lends confidence in Gothert's rule for the range
of moderate thickness ratios.
Fig. 13.10 shows the lower critical M, for the sphere to be 0.57. For
this speed, as well as for incompressible flow, the pressure distribution
over the surface of the sphere is shown in Fig. 13.11. These results,
which are typical for symmetrical bodies of revolution, show that over
practically all of the surface the effect of Mach Number is to augment
the absolute magnitude of the pressure coefficient.
Drag. Experimental measurements (I4) of the drag of spheres a t high
speeds are summarized in Fig. 13.12. The range of Reynolds Numbers
for these tests is in the region where the flow is separated over the
downstream part of the sphere, and hence the drag is substantially all
pressure drag.
Up to M, s 0.5, there is little effect of Mach Number on drag coef-
ficient, in conformity with the effect of Mach Number on peak disturb-
ance velocity shown in Fig. 13.10. Above the lower critical Mach
FIQ. 13.12. Experimental drag of smooth spheres of x6-inch diameter. Drag

coefficientis based on frontal area (after Charters).
Number of 0.57, however, there is a sharp rise in drag. Spark shadow-

graphs of the flow past the sphere show this rise in drag to be associated
with the appearance of shock waves.
13.6. Wings of Finite Span

Gothert's Rule. Consider a wing of span b, chord 1, thickness t, and
aspect ratio ~ l operating
, a t an angle of attack a in a stream of Mach
Number M,. The pressure coefficient a t any point in the flow may, by
Gothert's rule, be related to the pressure coefficient a t a corresponding
point for incompressible flow past an affinely related wing. The relation
may be written implicitly as
Since the lift and moment coefficients are found by integrating the
pressure coefficient over the entire surface of the wing, it follows from
Eq. 13.19 that the similarity rules for the lift and moment coefficients are
Lit-Curve Slope. This similarity relation, which holds for any plan-
form, compares wings of different aspect ratio and thickness ratio and
a t different angles of attack. To obtain a relation showing clearly the
Art. 13.6 WINGS OF FINITE SPAN 411
effect of Mach Number on a wing of $xed aspect ratio, fixed thickness

ratio, andsxed angle of attack, we introduce some approximate relations
obtained by lifting-line theory for incompressible flow past wings of
finite aspect ratio. The lift-curve slope in incompressible flow for a wing
of finite aspect ratio is approximately related to that for infinite aspect
ratio by the formula
But, for incompressible two-dimensional flow
where the factor 0.9 takes account of an average departure for existing
wings from the theoretical value of 27r. Combining the last two relations,
we get
Al
We may now write the identity
The first ratio on the right-hand side is found from Gothert's rule; the
second is found from the approximate rule that the lift of affinely related
profiles is proportional to the angle of attack; and the third is found
from Eq. 13.21~. Thus we get
Fig. 13.13, which is based on Eq. 13.22, shows how the aspect ratio
modifies the effect of compressibility on the lift coefficient. For infinite
-
aspect ratio, Eq. 13.22 reduces to the Prandtl-Glauert rule for two-
dimensional flow, namely: for a fixed profile, d C ~ / d a I/-.
At the other extreme of zero aspect ratio, the lift-curve slope is inde-
pendent of Mach Number. For aspect ratios in the neighborhood of 3,
the augmentation of lift coefficient by compressibility is only about half
what it would be for infinite aspect ratio.
In a high-speed airplane, the tail surfaces have a much smaller aspect
ratio than the wing, and hence become relatively less effective a t high
412 THREE-DIMENSIONAL. SUBSONIC FLOW Ch. 13
FIG.13.13. Influence of aspect ratio on the way in which the lift-curve slope varies
with Mach Number.
Mach Numbers. This may have an important bearing on aircraft

stability a t very high subsonic speeds.
Induced Drag. The minimum induced drag for linearized compress-
ible flow may be shown to be given by the same relation as for incom-
pressible flow, namely, /r-2
Combining this with Eq. 13.22, we have
which shows that the induced drag increases less rapidly with Mach
Number for wings of small aspect ratio than it does for wings of large
aspect ratio.
Experimental Results. The effect of Mach Number on the lift-curve
slope of a 12%-thickness profile with ~3 = 1.15 is shown in Fig. 13.14a.(15)
I t is seen that the Prandtl-Glauert correction factor for two-dimensional
flow is much too large and that the approximate theory developed in
this chapter agrees rather well with the measurements up to the point
where the lift-curve slope drops off as the result of local shock waves.
The moment-curve slope for the same wing is shown in Fig. 13.14b.
The change in the sign of the moment-curve slope a t M, E 0.75 has a
I
crucial bearing on aircraft stability a t high speeds.
Fig. 13.15 represents the results of measurements on wings having a
12%-thickness profile and various aspect ratios.(g) At low Mach Num-
Art. 13.6 WINGS O F F I N I T E SPAN 413
Mm Mm
(a) @)
F I ~ 13.14.
.
0012 profi le (after Gothert).
-
Ex~erimentalresults for wine havine- asuect
. ratio of 1.15 and NACA
ift-curve slope a t zero incidence versus Mach Number, showfing comparison
ith Gothert's ~ l e .
L
(4
NACA 0012 profile in wings of vari
bers, Fig. 13.15a1i t is aerodynamically desirable to have large aspect

ratios in order to avoid the induced drag that goes with wings of finite
span. As the Mach Number increases there is a t first little change in
FIG.13.16. Schlieren photographs of flow past wings with NACA 0012 profile at
4" angle of attack (after Lindsey and Humphreys). See also Fig. 13.3.
(a) a = m; M, = 0.78.
(h) = m; M, = 0.87.
(c) A( = 2; M, = 0.78.
(d) A( = 2; M, = 0.86.
the relative positions of the lift-drag polars because the relation between
lift and induced drag, Eq. 13.23, is substantially independent of Mach
Number. At higher Mach Numbers, and especially a t high lift coef-
ficients, local compressibility effects give rise to shock waves, but this
effect is smaller for wings of low aspect ratio because of the three-dimen-
sional "relief effect." Thus, a t M, = 0.85, Fig. 13.15d1 the order of
Art. 13.7 SWEPTBACK WINGS 415
aerody~amicefficiency for the wings of various aspect ratio<?is com-

pletely reversed as compared with the low-speed tests of Fig. 13.15a.
At high speeds, therefore, wings of low aspect ratio have some distinct
aerodynamic advantages over wings of high aspect ratio.
The changes in flow pattern corresponding to the data of Fig. 13.15
are illustrated by the schlieren photographs of Fig. 13.16. It is seen that
separation owing to shock waves is strongly dependent upon aspect ratio.
13.7. Sweptback Wings

A most effective method for delaying compressibility effects on lifting
surfaces, and one which is commonly employed for high-speed aircraft,
is the use of sweepback. A sweptback wing is one the mean chord line
of which is not perpendicular to the undisturbed flow.
Physical Considerations. The physical concepts underlying this
scheme may best be understood by considering a uniform wing of infinite
span with its leading edge swept back a t an angle u, from the normal t o
the free-stream velocity vector Vm (Fig. 13.17a). Now imagine the
(0) Compressible (b) Incompressible

FIG.13.17. Nomenclature for analysis of sweptback wing.
(a) Compressible flow.
(b) Incompressible flow past affinely related wing according t o lateral contrac-
tion required by Gothert's rule.
observer to move along the span of the wing a t the uniform speed
Vmsin u,. The approach flow then appears t o be normal to the leading
edge, and has the speed V mcos a,. The tangential velocity of the original
flow, Vmsin u,, does not influence the pressure distribution on the wing,
but is important only in the determination of the frictional stresses
on the surface. Since only the normal component of velocity is significant
in the potential flow, it follows that the effective Mach Number of the
flow is M, cos cr,. Thus, although the free-stream Mach Number M,
may be of the order of unity or even greater, the effective Mach Number,
M, cos v,, may, through sufficient sweepback, be made small enough
to avoid the adverse effects of shocks.
In the coordinate transformation mentioned above, the effective thick-
ness normal to the leading edge is also increased, which is unfavorable
as regards the maximum local velocity. This effect, however, is con-
siderably smaller than that of reducing the effective Mach Number
normal to the leading edge.
On the other side of the picture, sweepback has two disadvantages:
First, the lift coefficient is lessened by reducing the normal component
of velocity, thus leading to larger wing areas. Second, severe structural
problems are associated with sweptback wings.
Sweepback may also be used advantageously to improve the efficiency
of propeller tips a t high speeds and to ameliorate compressibility prob-
lems with compressors pumping gases of very high molecular weight
(i.e., low sound velocity).
Another way of picturing the physical effects of sweepback is to im-
agine a wing in a rectangular wind tunnel. If there is no sweepback,
the wing produces a blocking effect in complete planes normal to the flow.
If the wing is swept back, however, there is a comparatively small block-
ing effect in any plane normal to the flow. Hence, a three-dimensional
"reljef" effect is present, inasmuch as the air may avoid the wing by
flowing laterally in two dimensions. The corresponding reduction in the
perturbation velocity has a cumulative effect in compressible flow be-
cause the effects of Mach Number are decreased as the incompressible
disturbance velocity itself is decreased.
Gothert's Rule. I n applying Gothert's rule of similarity, we compare
the. original compressible flow, Fig. 13.17a, with an incompressible flow,
Fig. 13.17bJ past a body whose longitudinal dimensions are unchanged,
but whose lateral dimensions are all decreased by the ratio 8.
Assuming infinite aspect ratio, the geometric relations between the
two wings are
1
tan a< = = - tan U, (13.26)
edge edge edge
Then, at corresponding points in the two flows, the disturbance velocities

are related by
( ~ / V m ) c= ( ~ / v m > i / 8 ~ (13.27)
Influence of Sweepback on Compressibility Effects. From a practical

point of view, we are interested in knowing how sweepback alters the
influence of Mach Number on the maximum local velocity. To arrive
a t an approximate but general rule, let us suppose that the profile is
elliptical. For an elliptical profile with no sweepback, it is known that
in incompressible flow the maximum perturbation velocity divided by
the free-stream velocity is equal to the thickness ratio. Now let us
apply this result to the flow seen by an observer moving along the span
of the wing of Fig. 13.17b a t such a speed that the approach flow appears
to be normal t o the leading edge. For this observer the thickness is ti;
the chord is 1; cos a;; the free-stream velocity is i/', cos a;; and the dis-
turbance velocity is ui/cos ui, where ui is the perturbation velocity for
the incompressible flow seen by a stationary observer. Thus we may
write
ti
or
($)m max
=
ti
(~)icosa; = - 1
+
li 4 1 tan 2 ai
(13.28)
Then, by substituting Eqs. 13.25, 13.26, and 13.27 into Eq. 13.28,
simplifying, and dropping the subscript c for the compressible flow, we
get
tll -
- tll
(13.29)
+tan 1 - M,' +
tan 2 a
For the same profile in incompressible flow (b = I), we get
Finally, division of Eq. 13.29 by Eq. 13.30 yields
a x - (-Cp)
- max -
-
41 + tan a
2
(13.31)
m a ( C dl - Mm2 + tan 2
a
This formula, which may be expected t o apply a t least approximately

to conventional subsonic profiles, shows how sweepback modifies the
effect of Mach Number on the peak negative pressure coefficient, and
is plotted in Fig. 13.18. For zero sweepback, Eq. 13.31 reduces to the
Prandtl-Glauert rule in twodimensional flow. For SO0-sweepback, the
peak negative pressure coefficient is independent of Mach Number. T o
illustrate how sweepback reduces the effects of compressibility on pres-
sure coefficient, Eq. 13.31 shows that for an increase from M, = 0 to
M, = 0.8, the pressure coefficient is increased by 67% without sweep
back; by 39y0 for 30°-sweepback; by 21% for 45"-sweepback; and by

only 9% for 60°-sweepback.
An important fact shown by Eq. 13.29 is that whereas for unswept
wings the incremental velocity tends to become infinite as M, approaches
unity, for swept wings this tendency occurs a t supersonic speeds, accord-
ing to the relation M, = dl +tan 2 a. This result must be looked on
as only indicative, inasmuch as the linear theory fails for large perturba-
tion velocities and also near M, = 1.
F I ~13.18.
. Effect of sweepback on the way in which the maximum perturbation
velocity for thin elliptical profiles varies with Mach Number.
Effect of Sweepback on Lower Critical Mach Number. Suppose that

a wing is a "bent-back" type of swept wing, i.e., the wing as a whole is
rotated so that the profile in a plane normal to the leading edge remains
unchanged. This means that the profile in a plane parallel to V, becomes
thinner as u increases. Therefore, if 6 is the thickness ratio in the flight
direction, and 6~ is the thickness ratio in the direction normal to the
leading edge, 6 = 6N cos u, so that Eq. 13.29 may be written
With bentrback wings, 6~ remains constant as u increases. Further-

more, the value of u/V, is uniquely related to the local Mach Number
and M, through the isentropic relations. From Eq. 13.32, therefore,
we may plot, for a given thickness ratio, curves of maximum local Mach
Number versus M,, one curve being required for each sweepback angle.
The value of M, where each of these curves reaches a maximum local
Mach Number of unity defines the corresponding lower critical speed,
& , for the given 6N and u.
The results of such calculations (I6) for sweptback wings of infinite
span with elliptical profiles are summarized in Fig. 13.19. As an example,
for a profile of 10yo thickness ratio normal to the leading edge, a change
from an unswept wing to 45'-sweepback raises the M,

0.93.
..
from 0.82 to
If the wing is "sheared back" in such a way that the profile parallel
to V , remains constant, then Eq. 13.29 rather than Eq. 13.32 is to be
used. Since shear-back increases the thickness normal to the leading
edge, this type of sweepback is not as effective as bend-back. As an
example, for an elliptical profile of 10% thickness ratio parallel to V,,
-cff '40'; Mw=0.85; M, cos v = 0.65
0.720 , 30" 60' 90' M=O ILocd

05 1.0
Sweep - Bock Angle, u
FIG. 13.19. Lower critical Mach Num- FIG. 13.20. Experimental results for
ber as a function of sweepback angle and flow past sweptback wings of infinite
thickness ratio normal to leading edge, span with 9%-thick profile (normal to
for thin elliptical profiles (after Gothert). leading edge) at incidence of -2" (after
Lippisch and Reuschausen).
a change from 0'-sweepback to 45'-sweepback increases M, from 0.82

to 0.90, as compared with the change from 0.82 to 0.93 for the corre-
.,
sponding bent-back wing.
These estimates of lower critical Mach Number are based on the
criterion of the local total velocity becoming sonic. If, as an opposite
extreme, only the normal component of velocity were assumed to be
significant, the resulting lower critical Mach Numbers would be sub-
stantially greater than those shown in Fig. 13.19, and could easily
exceed unity for large angles of sweepback. The true state of affairs
probably lies between the two simplified extremes.
Experimental Results. To illustrate the principle that the pressure
distribution is determined primarily by the normal component of veloc-
ity, Fig. 13.20 shows the results of experiments (17)on a 9%-thick sym-
metrical profile, with maximum thickness a t 40% chord, a t an angle of
420 THREEDIMENSIONAL. SUBSONIC FLOW Ch. 13
attack of -2". Data are shown for three conditions, with different
values of u and M, but with the same value of M, cos u. The ordinate
represents the pressure coefficient based, in each case, on the corre-
sponding normal component of velocity. The principle under discussion
is, a t least approximately, verified by the fact that the three sets of
data are nearly superposed when plotted in this manner.
Fig. 13.21a, from the same set of experiments, demonstrates the same
principle by showing that curves of C~/cos' u versus M,cos u are almost
Q . 13.21 Experimental results for

wings of Fig. 13.20, but with a = +3'
(after Lippisch and Beuschausen).
(a) Curves of CL versus M, correlated
in terms of normal component of
velocity.
(b) Effect of sweepback on CL and on
liftdivergent Mach Number.
independent of sweepback angle a. I n Fig. 13.21b, where the effect of u

is omitted from the coordinates, we see that sweepback reduces the
lift coefficient significantly but also increases the Mach Number a t
which compressibility produces a sudden droop in lift-curve slope. For
example, 40°-sweepback reduces the lift coefficient by some 4070, but
increases the lower critical Mach Number from about 0.82 to ahout 0.90.
13.8. Sweptback Wings of Finite Span

In practice, of course, wings are of finite span, and the beneficial
effects of sweepback on compressibility are reinforced by the effects of
aspect ratio. Fig. 13.22 shows the results of experiments (la) on bent-
back wings of constant span along the 50%-chord line, formed by rotat-
ing the original unswept wing halves ?bout the intersection of the line
of symmetry and the 50%-chord line. The lift-drag polars of Fig. 13.2213
show the considerable improvement a t high subsonic speeds produced
by 45%-sweepback. Fig. 13.22~shows that, for a fixed CL, sweepback
has scarcely any effect up t o M, 2 0.7. At Mach Numbers between
Art. 13.8 SWEPTBACK WINGS O F F I N I T E SPAN 421
0.8 and 1.2, however, sweepback produces remarkable reductions in
drag coefficient.
An extensive comparison between the theoretical similarity methods
outlined here and a large number of experiments on wings of various
aspect ratios and sweepback is given in Reference 19.
FIG. 13.22. Experimental data on

sweptback wings of various sweepback
angles (after Ludwig).
(a) Change in planform produced @
sweepback.
Sweep- Back Angle, u (b) Lift-drag polars a t M, = 0.8.
(c) Drag coefficient versus sweepback
(4 angle, for CL = 0.2.
A generalized correlation incorporating the effects of both aspect

ratio and sweepback, and based on the methods outlined in this chapter,
has been proposed by Diederich.el) The correlation is based on the
parameter
Az
F z
-
1 ~CLN
(da)
2lr
cos .
where CLN is the section lift coefficient of the profile, measured with
respect to the speed and chord in a plane normal to the mean chord line;
FIG. 13.23. Correlation of lift-curve slopes for various planforms (after Diederich).
and a is the angle of attack (in radians) of the entire wing measured in a
plane parallel to the axis of symmetry of the wing.
Using CL to denote the lift coefficient of the entire wing, based on V,,
the total planform area, and the incidence a, the semi-empirical law
propos
- F
- - (13.33)
( d C ~ ~ / d cos
a) u
- 2- +, F-
4
1;
y - ' $
A comparison of this formula with accurate theoretical and exper-
imental results is shown in Fig. 13.23 for a wide variety of planforms.
The validity of Eq. 13.33 as a general correlation rule is seen to be
quite satisfactory.

1. GOTRERT,B. Plane and Three-Dimensional Flow a t High Subsonic Speeds,
NACA Tech. Memo., No. 1105 (Oct., 1946).
2. HESS, R. V., and GARDNER, C. S. Study by the Prandtl-Glauert Method of
Compressibility Effects and Critical Mach Number for Ellipsoids of Various
Aspect Ratios and Thickness Ratios, NACA RM, No. L7B03a (March,
1947).
3. SCHMIEDEN, C . , and KAWALKI,K. H. Contribution to the Problem of Flow at
High Speed. Part 11: Effect of Compressibility in Axially Symmetrical Flow
around an Ellipsoid, NACA Tech. Memo., No. 1233 (June, 1949).
P. LEES, L. A Discussion of the Prandtl-Glauert Method to Subsonic Compressible
Flow Over a Slender Body of Revolution, NACA Tech. Note, No. 1127 (Sept.,
1946).
6. HERRIOT,J. G. The Linear Perturbation Theory of Axially Symmetric Com-
pressible Flow with Application to the Effect of Compressibility on the P r e s
sure Coeficient a t the Surface of a Body of Revolution, NACA RM, No.
A6H19 (July, 1947).
6. LAITONE, E. V. The Subsonic Axial Flow About a Body of Revolution, Quar-
terly App. Math., Val. 5, No. 2 (1947), p. 227.
7. V A N DRIEST, E. R. Die linearisierte Theorie der dreidimensionale kompres-
siblen Unterschallstromung und die experimentelle Untersuchung van
Rotationskorpern in einem geschlossenen Windkanal, Mitt. Nr. 16, Institut
fur Aerodynamik, E.T.H., Zurich, 1949.
8. M UNK, M. M. Fluid Mechanics, Part 11, Division C, of Aerodymmzc Theory,
edit. by W. F. Durand, Val. I, P. 293. (Reprinted by Durand Reprinting
Committee, Calif. Inst. of Tech., Pasadena, 1943.)
9. STACK, J., and LINDSEY, W. F. Characteristics of Low-Aspect-Ratio Wings st
Supercritical Mach Numbers, N A C A Tech. Note, No. 1665 (Aug., 1948).
10. LAMB, H. Hydrodynamics. 6th ed. New York: Dover Publications, 1945, p. 141.
11. YON K ARMAN, T H. Calculation of Pressure Distribution on Air Ship Hulls,
N A C A Tech. Memo., No. 574 (1930).
12. GOTHERT, B. Comparison of Drop and Wind-Tunnel Experiments on Bomb
Drag a t High Subsonic Speeds, N A C A Tech. Memo., No. 1186 (May, 1948).
13. KAPLAN, C. The Flow of a Compressible Fluid Past a Sphere, N A C A Tech. Note,
No. 762 (May, 1940).
14. CHARTERS, A. C., and THOMAS, R. N. The Aerodynamic Performance of Small
Spheres from Subsonic to High Supersonic Velocities, Jour. Aero. Sn'., Val. 12,
No. 4 (Oct., 19451, p. 468.
15. GOTHERT, B. H. High-Velocity Measurements for a Wing of Small Aspect
Ratio, Air Materiel Command Report, No. F-TS-646-RE (Sept., 1946).
16. GOTHERT, B. Calculation of the Velocity Field of Swept-Back Wings a t High
Subsonic Speeds, Air Materiel Command Report, No. F-TS-3183-RE (Feb.,
1948).
17. LIPPISCH,A,, and BEUSCHAUSEN, W. Pressure Distribution Measurements a t
High Speed and Oblique Incidence of Flow, N A C A Tech. Memo., No. 1115
(Mar., 1947).
18. LUDWIG, H. Test Results on Arrow-Head Wings a t High Velocities, Air Materiel
Command Report, No. F-TS417-RE (Oct., 1946).
19. MURRAY, H. E. Comparison with Experiment of Several Methods of Predicting
the Lift of Wings in Subsonic Compressible Flow, N A C A Tech. Note, No. 1739
(Oct., 1948).
20. MAWHEWS, C. W. A Comparison of the Experimental Subsonic Pressure Dis-
tributions About Several Bodies of Revolution with Pressure Distributions
Computed by Means of the Linearized Theory, NACA RM, No. L9F28
(1949).
21. DIEDERICH, F. W. A Plan-Form Parameter for Correlating Certain Aerody-
namic Characteristics of Swept Wings, NACA Tech. Note, No. 2335 (1951).
PROBLEMS
13.1. A tunnel of circular cross-sectional area At is blocked by a model of

circular cross-sectional area A,.
Demonstrate by Gothert's rule that the incremental velocity produced by the
blocking is given approximately by
Compare this with the exact result

of one-dimensional isentropic flow,
namely,
dV
13.2. A certain wind tunnel of circular cross section has a contraction section
designed for incompressible flow to avoid at all points on the wall regions of
minimum pressure followed by adverse pressure gradients. Show that this same
contraction section will have no adverse pressure gradients for compressible
subsonic flow.
13.3. At low speeds the static-pressure hole of a pitot-static tube registers

an error given by
.
Estimate how large the corresponding error would be at M, = 0.85.
13.4. Assuming that for slender bodies of revolution in incompressible flow
the curve of peak negative pressure coefficient versus thickness ratio is the same
as for an ellipsoid of revolution, plot the lower critical Mach Number versus the
low-speed peak negative pressure coefficient.
13.5. Using Eq.13.21a, show that the compressible lift-curve slope for wings
of finite aspect ratio is related to the corresponding low-speed slope by
13.6. It has been shown by Bollay [Z.a.M.M., Bd. 19 (1939), p. 211, that
for incompressible flow past a flat plate of vanishingly small aspect ratio, the
lift coefficient is related to the angle of attack by the formula
Using Gothert's rule, find a similar relation for compressible flow. Compare
your results with the discussion in Art. 13.6.
13.7. Consider a wing of infinite span with 45"-sweepback and a 10%
elliptical profile. Assuming that the tangential component of velocity is of no
consequence, find the value of M, for which the local sonic velocity is first reached
at the surface of the wing. Compare your result with Fig. 13.19.
PART V
SUPERSONIC FLOW
NOMENCLATURE FOR CHAPTER 14
angle of attack Cartesian coordinates
speed of sound
drag coefficient Mach angle
lift coefficient thickness ratio; angle
moment coefficient coordinate normal to chord
pressure coefficient turning angle of streamline
drag (radians)
amplitude of wave-shaped wall 180 e / ~
ratio of specific heats coordinate along chord
wave length of wave-shaped mass density
wall; also chord of airfoil surface slope of profile with
lift respect to chord line
Mach Number perturbation velocity potential
counter-clockwise moment signifies free-stream condition
about leading edge signifies upper surface of airfoil
pressure signifies lower surface of airfoil
pressure index (Eq. 14.20)
surface area of profile signifies step along Mach wave
area between chord line and of family I, or across Mach
profile surface wave of family II.
perturbation velocity compo- signifies step along Mach wave
nents of family 11, or across Mach
velocity wave of family I.
Chapter 14
TWO-DIMENSIONAL, SUPERSONIC FLOW WITH SMALL

PERTURBATIONS

By virtue of a radical change in the properties of the differential
equations, the exact solution of problems in two-dimensional, friction-
less, supersonic flow is far easier than for subsonic flow. When the flow
is shock-free and irrotational the exact solution of supersonic problems
may even be said to be simple. Even with shocks and rotation present,
methods of solution are available which, although requiring extensive
numerical computations, are still manageable.
On the other hand, these exact solutions (Chapters 15 and 16) almost
always require either numerical or graphical, rather than analytical,
procedures. Therefore, in spite of the existence of such exact numerical
methods, there is a useful place for analytical, but approximate, methods.
Such approximate analytical methods provide the rapid means of
calculation which are so important for making preliminary estimates.
And, by showing clearly the relationships among the variables, they
provide a better insight into problems and permit general conclusions
to be drawn.
I n this chapter we shall discuss an approximate analytical method
which applies to two-dimensional, supersonic flows which are everywhere
shock-free and irrotational and in which there are only small perturba-
tions from a uniform, parallel flow. It is true that most supersonic
flows involve shocks, but, as will be seen in Chapter 16, the entropy
changes across weak oblique shocks are so small that they may often be
disregarded. Accordingly, in many problems, the magnitude of the
rotation is negligible.
The method of small perturbations yields useful results for such prac-
tical problems as the flow past thin airfoils with sharp leading and trail-
ing edges, the flow through turbine and compressor cascades with thin
blades and small turning angles, the flow inside of two-dimensional
ducts, and the flow at the exit of supersonic nozzles discharging into a
region of variable back pressure.
The solutions for linearized supersonic flow show, in especially simple
form, many of the important features of more complex supersonic flow
patterns. Hence it is often helpful to work out the approximate linear-
ized solution to a problem prior to carrying out the comparatively
laborious exact solution.
427
428 SUPERSONIC FLOW WITH SMALL PERTURBATIONS Ch. 14
A further advantage of the linearized procedure is that it brings home

clearly the vast differences between subsonic and supersonic flow with-
out a t this stage becoming lost in the mathematical complexities of a
more general approach.
Additional material relevant to the subject matter of the present chap-
ter may be found in Chapters 15 and 16 and in Volume 11, Chapter 17.
14.2. Linearization of the Equations

Perturbation Velocity Potential. Assuming that the flow is two-
dimensional, irrotational, and isentropic, and that there are only small
perturbations from a uniform, parallel flow along the x-axis, we may
follow exactly the procedure of Art. 10.2 for subsonic flow. In this way
we arrive at the linearized equation of motion, Eq. 10.21. We write the
where q~ is the perturbation velocity potential. The perturbation

velocity components are given by
u = aP/ax; v = acp/ay (14.2)
Restrictions on Linear Theory. In arriving a t the linearized equation,
it was necessary to assume that
Thus, as in subsonic flow, Eq. 14.1 loses validity in the transonic

region, i.e., near Mach Number unity.
In supersonic flow still another restriction appears, namely, that the
Mach Number may not be too high. Although no definite upper limits
can be given, it seems clear from the above assumptions that for prac-
tical purposes the maximum allowable order of magnitude of M m 2 is
about ten. Or, more precisely, since u / V m and v/Vm are of the same
order of magnitude as the thickness ratio 6 of thin bodies, we may say
that the linearized theory may be used when Mm26 is small compared
with unity.
Pressure Coefficient. Making assumptions similar to those of Art.
10.3, we find that the linearized pressure coefficient for supersonic flow
is the same as in subsonic flow, namely,
provided that u/Vm is small compared with unity.

Art. 14.3 SOLUTION FOR LINEARIZED SUPERSONIC FLOW 429
14.3. The General Solution for Linearized Supersonic Flow

Subsonic Versus Supersonic Flow. It might at first appear from the
similarity between Eqs. 14.1 and 10.21 that the linearized solutions for
subsonic and supersonic flow should be much the same. However, the
nature of the flow pattern proves to be entirely different depending on
whether M, is greater or less than unity. It may be recalled that in the
various one-dimensional analyses the sign of the term (Mm2- 1) was
a determining factor in the qualitative nature of the flow. If we write
Eqs. 14.1 and 10.21 in the form
and note that au/dx denotes longitudinal accelerations while a v / & ~is
connected with the convergence or divergence of the elementary stream
tube formed between two neighboring streamlines, we see that this
equation represents the already familiar fact that an acceleration re-
quires an increase in cross-sectional area for supersonic flow and a de-
-
crease for subsonic flow. Because av/ay and au/ax are of the same sign
for supersonic flow and of different sign for subsonic flow, the resulting
flow patterns are of different type. As in the one-dimensional simplifica-
tion, the long-familiar concepts of incompressible flow prove to be mis-
leading if applied to supersonic flows.
For subsonic flow Eq. 14.1 is said to be elliptic in type, and in fact
may - be reduced to the Laplace equation by the transformation given in
Chapter 10. Hence subsonic flows are, a t least qualitatively, similar to
incompressible flow. For supersonic flow Eq. 14.1 is said to be hyper-
bolic in type, and may be reduced to the classical wave kquation by a
simple transformation. Hence, linearized supersonic flows have many
properties akin to the properties of vibrating metal bars, air columns,
and strings. A fuller explanation of the mathematical properties of
elliptic and hyperbolic differential equations is presented in Appendix A.
Solution to the Differential Equation. We shall now show that for
supersonic flow the general solution to Eq. 14.1 is given by
c p = c p l ( x + m y ) + cp2(x- a?/ (14.4)
where cpl is an arbitrary function of the argument (z 1- +
y),
and cp2 is an arbitrary function of the argument (x - m y ) .
To demonstrate that this is indeed the solution, we first examine the
particular solution cp,. Taking derivatives,* we have
where v' is the first derivative of rp with respect to the argument Q.

By direct substitution, we see that the arbitrary function p1 satisfies

Eq. 14.1 exactly, and is therefore a particular solution. Similarly, it
may be shown that the arbitrary function cp2 is also a solution. Our
problem now is to find the particular forms of these functions corre-
sponding to any desired flow pattern.
Streamlines and Pressure Distribution. The arbitrary nature of the
functions pl and pz is a great simplification, for any number of solutions
may now be found a t will merely by specifying the forms of these func-
tions. Even better, we may easily determine the flow pattern corre-
sponding to specified boundaries. Let us write Eq. 14.4 in the following
equivalent form :
Then
u = ap/ax = v,(flf+ jzf)
For the differential equation of t,he streamline we write
where (to facilitate the subsequent integration) we have inserted in the

denominator the term Mm2u/V,, which has been assumed small com-
pared with unity. Rearranging, we have
= ',l/Mm" 1 (djl - dj2) '
where the last relation is obtained by noting that

ah
djl = - dx
ax
+ -,dy
afl
a~ i
= jlldx +m j l f dy, etc.,
the symbol flfdenoting differentiation of the function f1 with respect to

the argument (x 1- + y).
Integrating, we get for the equation of the streamline,
ystr = constant + VM,~- 1 (jl- j2) (14.6)

in which the magnitude of the constant identifies a particular stream-
line.
The pressure coefficient is now found to be
Within the assumptions of the linear theory, this may also be written a3
If the functions ji and ji are specified, the shapes of the streamlines

and the pressure distribution may be found from Eqs. 14.6 and 14.7. Or,
conversely, if the shape of a solid body immersed in the stream is known,
and if the functions jl and j2 are chosen to satisfy the body streamlines
according to Eq. 14.6 and also to satisfy the remaining initial conditions
necessary to determine (but not to overdetermine) the problem, then
Eqs. 14.5 through 14.8 give the linearized solution for the flow past the
solid body.
Flow with Waves of One Family. The general nature of supersonic
flow with small perturbations may best be brought out by considering
the two simple cases where either the function jl or the function ji is
degenerate, i.e., the function is constant throughout the entire flow field.
The particular solutions then obtained are referred to as simple wave
solutions, or solutions with waves of one jamily only. For the two classes
of simple-wave solutions, the foregoing results take the simple form
shown in the following table:
fi = constant fi = constant
I
Flow with Left-Running Mach Waves. Consider the simple-wave

flow for which the function fi is everywhere constant. We see from
Eq. 14.10 that the slopes of the streamlines are then constant along
lines on which j2' is constant. However, fi' is a function only of the
argument (z -1 - y). Consequently the slopes of the stream-
lines are constant on lines for which
x- d ~- 1,y =~constant
These form a family of parallel straight lines with slope dyldx =
1 Referring to the velocity triangle of Fig. 14.1, we see
FIG.14.1. Illustrates definition of Mach angle.
from the definition of the Mach angle (Chapter 3) that

cm 1 1
sin a, = - =- ; tan a, =
v m M, v'xF7
Hence the lines of constant j2' (and of constant j2) are inclined a t the
angle a, to the direction of flow. Thus the lines of constant fi are
identified with the left-running Mach lines of the flow. The left-running
Mach lines are those which, for an observer looking downstream,
appear to be going downstream in a generally leftward direction; con-
sequently the left-running Mach lines lie a t the angle or, above the
x-axis, while right-running Mach lines lie a t the angle a, below the x-axis.
The general relation between the streamlines must be as shown in
Fig. 14.2. All the streamlines are similar to each other and, moreover,
Left-Running Mach Waves
F I ~14.2.
. Flow with only left-running Mach waves present.
have equal slopes along each left-running Mach line. According to

Eq. 14.11 the pressure, velocity components, and all other fluid proper-
ties are also constant along the left-running Mach lines. Moreover, the
local pressure coefficient, according to Eq. 14.11, depends only on the

local inclination of the streamline, that is, only on the local direction of
the velocity vector.
In the case considered here, the changes in fluid properties may be
said to be propagated along the left-running Mach lines. A propagation
of this type is usually called a wave. It is customary, therefore, to
refer to the Mach lines as Mach waves. The latter may also be thought
of as pressure waves propagating along the Mach lines, according to the
simple descriptions of Chapter 3.
Similar results are obtained if we consider the other class of simple
flows, that is, with the function fi everywhere constant. In this case,
however, the Mach lines are inclined at the angle -a, to the direction
of flow (Fig. 14.3).
Choice of Wave Family. Consider the streamline ABCD, which is
the same in Figures 14.2a and 14.3a. Suppose the region below ABCD
Right -Running Mach Waves
F I ~14.3.
. Flow with only right-running Mach waves present.
(Fig. 14.2b) were made solid. Then ABCD might be considered the
surface of a body with unbounded flow above the surface. We now
inquire which of the two solutions (jl or f2) is valid, or whether some
combination of the two might be valid. The answer depends partly on
the boundary conditions. If we specify no upper boundary to the flow
and that the flow far upstream is uniform and parallel, then the function
fi must be everywhere constant, for otherwise the flow far upstream
.
could not be uniform and parallel. From another point of view, if we
consider the Mach lines as representing the lines of propagation of
pressure waves produced by changes in direction at the solid boundary,
these waves can physically be propagated only downstream in super-
sonic flow, and hence only the pressure waves corresponding to the
function f2 can appear in the solution. Accordingly, the flow pattern
must be as indicated in Fig. 14.2b.
By similar arguments, if the region above ABCD were made solid; the
flow pattern would be as shown in Fig. 14.3b.
If the curve BC represented a thin airfoil moving a t supersonic speed
through still air (or, equivalently, located in a supersonic wind tunnel
having a uniform, parallel air stream), the flow pattern would be as
shown in Fig. 14.4. The pressure distribution on the upper and lower
FIG.14.4. Flow past thin cambered profile. Above the profile the flow is affected
only by left-running Mach waves; below, only by right-running Mach waves.
surfaces of the airfoil (at each point of which the slope of the stream-
line is known) may easily be calculated from Eq. 14.11. That the flow
downstream of the airfoil must have the direction and pressure of the
upstream flow follows from the fact the two streamlines which join a t
the trailing edge must have the same pressure and direction; from the
two forms of Eq. 14.11, this will be so only when the flow is as shown
in Fig. 14.4.
14.4. Geometrical Interpretation of the General Solution

Physical Description of Flow. From the results of the preceding
article, we see that the general mathematical solution to linearized
supersonic flow may be interpreted physically by portraying the flow
as being influenced by two families of pressure waves which are inclined
a t the Mach angle to the velocity vector. Having arrived a t this con-
cept, i t is unnecessary to work with the potential functions directly.
Instead, i t is easier to proceed by constructing the wave pattern which
produces a flow corresponding to specified boundary conditions. In
order to do this, however, we must have a method for calculating the
changes produced in a streamline as the streamline crosses waves of
either family. The relations between changes in pressure and changes
in direction are given by Eq. 14.11. Here we shall derive the same
relations by a geometrical construction, due to Ackeret, which empha-
sizes the physical rather than the mathematical aspects of the problem.
From the form of the general solution, Eq. 14.4, it is evident that the
perturbation potential rp may be considered as the sum of the perturba-
tion potentials of two families of waves, denoted by the symbols I and 11,
and corresponding to the functions rpl and qZlrespectively. The waves
of family I are the right-running Mach lines, and those of family II are
Art. 14.4 GEOMETRICAT, INTERPRETATION OF SOLUTION 435
the left-running Mach lines. Now, since (pl depends only on the argu-
+
ment (z ' d ~-, 1 ~y), the lines of constant (pl are all inclined a t the
angle -a, to the flow;similarly, the lines of constant (p2 are all inclined
a t the angle or, to the flow. We may then consider the flow field to be
covered with these lines, as shown
in Fig. 14.5; the specific nature of
the flow pattern then depends on
(x -= y = const.)
Lines of Constant rp,
the numerical magnitudes of (pl n n n n n n
and 9, attached to each of these
lines.
Suppose that in a small region
only the function (p2 undergoes any x
change in numerical value, that is,

the function 91 is a constant, and
we have waves of only one family. '\'lYYYYY\ \
Then, since the gradient of the I I I l I I I I
Lines of Constant9
perturbation potential is equal to
(x + G y = const.]
the vector change in velocity, and
since the gradient of the perturba- FIG. 14.5. Two families of Mach lines.
tion potential is normal to the lines
of constant perturbation potential, it follows that as the streamline crosses
the specified region the change in the velocity vector must lie in the
direction normal to the lines of family I I . The velocities before and after
the change are, therefore, related geometrically as shown in Fig. 14.6.
FIG. 14.6. Velocity vector diagram for a streamline influenced by left-running pres-
sure waves.
Pressure-Turning Angle Relation for Mach Wave. To calculate the

relation between the change in pressure and the change in direction as
the streamline crosses a region where there are pressure waves only of
family IT, we start with the linearized form of the Euler equation,
P - P, = --PWV&
But, from the geometry of Fig. 14.6,
u = - v tana,
Furthermore, if A6 is the change in direction of the streamline, and if i t
4d6 SUPERSONIC FLOW WITH SMALL PERTURBATIONS Ch. 14
is assumed positive when measured counter-clockwise, then Fig. 14.6

shows that, to first-order terms,
v = vmA0
Hence. we find that
pm Vm2 ~ P , & O ~
(p - p , ) ~= pmVm2 A8 tan am= A0 =
4-
The pressure coefficient is accordingly given by the simple formula
The subscript I is used here to indicate that in crossing a Mach wave of

family 11 we are fcllowing a course of states for which the function
is constant or on which the argument of pl is constant. Or, more in line
with the generalization of these ideas by the method of characteristics
(Chapter 15), the subscript I signifies that the crossing of a Mach line
of family 11 may be accomplished by following a Mach line of family I.
Similarly, if we were to consider the influence of a system of waves of
family I, we would find that
It may be seen by comparison that these relations agree with Eq. 14.11.
Four Basic Waves. I n constructing solutions by means of waves, we
may synthesize a flow pattern corresponding to specified boundary
conditions out of four fundamental types of changes, depending (i) on
whether the wave is of family I or of family 11, and (ii) on whether the
stream is accelerated or decelerated. These four types of changes are
illustrated in Fig. 14.7. From these sketches and Eqs. 14.12 and 14.13,
we see that the pressure increases for a deceleration and decreases for an
acceleration, as required by momentum considerations. Furthermore,
during a compression the streamline bends tmard the Mach line,
whereas during an expansion the streamline bends away from the Mach
line; by considering the cross-sectional area of a stream tube formed by
two neighboring streamlines, it will be seen that in a compression the
area decreases while in an expansion the area increases, as might be
expected from simple one-dimensional considerations for supersonic flow.
When a streamline is influenced by a wave, i t may be seen from
Fig. 14.6 that the velocity component parallel to the wave is unaltered.
Only the velocity component normal to the wave is changed in magni-
"i .\.
Acceleration
\
./'
-A
\
Deceleration
\ -u
Acceleration Deceleration
\ -ne
/-
F I ~14.7.
. Four basic types of Mach waves.
tude. Or, stated differently, the vector change in velocity is in the direc-
tion for which the component of velocity is equal to the local sound
velocity.
Fig. 14.8 shows the compression Mach waves of family I and family
I 1 generated a t the nose of a thin wedge-shaped body.
FIG.14.8. Shadow pliotograrh oI flow (from left to right) past thin wedge-shaped
body (M.I.T. Gas Turbine Laboratory).
14.5. Flow Past a Wave-Shaped Wall

As an application of the foregoing ideas, let us consider the wave
shaped wall of Art. 10.4, having the equation
2n2,
Y , = h cos -
I
Perturbation Potential. Assuming that the wall has only a small
waviness (y, << x,) and that the flow is from left to right and above
the wall, we may eliminate pressure waves of family I on the grounds
that they cannot physically be propagated upstream. To satisfy the
equation of the streamline at the wall, the function j must, in accordance
with Eq. 14.9, be given by
Pressure Distribution. Using Eq. 14.11, we find that
Comparison of Subsonic and Supersonic Flow. The streamlines and

pressure distribution are shown in Fig. 14.9.a. How the maximum
pressure coefficient varies with Mach Number is shown in Fig. 14.9b
for both supersonic and subsonic flow. Many of the qualitative dif-
ferences between subsonic and supersonic flow are illustrated by com-
paring this solution with the subsonic solution of Art. 10.4. I n super-
sonic flow the effect of the wall is propagated out to infinity along the
Mach lines with undiminished intensity; in subsonic flow the effect is
propagated in all directions, but decays rapidly with increasing dis-
tance from the wall. In supersonic flow the flow pattern is unsymmet-
rical about a vertical axis through a crest or trough; in subsonic flow it
is symmetrical. In supersonic flow the pressure coefficient goes through
maxima and minima where the streamline has its greatest inclination;
in subsonic flow the pressure coefficient goes through maxima and minima
where the streamline has its largest curvature. In supersonic flow the
pressure coefficient decreases as the free-stream Mach Number increases;
in subsonic flow the opposite is true. In subsonic flow it was possible to
Art. 14.5 FLOW PAST A WAVE-SHAPED WALL 439
specify that the perturbation velocity shall vanish a t large distances

from the wall; in supersonic flow this is not possible, thus indicating that
the necessary and allowable amount of initial data (that is, boundary-
value information) are different in the two types of flow, corresponding
to the difference in the types of differential equation describing the
FIG.14.9. Flow past wave-shaped wall (after Ackeret).

(a) Streamlines and surface pressure distribution.
(b) Effect of Mach Number on peak pressure coefficient.
motion. This simple example thus demonstrates that the experience

gained from dealing with subsonic flow problems does not help in the
understanding of supersonic flow.
Solution by Construction of Wave Pattern. The solution given above

was obtained by seeking the proper form of the potential function. We
could also have proceeded by constructing the wave pattern. Suppose
that we were to start with the known shape of wall extending upstream
and downstream to infinity, with the flow unbounded on top of the
wall. Then, since there is no way of originating pressure waves of
family I, there can be present only waves of family IZ. From each
point of the wall of Fig. 14.8 an infinitesimal wave is propagated a t the
Mach angle into the flow, the turning angle of the stream across the
wave being determined by the dxferential change of angle a t the wall.
Since the waves are propagated with unchanged intensity (i.e., pressure
rise), i t follows that the velocity vector and all stream properties are
constant along each of the Mach lines. The streamlines may therefore
be constructed simply by transferring slopes along the Mach lines. Eq.
14.12a shows that for simple waves the pressure coefficient is linearly
proportional to the turning angle of the stream; hence, the local pres-
sure coefficient on a given streamline depends only on the !ocal slope of
r;)
the streamline. For the solid wall, as an example,
h 2nx,
= = - 2~ - sin -
S 1 I
and, using Eq. 14.12b, we get
which is identical with the result obtained previously.
14.6. Supersonic Airfoils

One of the most important applications of the method of small
perturbations is to bring out in an approximate but general may the
FIG. 14.10. Nomenclature for analysis of thin supersonic profile.
properties of supersonic airfoils of infinite span. Experience has shown

that the leading and trailing edges of supersonic airfoils should be sharp,
or a t most there should be only a slight rounding of the leading edge;
otherwise, there are large penalties in drag owing to the presence of de-
tached shocks which stand ahead of the airfoil. We shall consider,
therefore, profiles of the general form shown in Fig. 14.10, while keeping
in mind that the results will be valid only for small angles of attack and
for values of ~,'6 small compared with unit,y.
Flow Pattern. Using the concept of pressure waves generated a t the
surface of the airfoil, we construct the wave pattern and streamline
pattern of Fig. 14.10a. Since the pressure disturbances can be propa-
gated only downstream, the pressure waves on the upper surface must
be of family I 1 (left-running) and those on the lower surface of family I
(right-running). Thus the flow field is divided into several regions in
Art. 14.6 SUPERSONIC AIRFOILS 441
cach of which there are simple-wave patterns. From Eqs. 14.12 and
14.13, therefore, the local pressure at any point on the surface depends
only on the local inclination of the tangent to the surface.
Pressure Distribution. Let C;, 7 be a system of coordinates which lie
in the airfoil and which are therefore independent of angle of attack,
and for convenience let [ lie along the chord line. Also, let a denote the
local inclination of the surface measured from the chord line, and let
subscripts U and L refer to the upper and lower surfaces respectively.
Note that 7~ is measured upward and 7~ downward; similarly, uu is
measured counterclockwise and aL clockwise.
Referring to the geometry of Fig. 14.10b, the local directions of the
streamlines a t the surface, relative to the undisturbed flow, are
From Eqs. 14.12 and 14.13 we find the local surface pressures to be
Subtracting the second from the first of this pair of equations, we get,
for the pressure difference at any chordwise location,
L i t CoeEcient. The forces on the profile may now be found by in-

tegrating the pressure over the entire surface. Resolving the pressure-
area forces on the typical surface elements shown, and making use of the
approximations permitted by the linear theory, we have
~ L=
L p~ d s cos
~ (aL + a) 2 PL dt
dLu = -pu dsu cos (uU - a ) Z - p ~d t
so that
dI, = dLu + d L L = ( p L - p U ) d t = v k' pMd r~-n ~21 (UL - au f 2a) d(
Integrating this expression between C;'= 0 and E = 1, we find the
total lift per unit span to be simply
inasmuch as
and, likewise,
The lift coefficient and lift-curve slope are now computed as
Drag Coefficient. Proceeding similarly, we calculate the elementary

drags as
+
~ (UL -k a ) s ~ L ( U L a ) d t
DL = P L d s sin
dDu = p~ dsu sin (UU - a ) s pu(uu - a ) d[
Using the previous expressions for p i and pv, we find
Expanding this expression, integrating, and noting that, as before,
ituL
s lzuu
dl 0; d [ s O
- -
where UL' and uu2 are the mean squares of the slopes, as defined by
The drag coefficient is then
INDUCED DRAG. We see that the pressure drag may be thought of as

the sum of two parts. The first part is independent of the profile shape
and depends only on the angle of attack. Since the angle of attack is,
by Eq. 14.14a, directly associated with the generation of lift, this part
of the drag is known as the wave drag due to lijt, or, following the nomen-
clature of subsonic fluid dynamics, the induced drag:
However, the analogy to subsonic flow is somewhat overextended. In

subsonic flow the induced drag is associated with the kinetic energy in
the tip vortices which (according to Thomson's vortex theorem) are
shed from lifting wings of finite span. The term "induced drag" for
supersonic flow, on the other hand, is associated with the energy in the
ever-spreading wave system of wings of infinite span.
THICKNESS DRAG. The second part of the drag depends only on the
profile shape, and is known as the wave drag due to thiclcness:
SKIN-FRICTION DRAG. SO far we have calculated only the drag forces

resulting from pressure stresses acting on the surface of the profile. If
the concept of a thin boundary layer flow which does not interact with
the potential flow is applicable, it is clear that we may add the shear
drag directly to the pressure drag. Letting CDfdenote the skin-friction
coefficient, the total drag coefficient is then given by
It appears from experiment that CD,is approximately the same as in

subsonic flow, i.e., it is of the order of magnitude of 0.005.
Lift-Drag Ratio. The lift-drag ratio may now be found with the help
of Eqs. 14.14a and 14.15~. Including the skin-friction drag, we have
To find the maximum lift-drag ratio, we differentiate Eq. 14.16 with

respect to a and set the derivative equal t o zero. The angle of attack
a' corresponding to (LID),,, is thus found as
Inserting this into Eq. 14.16, we find that

Moment Coefficient. To obtain the counter-clockwise moment, 3n,

about the leading edge, we again employ the approximations of the
linear theory, and write
Integrating between 5 = 0 and 5 = I, we first evaluate the integrals
and
where SL is the area between the lower surface and the chord (positive
when the lower surface lies below the chord), and Su is the area above
the chord. With these integrals we obtain
and
Center of Pressure. Suppose that l,,,, denotes the distance aft of the
leading edge through which the resultant force is applied. To first-
order effects,
c . . - 32
3n = LI,.,.; or - - -
1 Ll
Then, using Eqs. 14.14 and 14.18, we get
Characteristic Parameters for Flat Plate and Double Wedge. Con-

sidering the special cases of a flat plate and a symmetrical double-wedge
profile of thickness ratio 6, for which the streamlines and surface pres-
F I G . 14.11. Pressure distributions at surface of flat-plate and symmetrical, double-

wedge proiiles.
sure distributions are shown in Fig. 14.11, we obtain the following

characteristic parameters:
Parameter Flat Plate

-
Double-Wedge
dCdda 4 / d M W 2- 1 -/4
CD; 4 a 2 / d M W 2- 1 4 a 2 / d M m 2- 1
--
cot, 0 4 6 2 / d ~ , 2- 1
af(for CD, = 0) 0 6
(LID),,, (for C D , = 0) m 1/26
--
dCm/da 2 / d M W 2- 1
L.,,./l 0.5
Performance Characteristics of Supersonic Proiiles. Let us review
O
'w
the results obtained thus far and draw some general conclusions.
LIFT. From Eq. 14.14b we see
that the lift-curve slope, d C ~ / d a ,
is the same for all thin profiles a t , c,
- V)
small angles of attack, irrespective 5~ 5

a z
of the specific shape of the profile.
Fig. 14.12 shows the variation of
> -
comparison, M,; for
dCL/da with there is also
convenience
shown the
in oo I 2 3
curve of dCdda '" 2*/wFIG.14.12. Lilt-curve slope for thin

for subsonic flow, based on the profiles.
Ma7
Prandtl-Glauert rule and on the well known value of 2~ for the theo-
retical lift-curve slope of a flat plate in incompressible flow.
DRAG. It may be seen from Eq. 14.15~that at small angles of attack
the thickness and skin-friction drag predominate, whereas a t large angles
of attack the induced drag predominates.
For a given profile a t a fixed Mach Number, Eqs. 14.14a and 14.15~
define a polar relation between C;, and CD with the angle of attack as a
parameter. As in subsonic flow, the lift-drag polar has the form of a
\
\
y\
\
\
6 \q\. ,
-
CL
CD
4
10% Double Wedge
4 6 8 I0
o. Degrees
F I ~14.13.
. Lift-drag polar for typical FIG.14.14. &ft-drag ratios for flat-
supersonic profile. plate and double-wedge profiles, show-
ing effect of skin friction.
parabola. Fig. 14.13 shows the lift-drag polar for a symmetrical double-
wedge profile of 10% thickness ratio operating a t M, = fi. Fig. 14.14
shows how the lift-drag ratio of the 10% double-wedge and of the flat
plate vary with angle of attack a t M, = 42. The angle of attack for
maximum lift-drag ratio, together with the value of (LID),,, are inde-
pendent of Mach Number and depend only on the shape of the profile,
provided of course that CD,= 0. It is clear from Eq. 14.15~that, from
the aerodynamic point of view, i t is desirable to use profiles with small
thickness ratios in order to minimize the thickness drag.
CENTER OF PRESSURE. The center of pressure is in the neighborhood
of the mid-chord point for profiles symmetrical about the chord, whereas
for subsonic flow the center of pressure is a t about 25% of the chord aft
of the leading edge. If an aircraft must fly a t both subsonic and super-
sonic speeps, therefore, the control problem will be seriously complicated
by the large changes in aerodynamic moments and in hinge moments of
the control surfaces as the aircraft passes from subsonic to supersonic
speeds. We also conclude from Eq. 14.19 that the control problem is
simplified if the profile is symmetrical about the chord line, for then the
center of pressure does not change with angle of attack. To first-order
effects the center of pressure is independent of Mach Number.
The foregoing generalizations, it should be emphasized, rest on the
linearized theory. Consequently they are not entirely correct. More
accurate methods of calculation are presented in Chapters 15 and 16.
For the present it will suffice to state that the linear theory is reasonably
correct as regards the lift and drag, but is unreliable as regards the
center of pressure.
Comparison with Experiment. Fig. 14.15 shows the theoretical and
experimental (1) surface-pressure distributions on a biconvex circular-arc
FIQ.14.15. Comparison of linear theory with measured pressure coefficient a t sur-

face of 10% biconvex profile (after Fern).
profile of 10% thickness ratio. The agreement is quite good for small
deviations from the free-stream direction and is very poor for large
deviations in direction, as might be expected from the assumptions of
the linear theory. Schlieren photographs indicate that the curious
variation of C, near the trailing edge of the lower surface is the result
of boundary-layer interaction with the oblique shock wave which,
according to a more exact theory, must be present near the trailing
edge (Chapter 16).
Although the pressure distributions are in poor agreement with the
linear theory, the errors are largely self-canceling as far as the resultant
forces are concerned, since the respective areas between the experimental
and theoretical curves for the upper and lower surfaces are nearly the
same. For this reason, the agreement as to lift and drag coefficients
is much better than the pressure distribution would indicate. For
\ example, (Figs. 14.16a and 14.16b show that the linear theory furnishes
a good approximation to the lift and drag coefficients. However, the
errors in pressure distribution are quite serious in respect to the moment

coefficient, as may be seen from Fig. 14.16~.
- 8.6.3 %
Mm=2,.13
0 ~ e r r i ' sData
- Linear Theory
o, Degrees
FIG.14.16. Comparison of linear theory with force coefficients of 6.3y0 single-

wedge profile (after Ferri).
To sum up, the linearized theory is useful for showing in a general way
the properties of supersonic airfoils. In addition, it furnishes a simple
method for approximating the lift and drag coefficients with fair accu-
racy.
Flat-Plate Airfoil in Subsonic and Supersonic Flow. I t is instructive
to compare a flat-plate airfoil in a frictionless fluid for incompressible
flow with the corresponding situation for supersonic flow. The stream-
line patterns are shown in Fig. 14.17. In supersonic flow the static
(0 )
/
/
/ - /
/ (b)
\ / -
b
\
b \. --
\ \
F I ~14.17.
. Streamlines for flat-plate airfoil.
(a) Subsonic flow.
(b) Supersonic flow.
pressure is uniform over the top and bottom surfaces; in incompressible

flow i t undergoes wide variations. There are stagnation points in sub-
Art. 14.7 REFLECTION AND INTERSECTION O F WAVES 449
sonic flow, but none in supersonic flow. Near the nose of the flat plate
in incompressible flow the velocities approach infinity and there are
infinite LLsuction
pressures," but no singularities of this sort appear in the
supersonic pattern. In incompressible flow the resultant force on the
plate is normal to the direction of the distant flow, that is, there is no
drag (the dlAlembert Paradox) ; in supersonic flow the resultant force is
normal to the plate, that is, there is a wave drag even though the fluid is
frictionless. The center of pressure is a t about 25% of chord for incom-
pressible flow, but a t about 50% of chord for supersonic flow. These
comparisons again illustrate the complete diierence in character be-
tween subsonic and supersonic flow.
14.7. Reflection and Intersection of Waves

All the examples of the linear theory thus far cited have referred to
simple-wave flows. We now consider flow patterns in which waves of
both families are present.
For flow with simple waves it is usually possible to obtain analytical
solutions, and hence the flow pattern may with no difficulty be thought
of as containing an infinite number of infinitesimal pressure waves. For
flow with mixed waves, on the other hand, it is usually necessary to
make stepwise calculations, and hence i t has been found convenient to
condense the wave system into a finite number of waves which divide
the flow field into regions of uniform, parallel flow.
Representation of Continuous Flow by Lumped Pressure Waves. To
illustrate this idea, suppose that an initially uniform parallel flow passes
over the curved surface AB of Fig. 14.18a. Then, by the theory of
FIQ.14.18. Replacement of a continuous wave system by lumped discontinuities.
simple-wave flow, the change in direction of the wall generates a t each

point an infinitesimal pressure wave which bends each streamline by
an equal amount. In Fig. 14.18b, as an approximation, the curved wall
has been replaced by a single sharp change in direction with the same
angle of turn, and the infinite system of pressure waves has been con-
densed intc' a single wave across which the entire change of stream
properties is concentrated. A better approximation is furnished by
Fig. 14.18c, and still better approximations could be had by dividing
the curved wall into a larger number of straight chords. According to
Eq. 14.12, the change in pressure (and of Mach Number) for a turn of
given A0 will be the same for the three patteins of Fig. 14.18; the approxi-
mation lies only in the construction of the wave pattern and stream-
line pattern.
In calculating the pressure change and wave angle for the flow pattern
of Fig. 14.18b, using Eqs. 14.12 and the definition of Mach angle, we
might in the first instance choose the Mach Number and direction in
zone 1 as the mean-stream conditions. However, after finding the
pressure in zone 2 the Mach Number in zone 2 may be calculated from
the isentropic relation. We could then make a better approximation
by taking the mean-stream conditions as those corresponding to the
average Mach Number and direction between zones 1 and 2. However,
since one of the functions of the linear theory is to provide answers
with a minimum of computation, such averaging refinements are gener-
ally not worth employing, especially since methods more accurate than
the linear theory are available (Chapters 15 and 16).
Unit Processes. Wave patterns may be synthesized from a few unit
processes, namely, (i) reflection of a wave from a straight wall, (ii)
cancellation of a wave a t a boundary, (iii) reflect.ion of a wave from a
boundary of constant pressure, and (iv) intersection of two waves.
Direction and Pressure Indexes. Before taking these up in turn, let
us for convenience define
180
@ = -0
r
where @ is the inclination of the streamline t o the x-axis, in degrees, and
0 is the corresponding inclination in radians. Also, let us define a dimen-
sionless pressure index
-1
180 VbArn2
p =-
kprnMrn2
p + constant
where the constant is so chosen that P has convenient numerical values.
Then i t follows from Eq. 14.12 that as a streamline crosses waves of
family I1 (left-running),
APr = A@ (14.20a)
and as it crosses waves of family I (right-running),
The local pressure index P of course determines the local Mach Num-
ber through the isentropic relationships.
Art. 14.7 REFLECTION AND INTERSECTION OF WAVES 451
Reflection of Wave from Straight Wall. I n Fig. 14.19a the flow is

assumed to be uniform and parallel in region I, with = 0 and say
o
PI = 100. A t the l bend in the lower wall, rarefaction wave a of
family I I is generated. In region 2, then, Oz = -1, and, from Eq.
14.20a1Pz = 99.
Now, what happens when wave a strikes the upper wall? It cannot
merely end on the upper wall, for then the flow would separate from the
upper wall. The continuity of flow a t the upper wall is maintained if
we assume that wave b, of family I , originates where wave a strikes the
FIG. 14.19. Reflection of expansion wave.

(a) Reflection from straight wall in like sense.
(b) Hodograph diagram for (a).
(c) Cancellation of wave by bend in wall.
upper wall. Wave b is called the reflection of wave a. In region 3, if

the flow follows the wall, it is necessary that Q3 = 0. Applying Eq.
14.20b to the flow across wave b, we get Pg = 98.
I n this example, the incident wave a is an expansion wave in the sense
that the fluid crossing the wave undergoes a drop in pressure and an
increase in velocity. The reflected wave b is also an expansion wave.
If a had been a compression wave, similar reasoning would show that b
would also be a compression wave. We therefore reach the important
conclusion that pressure waves are reflected from a straight wall in like
sense. Furthermore, for waves of small strength, to which the linear
theory is limited, the angle of reflection equals the angle of incidence,
and the strengths of the reflected and incident waves (as measured by
the respective pressure changes) are equal.
Fig. 14.19b shows the hodograph diagram for this example. In the
hodograph diagram all velocity vectors are drawn from the common
origin 0. A region in the physical plane of Fig. 14.19a is mapped as a
point in the hodograph diagram. After locating point 1 on the hodo-
graph, point 2 is found from the known direction of the velocity vector
in region 2 and from the fact that the vector change in velocity be-
tween regions 1 and 2 is normal to the direction of the pressure wave
of family I I . Point 3 may similarly be located in the hodograph diagram.
Fig. 14.20a illustrates the generation of a compression wave by a

wedge, together with two subsequent reflections.
FIG. 14.20. Schlieren photographs of wave reflection (M.I.T. Gas Turbine Labora-
tory).
(a) Compression waves generated by wedge are reflected in like sense a t duct
walls. Reflected waves are subsequently reflected in like sense a t wedge
surface.
(b) Compression wave generated by wedge is incident on a corner, the turn-
ing angle of which is slightly greater than that of the wedge. The compres-
sion wave is more than canceled by the expansion generated a t the corner,
and the net result is a very weak expansion emanating from the corner.
Cancellation of Wave. Referring again to Fig. 14.19, and following

the same reasoning as in the previous example, we see that wave a
would have no reflection if the upper wall were, a t the point of incidence
of wave a, bent in the direction of the lower wall. Fig. 14.19~shows the
resulting flow pattern.
Fig. 14.20b illustrates the near cancellation of a wave by a bend in
the wall.
Reflection of Wave from Boundary of Constant Pressure. Suppose
that in Fig. 14.21a a uniform, parallel jet leaves a nozzle and is exposed
on the upper side to an atmosphere of constant pressure. Assume that
FIG. 14.21. Reflection of a wave from a free (constant-pressure) boundary.

(a) Wave system, streamlines, and fluid properties.
(b) Hodograph diagram.
in region 1 the flow properties are = 0, P1 = 100 and that there is

a l o bend in the lower wall. Then, as in the example of Fig. 14.19, we
may construct expansion wave a and compute B2 = -1, Pa = 99.
The upper boundary of the jet by assumption remains a t constant
pressure. Hence, wave a must be reflected as the compression wave b.
In region 3 we have P3 = PI = 100, and, applying Eq. 14.20b across
wave b, we find that = - 2.
The upper boundary of the jet therefore undergoes a turning angle
twice as large as that of the bend in the lower wall. We concIude
furthermore that a pressure wave i s reflected from a boundary of constant
pressure in unlike sense.
The hodograph diagram for this example is shown in Fig. 14.21b.
Note that since regions 1 and 3 have the same pressure, they must also
have the same speed. Points 1 and 3 therefore lie on a circle with the
origin 0 as center.
Intersection of Waves. Fig. 14.22a shows the intersection of two
waves of opposite family. Our problem is to find the state in region 4
from the known state in region 1 and the known turning angles 6, and 6 b .
FIG. 14.22. Crossing of waves.

(a) Wave system, streamlines, and fluid properties.
Region 4 may be reached along a streamline from region 1 either through

region 3 or through region 2. We therefore apply Eqs. 14.20a and 14.20b
in the following manner:
If we first add and then subtract these two equations we obtain the
desired answers in the form
A general conclusion to be drawn from this example is that, for wave

strengths within the linear theory, intersecting waves do not interfere
with each other. This inference is reached by noting that waves a and c
have identical pressure changes and turning angles, and likewise for
waves b and d. We may therefore ignore the crossing of the waves and
simply assume that a certain wave has associated with it a constant
turning angle and pressure change, irrespective of the fact that it may be
crossed by waves of the other family.
Example of Overexpanded and Underexpanded Jets. In Chapter 5
it was remarked that when a supersonic nozzle discharges into a region
( b ) Underexponsion
FIG.14.23. Flow patterns a t exit of two-dimensional nozzle.

(a) Overexpanded nozzle.
(b) Underexpanded nozzle.
of variable back pressure, there is a range of back pressures for which

the one-dimensional approach yields no answers. Problems of this type
may now be solved, a t least approximately, by linear methods.
In Fig. 14.23a the pressure of the exhaust region is higher than the
pressure in the exit plane of the two-dimensional nozzle. On leaving
the nozzle the jet boundaries must be immediately adjusted to the back
pressure. Hence we assume that compression waves originate a t the
corners of the nozzle exit. The strengths of the waves are such as to
make-the pressure a t the jet boundaries equal t,o the back pressure. For
the specific case shown it was assumed that the pressure rise across the
first wave corresponds to a change in P of one unit. The subsequent
intersections and reflections of the original waves are handled accord-
ing to the unit process methods previously described. It may be seen
that the wave and flow patterns undergo cyclic changes which extend
to infinity in the absence of friction. In adjusting itself to the back
pressure, the jet in a sense "overshoot~s,"for there are zones in the jet
where the pressure is higher than the back pressure.
Fig. 14.2313 shows the wave pattern and flow properties for an under-
expanded jet, i.e., for the case where the exhaust-region pressure is less
then the exit-plane pressure.
Example of Supersonic Flat-Plate Biplane. Fig. 14.24 shows h,ow the
linearized theory may be used to determine approximately the aero-
dynamic properties of a supersonic biplane. The particular example of
(a) (b)
F I ~14.24.
. Flat-plate biplane a t 1" incidence.
(a) Wave pattern, streamlines, and pressure distribution. Dashed lines show
expansion waves; full lines compression waves.
(b) Ratio of lift coefficient of biplane (based on area of both plates) to that of
monoplane.
the sketch is for an angle of attack of one degree. From the pressure
distribution of Fig. 14.24a, it may be seen that over certain areas of the
plate there is no net force acting, whereas over the remaining areas the
force per unit area is the same as for a single flat-plate airfoil. From the
A
geometry of the wave pattern, it is found that the fraction of the area
over which there are pressure differences depends on the parameter
d m -21 Fig. 14.24b indicates how the ratio of lift coefficient
for the biplane to lift coefficient for the monoplane a t the same angle of
attack varies with the parameter t d M m 2- 1/1. Further consideration
shows that the location of the center of pressure also depends in marked
degree on the parameter t d M m 2- 1/1. From the viewpoints considered
here, the flat-plate biplane is inferior to the monoplane, but for thick
profiles the biplane has certain advantages over the monoplane.
The method of applying the linearized theory as given in this article
should be considered as a simple and rapid means for obtaining qualita-
tive results and rough quantitative results. Accurate numerical results
may be obtained by this method only when there are very small pertur-
bations from a mean paraJ.Ie1 flow. For .large changes in velocity and
direction t h e linear theory remains useful for portraying the flow p a t t e r n

i n a general way, but t h e calculations must be based om more exact
techniques, such as the method of characteristics (Chapter 15) and t h e
theory of oblique shocks (Chapter 16).

1. FERRI,ANTONIO. Experimental Results with Airfoils Tested in the High-speed
Tunnel a t Guidonia, NACA Tech. Memo., No. 946 (1940).
2. TAYLOR,G. I. Application to Aeronautics of Ackeret's Theory of Airfoils Moving
a t Speeds Greater than That of Sound, British A.R.C., R & M No. 1467
(1932).
3. ACKERET, J. Air Forces on Airfoils Moving Faster than Sound Velocity, NACA
Tech. Memo., No. 317 (1925).
4. ACKERET, J. Uber Luftkrafte bei sehr grossen Geschwindigkeiten insbesondere
bei ebenen Stromungen, Heluetica Physica Acta, Vol. 1 (1928), pp. 301-22.
5. LIEPMANN, H. W., and PUCKET~, A. E. Introduction to Aerodynamics of a Com-
pressible Fluid. New York: John Wiley & Sons, Inc., 1947.
R., Introduction to Theoretical Gas Dynamics. Ann Arbor: J . W. Edwards,
6. S A ~ E R
1947-
7. BONNEY, E. A. Engineering Supersonic Aerodynumics. New York: McGraw-Hill
Book Co., Inc., 1950.
PROBLEMS
14.1. Consider a nearly uniform, parallel flow with the following perturba-
tion velocity potential:
p=O for (Z - ~ M , -
Z 1 v) < 0
where A is a constant.
(a) Investigate the streamlines and pressure distribution.
(b) Show that when A approaches zero, the flow pattern is that corresponding
to flow from left to,right above a wall which has a sharp corner with a turning
angle c.
14.2. If the value of pz - p, across a weak pressure wave is calculated from
the linear theory, and if the values of Mz and of p,/pl are calculated from the
isentropic relations, it is found that the equation of continuity is violated across
p~~~~~
the wave. Explain why this is so.
,I
14.3. Consider the thin supersonic
foil
airfoil
is shown
symmetrical
in the sketch.
about the
Thechordair-
line.
PROB.14.3. Find expressions for the lift, drag,
and moment coefficients in terms of
the Mach Number, the constants a1 and a2, and the thickness ratio 6 = t / l .
Show that for a fixed thickness ratio, the wave drag due t o thickness is a
minimum when a1 = a2 = 0.5.
14.4. Consider a thin, curved profile of parabolic shape (see sketch), ex-
pressed by the curve, y = -h(x/t)'. The leading edge of the profile is tangent
to the direction of the oncoming air stream. Using the linearized theory,
(a) Find expressions for the lift and drag coefficients, for the lift-drag ratio,
and for the moment coefficient. all in
~ ~
terms of M i and the thickness ratio,

h/t.
(b) Plot the pressure coefficient for
the upper and lower surfaces against
per cent chord, for M, = 2 and
I--- ' ------I
PROB. 14.4.
h/t = 0.10.
(cj For M, = 2 and h/t = 0.10, compare the pressure drag of this airfoil
with the pressure drag of the flat-plate airfoil and of the 10% doublewedge
profile, assuming that, both the latter have the same lift coefficient as the para-
bolic profile.
14.5. A converging-diverging nozzle with straight divergent sides discharges
into a straight duct (see sketch). The total included angle between the straight
sides is 12". At the nozzle exit the Mach Number is 2.0. Two static pressure
taps are placed in the wall near the sharp corner which joins the nozzle to the
duct, one just upstream and one just downstream of the corner.
Estimate the pressure rise across the corner in (a) per cent of the pressure a t
the nozzle exit, and (b) per cent of the pressure a t the nozzle entrance.
Are your answers dependent on whether the nozzle is two-dimensional or
axially symmetric?
14.6. A certain tm-o-dimensional passage has one straight wall and one
curved wall (see sketch). The shape of the latter may, as a first approximation,
be replaced by two straight-line segments joined a t an angle of 2",as shown in
the sketch.
The Mach Number and pressure are initia!ly 2.0 and 1000 respectively.
Using linearized thcory, make a sketch to scale, showing streamlines, dis-
turbance lines, and pressures in the various regions of constant velocity. Carry
out the flow pattern until two complete reflections of the original wave have
occurred.
14.7. A frictionless, two-dimensional, converging-diverging nozzle is supplied
with air at 100 psia. The diverging section is shaped so that the stream passes
through the exit plane of the nozzle in parallel flow with a Mach Number of 2.0.
The pressure in the exhaust region is variable. Using linearized theory, make a
sketch to scale of the streamlines and disturbance lines in the region outside the
nozzle exit and indicate the pressures in the various regions of constant velocity,
(a) For an exhaust-region pressure of 14.7 psia
(b) For an exhaust-region pressure of 10.5 psia
14.8. A two-dimensional, supersonic stream of air is bounded by a solid wall
on one side only (see sketch). The wall is shaped so that the stream undergoes
a 20" change in direction along an arc of a circle. It is agreed to replace the
actual wall by a series of straight-line segments of equal length, and it is proposed
to investigate the effect of the number of segments on the calculated values of
the final Mach Number and pressure.
A modified linear theory is to be used, that is, linear theory will be used for
each disturbance line, but the Mach Number downstream of each disturbance
line will be modified so that it corresponds to the newly found pressure.
For an initial Mach Number of 2.0 and an initial pressure of 100 psia, calcu-
late the final Mach Number and pressure when the arc of the circle is replaced
in turn by 2, 4, and 6 equal line segments.
Estimate the asymptotic solution by graphically plotting the results against
the reciprocal of the number of line segments, and compare with the exact
solution (from Prandtl-Meyer theory) of M = 2.83, p = 27.4 psia.
14.9. Considering the wave pattern shown in the sketch, find ( 0 3 - 01)
and Pa - PI as functions of 6, and 6b. Investigate the special cases where
6, = 6 6 and 6, = -61,.
14.10. The sketch shows a two-dimensional duct having an offset formed by
o
l -turns a t the corners A , B, D, and E.
Assuming
-
that a t section AD the stream is uniform and parallel and that
M = 4 2 and p = 100 psia, determine for section CF the way in which the
pressure, Mach Number, and stream direction vary over the cross section, for
values of L/H of 1, 2, and 3.
14.11. The sketch shows a simple form of diffuser for decelerating a super-
sonic stream by very small amounts. Upstream of section AC the flow is uni-
Ch. 14 PROBLEMS 459
form and parallel, and it is desired to design the passage so that the flow down-
stream of section BD will be uniform and parallel.
(a) Show that for small changes in Mach Number, the correct ratio L/H is
related only to M, and find the relationship.
- c and AM.
(b) Find the relation between
(c) Suppose that M = 4 2 , L/H = 0.75, and c = 0.01 radian. Compute
(i) the ratio of the maximum variation in pressure over the section EF to the
average pressure diierence between AC and EF, and (ii) the ratio of the maxi-
mum difference in stream direction over section EF to the wall angle E.
14.12. Consider a supersonic biplane with the lower flat plate always at
zero angle of attack (see sketch). Explore the characteristics of such a biplane
and compare the lift coefficient and center of pressure with the corresponding
parameters for a flat plate. Demonstrate the significance of the parameter
t d ~ , 2 - 1/Z as suggested by Fig. 14.24b.
14.13. A two-dimensional supersonic airfoil of chord Z is to be tested in a

two-dimensional supersonic wind tunnel of height H (see sketch).
(a) Derive a general criterion among the variables I , H, and M, which will
insure that there is no effect of the tunnel walls on the performance of the airfoil,
for the case of very small angles of attack.
(b) Indicate qualitatively how this criterion would be modified for larger
angles of attack.
(c) Consider a flat-plate airfoil at an angle of attack ZO, with M, = 1/2,
H/Z= 2/3, and a test-section pressure of 10 psia. Calculate approximately
the lift coefficient, the drag coefficient, and the location of the center of pressure
(in fractions of the chord aft of the leading edge), using the method of small
perturbations.
14.14. One of the most important problems in supersonic wind tunnel design
is the design of the contours for producing in the test section a stream which is
exactly uniform and parallel.
The sketch shows a two-dimensional nozzle which is symmetrical about the
center line and in which the flow is intended to be uniform and parallel in the
region downstream of the line BAC, where B and C are the ends of the curved
contours, and AB and BC are the Mach lines corresponding to the test-section
Mach Number.
Because of errors in machining the contours, inaccuracies in the design, and
uncertainty as to boundary-layer behavior, it is necessary to calibrate every
tunnel.
(a) Assuming that the flow is two-dimensional, show that if the calibration
yields data on the velocity vector along the line A D , it is then not necessary
to do any calibration in the remainder of the region ABEDFC, where E D and
FD are Mach lines. That is, show that information as to the velocity vector
determines completely the state of the flow in the region ABEDFC.
(b) In the region ABEDFC the flow is very nearly uniform and parallel. If
the lines 1-3 and 1-2 are Mach lines, show that the pressure and stream direction
at a typical point 3 may be found from similar data a t points 1 and 2 from the
formulas
P3 =
PI +PZ + k p ~ 2 el - e2
2 2
where p and M are respectively the mean pressure and mean Mach Number in
the test section.
14.16. Show that the lift per unit span for a flat-plate airfoil at small inci-
dence in a supersonic stream is given by
L = - P, VJ'
where 'I is the circulation around the profile.
14.16. The sketch shows a turbine nozzle cut off at an angle of 30" to the
nozzle axis. The nozzle is designed for a Mach Number of g 2 . It operates
with a pressure of 100 psia in the plane shown and discharges into a region where
the pressure is 99 psia.
Ignoring friction, draw a sketch showing the wave pattern and the jet bound-
aries. Indicate compression waves by solid lines and rarefaction waves by
dashed lines. In each zone of flow show the pressure (psia) and flow direction
(degrees). Continue the sketch at least to the point where the jet pattern begins
to repeat itself, and indicate which portion of the pattern repeats itself infinitely.
Air
Moo:
&
E i. .
- .
"
!"
14.17. A supersonic biplane is constructed as shown from the two halves

of a symmetrical double-wedge profile.
It is desired to compute CD,,"/CD,', where CD; is the pressure drag of the
biplane a t zero angle of attack, and CD,' is the pressure drag of the symmetrical
double wedge a t zero angle of attack.
Assume that M, = 4 5 in both cases, that t / l is very small compared with
unity, and that the linearized theory is valid.
Ch. 14 PROBLEMS 461
Make a chart approximately to scale of CD,"/CD,' as a function of s/l, for

values of the latter between 0.25 and 2.
14.18. The sketch shows a two-dimensional, supersonic flow with wave a
incident on a surface of discontinuity which separates two regions of flow having
the same directions and pressures hut different velocities. The situation illus-
trated in the sketch might serve as a crude model of a shock entering a boundary
layer.
Surface of \
Me
P, = P,
e,, = e,
-
Discontinuity
0
\\ \
\
b
0
The problem is to determine the strengths of the transmitted wave b and of

the reflected wave c compared with the strength of the original wave a.
Carry out the analysis for the limiting. case where the turning angle 6 for
wave a is extremely small, so that the linearized theory for the flow across a
wave is valid. The Mach Numbers MI and M2 may, however, have any value
greater
- than unity.
Demonstrate that
t'6 - e2 - 2
--
6
and
Chapter 15
METHOD O F CHARACTERISTICS F O R TWO-

DIMENSIONAL, SUPERSONIC F L O W

The linearized approximation of Chapter 14 has its place in obtain-
ing simple analytical relations for thin supersonic profiles and for quickly
obtaining simplified stepwise solutions t o other problems of supersonic
flow. Of course, there are many instances in which more accurate
calculations are necessary, and for that purpose we shall discuss in this
chapter the method of characteristics.
Our analysis will a t first be limited t o two-dimensional, irrotational,
isentropic, supersonic flow, and we shall make the further assumptions
that the fluid is a perfect gas, that the flow is steady, and that gravity
forces are negligible. There are many problems involving supersonic
wings, supersonic compressor cascades, supersonic wind tunnel nozzles,
etc., where the method of characteristics, based on these assumptions,
can be and has been highly successful. In Art. 15.12 the method of
characteristics is extended to rotational flows of a nonviscous fluid.
The method of characteristics will be developed in two ways: first,
by an extension of the linearized theory, in which physical concepts are
emphasized; and second, by formal mathematical methods. The latter
development is more rigorous and leads to a better appreciation of the
concept of characteristic curves, a n appreciation which will be useful
in attacking other types of problems involving the method of charac-
teristics.
NOMENCLATURE
A cross-sectional area for isen- k ratio of specific heats

tropic flow M Mach Number
b slope of physical characteristic M* V/c*
c speed of sound P Dressure
L
c* critical speed of sound for isen- p, isentropic st,agnation pressure

tropic flow r radius measured outward from
c, specific heat at constant volum& sharp corner
h distance along Mach line (Fig. s entropy per unit mass
15.11a) T absolute temperature
462
Art. 15.2 WAVES OF ONE FAMILY
u x-component of velocity cp velocity potential

u y-component of velocity w defined by Eq. 15.6
V resultant velocity
x, y Cartesian coordinates I, I characteristic coordinates
( )I signifies characteristic curve of
a Mach angle family I
0 inclination of velocity vector ( )rr signifies characteristic curve of
to x-axis family I1
P density ( )E signifies exit of nozzle
15.2. Flow with Waves of One Family by Extension

of Linear Theory
Modified Linear Theory. From the linearized theory of Chapter 14,
we learned that small disturbances in a twodimensional, supersonic flow
are propagated in the form of pressure waves along the Mach lines,
n O/
0-
I
6
-----%.+
'4
v* .-
(a)
FIG.15.1. Propagation of disturbance produced by bend in wall at point a.
(a) Streamlines and Mach lines according to linear theory. Mean Mach line II,
lies midway between Mach lines of upstream flow (11,)1,and of downstream
flow (I1,)2.
(b) Hodograph diagram according to linear theory.
(c) Exact Prandtl-Meyer solution for corner flow.
that these lines are also lines of constant perturbation potential, and
that the vector change in velocity produced by a pressure wave must
accordingly have a direction normal to the Mach direction. From these
considerations, all the calculations can be reduced t o a single unit proc-
ess, illustrated by Fig. 15.1.
In Fig. 15.1a1an initially uniform parallel flow in region 1 is influenced
by a small turn in the wall a t point a. The effect of this turn is propa-
gated along the left-running Mach line II,, which makes the Mach
angle al to the initial flow. Fig. 15.lb shows how the velocity V 2 is
determined from the known turning angle 6 and from the rule that the
vector change in velocity is normal to the line II,. The inclination of
the latter depends on the Mach Number. As a first approximation
the Mach line I I , may be drawn a t the Mach direction corresponding
to the upstream flow, i.e., as line However, after V 2 and other
stream properties are tentatively found in region 2 by means of this
first approximation, the accuracy of the approximation may be improved
by taking the line IZ, a t the mean Mach angle E t o the mean flow direc-
tion. In this way, by several iterations, a better value of the stream
properties in region 2 may be computed with the help of Fig. 15.lb, a n d ,
the direction of the mean Mach line I I , may be more accurately drawn
in the physical plane. Fig. 15.la.
These ideas may be extended to the flow around the curved wall of
Fig. 15.2a. As an approximation, the continuous. wall is replaced by a
"a
/
I- /
/
3' .1
.-
(a) (b)
FIG.15.2. Flow past curved wall by modified linear theory.
(a) Streamlines and Mach waves.
series of straight-line chords with corners a t a, b, c, and d. From each

corner a pressure wave is propagated of such strength as to preserve
continuity a t the wall. Using the iteration process described above,
the conditions in region 2 and the direction of the Mach line IZ, are
found from the mean flow conditions between zones 1 and 2. Next, the
conditions in region 3 and the direction of the Mach line ZZb are found
by iteration from the mean flow conditions between zones 2 and 3. Thus,
by stepwise increments, the entire flow field is determined. This type
of flow, in which pressure waves of only one family appear, is known
variously as $ow with waves of one family, simple-wave $ow, Prandtl-
Meyer $ow, and corner-type $ow. It is distinguished by the features
that (i) all flow properties are uniform along each Mach line and the
Mach lines are straight, and that (ii) for given initial conditions, the
magnitude of the'velocity a t any point depends only on the flow direc-
tion a t that point. Later we shall see that these two features necessarily
go hand-in-hand mathematically.
The construction of the velocity vectors for the flow of Fig. 15.2a is
shown in Fig. 15.2b. If we denote +heend points of the velocity vectors
Art. 15.2 WAVES OF ONE FAMILY 465
representing the flows in fields 1, 2, etc., by the respective image points

l', 2', etc., Fig. 15.2b becomes the hodograph plane, or the velocity plane.
A line drawn from the origin to any point such as 1' gives the magnitude
and direction of the velocity vector pertaining to that point; hence the
velocity V and stream direction 8 are polar coordinates in the hodograph
plane. The Cartesian coordinates in the hodograph plane are u and v,
the x- and y-components of velocity, respectively. Since the velocity
a t each point of the physical plane may be represented by a correspond-
ing point in the hodograph plane, any line in the physical plane is in
general mapped into a corresponding line in the hodograph plane. For
example, the physical streamline AA of Fig. 15.2a is mapped into the
hodograph streamline 1'2'3'4'5' of Fig. 15.2b. We observe that in a
flow with waves of one family, all physical streamlines are imaged by a
single hodograph streamline.
Exact Solution for Simple Waves. The method of solution indicated
above for the curved wall of Fig. 15.2a becomes more and more exact
as we increase the number of chords by which the wall is replaced. We
obtain an exact solution in the limit by carrying out the process with
infinitesimal steps.
LEFT-RUNNING WAVE. Let us imagine that a Mach wave of family 11
(Fig. 15.3a) turns the flow through the positive angle dB, with correspond-
-- \, -du
V t dV
P
(a) (b)
FIG.15.3. Infinitesimal Mach wave.
(a) Physical plane.
ing infinitesimal changes in all stream properties. The hodograph dia-

gram of Fig. 15.3b is constructed with the rule that the vector change in
velocity produced by the wave is normal to the wave direction. From
the geometry of the figure, noting that du is in the limit the algebraic
increment in velocity, dV, we get
-du/dv = tan a = 1 1 4-~1 ~

466 METHOD OF CHARACTERISTICS ch. 15
Elimination of du and dv from this set of equations yields
Now, from the adiabatic relations of Chapter 4, khe relation between dT'
and d~ is
hence Eq. 15.1 may be written

.\/M2--1dM2
Integration of this equation by standard methods yields
6 =-
k-1
+ constant
arc tan ,/r (M' - I ) + arc tan 7
M' - 1
(15.4)
It is often more convenient to work with the dimensionless velocity
M* = V / c * rather than the Mach Number M. Using the adiabatic r e
lation that
Eq. 15.4 may be put into the form
where the function w(M*) is given by
-I.*) = & k+l

k - 1 arc
k
--+l
- arc tan J M * ~- I
k-1
I- -
!€-1 k + l M*.
(15.6)
The constant of integration, denoted by (21 - 1000) in Eq. 15.5, is
determined from the initial values of l3 and M* for any particular prob-
lem. Why the constant of integration is put in this particular form will
be made clear when we discuss waves of two families.
RIGHT-RUNNING
WAVE. If we had been dealing with a problem in
which there were Mach waves only of family I, the sign on the right-
hand side of Eq. 15.1 would have been reversed, and we would have
ended with
BIT = +w(M*) f 211 - 1000 (15.7)
NOMENCLATURE. The subscript I in the symbol BI of Eq. 15.5 signi-

fies that pressure waves of family I are absent, that the process occurs
a t a constant value of I, and that changes in flow properties take place
as the streamline crosses Mach waves of family 11, i.e., as it crosses left-
running Mach waves. Similarly for the nomenclature of Eq. 15.7. The
rationality of this nomenclature will be made more evident in connection
with the later derivation of these same results by the mathematical
theory of characteristics.
WORKING CHART. The curves of w versus M* and w versus M, corre-
sponding to Eqs. 15.6 and 15.4, are shown in Fig. 15.4. Since the Mach
(a) w versus M*.

(b) w versus M.
Number uniquely determines the pressure ratio, temperature ratio,

etc., for isentropic flow, i t follows that w is also directly related to the
pressure ratio, temperature ratio, etc.
SUMMARY OF CALCULATION RULES. From Eqs. 15.5 and 15.7 we
derive the convenient rules of computation,
B2 - B1 = - (wz - wl) for left-running Mach waves (15.8a)
Bz - B1 = wz - w l for right-running Mach waves (15.8b)
468 METHOD OF CHARACTERISTICS Ch. 15
TABULATED FUNCTIONS. TO facilitate numerical calculations, the

relationships between M, M* and w are tabulated in Table B.7 for k = 1.4.
Table B.7 also has certain other isentropic functions of Mach Number
which will be found useful.
The argument of this table is the variable w. It will be seen from sub-
sequent examples that it is convenient to select regular and equal inter-
vals of w for a problem, corresponding to equal turning angles across
each Mach wave.
Example of Simple-Wave Flow. Suppose that in the flow of Fig. 15.1,
MI = 2.059, O1 = 0, and O2 = -5". From Fig. 15.4, a t MI = 2.059,
we get w l = 28.0'. Then, since only left-running Mach waves are in-
volved, Eq. 15.8a gives w2 = & - & + wl = 0 - (-5) +
28 = 33'.
Then, using Fig. 15.4 again, for w2 = 33.0" we get M2 = 2.249. To find
the relation between the pressures in regions 1 and 2, we get, with the
help of the isentropic tables,
p z / p ~= (p/po)z/(p/po)~ = 0.0866/0.1165 = 0.743
These exact answers for the problem of Fig. 15.1 will of course be slightly
different from the linearized solution described earlier. In the exact
) ~ single pressure wave II, of the approximate
solution (Fig. 1 5 . 1 ~ the
solution (Fig. 15.la) is replaced by an infinite number of infinitesimal
Mach waves lying in the zone between (IIa)l and (11,)2, and the
streamline undergoes continuous changes as it passes through this
zone.
Hodograph Characteristics. If all velocities are divided by c*, the
hodograph diagram becomes dimensionless; with polar coordinates M*
and 0. Then, if Eqs. 15.5 and 15.7, which represent respectively the
hodograph streamlines for simple left-running waves and simple right-
running waves, are plotted in the hodograph plane, we get the two
families of curves of Fig. 15.5. The curves marked I are defined by
Eq. 15.5, each curve being identified by a particular value of the con-
stant, I, in Eq. 15.5; similarly for the curves of family T I defined by
Eq. 15.7. Obviously the two families of curves are mirror images of
each other, and any curve of a given family may be found from a
single curve of the same family merely by rotating the latter around
the origin.
The hodograph curves of family I in Fig. 15.5 are evidently a more
exact version of the curve 1'2'3'4'5' of Fig. 15.2b. They are known as the
hodograph characteristics. We then have the rule that withsimple
left-running waves the Mach line a t a given point in the physical plane
is normal to the hodograph characteristic of family I a t the image point
Art. 15.2 WAVES O F ONE FAMILY 469
in the hodograph plane; and, with simple right-running waves, the Mach
line a t a given point is normal to the hodograph characteristic of family
I I a t the image point.
FIG.15.5. Hodographic characteristic curves on M*, @-plane.
For a perfect gas i t may be shown that the hodograph characteristic

wrve on the M* - 0 plane is an epicycloid formed by rolling on the
+
circle of radius unity another circle of radius [ d ( k l)/(lc - I).-11.
470 METHOD O F CHARACTERISTICS Ch. 15
Fro. 15.6. Simple-wave flow past curved wall,

(a) Left-running expansion waves.
(b) Hodograph diagram for (a) and (c).
(c) Left-running compression waves.
(d) Schlieren photograph showing convergence of simple compression waves
(after Drougge). Flow from left to right in concave corner, with wall shaped
for centered Prandtl-Meyer compression. Note convergence of Mach lines
to form shock, and contact discontinuity (vortex sheet) extending from junc-
tion of continuous waves with shock (see Chapter 16).
Features of Simple-Wave Flow. To make the method clear let us

consider the flow past the curved wall of Fig. 15.6a. Suppose that the
curve begins a t point a, and that the flow upstream of a is uniform and
parallel, with properties 0, = 0 and Ma* = 1.66. From these data,
point a' may be located in the hodograph diagram of Fig. 15.6b. Then,
since the upstream flow was given as uniform and parallel, right-running
waves cannot be present, and we are dealing with a case of simple left-
running waves. All fluid states must therefore lie on the hodograph
characteristic of family I passing through point a'. Points b', c', and d'
may then be located on this characteristic because the flow directions a t
points b, c, and d are known. From the values of M* all other fluid
properties may be found a t these points. Furthermore, all fluid prop-
erties are constant on the Mach lines originating from points a, b, c, and d .
The Mach lines themselves are drawn normal to the hodograph charac-
teristic a t the appropriate points (Fig. 15.6b), and lie a t an angle
+
(0 a) to the horizontal. Thus the entire flow field is determined.
Each streamline is a transformation of the boundary streamline abcd,
as the slopes of the streamlines are constant on the Mach lines.
It may be noted from Fig. 15.6a that the Mach lines diverge. This
must always be so in an expansive flow: the streamline, from momentum
considerations, bends away from the Mach line, and, moreover, the
Mach angle becomes smaller; hence the Mach lines necessarily diverge.
Mathematically, since the left-running Mach wave is a t the counter-
clockwise angle a to the velocity direction, and the latter is a t the angle
e to the horizontal, the Mach wave is inclined a t the angle (0 a) to+
the horizontal. In the flow of Fig. 15.6a, both 0 and a decrease, and
the Mach lines therefore diverge. The pressure gradients for stream-
lines far from the wall are consequently smaller than for those near the
wall.
Conversely, as in Fig. 15.6~)the Mach lines always converge for a
simple-wave compressive flow. This gives rise to compression shocks
(Fig. 15.6d), as explained in Chapter 16.
Fig. 15.7 shows examples of right-running simple waves of both
expansion and compression type. Note that whereas the left-running
+
Mach lines are inclined a t the angle (0 a), measured counter-clockwise,
-
the right-running Mach lines lie a t the angle (0 - or), measured counter-
clockwise.
Many different simple-wave flow patterns may have the same image
in the hodograph diagram. For example, Fig. 15.8 shows the flo-NSalong
three curved walls, all with trhe same initial Mach Number, and all of
which are represented by the same hodograph characteristic. At each
point of each wall the flow properties are determined uniquely by the
streamline direction, and these flow properties are propagated out from
the wall along the corresponding Mach lines.
\
f
\
Mg
(0)
FIG.15.7. Right-running simple-wave flow.

(a) Expansion waves.
(b) Compression waves.
(c) Hodograph diagram.
/
6'
(dl
Fra. 15.8. Three simple-wave flows represented by the same hodograph charaoter-
istic.
(a, b, C) Physical plane.
(dl Hodograph plane.
473
Centered Waves. A special but im p ortant case of simple-wave flow is

that in which the streamlines are so shaped that all the Mach lines pass
through a common point 0 (Fig. 15.9a). I n Fig. 15.9b is shown a
special case of this special case, of even greater practical importance, in
which the boundary streamline itself passes through the center 0.
Either of these is a flow with centered simple waves, and is often called
(4 (4
FIG.15.9. Centered simple waves.
corner-type flow, or PrandtLMeyer flow. The corner-type flow is charac-

terized by the rules that (i) all flow properties are uniform on straight
rays emanating from the corner, and (ii) these rays make the local Mach
angle to the local flow direction.
A peculiarity of corner-type flow is that i t incorporates a rarefaction
shock a t the corner. However, this rarefaction discontinuity is confined
to a singular point, and hence there is no violation of the Second Law of
Thermodynamics. As the rarefaction wave propagates outward from
the corner, along the Mach lines, it becomes less steep, in accordance
with the considerations of Chapter 5.
A complete simple-wave $ow is one in which the Mach Number goes
from unity to infinity. Fig. 15.10~shows the hodograph representation
of a complete simple wave, and Fig. 15.10a shows the wave and streani-
line patterns if this complete simple-wave flow is of Prandtl-Meyer type.
From Eq. 15.6, for k = 1.4, w = 130.5" a t M*,. It follows from Eq.
Art. 15.2 .WAVES O F ONE FAMILY 475
15.8 that the maximum possible turning angle of the flow is also 130.5".
With this turning angle the stream would expand to zero pressure and
temperature. Since orl = SO0 and a2 = 0, the tangent to the hodograph
characteristic is along the radius a t the M* = 1 circle, and is normal to
the radius a t the M*,,, circle (Fig. 15.10~). This may also be deduced
from Eq. 15.3.
Query: what is the flow pattern when a flow with Mach Number unity
approaches's corner which turns through an angle of more than 130.5",
(4
FIG.15.10. Complete simple wave.
as in Fig. 15.10b? The flow is a t zero pressure when it has expanded

through 130.5", hence there will be a void, or cavitation ~ e g i o n adjacent
,
to the wall. This solution, however, is a t best only indicative. At very
low temperatures the thermodynamic properties may be very different
from that of a perfect gas, especially if liquefaction occurs. At very
low pressures, moreover, the assumption of a continuum is not realistic,
and, in order to get meaningful theoretical results, it is necessary to
employ the kinetic theory of gases.
Streamline Shapes. Suppose that in Fig. 15.1 l a we have simple-wave
flow along a curved wall, and that we let h. denote the distance, measured
along the Mach line, between a point on the wall and a point on a certain
streamline. Then the distance h* is proportional to the minimum cross-
I M
1 2 3 4 5 6 7 8
(a)
I
I
,
I
11'
rI \
/ '
. ,
(b)
0 40 80 120 160
v.90- ( a - w ) , Degrees
(c) (dl
FIG.15.11. Construction of streamlines for simple wave.
(a) Noncentered simple wave.
(b) Centered simple wave.
(c) Hodograph diagram.
(dl Chart for facilitating construction of streamlines.
sectional area for isentropic flow, and the distance h sin a is proportionai
to the cross-sectional area for flow a t any other section. We may there-
fore write
(h sin a)/h* = A/A*
where A/A* is the area ratio for isentropic one-dimensional flow, related
to the Mach Number by the formula (Chapter 4)
Then, since sin CY = l / M , we get

Art. 15.3 EXTENSION OF LINEAR THEORY 477
The Mach Number a t any point on the wall is uniquely determined

by Eq. 15.8a, and the Mach lines for this case, wherein 0 = -w, are
a t the angle ( a - w) to the horizontal. Since ( a - w ) depends only on
M, for each point on the wall streamline there may be plotted with the
help of Eq. 15.9 a corresponding point on any other streamline. Thus
the entire streamline pattern is easily determined.
If the initial flow is not a t Mach Number unity, the Mach line lies at
the angle [az- (w2 - w l ) ] to the initial direction, and the distances
between streamlines, measured along the Mach lines, are related by
The latter relation is most conveniently evaluated with the isentropic

tables, and is plotted in Fig. 15.11d.
When the simple-wave flow is of Prandtl-Meyer type, as in Fig. 15.11b,
we have
r h
-r*= - =h*& ( I +
3 M2 j& (15.11)
and the streamlines are easily plotted with r and (a- w ) as polar
coordinates.
15.3. Flow with Waves of Both Families by Extension

of Linear Theory
Modified Linear Theory. When waves of both families are present
in a problem, it is in general not possible to obtain an analytical solution.
However, a simple stepwise method may be devised which is founded
on the physical fact that changes in fluid properties in a supersonic flow
may be conceived as being brought about by pressure waves propagated
along the Mach lines of the flow.
The underlying idea is to imagine, for purposes of calculation, that
the physical flow plane is divided into many small regions, in each of
which all flow properties, including velocity, direction, pressure, etc.,
are considered constant. Considering two of these adjacent regions, or
fifields, which differ only slightly in properties, the linearized theory of
Chapter 14 may be applied to the changes in fluid properties as a stream-
line crosses from one field to the next. Now the linear theory shows that
the changes in properties occur as the streamline crosses a Mach line,
hence it follows that in general each field is a quadrilateral bounded
either by Mach lines or by physical boundaries. Occasionally the
Mach quadrilccteral takes on the limiting form of a triangle, as will be
seen in subsequent examples.
478 METHOD OF CHARACTERISTICS a.
15
Practical problems are then solved by constructing the wave pattern

which fits the given boundary conditions. The calculations are worked
out from field to field in stepwise fashion, using the linear theory for
each step in which a streamline passes from one field into another
across a Mach wave. In applying the linear theory to each Mach wave,
we employ the mean Mach Number between the two fields and draw
the Mach line a t the mean Mach angle to the mean flow direction.
Use of Hodograph Characteristics. The type of calculation outlined
above is facilitated by the use of the hodograph characteristic curves
of Fig. 15.5. From the analysis of the preceding article, we may set
down the following rules for using the hodograph chart:
(i) When a streamline crosses a left-running wave (Mach wave
of family 11), the hodograph of the streamline (that is, the
point in the hodograph plane representing the end point of the
velocity vector) moves along a hodograph characteristic of
family I. The Mach line is normal to the hodograph character-
+
istic of family I, and lies a t the angle (0 a) to the horizontal.
(ii) When a streamline crosses a right-running wave (Mach wave
of family I), the hodograph of the streamline moves along a
hodograph characteristic of family II. The Mach line is
normal to the hodographic characteristic of family 11, and
lies at the angle (0 - a) to the horizontal.
Each of the hodograph characteristic curves is identified by a partic-
ular value of I and IT. These values may be thought of as curvilinear
coordinates in the hodograph plane (mathematically they are the
characteristic coordinates), and may be used in place of M* and 0 to
specify the velocity vector. At a given point in the hodograph plane.
Eqs. 15.5 and 15.7 may be written:
First subtracting and then adding this pair of simultaneous equations,

we get
+
e = 1 II - 1000 (15.13a)
Thus, given 6' and w , 1 and 11 may be found from Eqs. 15.12. Or, con-
versely, given I and 11, 0 and w may be found from Eqs. 15.13. The
latter relations show also that in the hodograph plane, radial lines
(lines of constant flow direction) through the origin are lines of constant
+
(I 11), whereas concentric circles (circles of constant velocity and
pressure) are lines of constant ( I - 11).
Illustration of Method. The calculation procedure can best be described with

the help of a specific example. Let us consider underexpanded air flow at the exit
of a two-dimensional nozzle, the linearized solution for which was given in Fig.
14.23b. From symmetry considerations, the middle streamline must be straight,
and may be replaced by a wall (Fig. 15.12a), thus halving the calculations.
Suppose that at the nozzle exit, in field 1, M I = 1.4349, O1 = 0, and pl = 100.
Also, suppose that the surrounding pressure is 60.45.
From Table B.7 we get w l = 10.0" and (p/p0)1= 0.2991. Furthermore,
from Eqs. 15.12, we get 11 = 505 and 111= 495.
Now the boundary of the free jet must be everywhere at a pressure of 60.45.
The sudden drop in pressure at corner A from 100 to 60.45 can only take place
in an expansive corner-type flow. If 3 is the state after the corner flow, we have
Hence, from Table B.7,w3 = 20". Then, since states 1 and 3 lie on a common
I-characteristic, we have 1 3 = 11= 505. Eqs. 15.12 and 15.13 then give
03 = -10" and 1 1 3 = 485. Thus, there is a 10" deflection of the jet boundary
as it leaves the nozzle.
All the details of the corner flow between fields 1 and 3 are known analytically
from the simple-wave formulas of the preceding article. Subsequently, when
waves of both families are present, however, only stepwise methods can be used.
Therefore, we replace the actual Prandtl-Meyer continuous expansion by a
number of individual Mach waves into which the actual changes of state are
lumped. For the sake of illustration, we have here substituted two individual
waves, a and b, for the actual corner flow. Assuming equal angles of turn across
each of these waves, we have O2 = &/2 = -5O and 1 2 = I1 = 505. Then
w2 = 15.0 and ZZz = 490.
Points l', 2', and 3' may be spotted in the hodograph diagram of Fig. 15.12b.
Wave a is then drawn in the direction a', i.e., normal to the hodograph char-
acteristic of family I at the mean conditions between fields 1 and 2. Similarly,
wave b is drawn in the direction b', normal to the hodograph characteristic of
family I at the mean conditions between fields 2 and 3.
When wave a strikes the upper wall, it must be reflected as waye c in order to
have the flow follow the upper wall. If the region downstream of wave c is
called field 4, we have O4 = 0 and 1 1 4 = 1 1 2 = 490. Hence 1 4 = 510 and
wa = 20.0. Wave c is now drawn in the direction c', normal to the hodograph
characteristic of family I1 at the mean conditions between fields 2 and 4. The
intersection of wave c with wave b defines one corner of the Mach quadrilateral
for field 2.
Now assume that wave b is continued as wave d, and that wave c is continued
as wave e, and that they partially bound field 5. But field 5 is reached from
field 4 along a characteristic of family I, and is reached from field 3 along a
characteristic of family 11. Therefore Z6 = Zp = 510, and 1 1 6 = Z13 = 485,
whence ws = 25.0 and 196 = -5.0. Waves e and d are now drawn with the
respective directions e' and d'.
What happens when wave e strikes the free boundary of the jet? Since field 5
has a lower pressure than the surroundings, wave e must be reflected in the form
of wave g'to preserve the boundary condition of constant pressure. If the field
downstream of wave g is 7, we have 0 7 = 0 3 = 20.0 and 17 = 1 5 = 510, whence
117 = 490 and 0, = 0. Wave g may now be drawn with the direction g', which,
incidentally, coincides with the direction d'. The lower side of field 7 is the jet
boundary, which is drawn with the appropriate streamline direction, 07 = 0.
Fro. 15.12. Flow at exit of underexpanded nozzle.

(a) Physical plane. Full lines signify compression waves; dashed lines signify
expansion waves.
The remainder of the flow pattern may now be constructed by applying

successively the various methods outlined above. If an expansion wave is
denoted as one across which the pressure on a streamline decreases in the direc-
tion of flow, and conversely for a compression wave, in Fig. 15.12a expansion
waves are shown by dashed lines and compression waves by full lines. Various
streamlines may also be drawn from the known flow direction in each field, and
the pressure distribution plotted from the known Mach Numbers in each field.
The various properties in each field are summarized in the "Table for Fig.
15.12b." For each field, two of the four properties I, 11, u,and 0 are underlined,
thereby indicating the two properties in that field which are found first. The
remaining two properties in each field are of course found from the two under-
lined ones either algebraically from Eqs. 15.12 and 15.13, or graphically from
Fig. 15.12b.
Up- Down- UP- Down-

Wave stream stream stream stream Mean
+4 +4
Mean
(8 + a) (0
(@
(e - a) (e - (0 - a)
TABLE FOR FIG. 15.12b
Field I I1 o e a e + a e P/PO
+
The table also shows in each field the direction (0 a) for left-running waves,
and (0 - a) for right-running waves. These may be combined as in the "Table
for Fig. 15.12a" to get the inclinations of each wave. With the help of this
table, Mach wave directions may be determined without graphical operations
in the hodograph plane.
15.4. Application of Theory of Characteristics

The calculation method of the preceding article is based largely on the
physical concept of pressure waves being propagated a t the Mach angle
to the flow. We shall now show that the procedure has a rigorous ana-
lytical foundation which clarifies the significance and range of validity of
the calculation procedure usually termed the method of characteristics.
Differential Equation of the Velocity Potential. The physical condi-
tions of a two-dimensional, steady, isentropic, irrotational flow are
expressed mathematically by the differential equation of the velocity
potential (see Chapter 10)
+
(c2 - cpz 2)cpzz - 2rp,cpycpzPza,(c2 - cp,2)cp,, = 0 (15.14)
where u and v, which are respectively the x- and y-components of the
velocity vector, are given by
u = ~ z ;V = CP, (15.15)
and the local sound velocity, c, is related to the derivatives of the
. potential by the energy equation,
Definition of the Characteristic Curves. Our problem now is to find

solutions to Eq. 15.14 satisfying given boundary conditions. This
differential equation is of the general type
where the coefficients A, B, C, and D are functions generally of x, y, q,,

and rp,. Differential equations of this type are investigated in some
detail in Appendix A, where i t is shown that on the v(x, y) surface repre-
senting in cp, x, y-space the solution of Eq. 15.17, there lie certain charac-
teristic curves, so called, having special properties, namely:
(i) Real characteristic curves exist for supersonic, but not sub-
sonic flow.
(ii) The characteristic curves satisfy the differential equation,
Eq. 15.14, and hence may represent solutions of this differen-
tial equation.
(iii) On the characteristic curves there may be discontinuities in
the derivatives of the fluid properties. Regions of flow which
are analytically different may therefore be joined or patched on
the characteristics, the sole requirement being that there be
no discontinuities in the fluid properties themselves on the
pakhing lines.
Art. 15.4 APPLICATION OF THEORY OF CHARACTERISTICS 483
The characteristic curves of Eq. 15.14, as developed in Appendix A,

are defined by two pairs of simultaneous, ordinary differential equations
of first order:
The Characteristic Curves for Two-Dimensional, Supersonic Flow.

Comparing Eq. 15.17 with Eq. 15.14, we see that
Substituting these coefficients into Eqs. 15.18 and 15.19 we obtain,

after some rearrangement,
uv -1
Eq. 15.20 defines two characteristic directions at each point in the

physical plane. These directions give the slopes of the physical charac-
teristics a t each point. Likewise, Eq. 15.21 defines the slopes of the
corresponding pair of hodograph characteristics a t the hodograph image
point of each physical point. Shortly we shall see how these two sets
of '.differential equations taken together permit, in effect, numerical
solutions of the original Eq. 15.14 to be carried out.
It is evident from Eqs. 15.20 and 15.21 that for two-dimensional,

steady, irrotational, isentropic flow the characteristic curves are real
only for supersonic flow, that is, only when
u2 +v2
> c2, or V2 > c2
The Physical Characteristics as Mach Lines. We shall now show
that the physical characteristics are inclined a t the local Mach angle to
the local velocity vector. Using polar coordinates V and 8 for the hodo-
graph plane, we have
u = V cos 8; v = V sin 0 (15.22)
Introducing these into Eq. 15.20, remembering that V/c = M, and that
M = l/sin a and d~~ - 1 = l/tan a, we get
- cos 0 sin 8 f-1
sin2 cu tan a
-
cos28
I---
sin2 a
After considerable rearrangement and simplification wi t h the help of
the customary trigonometric identities, this becomes
+
( d y l d x ) ~= tan (0 - a); ( d y l d x ) ~=~tan (0 a) (15.23)
Referring to Fig. 15.13, Eq. 15.23 shows that each of the two physical
characteristics is inclined a t the Mach angle to the velocity vector, and
that the streamline direction a t a point in the physical plane bisects the
characteristic directions a t that point. Thus we see that the physical
FIG.15.13. Geometry of physical and hodograph characteristics.

Art. 15.4 APPLICATION OF THEORY OF CHARACTERISTICS 485
characteristics are identical with the Mach lines of the flow. The left-
running Mach lines are physical characteristics of family I1 and the
right-running Mach lines are physical characteristics of family I.
These results may be seen even more easily by arbitrarily placing the
z-axis parallel to the velocity vector a t a given point in t h e flow field.
At that point we would have u = V, v = 0, M = V/c, and 0 = 0 ;
Eq. 15.20 then becomes
lTIZ
Since, a t the point in question, the velocity vector coincides with the
x-axis, we conclude that the physical I-characteristic lies a t the Mach
angle below the velocity vector and the physical II-characteristic lies at
the Mach angle above the velocity vector.
Orthogonality Relations Between Physical and Hodograph Character-
istics. I t may be verified by algebraic simplification, using Eqs. 15.20
and 15.21, that
which means that the physical characteristic of one family is normal to the
hodograph characteristic of opposite family. This rule of reciprocal ortho-
gonality between the hodograph and physical characteristics is illus-
trated schematically in Fig. 15.13, and may be used to advantage for
graphical solutions.
Equations of the Hodograph Characteristics. The characteristic
curves for any particular problem are found by simultaneous integration
of Eqs. 15.20 and 15.21. Once these curves are established, the entire
problem is solved, for the characteristic curves relate points in the
physical plane to points in the hodograph plane, and hence the-Mach
Number and flow direction may be found for each point of the physical
flow field.
Now Eq. 15.20 shows that the differential equation of the physical
characteristics contains velocity terms, and must therefore be solved
simultaneously with the differential equation of the hodograph charac-
teristics. Quite differently, however, the differential equation (Eq.
15.21) of the hodograph characteristics contains only velocity terms, and
may therefore be integrated in advance of any specific physical problem
to give a set of universal hodograph characteristics which are independ-
ent of the flow pattern in the physical plane.
In obtaining the universal set of hodograph characteristics, it is con-
venient to employ polar coordinates, V, 8, rather than Cartesian coordi-
nates u, v. Applying the transformation of Eq. 15.22 t o Eq. 15.21, we
get after algebraic manipulation,
1
= 7 tan CY (15.25)
An even simpler method of getting Eq. 15.25 is to use the rule which
states that the normal to the hodograph 1-characteristic lies a t the angle
+a tu the velocity vector, while that for the 11-characteristic lies a t the
1 a . Applying this rule to the graphical construction of Fig. 15.14,
Eq. 15.25 becomes immediately
1
a
Eq. 15.25 is seen to be identical
th Eq. 15.1. Therefore, making
the substitution of Eq. 15.2, and
carrying out the integration of
Eq. 15.3, we arrive a t
01 = -w(M*) + 2 1 - 1000
(15.26a)
u 811 = +w(M*) + 211 - 1000
FIG.15.14. Geometrical interpretation
of the differential equation of the hodo- (15.26b)
graph characteristics.
where w(M*) is given by Eq. 15.6
and Fig. 15.4. Obviously Eqs. 15.26, representing the hodograph charac-
teristic curves, are identical with Eqs. 15.5 and 15.7. The two families
of curves of Fig. 15.5 are in fact the unique set of hodograph charac-
teristics which apply to any problem of two-dimensional, supersonic,
irrotational, isentropic flow. The symbols 1 and 11are the character-
istic coordinates, related to the state coordinates, w and 8, through Eqs.
15.12 and 15.13.
Example of Construction of Characteristics Nets. I t has already been pointed
out that, as a consequence of the special properties of the characteristic curves,
the original problem of finding a solution ~ ( xy), of Eq. 15.14 for given initial
conditions is replaced by the equivalent problem of constructing in the physical
and hodograph planes the characteristic curves of Eqs. 15.20 and 15.21. In a
sense this procedure gives the desired kinds of answers more directly, because
it is the velocity vector rather than the velocity potential which is of interest.
,The hodograph characteristics for all problems are given by Eq. 15.26. To
solve problems having specific physical boundary conditions, we construct
stepwise in the physical plane a characteristics net of whatever degree of fineness
is necessary to obtain the desired accuracy.
Art. 15.4 APPLICATION O F T H E O R Y O F CHARACTERISTICS 487
(a) Physical pla~lc. Solid lincs and

circles r c f r r t o Istrice- point
m e t h o d . Dash(.tl linc>s a n d
squares rdtrr t o ficld method.
(t)) Hodogr:~phdi;lgrnnl for (a).
(c) Str~:imlin(*s n r ~ d.\lac11 w:lvcs.
(d) Schliwvlr photograph of flow in
two-dimen~ionnl ~~ozzlc*wit11
s t r n i ~ l r t divergir~g w:tlls. T h c
llavlr lines arr. nitide visible by
srrlnll rougl~nc.sseson the noezle
walls (after hst:rr~arm).
(d)
Fro. 15.15. Source flow.
To illustrate the method, let us discuss the problem of a two-dimensional source
flow (Fig. 15.15). The exact solution to this simple problem is known from one-
dimensional considerations, and hence the only practical purpose of solving the
problem by the method of characteristics, other than for illustrative purposes,
is to investigate the effect of the fineness of the characteristics net on the accuracy
of the stepwise numerical solution.
We shall assume that we are given the straight, solid boundaries 1-8-15 and
611-18, inclined a t an angle of 15" to each other, and that along the circular
arc 1-2-34 the flow is radial and has a constant Mach Number of 1.348, corre-
sponding to a constant value of w = 7.5".
We begin by choosing the evenly spaced net points 1, 2, 3, and 4 along the
initial-value curve, thus arbitrarily determining the fineness of the net. Since
w and 8 are known a t each of these points, the corresponding values of I and 11
may be computed from Eqs. 15.12 and 15.13 and entered into the attached
table; or, graphically, the image points l', 2', 3', and 4' may be located in the
hodograph diagram, Fig. 15.15b. Of the four interconnected quantities I, 11,
w , and 8, of which two are independent and two dependent, those two which are
found first at each net point are underlined in the table.
Lattice
Point I I1 w 0 a 0+a 0-0 P/PO
Up- Down- UP- Down-

Wave * stream stream stream stream Mean
+ a) (0 +
Mean
(0 + a) (0 (0 - a) (0 - a)
(e - a)
* Items in parentheses refer to segments of flow boundary, and 0 rather than 0

is tabulated.
* a
Art. 15.5 SIMPLE WAVES BY THEORY OF CHARACTERISTICS 489
Since the hodograph characteristics are known to begin with, points 5', 6',
and 7' may be located immediately in the hodograph plane by means of the
hodograph characteristics passing through l', 2', 3', and 4'. Or, since the values
of I and II for 5', 6', and 7' are determined by the values of 1and 12for l', 2',
3', and 4' (e.g., 1 6 = I I and I I g = IIz),Eqs. 15.12 and 15.13 may be used for
filling in the values of w and 0 in the attached table.
To find point 5 in the physical plane, the physical I-characteristic 1-5 is drawn
normal to the hodograph 11-characteristic at the mean conditions between 1'
and 5'. Similarly the physical 21-characteristic 2-5 is drawn normal to the
hodograph I-characteristic at the mean conditions between 2' and 5'. The inter-
section of the two physical characteristics establishes the location of point 5.
If it is not desired to use the hodograph curves, wave 1-5 may be laid off at the
appropriate mean inclination (0 - a),and wave 2-5 a t the appropriate mean
+
inclination (0 a);these inclinations are shown in the attached table.
In locating points 8 and 8', we note that point 8 is on the physical boundary
and that the velocity direction a t 8' is known. This, together with the fact
that 8 lies on the 11-characteristic passing through 5, establishes points 8 and 8'.
Points 11 and 11' are found in the same way.
By continuing this process, the characteristics nets are established in both
planes. In the physical plane of Fig. 15.15a they are shown as solid lines con-
nected by small circles at the net points.
Figs. 15.15~and 15.15d show that the Mach wave pattern as determined by
this method is in good agreement with the pattern observed in schlieren photo-
graphs.
15.5. Simple Waves by Theory of Characteristics

Let us now investigate simple waves more rigorously b y means of
the theory of characteristics.
There are several interrelated ways in which flow with simple waves
may be defined. Let us use the definition that, in a region of flow with
simple waves, the velocity components u and v are not independent of
each other, b u t rather are unique functions of each other. What does
this restriction imply as t o other features of the flow?
We may write Eq. 15.14 in the form
Now, since b y assumption u = u(v), we may set
and
where the last equation is found by using the condition of irrotationality,

namely, that au/ay = av/ax. Inserting these expressions into Eq.
15.14a1it is found that av/ax vanishes, and we obtain
Solving for dv/du, we get
where either the positive or negative sign, but not both, is applicable,
because u and v are uniquely related. By comparison with Eq. 15.21,
it is seen immediately that the hodograph curve representing the unique
relation between u and v must in fact be one of the hodograph charac-
teristic curves.
Next, using the equation of irrotationality and the fact that u = u(v)
we carry out the following operations:
But, since dv = (av/ax) d x + (avlay) dy
so that dv
-- - -l/(d~/dx)~
du
Now dv/du is, as we have just seen, the slope of the hodograph charac-
teristic. Also, ( d y l d x ) , is the slope in the physical plane of lines of con-
stant v, and therefore of constant fluid properties, since u depends only
on v. Recalling now the reciprocal orthogonality relationships of Eq.
15.24, the equation immediately above proves that in a simple-wave
flow the lines of constant fluid properties are Mach lines of opposite
family to the hodograph characteristic on which the simple-wave flow
is mapped. Furthermore, since u and v are constant on each such Mach
line, the stream direction and Mach angle are likewise constant on each
Mach line. Hence the Mach lines under consideration must be straight
lines.
Art. 15.6 FIELD METHOD VERSUS LATTICEPOINT METHOD 49 1
A flow with "simple waves" may now be summarized as follows. I t

is a flow for which the Mach lines of only one family are physically sig-
nificant in the sense that the streamlines are influenced only as they
cross Mach lines of this particular family. The physically significant
Mach waves are straight, and all stream properties are constant along
each of these Mach lines. All the fluid states in a region of simple waves
are mapped on a single hodograph characteristic of family opposite to
that of the physically significant Mach waves.
In general, for flow with waves of both families, an area in the flow
field is mapped as an area in the hodograph plane. Flows with waves
of one family are singly degenerate because an area in the flow plane is
mapped as a line in the hodograph plane. A special case of simple-wave
flow is that of uniform, parallel flow, which is doubly degenerate because
an area in the flow field is mapped as a single point in the hodograph
plane.
15.6. Field Method Versus Lattice-Point Method
The reader has probably noticed that the procedure used for the
problem of Fig. 15.12 appears to differ from that for Fig. 15.15. The
former is termed the $field method, wherein the stream properties are
found in small regions or fields bounded by Mach lines. The latter is
known as the lattice-point method, wherein the stream properties are
found a t the lattice points of the Mach net.
These two methods are in substance identical. To illustrate this
point, the dashed lines of Fig. 15.15%illustrate the solution which would
be obtained by the field method, as compared with the solid lines ob-
tained by the lattice-point method. Midway between th? solid Mach
net, we imagine a dashed Mach net, a t the intersection of which are
small squares. As an example, we now consider the region surrounding
"lattice-point 9" to be one of uniform, parallel flow, and term this region
"field 9." The boundaries of field 9 are formed by the characteristic
quadrilateral e-g-n-h. In the field method the entire physical field of
flow is thus divided into Mach quadrilaterals, and the direction and
magnitude of the velocity vector are assumed to change only as a stream-
line crosses a Mach line joining two adjacent fields. An area in the
physical plane then corresponds to a point in the hodograph plane, e.g.,
field 9 corresponds to point 9'.
In using the field method, the solid Mach net of Fig. 15.15a would of
course not a t first be known, and it would be necessary to construct
independently the dashed Mach net dividing the fields of flow. To begin,
the initial-value line 1-2-3-4 is broken up into, say, four segments, l-a,
a-b, b-c, and c-4, along each of which the velocity is considered uniform
and parallel. In this case, the segment 1% is one of the bounding sides
of field 1, the latter covering only one-fourth, of a Mach quadrilateral.
Similarly, the initial-value segment a-b is a bounding side of field 2, the

latter comprising in this case one-half of a Mach quadrilateral. The
points l', 2', 3', and 4' are located in the hodograph diagram of Fig.
15.15b in accordance with the known velocity vectors in fields 1, 2, 3,
and 4, respectively.
To find the stream properties in field 6, say, we note that field 6 is
reachable from field 2 along a characteristic of family I, and is reachable
from field 3 along a characteristic of family II. Carrying out this con-
struction in the hodograph diagram locates 6', so that the properties in
this field are then known.
In addition to finding the stream properties in each Mach quadri-
lateral, we must determine the physical whereabouts of the Mach
quadrilaterals. To find a point such as e, we note that b-e is a Mach
wave of family 11, and is therefore normal to the hodograph I-character-
istic; the direction b-e is therefore drawn normal to 2'4'. By the same
argument, the direction a-e is drawn normal to 2'-5'. The intersection
of the two direction lines establishes point e. The remainder of the
Mach net is found by similar constructions.
The field method conveys the physical picture of streamlines being
influenced by pressure waves inclined a t the Mach angle t o the flow,
and is advantageous in this respect. However, when it is considered
that as a streamline crosses a Mach wave of one family, this is entirely
equivalent to the state point moving along the physical characteristic
of the opposite family, it will be seen that the field method and lattice-
point method are virtually identical except for small differences in pro-
cedure owing to the finite-difference feature of the method of charac-
teristics.
In general, the lattice-point method is slightly more convenient for
problems in which the boundaries of the flow are known, whereas the
field method is slightly more convenient when some of the physical
boundaries are t o be determined.
With either method, it is convenient to use equally spaced charac-
teristic coordinates if possible, for this leads to a simplifying regularity
in the calculations. In all the examples given thus far, for example, the
characteristics net was so chosen as to give equal spacings of the I and 11
coordinates between fields or between lattice points, and a glance a t
the tables for these examples will show the type of r e g u l a t y obtained.
With the field method, the choice of equally spaced coordinates is
equivalent t o the choice of equal turning angles across each Mach wave.
15.7. Unit Processes

All the various possible flow patterns can be synthesized in terms of the
corresponding wave patterns. The wave patterns themselves' are con-
Art. 15.7 UNIT PROCESSES 493
structed by repeated application of a relatively few operations, here

termed unit processes. They may be summarized as follows:
Generation of Wave. This occurs when a solid boundary of the flow
is curved, as in Fig. 15.6, or when the boundary of the jet is exposed to
a varying pressure, as near the nozzle corner of Fig. 15.12a. From
momentum considerations, we have the general rule that across a n ex-
pansion wave, the streamline bends away from the wave; whereas across a
compression wave the flow bends toward the wave.
Reflection of Wave at Plane Boundary. This is illustrated in Fig.
15.12a by wave a being reflected as wave c in order to preserve the given
flow direction at the wall. We observe that the incident wave a is an
expansion wave (because V2 > VI), as is the reflected wave c (because
V4 > V2), and hence we arrive a t the general rule that reflection of a
wave from a plane boundary i s in like sense.
Reflection of Wave from Free Boundary. This is illustrated in Fig.
15.12a by wave e being reflected as wave g in order to preserve the given
pressure a t the edge of the jet. Since wave e is an expansion wave, and
wave g is a compression wave, we infer the rule that reflection of a wave
from a constant-pressure boundary i s in unlike sense.
Cancellation of Wave. Especially in the design of supersonic wind-
tunnel nozzles, it is necessary to avoid reflection of waves incident on a
(a) (b)
FIG.15.16. Cancellation of a wave by appropriate curvature of wall.
wall boundary. This is done by shaping the boundary so that no reflec:,

tion is necessary to preserve continuity a t the wall. For example, in
Fig. 15.16 wave a of given strength is incident on the upper wall, and
it is desired that there be no reflection. This is easily accomplished if,
at point A where wave a is incident, the wall is bent so as to have the
same direction as the flow in field 2, for to satisfy this boundary condi-
tion i t is necessary that no wave be reflected.
Looked a t differently, if the upper wall were straight, wave a, which
is shown as an expansion wave, would be reflected as an expansion wave.
By bending the wall at A, however, by an amount equal to the turning
angle across wave a , a compression wave would be generated a t point A .

This compression wave exactly cancels the expansion wave which is the
reflection of wave a, with the net result that wave a merely ends at
point A . This line of argument underlines the concept of "canceling" an
incident wave by bending the wall through an appropriate angle.
Crossing of Waves. This is illustrated in Fig. 15.12a by waves b and c
crossing each other and emerging as waves d and e, respectively. Al-
though the waves are refracted as they cross each other, they are un-
changed in sense, and they retain their identity insofar as each has the
same values of Aw and A0 before and after the intersection.
15.8. Graphical Versus Numerical Method

In constructing the characteristics nets for any particular problem,
graphical and numerical methods may be used in several 'combinations,
depending on considerations of accuracy and time.
Purely Graphical Construction. The entire procedure may be carried
out graphically in both the hodograph and physical planes, using a large,
accurately drawn set of hodograph characteristics, as was illustrated in
connection with the problems of Figs. 15.12 and 15.15. This procedure
requires the drawing of normals to the hodograph characteristics, which
is difficult to do accurately. By use of the M*-a ellipse in conjunction
with the hodograph characteristics, the need for drawing normals is
eliminated, and the accuracy consequently improved.(5) Another tech-
nique for avoiding the troublesome task of drawing normals to a curve
is by use of the supersonic [email protected])
Semigraphical Construction. The graphical constructions involving
the hodograph plane may be replaced by such simple and rapid arithk
metical operations, based on Table B.7, that the hodograph plane may
well be reserved for illustrative purposes only. The types of calcula-
tions which replace the graphical use of the hodograph diagram are
shown in the "Tables for Figs. 15.12 and 15.15." I n the tables for
Figs. 15.12b and 15.15b are listed the flow properties in each field or a t
each lattice point. I n the "Tables for Figs. 15.12a and 15.15a" are listed
the slopes of each Mach wave in thc physical plane. The Mach nets
are then constructed graphically from the latter tables, using a protrac-
tor (preferably adjustable) or a drafting machine. The ease of the
method can only be appreciated by working out a specific case.
Purely Numerical Construction. When the utmost in accuracy is
required, it is best to eliminate graphical constructions entirely, and
employ numerical calculations for the physical plane as well as the
Art. 15.8 GRAPHICAL VERSUS NUMERICAL METHOD 495
hodograph plane. The numerical procedure may be formulated in

terms of an elementary step, which is described schematically by Fig.
15.17. Except at the boundaries of the flow region, the Mach net is
built up stepwise by repetitive use of an elementary step in which the
fluid properties and location of two initial points completely establish
the fluid properties and location a t a third point lying at the intersection
of the Mach waves passing through
the two initial points.
In Fig. 15.17, let us suppose that
we know wl,k x l , YI and ~ 2 &,, x2, I
y2, which is to say that we know
all the fluid properties a t points 1
and 2 and we also know the loca-
tions of points l and 2. Let point
3 lie a t the intersection of the I
I-characteristic passing through

point 1 and the 11-characteristic
passing through point 2. In the
stepwise procedure, the curved x
Mach lines are replaced by a set F I ~ .
15.17. Unit process of characteris-
method,
of straight-line chords extending tics
between the points of the nets. Let b~ be the average slope of the
chord 1-3 which approximates the Mach line 1-3; and, similarly, let bII
be the slope of the chord 2-3.
Now, from Eqs. 15.12, we may find the values of I and 11 a t points 1
and 2. Then, by virtue of the definition of point 3, we may find the
values of I3and 113through the relations
Having found I3and 113,Eqs. 15.13 may be used for computing w3

and e3,'thus determining fully the fluid properties a t point 3.
We are now in a position to compute the average slopes bI and bII.
From Eqs. 15.23, we may write approximately
b~ = tan +I(& - a3) + (01 - al)] = tan (0 - a)13 (15.27a)
= tan +[(83 + (~3)+ (02 + a d ] = tan (0 +

~ I I (15.27b)
and thus bI and bII are known.

Now, from the formulas of analytic geometry,
Solving for 23 and y3 from this pair of simulation equations, we have
and thus x3 and y3 may be solved for in terms of the initial data and the
values of bI and bII.
ORGANIZATION OF NUMERICAL CALCULATIONS. There are several ways
of organizing the calculations so as to put them on a routine basis suit-
able for computers. One type of organization, t o be used with the lat-
tice-point method, is illustrated in Fig. 15.18. The Mach net is twice
(sj, 0-4
FIG.15.18. Scheme of organization for purely numerical solution.
(a) Schematic characteristics net; computation of fluid properties.
(b) Schematic chvracteristics net; computation of coordinates of net points.
drawn schematically t o a very large scale; for example, a f45" net is

convenient. On one net, Fig. 15.18a, are listed the variables relating to
the fluid properties a t each point, and also the values of bI and bII on
each chord of the net. The latter are then transferred to the second net,
Fig. 15.18b1 where the remaining calculations required to establish the
physical location of each net point are made.
A similar organizational setup may easily be devised for use with the
Eeld method.
Art. 15.9 SOME SPECIAL FEATURES OF SUPERSONIC FLOW 497
At the boundary of the physical flow, one of the lines in Fig. 15.17
will be a segment of the boundary, and a modification of the computa-
tional technique is required. Either the flow direction or the fluid pres-
sure is then known along one of the lines in Fig. 15.17, and equations
similar to Eqs. 15.27 and 15.28 may be worked out for these special cases.
Effect of Grid Size on Accuracy. There is only one certain way of
determining the required Mach net for a specified final accuracy, and
that is to start with a very coarse net and obtain solutions with nets
successively finer, until two successive solutions agree t o the desired
accuracy. Occasionally it is possible to extrapolate to zero grid size.
A convenient rule is that the error owing t o the finite-difference type of
computation is proportional t o ( ~ w ) where
~, Aw is a measure of the
change in M* between fields or between lattice points.(3)
For example, Fig. 15.19 shows, for the sharp-cornered supersonic wind
tunnel nozzle of Fig. 15.34, the computed values of the exit coordinates
Aw, Deg. L\ w, Deg.

(a) (b)
FIG.15.19. Effect of net spacing on accuracy of solution for example of Fig. 15.34
(after Edelman).
(referred to the half-throat height) as a function of the magnitude of Aw

between fields. From one-dimensional considerations, the exit height
can be computed exactly; this value corresponds to Aw equal t o zero.
For Aw = 2", these curves show a 0.5% error in the final y-coordinate,
and a 7% error in the final x-coordinate.
For most engineering work not requiring extreme accuracy, a Aw of
1 or 2 degrees is satisfactory. The accuracy of an entire solution may
be improved considerably by using especially small intervals near Mach
Number unity, where certain derivatives vary rapidly and approach
infinity.
15.9. Some Special Features of Supersonic Flow
Initial-Value Problem. Having seen several sample problems, we
now ask, what initial data are necessary t o obtain solutions by charac-
teristics methods; and, for various types of initial data, within what
regions of flow may the solution be established?
This question is discussed in Appendix A, where it is shown that solu-

tions may be obtained in those flow regions within which the character-
istics nets may be constructed from the given initial data. For two-
dimensional, supersonic flow, the various important situations may be
summarized as follows:
FIG. 15.20a. The velocity vector is given a t all points of the non-
characteristic line AB. The solution is then completely determined in
FIG.15.20. Several initial-value problems.
the Mach quadrilateral ADBC. Note that the flow upstream of AB as

well as the flow downstream is established by the initial data.
This type of initial data is illustrated in the example of Fig. 15.15.
FIG. 15.20b. The velocity vector is given a t all points on the char-
acteristics AB and AC. The solution is then completely determined in
the Mach quadrilateral ABDC. As an illustration, see the example of
Fig. 15.23.
FIG. 15.20~. The velocity vector is given a t all points of the non-
&aracteristic line AB, and a t all points on the noncharacteristic lines
AC and B D there is given either the velocity magnitude or the velocity
direction. The solution is then completely determined within the figure
FACEDB. As an illustmtion, see the examplc of Fig. 15.12a.
FIG. 15.20d. The velocity vector is given a t all points of the char-
acteristic line EC, and along the noncharacteristic line AB there is given
Art. 15.9 SOME SPECIAL FEATURES O F SUPEWONIC FLOW 499
either the velocity magnitude or the velocity direction. The solution is

then completely determined in the region ECD.
Other combinations of initial data may be devised, but all the impor-
tant elements are included in the foregoing list.
Physical Significance of Characteristics. According to the initial-
value theorem for hyperbolic equations, if the velocity components are
known along - the noncharacteristic curve 1-2 of Fig. 15.21, then the en-
tire flow pattern in the Mach
quadrilateral 14-2-3 may be
solved for. Now, suppose that
the velocity components are al-
tered along 1-1'. This might be
thought of as a disturbance of
the flow in the neighborhood of
point 1. Since the values of the
velocity components remain un-
altered along 1'-2, it is evident
that the flow pattern in the char-
acteristic quadrilateral 1'-4'-2-3'
is not affected by disturbances
along 1-1'. Hence the effects of X
these disturbances are confined Fro. 15.21. Illustrates propagation of

disturbances along Mach lines, domain of
entirely t o the regions 1-44'-1' dependence, and range of influence.
and 1-1'-3'-3. We see, therefore,
that disturbances are propagated along the Mach lines, for if point 1'
approaches 1, the effects of a small disturbance in the interval 1-1' are
felt only along the characteristics 1-4 and 1-3.
These mathematical considerations have a very real physical counter-
part. If small disturbances are introduced into a supersonic flow, as by
roughening the walls of a passage or the surface of an airfoil, or by in-
serting a fine needle into the stream, schlieren and shadow photographs
show that these disturbances are propagated into the stream a t the
Mach angle to the flow.
Domains of Influence and Dependence. The initial-value theorem
indicates also that the specification of the velocity components along
the line 5-6 in Fig. 15.21 completely determines the flow $thin the
characteristic semi-quadrilateral 5-7-6. In other words, the flow pat-
tern in the region 5-7-6 is independent of the data along 1-5 and 6-2.
This means that conditions a t 7 depend only on the data along 5-6. We
may generalize this argument by stating that the "domain of dependence"
for a given point of the flow, or the region within which events can have
any effect on the conditions a t the given point, is the region bounded
by the upstream-running Mach lines passing ihrough the given point.
From similar considerations, i t may be demonstrated that the initial

data a t point 8 of Fig. 15.21 can affect the flow pattern only in the shaded
regions bounded by the characteristics passing through point 8. Al-
though both shaded regions appear in the mathematical argument, only
the shaded region pertaining to the Mach lines downstream of point 8
is physically significant, because pressure disturbances cannot be propa-
gated upstream in a supersonic flow. The downstream shaded region is
called, therefore, the range of influence a t point 8. Or, as von Karman
has put it, the region bounded by the downstream Mach lines through a
given point is the zone of action in which events a t the given point make
themselves manifest; while the entire surrounding region is the zone of
silence, within which events a t the given point are not felt. For exam-
ple, if a thin, cusp-edged profile is placed in a two-dimensional, super-
sonic flow a t zero angle of attack, the effects of the profile are confined
to the regions downstream of the Mach lines attached to the leading
edge.
Comparison Between Subsonic and Supersonic Flow. We have
already compared subsonic with supersonic flow on several different
bases. We may generalize these ideas by comparing flows defined by
hyperbolic differential equations, represented for example by steady,
supersonic flow, and flows defined by elliptic differential equations,
represented for example by steady, subsonic flow.
In supersonic flow, the phenomena are essentially those of wave
propagation, in which disturbances a t individual points of the flow are
propagated along the downstream Mach lines passing through these
points. Thus there exist in supersonic flow domains of dependence and
ranges of influence. These statements are connected, of course, with
the fact that in supersonic flow analytically different regions may be '/,
connected along the characteristics, so that the solutions of the differen-
tial equations are, in general, nonanalytic.
In subsonic flow, no real characteristics exist. The solutions t o the
differential equations must be analytic functions which apply over the
entire flow field. A disturbance a t any point in the flow field makes
itself felt a t all other points. Or, to put it differently, the domain of
dependence and range of influence for each point covers the entire field
of flow. Physically, this comes about because a pressure wave generated
a t any point in a subsonic flow can reach any other point of the flow.
Since no real characteristics exist for subsonic flow, no discontinuities
in any of the velocity derivatives may in general occur. For example,
if it is known that in any small finite region the flow is exactly uniform
and parallel, then it follows that the flow must be uniform and parallel
throughout the entire region of flow. Conversely, if the flow is non-
uniform in any region, then it must be nonuniform in all regions.
Art. 15.10 APPLICATIONS OF METHOD OF CHARACTERISTICS 501
15.10. Applications of Method of Characteristics

To illustrate how the method of characteristics may be used for prac-
tical calculations, a number of specific examples are discussed below.
Thin Profile. Fig. 15.22 shows a thin, curved profile whose leading
edge is parallel to the oncoming air stream. Above the profile there will
be only left-running Mach waves and below there will be only right-
running Mach waves. At any point on the profile the local pressure
(4 (b)
FIG.15.22. Flow past thin profile.
(a) Flow plane.
(b) Hadograph plane.
depends only on the local inclination of the surface, on the approach

Mach Number, and on whether the point is on the upper or lower
surface.
Note that the compression waves generated by the ,lower surface
converge. As described in the next chapter, they ultimately form an
oblique compression shock (see Fig. l5.6d).
When a profile has thickness, oblique compression shocks are usually
generated a t the profile itself, and the analysis of such profiles requires
the combined application of the method of characteristics and of ob-
lique-shock theory. Such profiles are discussed in the next chapter.
Underexpanded Nozzle. Fig. 15.23 shows the complete wave pattern
for a nozzle discharging into a region where the back pressure is lower
than the exit-plane pressure.
This example illustrates the role of the characteristics as patching
lines along which may be joined regions with analytically different types
of-fJows. Regions 1, 2, 3, 4, and 5 are doubly degenerate, since they
have uniform, parallel flows. Between 1 and 2, between 2 and 3, etc.,
are singly degenerate regions of simple waves. The remaining parts
of the. jet comprise regions in which waves of both families are present.
Each of the three types of regions is analytically different. They are
joined to each other, however, only along Mach lines, which are the
patching lines for supersonic flow. On the Mach lines the velocity is
continuous, but there are discontinuities in the first and higher deriva-
tives of the velocity.
The pattern shown in Fig. 15.23 is repetitive, since field 5 is identical
in state with field 1.
FIG.15.23. Flow a t exit of underexpanded nozzle (compare with schlieren photo-

graph of Fig. 5.24).
(a) Flow plane.
Schlieren photographs of jets (Fig. 5.24) confirm beautifully the

theoretical patterns of the type of Fig. 15.23, except of course that, as a
result of friction between the jet and the surrounding gas, the analytical
solution-breaks down after one or two cycles.
A Simple Supersonic Wind Tunnel. Fig. 15.24 shows a two-dimen-
sional passage, formed of two Prandtl-Meyer turns, which accelerates
a stream from Mach Number unity a t section 1 to some higher Mach
Number a t section 3.
Art. 15.10 APPLICATIONS OF METHOD OF CHARACTERISTICS 503
If it is desired to have the stream a t 3 parallel to the stream a t 1, it is

evident from the hodograph diagram that the turning angles must be
related to the final Mach Number by the relation
Of course, parallelism of the streams a t 1 and 3 is not a necessary

requirement, nor is it necessary that a centered simple wave be employed,
j -i7--
82-g3,
I f 500 1
(b)
Fro. 15.24. Design of a simple supersonic wind tunnel.
(a) Flow plane.
for similar results may be obtained with a noncentered simple wave.

However, if centered waves are used, the streamlines are Prandtl-Meyer
streamlines and may be laid out very easily.
Although this design of wind tunnel nozzle would be satisfactory for
an inviscid fluid, it is seldom employed in practice when a highly uni-
form, parallel stream is desired. The boundary layers on the passage
walls produce slight deviations from the theoretical flow pattern. For
the nozzle of Fig. 15.24 these deviations are not symmetrical for the
two walls, and hence lead to irregularities in the final flow. The con-
ventional type of supersonic wind tunnel nozzle is symmetrical, and is
discussed in Art. 15.11.
Supersonic Elbow. Fig. 15.25 shows a two-dimensional elbow which

turns a supersonic stream through the angle (01 - 03) without any net
change in Mach Number. Two simple-wave turns are employed, in
one of which left-running Mach waves expand t,he flow from 1 to 2, and
in the other of which right-running Mach waves compress the flow from
FIG.15.25. Design of a supersonic elbow.

(a) Flow plane.
2 to 3. In this illustrative example, the waves are assumed to be cen-

tered, and hence the streamlines are Prandtl-Meyer streamlines.
From symmetry conditions and the specification that w l = wa, it
follows that the elbow is symmetrical about A-A, and that
Impulse Turbine Blade. I n an impulse steam turbine or gas turbine,

a stream approaching a set of passages is turned through a certain angle
without any net change in pressure. If the flow relative to the blades
Art. .10 APPLICATIONS OF METHOD OF CHARACTERISTICS 505
is supersonic, then the type of passage illustrated by Fig. 15.26 is

theoretically suitable. The blades themselves are formed by having
one of the passage walls act as the concave surface, and the other passage
wall the convex surface.
B' P a r a l l e l Flow
__------
Fro. 15.26. Design of supersonic impulse blade.

(a) Flow plane.
Supersonic Compressor Cascade. Fig. 15.27 shows a pair of blades

which might be used for the stator cascade of a supersonic compressor.
Gas enters supersonically a t 1, is compressed by means of right-running
Mach waves in a centered simple wave to a Mach Number of about
1.2 to 1.5, and is then decelerated to subsonic speeds by a normal shock,
after which it is decelerated subsonically and also turned into the axial
direction.
Compressors employing this principle seem to hold great promise for

obtaining large pressure rises per stage with good efficiency, since the
losses in a shock a t M = 1.5 are very sma.11if separation can be avoided.
Two-Dimensional Supersonic Inlet. Fig. 15.28 shows an inlet dif-
fuser which might be placed in the leading edge of the wing of a super-
Fro. 15.27. Design of supersonic compressor cascade.

(a) Flow plane.
sonic airplane. Two Prandtl-Meyer turns and a weak normal shock

are used for reducing the incoming air speed to subsonic values.
Such a design might also be used for supersonic wind tunnel circuits
if the boundary layer on the tunnel walls could be removed so.as to
avoid separation of the flow in a region of positive pressure gradient.
Special problems arise in the starting of supersonic diiusers because
of the need for swallowing the starting shock. This problem may be
met in the design of Fig. 15.28 by making the central plug retractable,
so that the throat area may be temporarily enlarged.
Effects of Viscosity. The effects of viscosity are confined t o a thin
boundary layer unless the Reynolds Number is extremely low or unless
there is back flow in the boundary laver.
Art. 15.11 SUPERSONIC WIND TUNNEL NOZZLES 507
In an accelerating flow, the boundary layer tends to remain very
thin, and there is no tendency toward separation of the flow. Hence,
flows like those in Figs. 15.15, 15.23, and 15.24 in general exhibit very
good agreement between experimental results and results based upon
the 'method of characteristics for a frictionless fluid.
In a decelerating flow, the boundary layer tends to thicken, and, owing
to the adverse pressure gradient, there is a tendency to separation
because of back flow in the boundary layer. The former effect produces
an even larger pressure gradient in a supersonic flow, and thus tends to
aggravate the latter effect. If a slight amount of separation appears,
(a) Flow plane. (b) Hodograph plane.

FIG.15.28. Design of supersonic inlet.
the interaction between the resulting shock wave and the boundary
1a;yer makes the situation grow rapidly worse. There are likely to be
great disappointments, therefore, in the results obtained with passages
like those of Figs. 15.25, 15.26, and 15.27 unless due account is taken of
these problems. Shocks will usually appear and the general flow pattern
will be totally unlike what is expected. There is evidence, however,
that good results may be obtained with decelerating flows if three pre-
cautions are taken: (i) very large adverse pressure gradients should be
avoided, (ii) the passage walls should be displaced by the amount of
a carefully computed boundary-layer displacement thickness, and (iii)
any large initial boundary layer should be removed by suction. For
further details, the reader should refer to Article 16.12 and to Volume
11, Chapter 28.
15.11. Design of Supersonic Wind Tunnel Nozzles

The design of nozzles for two-dimensional, supersonic wind tunnels
constitutes one of the important practical applications of the method
of characteristics.
General Considerations a s to Form of Nozzle. The function of such a
nozzle is to accelerate a stream from Mach Number unity to some d e
sired final Mach Number. It is of the utmost importance, in order that

free flight be simulated, that the stream entering the test section be
uniform and parallel.
Referring to Fig. 15.29, if the nozzle is symmetrical, there must be a
general divergence from the throat to the test section. Also, because of
symmetry, the center line is a streamline, and may be considered a
Test S e c t ~ o n
[Uniform, Parallel
Flow 1
(a)
/
A\----
\
..'4
boundary of the flow for calcula-
W~.
/
:1---.
.,L -.
/- . -7
.-
tion purposes. Hence we need
consider only the upper half of
the nozzle.
e, T e = 0 5 ~ 7 * 4' I,,?' In order to have a net increase
e,=~,
in area without any net change in
(b) flow direction, the wall contour
FIG. 15.29. Nozzle for producing m i - must first curve outwards from
form, parallel flow.
7 to 3, and must then curve in
(a) Flow plane.
(b) Hodograph plane. again until a t the exit (point 1)
the wall is parallel to the initial
flow. Point 3 is an inflection point and is the point where the wall has
its maximum slope, Om,,. For the present we shall assume that the
subsonic inlet is designed to give a parallel, uniform, sonic flow a t the
throat, section 6-7.
ZONES I N NOZZLE. Three separate zones of flow may be identified in
Fig. 15.29a:
(i) The expansion zone, 6-7-3-2-6, bounded by the throat 6-7, by
the characteristic 2-3 which on the center line of symmetry attains the
required test-section Mach Number, and by the expansion portion of the
nozzle wall 7-5-3. This is a zone where waves of both families are
present.
(ii) The test section, which is the region downstream of 1-2, where the
flow is uniform and parallel a t the test-section Mach Number ME. Be-
cause of the uniform, parallel flow, the Mach line 1-2 is straight, and is
inclined a t the angle c u ~to the center line.
(iii) The straightening section, 3-2-1, which is bounded by the Mach
lines 3-2 and 2-1, and by the straightening portion of the wall, 3-1.
Since it is a general theorem that only a zone of simple waves can be
patched to a uniform, parallel flow, it follows that zone 3-2-1 must
have a simple-wave flow such that all flow properties are uniform on
straight left-running Mach waves in this zone. The wall contour is then
the streamline which begins a t point 3 and crosses the left-running
Mach waves until it emerges horizontal a t point 1. Since it is undesirable
for reasons of boundary-layer behavior to have compression waves in
the nozzle, the simple wave in the straightening section must be an ex-
pansion wave, and hence the straightening contour 3-1 must be concave
downward. Note that the determination of the flow in this zone from
initial data on the Mach lines 3-2 and 2-1 is an initial-value problem
of the type of Fig. 15.20b.
We may also discuss the main principles of the design using the physi-
cal concept of pressure waves propagating along the Mach lines. Assum-
ing sonic flow a t the throat, the wall is curved outward between 7 to 3.
This generates right-running expansion waves which are reflected from
the center line of symmetry as left-running expansion waves, which may
be reflected again and again a t the wall and the center line.
This process of generating expansion waves by curving the wall is
continued to the point where one of the waves (3-2) accelerates the flow
a t the center line (point 2) to the desired test-section Mach Number.
The left-running waves which cross line 3-2 would in general be reflected
from the wall. However, the purpose of the design is to avoid pressure
waves in the test section, and hence the straightening contour 3-1 is
curved so as to cancel the waves which are incident on it. Thus the re-
gion 3-2-1 is one of simple waves.
It is clear that if the expansion contour 7-5-3 is selected arbitrarily,
thus allowing the Mach line 3-2 to be constructed, it is possible then to
construct the streamline 3-1 and thus complete the nozzle design. We
now ask, what are the considerations that go into the selection of the
expansion contour, 7-5-3?
Design of Expansion Contour, 7-5-3. The design of the expansion
contour depends somewhat .on the shape of the sonic line a t the throat.
For illustrative purposes, and to make the discussion concrete, let us
suppose that the sonic line is straight and normal t o the axis, as in
Fig. 15.29a.
Furthermore, also for the sake of concreteness, let us assume for the
present that the first test-section Mach line (2-I), when extended back-
ward, is reflected a t the wall contour twice (at points 3 and 5) before
reaching the sonic line. Now, if the radius of curvature of the contour
7-5 were finite, the waves originating a t 7 would be reflected infinitely
many times a t the axis and the contour before the Mach Number in-
creased above unity, because the Mach line running rightward from 7 is
normal to the axis, and likewise the Mach wave reflected from 6 is also
normal to the axis. Therefore, to meet the stipulation that the back-
wardly extended Mach line 1-2 be reflected only a t 3 and 5, it is neces-
sary that the contour 7-5 have zero radius of curvature. This means
that points 5 and 7 coincide in physical location, that there is a corner
a t 7, and that the lines 6-7 and 5-7 are identical.
Although in the triangular region 7 - 4 4 there are waves of both
families, the corner 7 is a singular point a t which there is locally a cen-
tered simple expansion wave, so that the series of states between 7 and
5 map on a single hodograph characteristic, as in Fig. 15.2913. The lat-
ter figure also shows the hodograph characteristics corresponding to the
backwardly extended Mach line 1-2.
Now, from Eqs. 15.26, since B6 = Bp = el = 0,
Furthermore, since U6 = 0,
Therefore,
wl = 193) + le,l + b5l 4- lO5l = 2l9,l + 21e51
1931
- - bsl -
- 1 - ---- W5
-1-2- (15.29)
4 2 4 2 0 1
Since the expansion contour 5-3 is concave outward,

1931 - 5 20
But, from Eq. 15.29,
Hence, since 1951 is also greater than zero,

W1 a1
OI1951 I - ;0 _ < 0 5 5 -
4 4
and, putting this into Eq. 15.29, we find that
Thus, equations 15.30 and 15.31 put certain limitations on the design.
It is instructive to examine the two extremes:
Art. 15.11 SUPERSONIC W I N D TUNNEL NOZZLES 511
(i) 4 0 ,= 1/4 and 1831 = ul,!4. In this case, = 10.71,SO that the
contour 5-3 is a straight line whose slope is related to the test-section
Mach Number by
1031 = lesl = 01/4
The corresponding nozzle contour and characteristics diagram are shown
in Figs. 15.30a and b.
-
Test
-.
6 4
('J)
FIG.15.30. Nozzle with w b / w l = % and 1031 =4 4 .
(a) Flow plane.
(ii) 05/01 = 0 and 14= w1/2. In this case, points 5 and 7 are iden-
tical as to both physical location and fluid properties, but, since a cor-
ner a t 7 is still necessary, point 3 is coincident with 7, although it has
different properties. The maximum slope of the contour is immediately
FIG.15.31. Nozzle with w ~ . / w l = 0 and (6'31 =d 2 .

(a) Flow plane.
(b) I-Iodograph plane.
downstream of the corner, and is related t o the final Mach Number by

= w1/2. Figs. 15.31a and b show the nozzle contour and character-
istics diagram for such a nozzle.
GENERALIZATION FOR n REFLECTIONS. The relations given above may
now be generalized for the case where there are n backward reflections
of the first test-section Mach line on the wall contour (n = 2 for Fig.
15.29a). Referring to Fig. 15.32, let A be the downstream side of the
(b)
FIG. 15.32. Nomenclature for generalized nozzle-design relations.
corner, let B denote the infle tion point of the contour, and let EE r e p
resent the first Mach line in the test section. Then
At the right-hand limits of the latter two equations,
Therefore, a straight line of inclination wE/2n connects points A and B,

the contour BE is curved, and the nozzle contour looks like that of
Fig. 15.30a.
At the left-hand limits, w~ = 0, which means that point B must be
at the downstream side of the corner. Furthermore, gB = wE/2. The
nozzle is therefore identical with that of Fig. 15.31a, and its maximum
wall inclination, of magnitude WE/^, is a t the corner. All n reflections
then occur at the corner.
Sharpcorner Nozzle. As we have just seen, a limiting design is
that for which the entire expansion contour is contracted to a sharp
corner a t the throat, of angle wE/2, and for which the straightening con-
tour extends from the corner to the exit (Fig. 15.31).
Art. '15.11 SUPERSONIC WIND TUNNEL NOZZLES 513
The construction of the wave pattern and wall contour for such a
nozzle, using the field method, is illustrated graphically in Figs. 15.33a
and b.
7'
FIG. 15.33. Construction of contour of sharpcorner nozzle.

(a) Wave diagram.
A scale drawing of the Mach net for W E = 28' (ME = 2.059, BB = 14O),
using lo-intervals, is shown in Fig. 15.34. I t is to this design that Fig.
15.19b, illustrating the effect of Mach-net fineness on accuracy, pertains.
No nozzle not conta.ining compression waves can have a smaller ratio
of length to throat width nor a larger value of 8,- than the sharp-corner
FIG. 15.34. Mach lines for sharp-corner nozzle with exit Mach Number of 2.059
(WE = 2g0), based on lo-intervals of w . The origin of the x, y-coordinate system is
in the plane of the throat, on the line of symmetry, and y~ denotes the half-height
a t the throat (M.I.T. Gas Turbine Laboratory).
nozzle with Om,, = q / 2 . This type of nozzle is also becoming increas-

ingly popular because it has the most simply defined expansion contour.
If the sharp corner is thought undesirable because of boundary-layer
considerations, one of the interior streamlines may be employed for the
wall contour.
The region with waves of both families (zone 74-6 in Fig. 15.31), if
drawn for a very high exit Mach Number, contains within it the mixed-
wave zone for all lower Mach Numbers. This is illustrated in Fig. 15.35,
Contour For w~ = I 2
------r---
-.
FIG. 15.35. Illustrating construction of contour for sharpcorner nozzle from "kernel."
where it is shown how a portion of the mixed-wave zone for UE = 12"

may be used to determine the wall contour for UE = 8". Thus, if the
mixed-wave zone, or "kernel," is very carefully worked out once for some
particular Mach Number, the calculations for the kernel need never be
repeated for lower Mach Numbers.
Coordinates of carefully calculated nozzle contours up to WE = 28"
are tabulated in Reference 9, and the kernel coordinates for values of
ME up to 10 are given in Reference 10.
The nozzle contours (g) are shown in Fig. 15.36. In Fig. 15.37,(1°'
the ratio of nozzle length to test-section height is shown as a function of
final Mach Number.
COMPARISON WITH EXPERIMENT. The very close agreement between

theory and experiment for an accelerating flow is illustrated by Figs.
15.38 and 15.39.(9) Fig. 15.38 shows the theoretical and experimental
pressure distributions on the axis of the nozzle. Fig. 15.39a shows the
-- -...
0.6
0 Measured
Theoretical
FIG. 15.37. Length of sharp-corner noz- FIG. 15.38. Comparison between meas-
zle (after Shames and Seashore). wed and theoretical pressure distribu-
tion on axis of sharp-corner nozzle of
Fig. 15.34 (after Edelman).
FIG.15.39. Contours of constant pressure for sharp-corner nozzle of Fig. 15.34

(after Edelman).
(a) From method of characteristics.
(b) From interferogram analysis.
(c) Typical interferogram of zero band width, showing contours of constant.
density.
theoretical curves of constant density and is to be compared with the

measured curves of constant density of Fig. 15.3913. The latter chart
was c'onstructed with the help of interferograms similar to Fig. 15.39~.
The agreement between theory and experiment is particularly good
when it is considered that the test section of the nozzle was less than
1 inch by 1 inch, thus making boundary layer effects comparatively
prominent.
The Subsonic Inlet. The shape of the sonic line, from which the
entire nozzle design begins, depends on the shape of the subsonic inlet.
There is no theoretical reason why any particular shape of sonic line is
best, but the important thing is to know the shape. There are several
possibilities:
(i) Subject to limitations of boundary-layer thickness, the subsonic
inlet might be made so long and have such gentle curvature as to insure
that the sonic Line will be almost exactly a straight line normal to the
axis.
(ii) For a given shape of inlet the shape of the M = 1 line might be
determined by analytical methods (see Chapter 21, Volume 11).
(iii) For a given shape of inlet, preliminary tests, perhaps on a model,
might be used to determine the sonic line. With this procedure i t
would probably be preferable to determine the conditions at some line
downstream of the throat, so as to avoid the difficulty of constructing
the Mach net near Mach Number unity. .
In any case, to eliminate separation or thick boundary layers, the
subsonic contours should be designed so that there are no local regions
of rising pressure.
Construction Near Mach Number Unity. Because dM/& approaches
infinity as M approaches unity, the Mach net must be very fine in this
region in order to have good accuracy.
Boundary Layer. The effects of friction are, for an accelerating flow
with high Reynolds Number, confined to a thin boundary layer. In
most nozzles, therefore, friction has only a perturbation effect. It acts
to decrease t,he cross-sectional area available for flow, and thus to reduce
slightly the exit Mach Number.
For best results it is desirable to correct the nozzle contour a t each
point by the amount of the local boundary-layer displacement thick-
ness.
15.12. Adiabatic, Nonviscous Flow with Rotation
As shown in Chapter 16, an adiabatic nonviscous flow which is orig-
inally irrotational may become rotational as the result of passing through
a curved shock. If there are no viscous or heat conduction effects, except
Art. 15.12 FLOW WITH ROTATION 517
within the thin zone which we think of as the shock discontinuity,

the stagnation enthalpy remains constant on each streamline. More-
over, the entropy on each streamline remains constant both upstream
and downstream of the shock but undergoes an increase across the shock.
Since the entropy increase for each streamline depends on the correspond-
ing shock strength, the flow downstream of a curved shock is rotational
(the rotation in a fluid is related to the rate of change of entropy and of
stagnation enthalpy in a direction normal to the streamlines).
We shall outline in this article, therefore, the method of characteristics
for a steady, two-dimensional, adiabatic, supersonic, nonviscous, rota-
tional flow in which the stagnation enthalpy is the same for all points
and in which the entropy is constant on each streamline but varies from
one streamline to the next.
Governing Equations. The equation of continuity may be written
Since there are no viscous forces, Euler's equations may be written
Since we are considering regions of flow from which shocks are ex-
cluded, the entropy is a continuous function of x and y. Hence
Now, the definition of a streamline is dy/dx = v/u. Furthermore, along

a streamline, ds = 0, by assumption. Making these substitutions
simultaneously, we get
+
u ( ~ s / ~ x ) v(as/ay) = o (15.38)
Since the pressure is a continuous function of entropy and density,
For a perfect gas,

ds/cv = dp/p - ?I ~P/P (15.40)
so that (ap/as), = p/c,. Furthermore, the definition of sound velocity
applied to a .perfect gas gives
Thus Eq. 15.39 may be expressed as

and the partial derivatives ap/ax and a p / d y which appear in Eqs. 15.36
and 15.37 may therefore be written
a ~ / a z= (PC +
/ k ) ~ ( s / c , ) / ~ x c2 a p / a x (15.41)
aday = ( P C ~ / ~ ) ~ ( s / c , ) / ~+
Y c2 splay (15.42)
Since u, v, p, and s are all continuous functions of x and y, we may
write
+
du = ( a u / a x ) d x (aulay)dy (15.43)
dv = (av/ax) d x + (av/ay) d y (15.44)
dp = (ap/ax) dx + ( ~ P / ~ d~Y I (15.45)
= (as/ax) dx + (aslay) d y (15.46)
Solution of Equations. The expressions for dp/dx and a p / a y given by
Eqs. 15.41 and 15.42 are now substituted into Eqs. 15.36 and 15.37,
respectively. We then have eight quantities, namely, au/ax, du/ay,
dv/ax, av/ay, ap/ax, ap/ay, &/ax, and aslay, which are connected by
eight independent, linear, nonhomogeneous equations, namely, Eqs.
15.35, 15.36, 15.37, 15.38, 15.43, 15.44, 15.45, and 15.46. The solution
for each of the eight quantities may be written symbolically in terms of
the augmented determinant of the eight equations:
Equations of the Characteristics. In general, the characteristic curves

are defined as those curves on the integral surfaces comprising the
solution to the eight equations on which the eight derivatives au/ax,
&lay, av/ax,. . ., etc., may be either indeterminate or discontinuous.

The condition for this to be so is that each of the derivatives be of the
form 0/0. Indeed, by selecting any one quantity, such as au/ax, and
setting equal to zero the determinants in both numerator and denomi-
nator of the solution for this quantity, we get after simplification the
complete systems of characteristics:
u dy - v dx = 0 (15.48a)
(udy - vdx)(udv - vdu) +c 2

k
(udx + vdy) = 0
Eq. 15.48a is the equation of the streamline. Likewise, Eq. 15.48b
refers to a line of constant entropy, and hence also identifies a particular
streamline. Furthermore, using the relations that
udu+vdv = VdV
it is clear that Eq. 15.4% represents Euler's equation for changes in-
state along a streamline. Hence, Eqs. 15.48 state in sum that the
streamlines are characteristic curves in a rotational adiabatic flow.
Physically, this means that the streamlines are possible slip lines, or
vortes sheets in an infinitesimal sense, on which there may be disconti-
nuities in the derivatives of the velocity, density, entropy, and pressure.
If Eq. 15.49 is solved for dy/dx, the resulis will be recognized as being
identical with those of Eq. 15.20. Hence Eq. 15.49 gives the differential
equations of the physic31 characteristics, and can be reduced to the form
of Eq. 15.23, namely,
This shows that these physical characteristics make the Mach angle t o
the velocity vector and are, therefore, identical with the Mach lines of
the flow.
Turning to Eq. 15.50, it is convenient to express the velocity vector in
terms of polar coordinates V and 8, by means of the relations u = V Cot3 e
and v = V sin 8. Making this transformation, and eliminating dy/&
from Eq. 15.50 with the help of Eq. 15.51, we get the differential equa-
tion of the hodograph characteristics in the form
dB dp
3=
sin a! cos a
- +- + 1- ~ ( s / c , )= 0
p k
NOW,since all states have the same stagnation enthalpy, it follows

that
+
d ( V 2 / 2 ) c, dT = 0 ; or V dV +.
( 2 / k - 1)c dc = 0
Furthermore, from the relation for the entropy of a perfect gas,
T ds = c , dT + p d(l/p)
we have
ds/c, = 2dc/c - (k - 1 ) dp/p
Combining these relations, we obtain
dp - 1 ds VdV
- - - -. ---
P k-lc, c2
Substitution of Eq. 15.53 into Eq. 15.52 yields, after rearrangement,
7df.I----
1 dV sin a cos a!
d
(;-") =O (15.54)
tan a! V k(k - 1 )
This may be solved for dV/dO to give
By comparing Eq. 15.55 with Eq. 15.25 we see that the effects of
vorticity are contained in the third term of Eq. 15.55. This term depends
in part on the physical variables x and y, because the entropy is deter-
mined by the streamline rather than by the Mach Number. Conse-
quently it is not possible to integrate immediately the hodograph charac-
teristics, as was possible for irrotational flow. Instead, for each specific
problem, both the physical and hodograph characteristics must be
simultaneously constructed in a stepwise manner. Another consequence
of the presence of the third term in Eq. 15.55 is that i t eliminates the
possibility of there being regions of simple waves in a rotational flow.
Numerical Method of Solution. The solution of practical problems is
carried out by constructing the physical and hodograph characteristics
nets piecewise by simultaneously solving Eqs. 15.51 and 15.55 in finite-
Art. 15.12 FLOW W I T H ROTATION 521
difference form, replacing an element of arc on one of the characteristic
curves by a straight-line chord. Several related methods for carrying
out such calculations are described
in detail in Reference 3. A brief ,
outline of one line of attack is
given below in terms of the unit
process required to construct the
characteristics net.
Suppose that a t points 1 and 2
in Fig. 15.40 we know X I , yl, 22,
Y 2 , v 1 , v 2 , 81, 8 2 , S l , s 2 , thus corn-
pletely defining the location and
fluid properties of these two
points. Let point 3 lie at the in-
tersection of the physical ZI-char-
acteristic passing through point 2
and of the physical Z-characteris-
tic passing through 1. The prob-
2 -X
FIG.15.40.Determination of properties
lem is to determine the bcation a t point 3 from data a t point*^ 1 and 2.
and the fluid -ropert - ties of point 3.
Equations 15.51 may be written in finite-difference form
where gl3 represents the arithmetic mean value of 8 between points 1

+
and 3, i.e., gl3 = (el e 3 ) / 2 ;and similarly for the other quantities with
a bar.
The finite difference forms of Eq. 15.55 are
From the adiabatic relation between M and V,
and V are uniquely connected with each other. Moreover, the entropy
depends only on the particular streamline on which a given point lies.
Eqs. 15.56 to 15.59 may be solved simultaneously by an iteration
method. The values of V3 and 8 3 are first chosen tentatively. Using
these values, simultaneous solution of Eqs. 15.56 and 15.57 yields ap-
proximate values for x3 and ys By extrapolating the streamlines toward
point 3, the entropy of point 3 may be tentatively determined. Then
Eqs. 15.58 and 15.59 may be solved simultaneously for V3 and 83. If
these values depart too greatly from the values originally assumed, the
process of calculation is repeated again and again until satisfactory
convergence is obtained.
Successive application of this type of calculation allows the entire
characteristics net to be determined in a manner exactly like that for
irrotational flow, with similar considerations applying as to the way in
which various types of initial data determine the regions in which solu-
tions may be obtained.
,
Semigraphical Method of Solution. The tediousness of the iteration
method of calculation can be partly relieved by employing a semi-
graphical procedure which employs the irrotational hodograph charac-
teristics. Noting that
dV/V = dM*/M*; sin a = 1/M; tan a = I/-
and using the adiabatic relation between M and M*, we may put Eq. 15.55
in the form
(15.60)
Expressed in finite-difference form, this becomes
i
k-I- k-l-
I--- M * ~ I--- M *2
(AM*)I, 11 = TM*
Ic +1 A0 -
k+l
2k(k - I ) -
M *
- -
~ 1 M*
k+l
(15.61a)
where M* refers to the mean va!ue of M* over the interval to which
Eq. 15.61a is applied. This equation is also expressible as
where the differential coefficients in the latter equation are functions of

M* and k as indicated by Eq. 15.61a.
Returning now to the unit process of Fig. 15.40, we may from the
given initial data locate the image points 1 and 2 in the hodograph dia-
gram of Fig. 15.41. The mean Mach lines I and II are now tentatively
drawn in Fig. 15.40, thus establishing an approximate position for
point 3 and allowing the entropy a t 3 to be determined tentatively.
The second term on the right-hand side of Eq. 15.61b may then be
approximately calculated for each characteristic, and the correspond-
ing partial values of AM* laid off radially in Fig. 15.41 as the lines 1-la
FIG.15.41. Graphical construction in hodograph plane for flow with vorticity.
and 2-2a, inasmuch as this part of A M * represents the change in M * a t

constant 8. Thus the intermediate points l a and 2a are tentatively
established.
The first term on the right-hand side of Eq. 15.61a gives the change
of M * for a given change of 9 a t constant entropy, and corresponds to
irrotational flow. This term is represented graphically by the hodo-
graph characteristics for irrotational flow, i.e., by the epicycloid curves
- 15.5. The intersection of the irrotational 11-characteristic
of Fig.
passing through 2a with the irrotational I-characteristic passing through
l a therefore establishes the hodograph point 3, a t least approximately.
The mean Mach lines I and 11 may now be redrawn t o establish the
physical point 3 more accurately, and the entire process of calculation
repeated until the desired accuracy is obtained.
It may be shown 0) that if iteration is not used, i.e., if the initial
values rather than the mean values for each step are employed to
evaluate the coefficients in Eq. 15.61a or in Eqs. 15.56 to 15.59, the
error in the stepwise integration is of the order of (A@', where the latter
is a measure of the net spacing; whereas, if iteration is employed, the
error is only of the order of (~8)~.
METHOD OF CHARACTERISTICS Ch. 15

FERRI, A. Elements of Aerodynamics of Supersonic F1ow.n. New York: The
Macmillan Co., 1949.
LIEPMANN. H. W., and PWCKETT.A. E. Introduction to Aerodynamics of a Com-
pressibie ~ l u i d . ' New York: John Wiley & Sons, Inc., 1947.
ISENBERG, J. S., and LIN, C. C. The Method of Characteristics in Compressible
Flow. Part I: Steady Supersonic Flow, Air Materiel Command Report, No.
F-TR-1173A-ND (1947). Part IA: Tables and Charts, Air Materiel Com-
mand Report, No. F-TR-1173B-ND (1947).
CRONVICH, L. L. Numerical-Graphical Methods of Characteristics for Plane
Potential Shock-Free Flow Problems, Jour. I m t . Aero. Sci., Vol. 14, No. 4
(1947), p. 237.
SHAPIRO, A. H., and EDELMAN, G. M. Method of Characteristics for Two-
Dimensional Supersonic Flow-Graphical and Numerical Procedures, Jour.
A PP . Mech.. Vol. 14. No. 2 11947). D. A-154.
T E M ~ E , G. The ~ e t h o d of ~hara&istics in Supersonic Flow, British Rep.,
No. S.M.E. 3282, R.A.E. (Jan., 1944).
C ROWN, J. C. Supersonic Nozzle Design, N A C A Tech. Note, No. 1651 (1948).
PINKEL,I. I. Equations for the Design of Two-Dimensional Supersonic Nozzles,
N A C A R M , No. E8B02 (1948).
EDELMAN, G. M. The Design, Development, and Testing of Two-Dimensional
Sharp-Cornered Nozzles, S.M. Thesis, Dept. of Mech. Eng., M.I.T. (1948).
SHAMES, H., and SEASHORE, F. L. Design Data for Graphical Construction of
Two-Dimensional SharpEdge-Throat Supersonic Nozzles, N A C A R M ,
No. E8J12 (1948).
PROBLEMS
15.1. Consider the expansive flow around a curved wall (see sketch).
I
Compare the exact theory with the linear theory by plotting curves of p,/p,
versus MI for two values of 6, namely, 2O and 10". Use a range of M1from 1 to 3:
15.2. Compare supersonic flow past a convex corner with (a) incompressible
flow, (b) subsonic flow. If the fluid is compressible, is it possible to have a purely
subsonic flow in the neighborhood of the corner? How would a boundary layer
near the wall influence the supersonic flow past a convex corner?
15.3. Plot a Prandtl-Meyer streamline from M1 = 1 to Mt = 5.
15.4. I n simple-wave flows what are the maximum stream turning angles for
(a) Decelerating flows with initial Mach Numbers of (i) 1.0, (ii) 2.0, (iii)
5.0, and (iv) 10.0?
(b) Accelerating flows with initial Mach Numbers of (i)' 1.0, (ii) 2.0, (iii)
5.0, and (iv) 10.0?
Ch. 15 PROBLEMS 525
15.5. A two-dimensional jet leaves a convergingdiverging nozzle in parallel
flow (see sketch) and discharges into the atmosphere, where the pressure is
14.7 psia. The area ratio of the nozzle
is 2.0.
Calculate the angle 6 in degrees, if
(a) po = 200 psia
(b) po is infinite
15.6. Suppose that in Fig. 15.118
the shape of the wall is given by
Y = f(x1, Yl, 4
Find the equation of the streamline
passing through point l', in terms of
XI,, yl,, and the functional relation PROB. 15.5.
above.
Generalize your results to any arbitrary initial Mach Number, introducing
MI as a parameter.
15.7. Consider a corner-type flow beginning a t Mach Number unity. .Derive
the following equation in polar coordinates for a streamline:
where r is the radius vector to the streamline; r* is the radius vector to the same
streamline a t M = 1; and v is the angle measured clockwise from r* to r .
15.8. Using the theory of characteristics as outlined in Appendix A, deter-
mine the shapes of the characteristic curves in the physical and hodograph
planes for the differential equation corresponding to supersonic, uniform,
parallel flow with small perturbations.
15.9. Consider a perfect gas with k = 1.0
(a) Investigate the shape of the hodograph characteristics.
(b) For a complete corner-type flow determine in polar coordinates the equa-
tion of a streamline in terms of r , the radius vector to the streamline; r*, the
radius vector to the same streamline a t M = 1 ; and v,-the angle measured clock-
wise from r* to T .
15.10. Show that a flow region adjacent to a region of uniform, parallel flow
must be a region of simple waves.
15.11. (a) Demonstrate that when a pressure wave is reflected from a plane
wall,
(b) Show that when two pressure waves of opposite family penetrate each
other, the values of A0 and of Aw for each wave are the same after penetration
as they are before penetration.
(c) Derive a corresponding rule for reflection of a wave from a constan[-
pressure boundary.
15.12. (a) Suppose that, in Fig. 15.17, the line 2-3 is a solid boundary, and
that 0 is known on this boundary. Work out the computation formulas for
determining conditions a t 3 from those a t 1 and 2, given wl, 01,xi, yi, oz,02,
xz,y2, and given also 83 and ya as functions of x3. Note that iteration is required.
(b) Suppose that in Fig. 15.17, the line 2-3 is a constant-pressure boundary
of the flow, as for a free jet. Work out the computation formulas for determining
conditions a t 3 from those a t 1 and 2, given wl, 01,XI, yl, WZ,02,52,y2, and wa.
15.13. Demonstrate that if in a given region of the flow all states are mapped
on a single hodograph characteristic of family I , then in that region all the
left-running Mach lines are straight lines, and on each left-running Mach h e
all fluid properties are constant.
Are the right-running Mach lines straight lines for this case?
15.14. Consider a plane in which the Cartesian coordinates are the charac-
teristic coordinates, Z and ZZ (see sketch).
(a) Sketch in this plane lines of constant Mach Number and of constant flow
direction.
(b) Plot the fields of Fig. 15.128 in this plane.
(c) Plot the net points of Fig. 15.15a in this plane.
15.15. A thin profile has the shape of a parabola, with its leading edge parallel
to the oncoming air stream (see sketch).
Assuming that t / l = 0.10 and M l = 2.059, find the lift and pressure drag
coefficients.
<
15.16. A nozzle is to be designed to deliver a parallel, uniform stream of a,ir
a t a Mach Number of 2.059. The general arrangement of the nozzle is as shown
in the sketch.
- FI,
,;' it.'
The inlet section will be assumed to have a shape which produces a hori-
zontal parallel flow a t the throat section AB.
Ch. 15 PROBLEMS 527
In the region AFDB the flow is an expansive "corner flow" around the
corner C. The length AC is twice the length AB.
In the region FED the flow is uniform and parallel.
In the region FHIE the flow is an expansive "corner flow" around the corner
G. The length GE is twice the length FE.
Region J is a uniform, parallel flow with a Mach Number of 2.059. The
direction of the stream at J is the same as the direction at the throat AB.
Assuming that the throat height AB is one inch, find the lengths and inclina-
tions to the horizontal (in degrees) of the following lines: FD, FE, DE, and HI.
Using these principal dimensions and directions, make a sketch to scale of the
nozzle.
16.17. Work out the problem of Fig. 15.12 by the lattice-point method.
16.18. Consider a two-dimensional, source-type flow with an initial Mach
Number of 1.0 and a final Mach Number of 2.0. Using the method of char-
acteristics, plot pressure versus radius for stepwise solutions with turning angles
across each wave of 8°, 4°, and 2°, and compare with the exact solution, corre-
sponding to a turning angle across each wave of 0°.
16.19. Design a symmetrical blade for an impulse turbine with a turning
angle of 120° and an entering and leaving Mach Number of 2.059.
16.20. Design a stationary supersonic compressor cascade of the type shown
in Fig. 15.27, assuming that M1 = 2.059, P1 = 10, that the air entering is in-
clined 60° to the axial direction, that a normal shock occurs at M = 1.4, and that
the flow leaves in the axial direction.
Lay out a pair of blades, and determine the leaving Mach Number, pressure,
and the over-all ratio of stagnation pressures.
16.21. A uniform, parallel flow with M1 = 2.059 and P1 = 100 approaches
the two-dimensional 90°-bend shown in the sketch. Determine the flow pattern
leaving the bend.
~~~--- 2h - - - - - ;
1
PROB. 15.21. PROB. 15.22.
16.22. The two-dimensional passage shown in the sketch has parabolically

shaped walls. At section 1- 2- 3 the flow is uniform and parallel, and M = 1.33.
Using the coarse characteristic system suggested by the sketch, plot the pressure
distribution on the walls and along the center line for a distance of 4h down-
stream of the initial section.
15.23. A converging-diverging nozzle for a wind tunnel (see sketch) is
to be designed to deliver a supersonic stream a t a Mach Number of 1.915.
The nozzle will draw air from the atmosphere, where the pressure and tem-
perature are 14.7 psia and 70°F, respectively. A flow rate of 1 lb/sec is to be
accommodated.
The design is to be based on sharp corners a t the throat section aa', a t which
section it may be assumed that the flow is uniform and parallel. The contours
a'b'c'd' are to be designed so as to produce a uniform, parallel stream in region 4.
In carrying out the design, use the wave pattern suggested in the sketch, selecting
the Mach lines between regions 1 and 2 to correspond to equal deflection angles
across each of the Mach lines.
Assuming that the height of the nozzle a t the throat (distance aa') is 2 inches,
present the following design data:
(a) A sketch to scale of the nozzle contours.
(b) The height of the test section, dd', and the width of the tunnel a t the
throat and a t the test section.
(c) A plot of pressure versus distance along the axis of the tunnel, and
along the wall contour.
(d) The maximum half-angle of divergence of the expanding portion of the
nozzle.
(e) Draw the streamline which a t the throat is midway between the axis
and wall. Compare the length-height ratio of a nozzle using this
streamline as a wall contour with the length-height ratio of the sharp-
corner nozzle.
15.24. Determine the form of the hodograph characteristics for the tangent
gas of Chapter 11.
16.26. Show, for adiabatic flow, that a plot with polar coordinates M* and
a traces out an ellipse.
Demonstrate how an M*-or ellipse, drawn on a sheet of tracing paper to
the same scale as the M * 4 characteristic curves, may be used in conjunction
with the latter for graphically determining wave directions without drawing
normals to the hodograph characteristics.
Chapter 16
OBLIQUE SHOCKS

Normal shocks (Chapter 5) are only special forms of pressure discon-
tinuities within the fluid. More generally, the discontinuities observed
in practice are inclined to the direction of the oncoming flow. They are
called oblique shocks.
Oblique shocks tend t o occur in supersonic flow because, as shown in
Chapters 5 and 15, continuous compression waves tend to merge until
they form a discontinuity. Sometimes the oblique shock is attached to
the surface of a body, but this occurs only when the stream is forced to
change direction suddenly, as i:i a sharp concave corner. Otherwise,
the oblique shock is usually detached.
Oblique shock waves occur in almost all supersonic flow patterns of
practical significance, although it should be pointed out that the mere
existence of supersonic flow does not imply that shock waves must neces-
sarily appear somewhere in the flow field. The subject matter of this
chapter is, therefore, essential to the study of supersonic flow, and is
relevant to such varied applications as nozzles, diffusers, projectiles and
missiles, lifting surfaces, supersonic compressors, etc.
chapter may be found in Chapters 5, 14, 15, and in Volume 11,Chapters
19 and 28.
NOMENCLATURE
angle of attack shape integrals defined by

critical speed of sound Eqs. 16.24
specific heat at constant ratio of specific heats
pressure chord length of profile
C1, cz, c3 coefficients defined by distance from leading edge
Eqs. 16.21 to center of pressure
lift coefficient Mach Number
drag coefficient pressure
moment coefficient. reck- gas constant
oned about leading edge s entropy per unit mass
coefficient defined by Eq. t maximum thickness of
16.23 profile
529
OBLIQUE SHOCKS Ch. 16
NOMENCLATURE-Continued
absolute temperature P mass density

x-component of velocity u , angle of shock to incoming
y-component of velocity flow
resultant velocity T slope of profile surface
component of velocity with respect to chord
normal to shock line
component of velocity
parallel t o shock ( 11 signifies conditions up-
stream of oblique shock
Mach angle ( 12 signifies conditions down-
turning angle across stream of oblique shock
shock; thickness ratio ( 10 signifies stagnation state
of profile ( )Z signifies conditions up-
turning angle for which stream of normal shock
flow downstream of ( )V signifies conditions down-
shock is sonic stream of normal shock
maximum turning angle ( )m signifies free-stream con-
for oblique shock ditions
coordinate normal t o ( )L signifies lower surface of
chord of profile profile
flow direction ( 1" signifies upper surface of
coordinate along chord of profile
profile
Transformation of Normal Shock into Oblique Shock. A logical means

for introducing the oblique shock is b y way of the normal shock. Refer-
ring t o Fig. 16.la, suppose t h a t a stationary observer sees the gas in
state 1 suddenly decelerated a n d compressed t o state 2 by a normal
shock. Now imagine t h a t the observer moves along the shock front in
a downward direction with the speed V t . T h e moving observer then
perceives the flow pattern of Fig. 16.lb, in which the shock front is in-
clined t o the initial direction of flow, and in which the stream undergoes
a sudden change of direction a s i t crosses the shock. The oblique shock
flow pattern thus constructed is characterized by equal tangential com-
ponents of velocity on the two sides of the shock. By allowing one of
the streamlines of Fig. 16.lb t o represent a solid wall, we obtain the
solution for supersonic flow in the neighborhood of a concave corner
(Fig. 16.1~). Obviously, by the same device of imparting a uniform
velocity t o the entire flow field along the shock, a n y straight segment of
an oblique shock may be transformed back into a normal shock.
The relations between the fluid properties on the two sides of the
oblique shock may be obtained from the familiar normal shock relations
by employing the transformation equations between the flow picture of
Fig. 16.la and t h a t of Fig. 16.lb. I n this connection, note t h a t such
thermodynamic properties as temperature, pressure, density, sound

speed, and entropy are unchanged by the motion of the observer. On
the other hand, the fluid velocity, Mach Number, and stagnation tem-
perature and pressure are altered when the tangential velocity V t is
added to the normal-shock flow pattern. The magnitude of V t is arbi-
trary, and hence there is an additional degree of freedom in the oblique
shock relations. This means that although only one independent param-
eter (the approach Mach Number, say) is required for the normal shock,
two independent parameters will be required for oblique shocks.
FIG.16.1. Transformation of normal shock to oblique shock.

(a) Stationary observer sees normal shock.
(b) Observer moving along shock line sees oblique shock.
(c) Flow of (b) interpreted as pattern produced near concave corner in wall.
Rather than use the transformation approach outlined above, we shall

derive the oblique shock relations by returning to the principles of con-
tinuity, momentum, and energy. It will be assumed throughout this
chapter that we are dealing with a perfect gas.
Distinction Between Shocks and Mach Waves. It will be recalled
from Chapter 15 that when the velocity is everywhere continuous, re-
gions having different types of flow may be "patched" together only on
the characteristic curves. That is, along the characteristics there is no
discontinuity in the velocity or any other fluid property, but there may
be discontinuities in the derivatives of the velocity or of other fluid
properties. When we deal with shocks, however, we are no longer bound
by this rule, because we admit the possibility of discontinuities in all
fluid properties across the shock line. Hence the shock is also a patch-
ing line which joins regions having different types of flow. But it is
essential to keep in mind the distinction that the physical characteris-
tics, or Mach lines, are patching lines for continuous flows, whereas
shocks are patching lines for discontinuous flows.
532 OBLIQUE SHOCKS Ch. 16
16.2. Oblique Shock Equations

Governing Physical Relations. The oblique shock will be analysed
by writing the continuity, momentum, and energy relations for the flow
through the control surface of Fig. 16.2. Noting that the two stream-
Streamline
control/
Svrfgre
FIG.16.2. Nomenclature for oblique-shock analysis.
lines forming part of the control surface are exact duplicates of each
other, we may write
Continuity.
P I J ' ~ ~= PZV~Z
Momentum in t-direction.
(PIVnl)Vtl= ( P Z V ~ ~ ) V-'.~Vt2
Z ;= Vtl = Vt (16.2)
Momentum in n-direction.
2
PI - ~2 = P Z V ~-
Z~ 1 V n l ~
Energy.
cP(T1- T2) = (VZ2- V I 2 ) / 2
The absence of change in the tangential component of velocity
(Eq. 16.2) bears out the change-in-coordinate point of view of Art. 16.1.
From Eq. 16.2 and the geometry of the velocity triangles,
Furthermore, the perfect gas relations yield
hence the energy equation may be expressed as
Eqs. 16.1, 16.2, 16.3, and 16.5, together with the perfect gas relations,
define completely the relations between the states on the two sides of
Art. 16.2 OBLIQUE SHOCK EQUATIONS 533
the shock. For example, suppose that state 1 is completely specified hg-
the values of pl, pl, Vnl, and Vtl. Then the four equations mentioned
are suffcient to determine p2, p2, Vn2, and Vtz. Since Vnl and Vtl to-
gether specify the shock angle u, and since Vn2and Vt2together specify
the angle (a - a), the solutions also yield the shock angle u and the
turning angle 6. Furthermore, the values of p and p define the sound
speed c, so that the Mach Numbers fore and aft of the shock, MI and M2
respectively, may also be found.
The basic shock relations of Eqs. 16.1: 16.2, 16.3, and 16.5 may be
combined in many ways to yield a great many algebraic relations relat-
ing various shock parameters.'', 3 , Practically, we wish to obtain
2n
those relations which lead to simple numerical calculations, or which by

their forms lead to general conclusions concerning the nature of oblique
shocks. We shall in the remainder of this article derive or set down
some of the more useful relations of these two types.
The Rankine-Hugoniot Equations. Let us eliminate all velocity terms
from the basic equations and arrive a t a relation in terms of pressure and
densities.
Rearranging Eq. 16.3 and employing Eq. 16.1, we have
P2 v n 2
= plvn12(l- );
from which
Similarly, we may solve for Vnz from Eqs. 16.3 and 16.1,
Multiplying these two relations, we get
Substituting Eqs. 16.6a and 16.6b into Eq. 16.5, and rearranging, we
obtain the prcssurc ratio as a unique function of the density ratio, or the
density ratio as a unique function of pressure ratio:
These are the Rankine-Hugoniot relations and are represented graphi-

cally in Fig. 16.3. They show that the density
100
ratio across the shock depends on the initial
60
Mach Number and on the shock angle only
60
insofar as these quantities influence the pressure
40 ratio. A shock with a given pressure ratio has
associated with it a density ratio which does
not depend on the obliquity of the shock, and
20
hence the Rankine-Hugoniot relations are valid
-
p2 for normal shocks as well as for oblique shocks
Dl
10 of any shock angle. Indeed, this might have

B been set down from the beginning by arguing
6
that, in a coordinate transformation which
changes a normal shock to an oblique shock,
4 the relations connecting the pressure and den-
sity changes are unaltered.
I t follows further from the perfect gas laws
that across shocks of any obliquity there is a
unique relation between temperature ratio and
pressure ratio, and a unique relation between
I 2 3 4 5 6
temperature ratio and density ratio.
P, / p ,
16.3. ~ ~ ~ k iThe ~ Prandtl
~ - ~Relation.
~ - An interesting and
g?niof, relation compared revealing relation between the velocity com-
wlth lsentropic law, for k
= 1.4. ponents may be found by eliminating the pres-
sures and densities from the basic equations.
Since the flow is adiabatic, we may rewrite Eq. 16.4 in terms of the
stagnation temperature, To:
cpTl + (Vn12+ Vt12)/2 = cPT2+ (Vn2' + Vr22)/2 cpTo
But,, since cpT = (k/k - l ) ( p / p ) ,this becomes
k PI
--
k- PI
i-
VnI2+ Vt12 -
2
k P2
k - 1 ~ 2
+ Vnz2+2 Vtz2 - k-5 POP O
However, from the adiabatic relations of Chapter 4,

k Po - co2 - + l C*2
k-1po k - 1 2(k-1)
We may now solve for pl and pz from the previous equation in the form
Substituting these expressions for p, and p2 into Eq. 16.3, we get
Introducing Eqs. 16.1 and 16.2, and reducing algebraically, we finally get
which is known as the Prandtl relation.

For a normal shock, with V t = 0, this takes the special form
v1v2= c * ~ .
Another useful form is found through combination of Eqs. 16.9 and
16.7,
Working Formulas for Oblique Shocks. We shall now derive a con-

venient set of formulas for numerical calculations.
From the geometry of Fig. 16.2,
V t l = V 1cos a ; Vtz = Vz cos (u - 6) (16.11)
Vnl = V 1sin u; V n 2= V 2sin (u - 6) (16.12)
Since V t l = Vt2,i t follows from Eq. 16.11 that
v1/v2
= cos (a - q/cos u
With Eq. 16.12, the continuity equation becomes
P I V lsin a = pzV2 sin (a - 6)
Therefore,
p2 V 1 sin a
-- cos (a - 6)
- -
sin u - tan u
- (16.13)
pl V 2sin (u - 6) cos u sin (a - 6) tan (a - 6)
We now rearrange Eq. 16.3 with the help of Eqs. 16.1 and 16.12, and get
However, since plV12= k P 1 ~ 1 2we

, may solve for the pressure ratio in
the form
If we had factored out p2V,22 in Ea. 16.3, we would have arrived in-
stead a t
Eqs. 16.8, 16.13, 16.14, and 16.15 constitute four relations among the
variables u, 6, MI, M2, p2/pl, and p21P1. Any two of these variables,
therefore, may be taker, rts independent parameters, the values of which
completely determine the values of the four remaining variables. The
temperature ratio, entropy change, and ratio of isentropic stagnation
pressures may of course be calculated readily after the variables named
above are specified.
M,, I n i t i a l Mach Number

FIG.16.4. Shock angle versus inlet Mach Number and turning angle (k = 1.4).
Tables and Charts. I n Reference 8, tables representing the shock

relations are given with a and 6 as independent arguments. In Refer-
ence 5, one set of tables has M1 and a as independent arguments, and
mother set has MI and 6 as independent arguments. Reference 4 has
zonvenient large-scale working charts with M1 and 6 as independent
variables. Additional convenient tablcs and charts may be found in
References 6, 7, 1.5, 16, and 17.
Figs. 164, 16.5, and 16.6 show the general orders of magnitude repre-
sented by the various shock formulas. In these charts, MI and 6 are
taken as independent, and u, M 2 , p 2 / p 1 , and po2/pol are the dependent

quantities. The corresponding values of p2/pI may be read approxi-
mately from Fig. 16.3, or may be found from Table B.3 a t the appro-
priate value of p 2 / p l .
-
Deflection
M,, I n i t i a l M a c h Number
FIG.16.5. Exit Mach Number versus inlet Mach Number, with turning angle as
parameter (k = 1.4).
ADAPTATION OF NORMAL SHOCK TABLES TO OBLIQUE SHOCKS. Table

/ B.3, representing the normal-shock relations, may be adapted to oblique-
shock calculations. An oblique shock with an approach Mach Number
M 1 may, by a transformation of the coordinate system, be reduced to a
normal shock with an approach Mach Number M 1 sin u. Consequently,
if M, of Table B.3 is taken to represent MI sin u, then p,/p,, p,/p,, T,/T,,
M , , Initial Mach Numbel
(a)
M,, Initial M a c h Number

(b)
- -
FIG.16.6. (a) Pressure ratio versus inlet Mach Number. with turnine anele as
(b) Stagnation-pressure ratio versus inlet Mach Number, with turning angle as
538
Art. 16.3 SHOCK GEOMETRY 539
and po,/poZ of the table are respectively the values of p2/pI, P2/Pl, T2/T1,
and p~,/pop,,for an oblique shock with inlet Mach Number M1 and shock
angle u. From the geometry of Fig. 16.2 it follows further that M2 =
M,/sin (a - 6).
16.3. Shock Geometry
Application of Second Law of Thermodynamics. The shock formulas
as previously developed do not distinguish between a compression dis-
continuity (pressure increase) and a rarefaction discontinuity (pressure
decrease). However, since the second law of thermodynamics allows
only entropy increases for adiabatic processes, it follows that s2 2 sl.
Remembering that an oblique shock may be reduced to a normal shock
by means of a coordinate transformation not involving entropy changes,
and that in a normal discontinuity the entropy decreases for a rarefac-
tion shock, we conclude that the only type of oblique discontinuity
which is allowable is one in which there is a pressure rise.
BENDING OF STREAMLINE. Since p2 > pl, it necessarily follows from
the continuity relation (Eq. 16.1) and the momentum theorem (Eq.
16.3) that Vnl > Vn2. Then, since Eq. 16.2 requires that Vtl = Vt2,
the velocity triangles of Fig. 16.2 show that 6 must always be positive
in magnitude. In other words, we may state the general rule that in
crossing a n oblique shock the streamline always bends toward the shock line.
DECAY OF RAREFACTION SHOCK AT CONVEX CORNER. The impos-
sibility of a rarefaction shock is of course also connected with the fact
Dl/ / Rarefaction /
/ / /
Streamline ,'/ / /n2
/
/
Possible
FIG.16.7. Impossibility of oblique rarefaction shock.

(a) Rarefaction discontinuity a t corner rapidly attenuates to a continuous
Prandtl-Meyer rarefaction wave.
(b) Obliquc rarefaction discontinuity would violate Second Law of Thermo-
dynamics.
that rarefaction waves always tend to become less steep. Thus, in the
expansive flow around a corner, Fig. 16.7, there is a rarefaction discon-
tinuity confined to the singular point of the sharp corner, but this dis-
continuity rapidly decays to a continuous Prandtl-Meyer expansion
which fills the entire expansion region.
Impossibility of Prandtl-Meyer Compression in Concave Comer. If

we were to examine the possibility of a Prandtl-Meyer compression in a
sharp corner (Fig. 16.8), following the analysis of Chapter 15, we would
/
/
Impossible
FIG. 16.8. Flow near concave corner.

(a) Prandtl-Meyer flow is not physically possible.
(b) Oblique shock is physically possible.
find that each streamline reverses itself twice a t cusps. This means that
a t each point in the compression region there would exist simultaneously
three different pressures and velocities! Mathematically this is con-
ceivable with the aid of multi-sheeted surfaces branched a t the Mach
lines, but it is physically impossible. The oblique compression shock
provides a means for carrying out a sudden change in direction without
involving this difficulty. The reversal of the hypothetical Prandtl-
Meyer compressive flow occurs a t cusps whose loci are known as limit
lines. This example, and the example of flow reversal in a source-type
flow a t a limit line where M = 1,
are perhaps the simplest illustra-
tions of the limit line phenomena
/ which are discussed more fully in
Volume 11, Chapter 20.
Relation of Shock Line to Mach
L i e s . In Fig. 16.8 the shock line
is shown as lying between the
Mach lines of the upstream and
downstream flows. We shall now
demonstrate that this is always
FIG. 16.9. Illustrates that left-running SO.
shock "overtakes" upstream left-running Referring to Fig. 16.9, let 11,
Mach wave, and is overtaken by down-
stream left-runniner- Mach wave. and 112 represent the respective
directions of the left-running Mach
lines in regions 1 and 2, making the angles a, and a2 to the upstream
and downstream flows, respectively.
Art. 16.3 SHOCK GEOMETRY 541
SHOCK OVERTAKES STREAM M ACH LINES. From Eq. 16.12 and

the definition of Mach angle,
sin a = Vnl/V1; sin a1 = cJVl
But, since the flow in region 1 must be supersonic to an observer who
sees the discontinuity as a normal shock, and since the sound velocity in
region 1 does not depend on the motion of this observer, it follows that
Vnl 2 el, whence
a 2 ffl
which demonstrates that the shock line i s more steeply inclined to the $ow
than a Mach wave in region 1.
SHOCK IS OVERTAKEN BY DOWNSTREAM M ACH LINES. We note from
Eq. 16.12 and the definition of cuz that
But the flow in region 2 must be subsonic to an observer viewing the

discontinuity as a normal shock. Hence VC25 cz, from which it fol-
lows that (a - 6) <a ~or,
u I f f z f 6 (16.17)
This, together with Fig. 16.9, shows clearly that a Mach wave in region 2
i s more steeply inclined to the $ow than the shock line.
RULES OF INTERACTION. From the two rules expressed by Eqs. 16.16
and 16.17 we may reach the following conclusions as to the possibility
for interaction of waves of the same family:
(a) A shock wave always tends to interact with a continuous wave
which is further downstream (Fig. 16.10a).
(b) A shock wave always tends to interact with a continuous wave
which is further upstream (Fig. 16.lOb).
(c) Two shock waves of the same family always tend to meet
(Fig. 16.10~).
. FIG. 16.10. Interaction of waves of same family.

(a) Shock always interacts with downstream Mach waves.
(b) Shock always interacts with upstream Mach waves.
( c ) Two shocks always tend to intcrsect.
OBLIQUE SHOCKS Ch. 16
16.4. Shock Polars

Two types of graphical representation of the oblique shock relations
which have been devised are especially useful because they include some
of the geometric features of the shock.
Hodograph Shock Polar. Let the x- and y-components of velocity be
denoted by u and v, respectively, and let the flow approaching the shock
be in the x-direction, so that ul = Til, and vl = 0. Then the states
before and after the shock may be located in the hodograph diagram ol
Fig. 16.11a. For a shock of given turning angle 6, all the information
/
/-Shock Line
(a) (b)
FIG. 16.11. Hodograph shock polar.
(a) Configuration of velocity vectors.
(b) Construction for determining shock angle and final velocity.
required to lay off the vector V 2 as well as the shock angle (T may be
found from the preceding formulas, and thus point 2 may be located in
the hodograph plane. The upstream and downstream velocity com-
ponents along and parallel to the shock line must be related t o each
other as shown in Fig. 16.11a, with VL1= V t 2= V t . It is to be noted
that Fig. 16.11a may be found by merely superposing the velocity tri-
angles of Fig. 16.2.
Now, for each value of Vl/c* there are an infinite number of end states
(corresponding to different values of 6) in the hodograph plane which
can be reached from state 1. The locus of these points is known as the
hodograph shock polar for MI*. A typical shock polar is shown in Fig.
16.11b. It is symmetrical about the horizontal axis, since there are two
possible shock waves for given values of M 1 and 6, one running leftward
relative t o the flow, and the other rightward. Having once constructed
the shock polar for a certain value of V l / c * , the solution for a given
turning angle 6 is found by laying off from the origin a line a t the angle 6.
Where this line intersects the shock polar is the end point of the velocity
vector Vz/c*. The normal OA to the extension of the straight line ?-2
then gives the direction of the shock line.
Art. 16.4 SHOCK POLARS 543
The complete family of hodograph shock polars, with VJc* as param-

eter, is shown in Fig. 16.12.
The approximate shape of the shock polar for a specified value of
Vl/c* may easily be sketched by using Eq. 16.9 in conjunction with the
construction of Fig. 16.lla, especially when it is realized that the second
term on the right hand side of Eq. 16.9 is usually small compared to the
first term.
Fro. 16.12. Family of hodograph shock polars (k = 1.4).
EQUATION OF SHOCK POLAR. TO obtain the equation of the shock

polar in u-, v-coordinates, we find from the geometry of Fig. 16.11a that
Vnl = ul sin a; Vt = ul cos u
Substituting these into the Prandtl relation, Eq. 16.9, we get
ul 2 sln + (ul - ~ 2 ) '

- 2 a - ul sin u d v Z 2 =
k-1
c*' - -u1 cos2 a
,
k+l
Then, eliminating u with the additional relation obtained from Fig. 16.11,
namely,
u1 - u2
tan u =
v2
we finally obtain after considerable algebraic rearrangement,
This formula allows us to plot v2/c* versus u2/c* for any specified value
of the parameter ul/c*.
The Pressure-Turning Angle Shock Polar. Later we shall see that

surfaces of discontinuity are usually fortned when two shocks interact
with each other, or when a ~hocli interact,^ with a continuous wave.
These surfaces of discontinuity,
I
usually called slip lines, or vortex
sheets, are also streamlines, and
divide two regions having equal
pressures and flow directions but
different velocities. For problems
of this nature it is convenient to
represent the oblique shock rela-
tions by a shock polar for which
the coordinates are turning angle
and pressure.ratio.
For any given initial Mach
Number, the formulas of Art. 16.2
allow us to find the pressure ratio,
pz/pl, as a function of turning
angle, 6. Plotting these as in Fig.
16.13, we get the family of 6-p
shock polars with M Ias parameter.
-
"0
S Convenient large-scale representa-
FIG. 16.13. Pressure-deflection shock
tions of these shock polars are
~olars. given in Reference 15.
16.5. Some Special Aspects of Oblique Shocks

Strong Versus Weak Shocks. A curious fact revealed by the graphical
representations of Figs. 16.4, 16.5, 16.6, 16.13, and 16.14 is that for a
given initial Mach Number MI and a given turning angle 6 there are
either two solutions to the oblique shock relations, or none a t all. Let
us consider specifically a uniform flow of Mach Number MI approaching
a wedge of half-angle 6 or approaching a concave corner of turning angle
6 (Fig. 16.14). From the shock polars, we see that if a solution exists
there may be either (i) a strong shock (2s), with a relatively large shock
angle, a relatively large pressure ratio, and with the downstream state
usually subsonic, or (ii) a weak shock (2w), with a relatively small shoch
angle, a relatively small pressure ratio, and with the downstream state
usually supersonic. Which of these two solutions will occur in practice?
Unfortunately, there is no simple and clear-cut answer to this question,
but certain general considerations serve as a guide.
ONE-DIMENSIONAL CASE. AS a start toward understanding this prob-
lem, let us consider the special case of one-dimensional, supersonic flow
Art 16.5 SPECIAL ASPECTS OF OBLIQUE SHOCKS 545
-
pE,; 2w
for M , -
o 6
FIG. 16.14. Illustrates strong and weak shocks.
in a duct of constant area. For this case the turning angle 6 is zero, and
the strong solution on the shock polar of Fig. 16.15 is a normal shock,
while the weak solution is an infinitesimal wave of zero pressure rise.
From our previous discussions of one-dimensional flow in ducts, i t is
clear that the solution which occurs in practice depends on the boun-
dary conditions, in particular on the ratio of downstream pressure to
upstream pressure. If the downstream pressure is sufficiently low, thc
weak solution occurs, which is to say that there are no shocks a t all.
I
(Normal (Plane
Shock) Sound Wave)
FIG.16.15. Strong and weak shocks for 8 = 0.
If the downstream pressure is sufficiently high, a normal shock stands

in the duct, which is to say that the strong solution occurs.
TWO-DIMENSIONAL CASE. Returning now to the general case of
oblique shocks, we may generalize the argument of the preceding para-
graph by stating that whether the oblique shock is strong or weak de-
pends, a t least in part, on the boundary conditions as to pressure. If
the downstream pressure for the two-dimensional flow, Fig. 16.16a, is
sufficiently small, a weak shock will occur a t corner A and the flow
approaching corner B will usually be supersonic. If the back pressure
is sufficiently high, on the other hand, the shock emanating from corner
0 @ Subsonic H
Fra. 16.16. Effect of back pressure on shock form.

(a) Weak shock (low back pressure).
(b) Strong curved shock (high back pressure).
(c) Hodograph shock polar.
(d) Pressure4eflection shock polar.
A will be of the strong type, and the flew approaching corner B will
usually be subsonic.
As will be seen later, however, an oblique strong shock cannot extend
to a wall, but rather curves as shown in Fig. 16.16b until it is a normal
shock at the upper wall. Each streamline therefore passes through a
shock of different strength, and each point on the shock line is repre-
sented by a corresponding point on the segments between 2s and 4 of
the shock polars of Figs. 16.16~and 16.16d. The flow behind the shock
is subsonic, the streamlines are curved, and there is vorticity or rotation
in the flow downstream of the shock because of entropy variations from
streamline to streamline.
When a symmetrical wedge of half-angle 6 is moved through the
atmosphere at supersonic speed, only the weak shock is observed in
practice. This may be rationalized in the following way. Suppose that
Art. 16.5 SPECIAL ASPECTS OF OBLIQUE SHOCKS 547
the upper wall of Fig. 16.16a is moved infinitely far upward. Then in
the limit, we obtain the upper half of the flow pattern for the wedge
moving a t supersonic speed. In the atmosphere, however, the pressures
far downstream of the wedge can differ from those far upstream by only
infinitesimal amounts, and hence the boundary conditions are such that
the weak shock occurs. Thus, with supersonic aircraft, the obliquc
shocks seem to be invariably of the weak type, except under conditions
described below where the geometry of the aircraft does not allow of
attached shocks.
The schlieren photographs of Fig. 16.17 support the view that the
boundary conditions determine whether the weak or strong shock a p
pears in practice. All three photographs show the flow near the entrance
FIG. 16.17. Schliereu pl~otograpllsshowing influcncc of back pressure on type of

shock. All three photographs are for the internal flow geometry (two-dimensional)
of the right-hand photograph (after Busemann, NACA).
of a two-dimensional, supersonic inlet diffuser which has a symmetrical

two-dimensional wedge as a core. In the left-hand photograph the back
pressure for the internal flow is comparatively high, and strong shocks
are attached to the nose of the wedge. In the right-hand photograph,
corresponding to a comparatively low back pressure, the customary
weak shocks are attached to the nose of the wedge, but suddenly change
into strong shocks a t some distance from the nose. In the center photo-
graph, corresponding to an intermediate back pressure, the flow is un-
symmetrical, with a weak shock on one side and a strong shock on the
other side of the wedge.
Another factor which may influence the occurrence of strong or weak
shocks is that concerning stability. It often seems possible in fluid dy-
namics to have two or more solutions which satisfy given boundary
conditions. I n such cases the solution which occurs in practice may de-
pend on the direction from which the existing boundary conditions are
approached and on disturbances which may be present in the steady-
state flow. This situation is analogous to the hysteresis cycles which
supersonic diffusers may exhibit under conditions of variable Msch
Number, variable area ratio, or variable pressure ratio. This aspect of
the weak-versus-strong shock question has not as yet been explored

sufficiently.
Detached Shocks. The various graphical repre~entat~ions of Figs. 16.4,
16.5, 16.6, 16.12, and 16.13 show 'that for a specified initial Mach num-
ber, MI, there is a maximum value of the turning angle, 6, for which
there exists an oblique shock solution. Or, conversely, for a specified
turning angle, 6, there is a minimum initial Mach Number, Ml, ,in, for
which there is an oblique shock solution. From Figs. 16.18a and 16.18b
,5 10 20 30 40 50
8 ( o r 81, Degrees
Fro. 16.18. Maximum turning angle For a given initial Mach Number.
(a) Pressure-deflection shock polar.
(b) Hodograph shock polar.
(c) Chart for k = 1.4.
it is evident that a t this limiting condition the weak and strong oblique
shocks become identical. The relation between MI and 6,, or between
MI, ,in and 6, is shown in Fig. 16.18~. No solutions may be obtained
in the shaded area below the curve in the graph. (See Problem 16.11a.)
Consider the case of uniform, parallel flow a t Mach Number M1 past
a wedge of half-angle 6. If 6 is less than 6,, the shock is attached to
the wedge (Figs. 16.19a and 14.8). However, if 6 is greater than 6,,
the shock cannot be attached to the wedge, for this would require the
streamline approaching the point of the wedge to turn through an angle
greater than 6,. Under these circumstances, we observe in practice
that the shock is detached from the wedge (Figs. 16.19b and d). Simi-
larly, a detached shock always stands in front of a blunt body (Fig.
16.19~).
Art. 16.5 SPECIAL ASPECTS OF OBLIQUE SHOCKS 549
When the shock is detached, it is invariably curved. If the body of

Fig. 16.19~represents the leading edge of the wing of a supersonic air-
craft, we can see intuitively that the curved shock which begins a t A
Fra. 16.19. Attached and detached

shock.
(a) Attached shock on wedge of
small angle.
.(b)
~, Detached shock on wedge of
large angle.
-
(c) Detached shock in front of blunt
body.
(d) Interferogram of detached shock
in front of wedge of large angle
(NACA).
as a normal shock must gradually weaken until a t B, which is a t a great

distance from the body, it is of vanishing strength, i.e., it is a Mach wave.
If this were not so, the air passing the wing would undergo an infinite
change of entropy and this would result in infinite drag. With a curved
detached shock as in Fig. 16.19c, therefore, there is a segment of the
shock to corr~spondto every point on the shock polar curve pertaining
to the given MI. Behind the detached shock the flow'is in part super-
sonic and in part subsonic, leading to great difficulties in analysis be-
cause of the radically different properties of the differential equations
for subsonic and supersonic flow.
Similar considerations apply to the flow past a concave corner in a
duct (Fig. 16.20).
F IG. 16.20. Attached and detached shocks at concave corner in duct.
If the abscissa of Fig. 16.18~is regarded as the actual half-angle of

the wedge or the actual turning angle of the concave corner, the region
above the curve represents attached shocks, and the region below repre-
sents detached shocks.
Sonic Flow After Shock. As shown in Fig. 16.21, for each value of
MI there is a certain turning angle, 6*, for which the flow downstream
of the shock is exactly sonic. The relation between 6* and a,, is as
illustrated in Fig. 16.21, that is, 6* is always less than .,,6 However,
for k = 1.4, 6* does not differ from 6,, by as much as 0.5" over the
, ,a
FIG. 16.21. Relation between and A*.
entire range of Mach Numbers from unity to infinity. This means that,
except in a very narrow range, the flow behind a weak shock i s supersonic
and the flow behind a strong shock i s subsonic. (See Problem 16.11b.)
Nearly Normal Shocks. From the geometry of the hodograph shock
polar, Fig. 16.11b, it may be seen that a strong shock of small turning
angle will have a shock angle a of nearly 90'. From the 6--p shock polar
of Fig. 16.13, it may be seen furthermore that the pressure rise across
such a shock is substantially the same as that across a normal shock.
Art. 16.6 VERY WEAK SHOCKS 55 1
Thus we have the approximate rule that strong shocks of small turning
angle are nearly equivalent to normal shocks.
16.6. Very Weak S h o c k s

In practical design problems, an approximate treatment of weak
shocks may often be used with advantage to simplify the calculations.
Shock Strength. We shall specify the shock strength in terms of the
pressure ratio, pz/pl, across the shock. The Rankine-Hugoniot relation,
Eq. 16.8, then shows that the density ratio, pz/pl, may also be used as a
measure of shock strength. Inspection of Eq. 16.14 further reveals that
M1 sin u is a direct measure of shock strength. Since the entropy change
depends only on the pressure ratio and density ratio, the entropy change
is also a unique measure of shock strength. Finally, the stagnation
-
pressure ratio, poz/pol is a direct measure of shock strength because it

is determined uniquely by the entropy change in an adiabatic process
through the relation
s 2 - 81 = -R in ( ~ 0 2 / ~ 0 1 )
Mach Wave as the Limit of a Shock Wave. By means of algebraic

rearrangement, Eq. 16.8 may be written in the form
From this we see that as Ap goes to zero, Ap also goes t o zero. The
limiting relation between Ap and Ap, however, as Ap/pl goes to zero, is
evidently
AP/AP = k ~ l / =
~ lc12
Thus we see that for weak shocks, Ap/Ap = c12. But, since c2 =
(ap/ap),, it follows that the changes in pressure apd density in a very
weak shock are connected in the limit through the isentropic relation
(see Fig. 16.3).
Combining this result with Eq. 16.7, an'd noting that, for very weak
shocks, Eqs. 16.1 and 16.3 predict that 17,~approaches Vnl, we get
In other words, the component of velocity normal to the wave becomes

equal to the sound velocity. Therefore,
tan u --, ( c / V ) = tan a
so that in the limit the shock angle u becomes identical with the Mach
angle a. We have thus shown that in all respects an oblique shock wave
Ch. 16
552 OBLIQUE SHOCKS
of vanishing strength is identical with a Mach wave. For a given initial

Mach Number MI, the strength of the shock gocs to zero as the turning
angle 6 goes to zero. If, in the geometric construction of Fig. 16.11b we
allow point 2 to approach point 1, we obtain in the limit the construction
of Fig. 16.22, where the direction of the Mach wave is given by the nor-
mal to the shock polar a t point 1.
Tangent to
*
Shock Polar a t I
.
Shock
Polor
\
\
FIG. 16.22. Normal to hodograph FIG. 16.23. Hodograph shock polar and
shock polar at 1 has same direction as hodograph characteristic p'assing through
Mach line. point 1 have equal slopes and radii of
curvature.
Entropy Change Across Weak Shock. I n Chapter 5 we found that

the entropy change across a shock could be expressed in terms of shock
strength through the series expansion
thus demonstrating that the entropy change is of third order in shock

strength and is extremely small for shocks of small pressure rise. In
the light of our previous discussion, the entropy change is also of third
order in such quantities as MI sin a or (p02- pol)/$ol (Problem 16.12).
Since the isentropic relation and the Zankine-Ilugoniot relation are
identical up to and including terms of second order in shock strength,
the curves of Fig. 16.3 are not only tangent as p2/p1 1, butlhave the
same radius of curvature as well.
Although MI sin o is a unique measure of shock strength, in pract.ica1
problems we are more likely to be interested in M1 and 6 as primary
variables. There is no simple relation giving shock strength in terms of
M, and 6. However, the curves of Fig. 16.6 show this relation gmphi-
cally, and Fig. 16.Gb demonstrates clearly that the entropy changcs
across shocks of small turning angle are extremely small.
Approximate Treatment of Flows Containing Weak Shocks. Re-
membering that the normal to the hodograph characteristic gives the
direction of the Mach line (Fig. 15.13), it is evident from Fig. 16.22 that
for a given initial state the shock polar and characteristic curves in the
.4rt. 16.7 REFLECTION AND INTERACTION OF SHOCKS 553
hodograph plane are tangent to each other. Furthermore, since the

shock relations and isentropic relations differ only in terms of third order
of shock strength, it may be shown that the shock polar and character-
istic curves have the same radius of curvature a t point 1. Thus, as
illustrated by Fig. 16.23, the shock polar and characteristic curves are
very nearly identical for small angles of turn.
The preceding considerations suggest a greatly simplified procedure
when dealing with two-dimensional flows involving weak shocks.
Rather than use the exact oblique shock relations, we may with little
loss of accuracy ignore the entropy change across the shock and treat
the latter as a reversible compression wave. This is equivalent to using
the hodograph characteristic curve rather than the h~dographshock
polar. Assuming that there are no strong shocks present, the entire
two-dimensional flow may then be treated by the characteristics methods
of Chapter 15, without the need for changing the calculation procedure
as weak shocks appear. This is an especially important simplification
when it is recalled that shocks usually produce rotation in the flow, be-
cause if this rotation were taken into account the simple methods of
Arts. 15.3 and 15.4 would have to be replaced by the complex methods
of Art. 15.12.
In using the characteristics theory with the field method, continuous
changes in stream properties mere lumped into fictitious discontinuities
which were assumed to be concentrated on certain discrete Mach lines.
This is essentially the approximate procedure outlined above for weak
shocks. However, there is a basic difference which should be kept clearly
in mind. The analysis of continuous changes by the field method may
be made as exact as we please merely by increasing the number of fields
(or increasing the fineness of the Mach net), whereas no such improve-
ment in accuracy is possible in the approximate treatment of shocks
because the salient feature of shocks is their discontinuous nature.
Finally, it is interesting to note that the method of characteristics,
the analysis for oblique shocks, and the linearized supersonic theory all
give identical results for the limiting case where a supersonic stream is
turned through a vanishingly small angle.
16.7. Reflection and Interaction of Shocks

In practical problems, oblique shock waves are often incident on solid
boundaries or on free boundaries of jets; sometimes waves of opposite
family cross each other; and, as demonstrated in Art. 16.3, two oblique
shocks of the same family always tend to intersect. We shall now dis-
cuss the analytical treatment of such processes. The reader will find it
instructive to compare the exact methods described here with the ap-
proximate methods of the linearized theory (Art. 14.7).
Reflection of Shock from Free Boundary. Shock A (Fig. 16.2413) is

incident on the constant-pressure boundary of a two-dimensional
Expansion ( 8 )
FIG. 16.24. Reflection of shock from constant-pressure boundary (see also Fig.
14.21).
(a) Hodograph diagram.
(b) Streamlines and waves.
stream. Since the pressure in region 2 downstream of the shock is

greater than that of the free boundary, it is evident that an expansion
wave B must be reflected where wave A is.incident on the free boundary.
Shock P o l a r
.<
I
3 "/c*
j \shock polar
For S t a t e 2
S h o c k Polar Furthermore, rarefaction wave B must be a

Shock centered Prandtl-Meyer wave so that, a t the
free boundary, the sudden rise in pressure from
pl to pz may be instant!^ canceled by the
State
2
sudden drop in pressure from p2 to pa The
strength of the simple wave B is related to the
strength of shock A through the requirement
that P2 - PI = P2 - Ps
Regular Reflection from Solid Boundary.
FIG.16.25. Reflection of In Fig. 16.2513 shock A is incident on a
shock from solid boundary.
straight wall. Since the gas flow is assumed
(a) Hodograph dia- to fo1l.o~the solid boundary, it is necessary
gram.
(b) Streamlines and for shock B to be reflected so that the stream
shock lines. in region 3 will have the same direction as
(c) Pressure-deflec-
tion diagram. in region 1. The strength of shock B is
Art. 16.7 REFLECTION AND INTERACTION OF SHOCKS 555
related to that of shock A by the requirement that they have equal

turning angles.
Fig. 16.25b shows how the problem may be solved graphically with
the help of the hodograph shock polars, and Fig. 16.25~ shows the solu-
tion with the help of the 6-p shock polars. In both these graphical con-
structions the shock polar corresponding to M2 is displaced through an
angle 6 so as to correspond to the actual direction in region 2.
It is evident from this example that an oblique shock is reflected from
a solid wall in like sense, i.e., it is reflected as an oblique shock.
In practice, the presence of a boundary layer near the upper wall
makes the process of reflection
25
much more complicated than the
U
simple solution given here, as will g 20
be seen in Art. 16.12. w
0
A study of Figs. 16.25a and I5
16.25~will show that for a given -k

M1 there is a maximum value of :l o
E
turning angle 6 for which the
5
type of reflection in Fig. 16.25b1
known as regular reflection, is o
1.0 2.0 3.0 4.0
possible. If 6 is too large, then it M I ( 0 1 Ml",,"~
will exceed the value of 6,,, cor-
FIG.16.26. Limiting values of MI and 6
responding to Mz (see Fig. 16.18). for regular reflection ( k = 1.4).
It may also be seen that for a
given 6, there is a minimum value of MI below which regular reflection
is impossible. Fig. 16.26 summarizes the ranges of M1 and 6 within
which regular reflection may occur.
When regular reflection is impossible, a Mach reflection, described
later, takes place.
Crossing of Shocks of Opposite Family. Referring to Fig. 16.27a,
suppose that oblique shocks A and B of known strength are propagated
into region 1 and ultimately meet. How will the shocks influence each
other? Let C be the continuation of B, and D be the continuation of A.
Now consider two nearly adjacent streamlines which pass on opposite
sides of the junction point, one streamline passing through A and C,
and the other through B and D. After these streamlines have passed
through the shock system, they must have the same pressure and direc-
tion. As they have passed through shocks of different strength, how-
ever, they have incurred different entropy increases, and hence they
have different speeds. A surface of discontinuity (slip line) which is
also a streamline, therefore emanates from the junction point. Regions
4' and 4" have the same direction and pressure, but different speeds.
The slip line is therefore a very thin region of concentrated vorticity,
and is sometimes called a vortez sheet.
The graphical solution to this problem with the aid of the shock polar
diagrams is shown in Fig. 16.27b and 16.27~. The latter diagram is
especially convenient for this type of problem.
Since fields 4' and 4" have the same pressure and same stagnation
temperature, the adiabatic and isentropic relations require that the
field with the lower s t a g n a t i o a y (higher entropy) have the lower
(c) (4
Fro. 16.27. Crossing of shock waves.
(a) Streamlines, shock lines, and slip line.
(c) Pressure-deflection diagram.
(d) Schi~erenphotograph showing refraction of two shocks of equal strength;
may also be interpreted as reflection of shock from straight boundary. (Re-
produced by permission of The Macmillan Co. from Elements of Aerodynam-
ics of Supersonic Flows, by A. Ferri, 1949.)
velocity. In this case shocks A and C are stronger than B and D, hence
the gas speed is larger in field 4" than in field 4'. This same rule may be
applied whenever vorticity appears in an adiabatic flow.
If shocks A and B are of equal strength, i.e., if 62 = 83, then the shock
configuration will be symmetrical, and no slip line will appear (Fig.
16.27d). The streamline passing through the junction point will undergo
no change in direction, and the entire process will therefore be equivalent
to that of Fig. 16.25, that is, to shock reflection from a plane wall.
The type of regular interaction shown in Fig. 16.27 may prove to be
impossible if MI is too small or if 62 and 63 are too large, for the same
reasons that regular reflection may be impossible with the incident wave
pattern of Fig. 16.25. A Mach reflection then occurs.
Art. 16.7 REFLECTION AND INTERACTION OF SHOCKS 557
Mach Reflection. When MI is so small or 6 is so large as to precIude
the possibility of regular reflection in Fig. 16.25, a Mach reJEection (Fig.
16.28) occurs. Near the upper wall the shock must necessarily be nor-
mal, for this is the only way in which a shock can occur without a change
in direction. Shock C, of opposite family to shock A , emanates from
lrrotationol
Flow Rotational
FIG. 16.28. Mach reflection.

Explanatory diagram.
Schlieren photograph of Mach reflection from wall, obtained with a moving
- -
wave svstem in a fluid a t rest. thus obviating distortions due to boundarv
layer (ifter Bleakncy and ~ a u b ) .
Schlieren photograph of Mach reflection of two intersecting shocks, flow
from right to left (reproduced by permission of The Macmdlan Co. from
Elements of Aerodynamics of Supersonic Flows, by A. Ferri, 19 $9).
the juncture of incident shock A and normal shock B. However, shocks

A , B, and C cannot be straight to the point of mutual intersection, for
the same reason that regular reflection of A was impossible to begin
with. It is observed in practice that legs A , B, and C are curved near
the junction point. A slip line emanates from the junction, and the
flow behind all the shocks is rotational because of the curvature of the
shocks. The curvature of the shocks also gives rise to a continuous
wave system. Thus, the flow pattern is exceedingly complicated, and
a t present our knowledge is limited to the qualitative aspects of Mach
reflection.
When regular crossing of waves A and B in Fig. 16.27 is impossible, a
Mach reflection occurs (Fig. 16.28~). Waves A and B of Fig. 16.27a
do not meet, but are bridged rather by a nearly normal shock.
Intersection of Shocks of Same Family. Referring t o Fig. 16.298,

suppose that MI, 02, and O3 are given, thus defining the angles and
strengths of shocks A and B. For

reasons given previously, wave A
will always overtake wave B.
Let us first ask whether waves A
and B can be continued only in the
form of wave D. If such were the
case, regions 3 and 5 would have
the same pressure and direction,
but different velocities. However,
c-2,
the pressure and dilrctiori in region
.??
... . . .... -. -. . . . . . ,... 3 arch tletcrrnined by tl~c: initial
(4 data. This means that wave D
FIG. 16.29. Intersection of two left- must be of such strength as to
running shock waves.
satisfy simultaneously both the
(a) Physical plane. pressure and direction require-
(c) Schlieren photographof flow cor- ments. However, if the turning
responding to (a). Note Mach angle of D is selected so that
lines generated from roughened
surface and their inkract,ions O5 = 83, all other properties of D
with both shocks. SeeFig. 16.10 are immediately determined, and
(after Johannesen).
it is to he expected that. but for
exceptional circumstances, the pressure in region 5 will not match
that in region 3.
These arguments serve t o prove that something more than wave D is
required. One way of resolving the difficulty is to suppose that wave C,
which may be either a Prandtl-Meyer expansion or an oblique shock,
also emanates from the junction point. Then, by suitably adjusting the
strengths of waves C and D it is possible to obtain a solution for which
Art. 16.8 CURVED SHOCKS 559
p5 = p4 and O4 = 05. Of course a slip line divides regions 4 and 5. The

hodograph solution, with wave C shown as a rarefaction, is illustrated in
Fig. 16.29b.
16.8. Curved Shocks

Applicability of Shock Equations. In most practical problems curved
shocks appear either because the shock is necessarily detached (as in
Figs. 1 6 . 1 9 ~
and 16.20b), or because an initially straight shock interacts
with a continuous wave.
Are the shock equations of Art. 16.2 valid for shock surfaces which
are curved in two or three dimensions? The answer is yes, since the
analysis of Art. 16.2 may be applied to two streamlines which in the
limit approach each other. The shock relations are, therefore, applicable
to each individual streamline, provided that the shock.angle a is meas-
ured between the streamline entering the shock surface and the tangent
to the shock surface a t the point of entry, in the plane containing the
approach streamline and the normal t o the tangent plane.
Interaction of Shocks with Continuous Waves. There are four pos-
sible types of such interactions, depending on whether the continuous
wave is a com~ressionor rarefaction, and on whether the continuous
wave is of the same or opposite family to the shock wave.
As an illustration of the method of analysis, let us consider the flow
past the triangular airfoil profile of Fig. 16.30a, in which the oblique
shock generated a t point A interacts with the continuous Prandtl-Meyer
rarefaction wave of the same family originating a t point C. From the
profile geometry and initial Mach Number, and with the help of the
shock relations, we may find conditions in zones I and 2. Then, employ-
ing the field method of calculation, we may split the centered rarefaction
wave into several zones, the first of which is zone 3. Applying the char-
acteristics theory to the expansion between 2 and 3, we may find the
stream properties in zone 3 and the direction of the first rarefaction
- -
wave CB. Thus, point B, representing the beginning of the interaction

process, may be located. Now consider the two streamlines immediately
above and below B. Because of the pressure drop from 2 t o 3, the shock
above B will not be as strong as that below B, and the two streamlines
referred to will undergo different entropy changes. In order to have
equal pressures and flow directions in fields 4' and 4", a small rarefac-
tion wave B E must be reflected a t point B. Since the speeds are dif-
ferent in fields 4' and 4", a slip line BF emanates from point B. The
strength of the shock between B and D and the strength of the reflected
rarefaction BE are determined by the two simultaneous conditions that.
the pressures and flow directions must be the same in zones 4' and 4".
as illustrated in Fig. 16.30b.
The calculation procedure is continued by similar methods. Where

the second rarefaction CD strikes the shock BD, the shock again changes
strength and direction, and n slip line DN and reflected wave DG are
gcnerated. A reflected wave such
as R E interacts with the waves BC
and CD, and this interaction may

be computed by the method of
characteristics. It should be noted
also that when a wavelet such as
CD crosses the slip line BF, the
wavelet is in part transmitted with
changed strength, and is in part
reflected; this follows from the re-
quirements that the pressures and
flow directions must be the same
for the adjacent streamlines form-
ing the slip plane.
In the limit, if the stepwise pro-
FIG. 16.30. Calculation of wave system cedure outlined above is made ever
produced by single-wedge profile.
finer, i t may be seen that the region
(a) Flow pattern.
(b) Hodogrnph diagram. of interaction extends between the
(c) Simplified flow pattern, neglect- first reflected wave BE and the
ing vorticity.
portion of the shock downstream of
point B. The shock is straight from A to B, and is continually curved
downstream of B. In the limiting continuous solution, furthermore, no
individual slip lines appear as such, but the entire region bounded by
the streamline BF and by the curved shock downstream of B has vor-
ticity. The slip lines are merely artifices of the stepwise solution which
allow us to lump a continuously distributed vorticity into discrete dis-
continuities for convenience in calculation.
The method of characteristics for isentropic flow (Arts. 15.3 and 15.4)
may not be applied t o the region'where vorticity exists, except as ar,
approximation, because this region is not isentropic. However, it turns
Art. 16.9 EXPLICIT SOLUTIONS BY SERIES EXPANSIONS 561
out that this approximation very often gives acceptable results. I n es-
sence, the approximate method ignores the variations in entropy change
across different parts of the oblique shock, and, therefore, the slip lines
and the waves reflected from the shock do not appear. Fig. 16.30~ shows
the simplification in wave pattern afforded by this approximation.
I n the region of vorticity where continuous waves are present, the
entropy represents an additional variable. The method of character-
istics may be worked out formally in this case, as in Art. 15.12.
Without formally working out the method of characteristics for non-
isentropic flow, however, we may make calculations for the region of
rotational flow by extending the ideas already presented. In isentropic
flow the pressure is a unique function of velocity, whereas in nonisen-
tropic flow this is not true. This difference explains why the hodograph
characteristics are fixed for two-dimensional, isentropic flow but are not
fixed for nonisentropic flow. These remarks suggest further that we
apply the method of characteristics stepwise to the rotational flow pat-
terns, using not the hodograph characteristics diagram, but rather work-
ing out the physical characteristics in terms of the relations between
pressure change and direction change for each individual wavelet, taking
note of the actual pressures, velocities, and Mach Numbers in each field.
This method is in essence equivalent to the more formal mathematical
methods.
SHOCKS CROSSING SLIP LINES. When a shock enters a region of
parallel flow with varying speed, it is refracted and suffers partial re-
flection. This problem is discussed in Reference 14 (see Problem 16.36),
and, more generally, in Reference 18.
16.9. Explicit Solutions by Series Expansions

The shock relations are so complex algebraically that i t is seldom
possible to get analytical solutions even for such simple cases as the flow
over a flat-plate airfoil at an angle of attack. In Chapter 14 simple
analytical results were achieved by means of the linearized theory. The
formulas of the linear theory may be thought of as the first term of a
series solution, thus suggesting that more accurate analytical results
might be found by taking terms of higher order than the first.
Series Solution for Continuous Waves. Returning for the moment
to isentropic waves of one family, the characteristic relations of
Chapter 15 are -
*O = J-
k + l
k-1
arc tan J-
k-1
k+l
@.A2 - 1)
-arc tan dtA2- 1 + constant

For isentropic flow, moreover, L
Eliminating M from this pair of equations, and then expanding

p - p, in terms of M, and Ae, where p, is the initial pressure, M, is the
initial Mach Number, and A0 is the turning angle of the streamline from
the original direction, there is obtained after tedious calculations (lo'
2 C 3 ( ~ e )+
~ c(ae), + C 2 ( ~ e )=t
P - Pw
1
= 3 ... (16.20)
zkpmMm2
where A0 is positive when counterclockwise. The upper and lower signs
refer respectively to left-running and right-running Mach waves, and
the coefficients are as follows:
(16.21~)
It will be noted that the first term of Eq. 16.20 represents the linear-
ized solution.
Series Solution for Oblique Shock. If the shock relations of Art. 16.2
are expanded in a similar way, letting subscript 1 refer to conditions
upstream of the shock, we find (ll.*) that the first two terms are the
same as for isentropic waves of one family, with the difference between
the two types of flow appearing initially in the third term:
where
Although the coefficients C1, Cz, C3, and D appear complex, they
depend only on M, and k, and need be computed only once. The coeffi-
cients C1 and Cz are tabulated in Table B.8 and are plotted in Fig. 16.41.
Art. 16.10 EXAMPLES OF FLOWS CONTAINING SHOCKS 563
Equations 16.20 and 16.22 may then be uscd for obtaining analytical
results, provided of course that waves of opposite family do not cross
each other. Examples are given later.
If only terms up to second order in A0 are employed, it is clear from
these equations that the oblique shock theory and the continuous wave
theory give identical results.
16.10. Examples of Two-Dimensional Flows Containing Shocks

Most supersonic flows include both shocks and regions of continuous
waves. A number of illustrative problems will now be discussed in the
light of the characteristics theory of Chapter 15 and of the oblique shock
theory of this chapter.
Formation of Shock from Continuous Wave. When a pointed body
with a profile such as A-B-C-D-E-F in Fig. 16.31a is placed in a super-
sonic stream, compression wavelets are generated a t each point on the
concave surface. These wavelets converge - and ultimately- form the
beginning of an envelope a t G, a t which point an oblique compression
shock begins. As additional compression wavelets starting a t C, D, etc.,
reach the initial wave at H, J, etc., they merge with the compression
shock and strengthen the latter. Finally all the compression wavelets
are coalesced into a single oblique shock beginning a t point L.
Consider (Fig. 16.31b) the merging of the first two wavelets a t point
G. The pressures, velocities, and flow directions in fields 2 and 3 are
found by applying the characteristics theory to waves AG and BG. The
subsequent wave pattern near G is determined by the requirements
that the streamline passing through the shock immediately above G end
with the same pressure and flow direction (in field 4") as the streamline
passing through the two compression wavelets immediately below G
(field 4'). I n general, these conditions can be met only if a weak re-
flected wavelet GM is generated; furthermore, since field 4' has a higher
stagnation pressure than field 4", but the same static pressure, the gas
speed in field 4' will be larger than that in 4" and a slip line will divide
fields 4' and 4". (See Problem 16.13.)
If the calculation procedure of the preceding paragraph is continued,
i t is found that the shock GHJKL is curved; that there is a weak system
of reflected waves, GM, HN, etc.; and that slip lines GS, etc., cover the
region downstream of GL, thus making this region one of vorticity.
Furthermore each wavelet crossing a slip line suffers a partial reflection.
To avoid complicating the figure, these very weak reflections are not
shown.
-. the reflected waves GM, H N , etc.,
Unless the shock is Quite strong,
A
are extremely weak and the velocity difference across the slip lines i s
OBLIQTJE SHOCKS Ch. 16
FIG. 16.31. Formation of shock from continuous compression wave.

(a) Exact solution, taking account of vorticity behind curved shock. Slip line*
emanating from H , J, K , L are omitted. Reflected wavelets generated OII
GS are also omitted.
(b) Details of pattern near G.
(c) Approximate solution, ignoring vorticity.
(d) Schlieren photograph of shock formed from continuous wave generated a t
circular-arc surface. Turbulent boundary layer. Note vortex sheets (after
Johannesen).
Art. 16.10 EXAMPLES OF FLOWS CONTAINING SHOCKS 565
very small, thus suggesting the simplified solution of Fig. 1 6 . 3 1 ~in

which the entropy changes across the shock are ignored. In many
practical problems, the simple construction of Fig. 1 6 . 3 1 ~is quite ade-
quate. Sometimes it is desirable to use the simple construction of Fig.
16.31~ for the weaker portions of the shock and the more rigorous solu-
tion of Fig. 16.31a for the stronger portions of the shock.
Overexpanded Jet from Nozzle. Consider a uniform, parallel, super-

sonic, two-dimensional jet which
leaves a nozzle a t section 1 with a
pressure less than that of the ex-
haust region into which the nozzle
discharges (Fig. 16.32a).
-
The pressure of the jet bound-
aries must be equal to that of the
exhaust region. Accordingly, ob-
lique compression waves A E and
CE are generated and the jet
boundaries are turned through an -
angle 6 such that pz and p, are
equal to the exhaust pressure.
Because of symmetry condi-
F IG. 16.32. Jet leaving overexpanded
tions, the interaction of oblique
shock waves A E and CE is equi-
(a) Regular reflection of shocks.
valent to the reflection of a shock (b) Mach reflection of shocks.
from a plane wall. The strengths
of waves .EDand EB are determined by the condition that the stream
has the same directions in fields 1 and 4.
The pressure in field 4 is greater than the exhaust pressure. In order
for the jet boundary to remain a t constant pressure, shock waves ED
and Ei? must be reflected as centered Prandtl-Meyer rarefaction waves
where the former strike the jet boundaries a t D and R. The strengths of
the rarefaction waves are determined by setting p5 = p~ = pz = p s
The turning of the jet boundary a t B and D is, approximately, 26, as
given by the linear theory.
At secticn BD there is a uniform, parallel flow a t a pressure greater
than the exhaust pressure. Beginning with this section, therefore, the
situation is equivalent to that of an underexpanded jet leaving a nozzle.
All further reflections a t the boundary are either centered ra.refactions
or centered compressions, and the jet goes through a series of repeated
cyclic variations (Fig. 15.23a). Schlieren photographs (Figs. 5 . 2 4 ~and
16.27d) confirm the flow pattern constructed here except for the effects
of viscous mixing a t the boundaries.
Although both overexpanded and underexpanded jets exhibit a

repeated pattern of cyclic variations, the first half-cycle of the over-
expanded jet (the region between sections AC and BD) is never repeated.
These results may be compared with the approximate solutions based
on the linearized theory (Fig. 14.23).
If the exhaust pressure is too much larger than the pressure in the
exit plane, regular reflection is impossible, and the Mach type of reflec-
tion is observed (Figs. 16.3213 and 16.28~). The subsequent flow pattern
is very difficult to analyze because it consists of mixed subsonic and
supersonic flow and contains regions of vorticity.
Jet Leaving Turbine Nozzle. The exit plane of a turbine nozzle is
usually cut off a t an angle to the direction of flow. Fig. 16.33 shows the
FIG.16.33. Jet leaving converging turbine nozzle.
nature of the flow pattern for such a jet when the nozzle is of the con-
verging type and the flow is underexpanded. The supply pressure is
assumed to be 100 psia, and the exhaust region pressure 27 psia. Hence
sonic flow exists a t the throat with a pressure of 53 psia. The wave
pattern is worked out for a 2" turning angle across each wave, except
for the shock. Entropy increases across the latter are ignored. Expan-
sion wavelets are shown by dashed lines, and compression wavelets by
solid lines.
An important result demonstrated by Fig. 16.33 is that the average
jet direction differs significantly from the nozzle direction because of
the general bending of the jet. This bending effect reduces the work
available from the turbine wheel.
Oblique-Shock D i h s e r . Fig. 16.34 shows an oblique-shock diffuser,

the upper boundary of which could be considered either as a &reamline
of symmetry or as a solid wall.
The oblique shock originating a t the bend (point A ) is reflected from
each of the walls in turn. Each reflected shock has the same turning
angle as the original shock AB. Because of the deceleration through
each shock, a point is reached, depending on the initial Mach Number
i
Art. 16.11 TWO-DIMENSIONAL PROFILES 567
and turning angle, beyond which regular reflection is impossible. The

last shock, DE, is, therefore, curved, and reduces virtually the entire
stream to subsonic speeds.
The potential advantage of such a diffuser is that the supersonic
deceleration occurs across several oblique shocks of small turning angle.
Calculations indicate that this leads to comparatively small losses in
stagnation pressure. The normal, or nearly normal, shocks DE and F G
appear only after the Mach Number has been reduced to nearly unity.
Fro. 16.34. Oblique-shock diffuser.
In practice the inviscid flow pattern of Fig. 16.34 is seriously modified

by the interaction between the shock waves and the boundary layers on
the walls.
16.11. Two-Dimensional Profiles
Shock-Characteristic Analysis for Two-Dimensional Profiles. The
"exact" solution for the lift, drag, and moment produced by a super-
sonic stream on a two-dimensional profile is obtained by combining the
shock analysis with the method of characteristics, on the assumption
that viscous effects can be ignored.
Consider, for example, the flow past a symmetrical double-wedge
profile a t an angle of attack. The boundary conditions a t the profile
are satisfied by the wave pattern of Fig. 16.35. Each of the four faces
of the diamond profile has a constant pressure over its surface. The
wave between regions 1 and 2 is an oblique shock for small angles of
attack, and a centered Prandtl-Meyer rarefaction wave for large angles of
attack. For the case illustrated there is a shock between fields 1 and 2.
The values of p2 and M2 may accordingly be found with the help of the
oblique-shock theory from p,, M,, and the turning angle between 1 and 2.
Next, the values of p4 and M4 may be found by applying the method of
characteristics to the Prandtl-Meyer flow originating a t point B, using
the known turning angle between fields 2 and 4 and the previously found
values of p2 and M2. Similar calculations yield the pressures and Mach
Numbers in fields 3 and 5.
Having found the pressure distribution over the profile, the resulting
forces and moments may be calculated, leading finally to the lift, drag,
and moment coefficients. Although the "exact" theory as given here is
more accurate than the linear and second-order theories, it has the dis-
advantage that a separate set of numerical calculations is required for
each combination of profile shape, angle of attack, and Mach Number,
and that it is not possible to arrive a t simple analytic formulas showing
the effects of these variables.
Eggers and Syvertson ( ' 9 ) have formulated a comparatively simple
approximation to the shock-expansion method for thin profiles, which,
FIG.16.35. Double-wedge profile.
for Mach Numbers greater than 3 and flow deflection angles less than
25", yields pressure distributions in error by less than 10yo.
In order to determine the nature of the wake immediately behind the
profile, we note that the two shocks emanating from D must be of such
strengths that the pressures and flow directions in regions 6 and 7 are
the same. The two boundary streamlines passing through D have
passed through shocks of different strength, and, consequently, have
different stagnation pressures. Accordingly, the double streamline trail-
ing from point D is a slip line, or vortex sheet. The common flow direc-
tion of fields 6 and 7 is in general different from that in region 1, and,
furthermore,'the common pressure in these fields differs from that of
region 1. These differences, however, areusually minute.
The "exact" calculation is very much more difficult if the waves
generated a t the profile interfere with each other so near the profile
that the reflected waves reach the profile further downstream, as is the
case with the circular-arc profile of Fig. 16.36. Consider the flow over
the upper surface (Fig. 16.3Ga). An oblique shock is generated a t the
nose in accordance with the turning angle associated with the tangent,
at the nose. Because of the continuous curvature of the profile, however,
expansion wavelets are continually generated. These intersect and

weaken the shock. At the intersection of each wavelet with the shock
there are generated a reflected wave and a slip line. Some of the re-
flected waves reach the profile and thereby modify the surface pressure
distribution. The computations are evidently much complicated by
the vorticity in the fluid downstream of the shock.
FIG.16.36. Biconvex profile.

(a) Some details of exact wave pattern, showing that wavelets reflected at curved
shock influence surface pressure distribution.
(b) Approximate wave pattern, ignoring vorticity behind shock.
(c) Schlieren photograph for flow of (b) (after Ackeret).
As pointed out previously, however, it is often acceptable to ignore

the variation in entropy change over a fairly large region. If this is done,
there are no reflected wavelets to deal with, and we get the easily calcu-
lated flow pattern of Fig. 16.36b. On the upper and lower surfaces the
flow is of simple wave type, and the curvatures of the shocks do not
enter into the calculation of the lift, drag, and moment. Schlieren pho-
tographs (Fig. 16.36~)show wave patterns in good agreement with the
approximate pattern of Fig. 16.36b, and the calculations of Reference
19 further confirm that there is little loss in accuracy incurred by ignor-
ing the entropy variations.
Relation Between Wave Drag and Shock Losses. I n Chapter 14 we
saw from the linearized analysis that the pressure drag of supersonic
570 OBLIQUE SHOCKS Ch.16
profiles is associated with the thickness of the profile and with the pro-
duction of lift. We shall now look at the matter from another point of
view.
Fig. 16.37 shows a control surface all the boundaries of which are at
great distances from the profile. Although the flow pattern near the
profile is marked by large variations in pressure and direction, the wave
systems ultimately interact with each other and tend to wipe out non-
uniformities. As described in more detail later, the net effect of the wave
interactions is that a t very great distances from the profile the flow is
uniform and parallel and differs from the original direction and pressure
----------
"--( I
I
I Control
I
i isurface
I
I
8-rj
7
I
I
FIG.16.37. Control surface for momentum analysis.
by only infinitesimal amounts. The approach flow is uniform and paral-

lel over section AB, and, from the foregoing considerations, the stream
pressure and direction over sections CD, BD, and AC are the same as
over AB. Hence the drag force on the profile is equal to the momentum
flux through AB minus the momentum fluxes through CD, BD, and AC.
Within the control surface the streamlines suffer losses of stagnation
pressure (or increases of entropy) wherever they cross shocks. In gen-
eral, therefore, the stagnation pressure along AB is greater than that
along CD, BD, and AC. But, since the pressures on these planes are
equal, the velocities are smaller along CD, BD, and AC than along AB.
Consequently there is a drag on the profile, a drag which is directly
related to the losses in stagnation pressure through the shock system.
Translated into practical terms, these considerations show why super-
sonic profiles have sharp leading edges, for a rounded leading edge pro-
duces detached shocks which are nearly normal and hence have large
losses in stagnation pressure. Furthermore, it is plain that with thin
sections the leading edge shocks are relatively weak and produce rela-
tively small losses in stagnation pressure; hence thin sections have rela-
tively small drag. I n order to obtain large lift coefficients it is necessary
to have large angles of attack. The latter, however, are accompanied
by relatively strong shocks which in turn give rise to a relatively large
pressure drag. In supersonic two-dimensional flow, therefore, the pro-
duction of lift is always accompanied by the production of drag.
The importance of a sharp leading edge for supersonic flow is illus-

trated strikingly by experimental data (Fig. 16.38) for the flow past a
strut formed of a cylinder and a fairing.(I2) The lift-drag polars show
clearly that in supersonic flow i t is better to make the sharp end of the
strut the leading edge and the rounded end the trailing edge. This is
exactly the opposite of what is desirable for subsonic flow! In super-
sonic flow a blunt leading edge produces enormous increases in drag
because of the strong detached shock wave in the flow field. Although
FIG. 16.38. Experimental drag polars FIG. 16.39. Experimental drag polars
for strut (after Busemann and Walchner). for subsonic profile (after Busemann and
Walchner).
i t is desirable to have both the leading and trailing edges sharp, it is

more important to avoid a blunt leading edge than a blunt trailing
edge.
This point is demonstrated again in the experimental data (I2) of
Fig. 16.39 for a conventional subsonic profile. In supersonic flow the
profile is more efficient when the leading and trailing edges are reversed.
Second-Order Theory for Two-Dimensional Profiles. To obtain
analytical results somewhat more accurate than the formulas of the
linear theory (Chapter 14), we may employ the series expansions of
Eqs. 16.20 and 16.22.
If third-order terms are taken into account, the calculations become
so involved that i t is easier t,o use the exact shock-characteristics theory.
Hence, since simplicity is one of the chief aims of an approximate anal-
ysis, we shall consider only terms up to the second order in Eqs. 16.20
and 16.22. Since these two series expansions are identical up to second-
order terms, the second-ord~rtheory takes no account of the entropy
changes which accompany oblique shocks in the flow pattern.
We refer a system of axes to the profile (Fig. 16.40), with the f-axis
passing through the sharp leading and trailing edges. The upper sur-
face of the profile has the ordinate ~ C J taken
, positive when above the
chord line; similarly the ordinate V L of the lower surface is taken posi-
tive when below the chord line. The angle of attack, a, is positive when
directed as shown in the sketch.
The method of analysis is similar to that of the linearized theory
(Art. 14.6), but the second-order terms of Eqs. 16.20 and 16.22 are
Fro. 16.40. Nomenclature for second-order thin profile theory.
taken into account. During the course of the calculations several inte-
grals appear which depend only on the geometry of the profile. These
are
Each of these definite integrals must be evaluated separately for the

upper and lower surfaces. The corresponding values are denoted by
Jiu, J I L , etc.
The force and moment coefficients which result from the analysis
are (1,6)
Only pressure drag is included in the drag coefficient given above. The
skin-friction drag coefficient must therefore be added to the right-hand
side of Eq. 16.26 in order to get the total drag coefficient.
The moment coefficient of Eq. 16.27 is reckoned about the leading
edge, and is taken to be positive when the moment acting on the profile
-
"m E Chord-' E
F IG. 16.42. Cambered single-wedge
profile.
(c)
FIG. 16.43. Typical supersonic profiles.

Mm
FIG. 16.41. Coefficients of the second-
order theory (k = 1.4).
is counterclockwise, with the direction of flow from left to right as in

Fig. 16.40.
By setting Cz equal t o zero in Eqs. 16.25, 16.26, and 16.27, the second-
order theory is reduced to the first-order, or linear, theory, and the re-
sults are then identical with those previously found in Chapter 14.
The orders of magnitude of the coefficients C1 and C2 are shown in
Fig. 16.41. These cocfficients are always positive.
All the J-integrals are zero for a flat plate, and Eqs. 16.26, 16.26, and
16.27 then show that the first-order and second-order theories are identi-
cal for a flat plate because only terms in C1 remain. We may therefore
state the general rule that the results of the second-order and linear
theories approach each other as the profile shape approaches a flat plate.
T RIANGULAR PROFILE. TO illustrate the difference between the
second-ordcr and linear theories, consider the triangular profile of Fig.
16.42 a t zero angle of attack. The various J-integrals are
For the force coefficients we then obtain
The first-order theory predicts zero lift, but the second-order theory
predicts a negative lift. The latter fact is especially interesting in that
it shows still another surprising difference between subsonic and super-
sonic flow. In subsonic flow at zero angle of attack with a positive cam-
ber as in Fig. 16.42, the lift is also positive; in supersonic flow the lift
for the same profile is negative.
At zero angle of attack, the thickness drag for this profile is the same
for both the first-order and second-order theories.
If lc.p, is the distance from the leading edge to the center of pressure,
we may show by taking moments that
Thus, the second-order theory predicts that the c.p. is a finite distance
forward of the mid-chord point, whereas the linear theory predicts it
to be infinitely forward. As a general rule the most significant improve-
ment of the second-order theory over the linear theory lies in a more
accurate prediction of the center of pressure.
COMPLETELY SYMMETRICAL PROFILE. If the profile is symmetrical
both laterally and longitudinally, we may set Jr. = J uand we may also
set J2= 0. We then obtain
CL = 2Cla (16.28a)
+ 2C1Jlu
CD = 2Cla2 (16.28b)
Cm = Cla + 4C2J3ua (16.28~)
No second-order terms appear in the expressions for CL and CD. We
may write, for the location of the center of pressure,
Now J S U is equal to ( - S v / l 2 ) , where SU is the cross-sectional area

between the chord line and the upper surface of the profile. Therefore
J B u is always negative for such a symmetrical profile. Eq. 16.28d shows
that the center of pressure is forward of the mid-chord, and, within the
approximations of the second-order theory, i t is independent of the angle
of attack.
Some Types of Two-Dimensional, Thin Profiles for Supersonic Flow.
The common forms of supersonic profiles are modifications or combina-
tions of two basic forms: (1) a profile formed of straight lines (Fig.
16.43a), and (2) a profile formed of circular arcs (Fig. 16.43b).
Many practical considerations may enter into the choice of a profile:
drag a t low and high angles of attack; stability of the center of pressure
location over a wide range of angle of attack and Mach Number; per-
formance a t subsonic speeds; maximum stresses; maximum deflections;
thermal stresses; vibrations; and weight and volume.
To avoid high drag, the thickness ratio must not be large, and the
turning angles for shocks must not be large. For a given thickness
ratio the symmetrical double wedge (Fig. 16.43a) has the smallest turn-
ing angles for shocks. It may in fact be demonstrated (') from the sec-
ond-order theory, with the help of the calculus of variations, that (i) for
a given thickness ratio, the profile of minimum drag is diamond-shaped,
i.e., it is formed of four straight lines, (ii) for a given thickness ratio and
minimum thickness drag, the diamond-shaped profile is symmetrical
about the chord, but has its position of maximum thickness slightly a f t
of the mid-chord point, and (iii) for a given thickness ratio and maximum
lift-drag ratio, the diamond-shaped profile has a slight positive camber
and the position of maximum thickness is slightly aft of the mid-chord
point.
From a practical point of view, it may be stated generally that an
aerodynamically optimum profile shape seldom differs significantly from
the symmetrical double wedge, and that slight variations in profile from
this shape do not produce significant changes in aerodynamic character-
istics. The actual choice of profile in any given instance will, therefore,
depend heavily on the other factors listed above.
Comparison of Theoretical with Experimental Results for Thin
Profiles. Figs. 16.44 and 16.45 show a comparison of typical experi-
mental data ('3) with the results of the linear, second-order, and exact
theories. These same data were previously compared with the linear
theory in a somewhat different way (Figs. 14.15 and 14.16).
From Fig. 16.44 we see that the second-order theory is a great improve-
ment over the first-order theory in predicting the pressure distribution,
and is almost as accurate as the "exact" theory. I n comparing the
latter with the experimental data, it should be noted that the "exact"
theory was applied only approximately, in that the type of solution of

Fig. 16.36b, rather than that of Fig. 16.36a, was employed. At about
60% of chord, the measured pressure distribution on the lower surface
departs markedly from the theoretical. This is the result of an inter-
action between the trailing-edge shock and the boundary layer, as will
be discussed more fully later (Art. 16.12).
0 Ferri's Experiments
- "Exact" Theory
-.- Second O r d e r Theory
--- Linearized Theory
% Chord
FIG. 16.44. Comparison of experimental pressure distribution with results of exact,
second-order, and linear theories (after Ferri).
Fig. 16.45 shows that the second-order and exact theories give virtu-
ally identical results for lift and drag up to angles of attack of the order
of 10". Both provide a significant improvement over the linear theory,
and compare very well with the experimental data. The agreement
with the moment data is only fair. It is a t first surprising that the
actual drag is slightly less than the theoretical pressure drag of the exact
theory, since the latter does not include skin friction. This effect is
associated with the shock-boundary layer interaction near the trailing
edge (Art. 16.12).
Downwash from Thin Profile. The flow aft of a profile is of interest
if other bodies should have to be placed in the wake. For instance,
Art. 16.11 TWO-DIMENSIONAL PROFILES
Theory Includes
a, Degrees
6.3% Thick
-
-.-
0 Ferri's Experiments
Exact Theory
--- Second O r d e r T h e o r y
Linear Theory
FIG. 16.45. Comparison of experimental force coefficients with results of exact,

second-order, and linear theories (after Ferri).
control surfaces usually are in the wake of the main lifting surface of an
aircraft.
If not for entropy increases across the shock waves (Fig. 16.46), there
would be no slip line between zones 4 and 5 and the upwash angle @
would be zero. It is therefore necessary to include terms of third order
in Eqs. 16.20 and 16.22 in an investigation of the upwash angle. Such
an analysis (lo' yields the following results for a flat-plate airfoil when
These results were found to check within 1% of the exact theory for
a = 10' and MI = 5. The magnitude of 0 is generally small, being al-
ways less than 0.06' for a = 5.7' over the range of M, from 1.2 to 5.0.
A surprising feature is that, except for M, less than about 1.3, there is
an upwash behind the wing, rather than a downwash as would be ex-
pected for subsonic flow. Of course momentum considerations require
that a lift on the profile must be accompanied by a net downward change
of momentum flux of all the air influenced by the profile; hence if the
part of the airstream near the profile suffers an upwash, other parts of
the airstream must acquire a downwash.
F I ~16.46.
. Nomenclature for upwssh relations.
(a) Flabplate profile.
(b) Doubly symmetric profile.
When (i) the profile has thickness, (ii) there are four oblique shocks a t
the nose and tail (Fig. 16.46b), and (iii) the airfoil is completely sym-
metrical, it may be shown ( l o ) that
This formula indicates that the flow deflection for a profile with thick-
ness is considerably larger than for a flat plate. For example, when
E = a, the ratio between the two is 16.
Wave System in Wake. Let us consider the wake of a flat-plate

profile a t an angle of attack (Fig. 16.47). In the immediate neighbor-
hood of the profile, owing to oblique shocks and Prandtl-Meyer waves,
there are large variations in pressure and flow direction. However, the
compression and expansion waves merge and thereby are mutually
weakened. Vorticity is generated as a result of the curvature of the
shocks. I n the discrete wave pattern this phenomenon is represented
by slip lines. Wavelets striking slip lines are in part transmitted and
part reflected. Thus, more and more waves appear in the flow field, the
net result being to distribute nonuniformities on an ever finer scale. At
very large distances from the profile the flow is, for all practical purposes,
again uniform and parallel, with only infinitesimal deviations from the
free-stream condition and with all waves a t the free-stream Mach angle.
A large part of the wave system for the lower side of the profile is omitted
from Fig. 16.47 so as not to confuse the pattern unnecessarily.
It is this type of reasoning which justifies the analysis accompanying
Fig. 16.37.
FIG. 16.47. Wave pattern downstream of flat-plate airfoil. Most of the wave
system on the underside is omitted in the interest of clarity.
Supersonic Biplane. We have seen that in supersonic flow profiles

must be very thin if large drags are to be avoided. This imposes severe
practical difficulties in construction and makes it all but impossible to
put propulsion plants inside the wings of supersonic aircraft.
An arrangement for avoiding these difficulties is the Busemann bi-
plane (Fig. 16.48). The contours ACE and BDF are so shaped that, a t
zero angle of attack, the entering flow a t section AB is decelerated re-
versibly to section CD, and is then reversibly accelerated t o the initial
Mach Number a t section EF. Since there are no shock waves, there is
no drag, and hence we see that the scheme permits a wing section of
finite thickness without any accompanying thickness drag. If the bi-
plane were inclined, the drag owing to angle of attack would be approxi-
mately the same as for a conventional supersonic profile, so that the
advantage of the biplane (not taking into account the increased skin
friction for the biplane) is in the range of low angles of attack.
The contours may be designed by the method of characteristics.

Since the forward half of the biplane is a supersonic diffuser, care must
be taken in adjusting the area ratio C D to AB, for otherwise the diffuser
will not start, but rather a detached bow wave will stand ahead of the
entrance and subsonic flow will prevail in the passage. From these
considerations, i t is clear that the allowable thickness of each half of
the biplane increases with Mach Number.
(a) (b)
FIG. 16.48. Busemann biplane.
(a) Zero thickness drag.
(b) Small thickness drag.
Even when shocks are present, as in the biplane design of Fig. 16.4813,
they are confined to the space between the profiles and hence produce
only a fraction of the drag which they would otherwise produce if allowed
to propagate out to great distances from the profile.
16.12. Interaction of Shock Waves with Boundary Layer

We have thus far ignored the effects of viscosity on the nature of the
flow pattern, but it is particularly important not to lose sight of viscous
effects when shocks are present.
One of the great inventions in modern fluid mechanics is Prandtl's
concept of the boundary layer. The underlying idea is that viscous
stresses in a fluid of low viscosity are everywhere small compared with
other terms in the momentum equation except in comparatively thin
layers near solid boundaries where large velocity gradients, and hence
appreciable viscous stresses, must necessarily exist. Viscous effects are
therefore assumed to be important only in "boundary layers" adjacent
to solid surfaces, and the exterior flow is assumed to be frictionless.
This approach is clearly a great simplification when the boundary
layer flow does not significantly affect the external flow, for then Lhe two
regions of flow may be calculated separately, assuming for a first approx-
imation that the boundary of the frictionless flow coincides with the
solid boundary. Such is often the case for problems of subsonic flow,
especially when the boundary layer is relatively thin compared with
other significant length dimensions, e.g., the thickness of a profile or the
diameter of a duct. The presence of the boundary layer then has only
a second-order effect on the exterior flow.
Art. 16.12 SHOCK-BOUNDARY LAYER INTERACTION 581
When separation of the boundary layer occurs, however, due to ad-

verse pressure gradients, as in the stalling of a wing profile, the boundary
layer thickens to such an extent that it materially affects the cxterior
flow, and it is no longer possible to assume that the boundary of the
exterior flow coincides approximately with the solid boundary. When
this occurs the simplicity of the boundary-layer concept breaks down,
for the boundary layer then affects the exterior flow in a manner which
is by no means negligible.
When shock waves occur near solid bodies, the interaction between
the shock waves and the boundary layer usually produces first-order
effects on both flows. The extraordinary pressure gradients in the
neighborhood of the shock wave violently distort the boundary-layer
flow. These distortions, which are propagated both upstream and down-
stream in the subsonic part of the boundary layer, in turn influence the
conditions outside the boundary layer by means of expansion and com-
pression waves generated by the disturbed boundary layer. Thus there
is a mutual interaction between a shock wave and a boundary layer, an
interaction which may lead to a flow pattern radically different from
what might be anticipated if no boundary layer were present. A more
complete discussion than is given here is presented in Volume 11, Chap-
ter 28.
Oblique Shock Wave Incident on Flat Plate. In Fig. 16.49a the ob-
lique shock AB is incident on a fla which there is a boundary
layer. If the boundary layer were LLvupLGaGLLt,
the incident shock would
be reflected as an oblique shock of equal turning angle, as in Fig. 16.25b.
When a boundary layer is present this simple type of reflection cannot
occur because-'(i) the shock bends as it enters the region of changing
velocity in the boundary layer, and must terminate a t the sonic line in
the boundary layer, and (ii) the effects of the shock can be propagated
upstream in the subsonic boundary layer.
As a result of the latter effect, the pressure in the boundary layer
begins to rise before the point of incidence of shock AB. With this
pressure rise goes a velocity decrease and hence a thickening of the
boundary layer along DB. Oblique compression wavelets are generated
along DB as a result of the turning of the streamlines. These wavelets
coalesce into the shock EF, which may be thought of as part of the re-
flected shock system.
As the original shock AB enters the boundary layer, it is refracted
along BC because it is in a region of changing Mach Number. As a re-
sult of this refraction, wavelets are generated along BC; some of these
are rarefaction wwelets, but often they are compression wavelets which
coalesce into the shock MN.
The pressure a t F is often.greater than that a t G because the stream
a t F has passed through a shock of comparatively great strength. Such
a pressure gradient, if present, causes the streamlines to turn back

toward the wall. This streamline curvature in turn produces expan-
sion waves in the supersonic flow.
If the pressure difference between F and G is quite large the flow may
be turned toward the wall. The latter, however, constrains the flow in
(tj) (c)
FIG.16.49. Reflection of oblique shock from boundary layer on flat plate.
(a) Schematic representation.
(b) Schlieren photograph for turbulent boundary layer (after Liepmann, NACA).
(c) Schlieren photograph for laminar boundary layer (after Liepmann, NACA).
such a way that compression wavelets are generated along the concave
part of the boundary layer HJ. These wavelets coalesce into the shock
KL.
Experimental observations are in qualitative accord with the fore-
going rationalization of the wave pattern. The nature of the interac-
tion depends upon Mach Number, shock strength, velocity profile in
the boundary layer, and especially on whether the boundary layer is
laminar or turbulent. In all cases observed, however, the flow patterns
contain in greater or lesser degree all or part of the elements shown in
Fig. 16.49a (see also Chapter 28 of Volume 11).
Art. 16.12 SHOCK-BOUNDARY LAYER INTERACTION 583
"Normal" Shocks in Ducts. When discussing one-dimensional super-

sonic flows in ducts (Chapter 5 ) we often speak of a normal shock stand-
ing in the duct when the back pressure is high. I n practice it is observed
that the pressure rise through the shock occurs over an axial distance of
several duct diameters, and schlieren photographs show patterns like
those of Fig. 16.50. This pattern may be rationalized by noting that the
FIG.16.50. "Normal" shock in duct.
normal shock could not possibly extend to the wall when a boundary
layer is present and that the rapid thickening of the boundary layer up-
stream of the shock gives rise to an oblique shock. The reflected shock
a t the juncture of the normal and oblique shocks is necessary as a result
of continuity requirements. Thus, a series of forked normal shocks ap-
pears, with the length of the normal shock becoming progressively
smaller as the boundary layer thickens. Finally the sonic velocity is
reached and the stream then decelerates subsonically as it refills the'
passage. In place of a normal shock, therefore, the stream contracts by
means of a combination of normal
and oblique shocks to Mach Num-
ber unity and then diverges to
subsonic speeds.
Shock Propagated at a Bend.
When no boundary layer is present
a shock is propagated in a straight
line from a sharp concave bend, as
in Fig. 16.1~.
The corresponding pattern for
supersonic flow past a sharp con-
cave corner with a FIG. 16.51. Generation of shock in con-
layer cave corner with boundary layer present.
present is shown in Fig. 16.51.
The large pressure rise near the corner propagates upstream through the
subsonic boundary layer. As the boundary layer decelerates, it also
thickens, and thus continuously propagates compression wavelets into
the supersonic flow. These wavelets merge and ultimately form a shock
across which the turning of the stream occurs.
Flow Near Trailing Edge of Supersonic Profile. The concepts of the

preceding paragraph help to explain the flow near the trailing edge of a
supersonic profile a t an angle of attack.
Without a boundary layer, the wave pattern, flow pattern, and pres-
sure distribution on the upper surface near the trailing edge are as shown
in Fig. 16.52a.
With a boundary layer, the situation is. much like that of Fig. 16.51,
and the resulting wave pattern, flow pattern, and pressure distribution
Expansion Waves Exp.Waves Camp. W a v e s
FIG.16.52. 1 % ~
rwar trailing edge of supersonic profile.
(:I)Boundary layer abscnt.
(b) Boundary layer prescrlt.
are as shown in Fig. 16.5213. The difference in flow patterns between

Fig. 16.52a and Fig. 16.52b is the result of the rapid thickening of the
boundary layer as it approaches the trailing edge.
The pressures on the upper surface near the trailing edge are higher
than those which might be expected if viscosity were neglected. This
conclusion is clearly confirmed by the typical experimental results of
Fig. 16.44. Over the rearward half of the profile the measured pressures
on the lower surface are markedly higher than those computed by the
inviscid theory.
From a comparison of the pressure distributions of Figs. 16.52a and
16.5213 it may be concluded that the effect of the shock boundary-layer
interaction is to reduce the lift coefficient, and, surprisingly, to reduce
the drag coefficient also. This latter effect explains why in Fig. 16.45
the measured drag is less than the calculated drag even though the lat-
ter does not include skin-friction drag.

1. FERRI, A. Elements of Aerodynamics of Supersonic Flows. New York: The
Macmillan Co., 1949.
2. C OURANT , R., and FRIEDRICHS, K. 0. Supersonic Flow and Shock Waves. New
3. SAUER, R. IRtroduction to Theoretical Gas Dynamics. Translated by F. K. H ILL
and R . A. ALPHER,Ann Arbor: J. W. Edwards, 1947.
4. D AILEY , C. L., and WOOD, F. C. Computation Curves for Compressible Fluid
Problems. New York: John Wiley & Sons, Inc., 1949.
5. NEICE, M. M. Tables and Charts of Flow Parameters Across Oblique Shocks,
N A C A Tech. Note, No. 1673 (1948).
6. AMESAERONAUTICAL LABORATORY. Notes and Tables for Use in the Analysis of
Supersonic Flow, NACA Tech. Note, No. 1428 (1947).
7. ISENBERG, J. S. The Method of Characteristics in Compressible Flow, Part IA:
Tables and Charts, Air Materiel Command Report, No. F-TR-1173B-ND
(19473.
8. KEENAN, J. H., and KAYE, J. Gas Tables. New York: John Wiley & Sons, Inc.,
1948.
FI. ISENBERG.
. - J. S.. and LIN. C. C. The Method of Characteristics in Com~ressible
Flow, Part I: Steady Supersonic Flow, Air Materiel Command Report, No.
F-TR-1173A-ND (1947).
10. K AHANE , A., and LEES, L. The Flow a t the Rear of a Two-Dimensional Super-
sonic'~irfoi1,Jour. kero. Sci., Vol. 15, No. 3 (1948), p. 167.
11. LAITONE, E. V. Exact and Approximate Solutions of Two-Dimensional Oblique
Shock Flow, Jour. Aero. Sci., Vol. 14, No. 1 (1947), p. 25. ,
12. BUSEMANN, A,, and WALCHNER, 0. Profileigenschaften bei Ueberschallgeschwin-
digkeit, Forschung aus dem Gebiet des Zngenieur-Mresens, Vol. 4 (1933), p. 87.
13. FERRI, A. Experimental Results with Airfoils Tested in the High-speed Wind
Tunnel a t Guidonia, N A C A Tech. Memo., No. 946 (1940).
14. B UHLER, R. D. Reflection and Refraction of Oblique Shock Waves a t Contact
Discontinuities (Vortex Sheets), Report No. AL944, North American
Aviation, Inc. (1949).
15. ROCKEW,J. A., and HAYES, W. D. Monograph 11: The Method of Character-
istics in Comoressible Flow. Part IC: Two-Dimensional Flow with Larize
Entropy Changes, Air Materiel Command Report, No. 102-A~49/6-1&
(1949).
16. IVEY,H. R., STICKLE, G. W., and SCHUE'ITLER, A. Charts for Determining the
Characteristics of Sharp-Nose Airfoils in Two-Dimensional Flow a t Super-
sonic Speeds, NACA Tech. Note, No. 1143 (1947).
17. MOECKEL. W. E.. and CONNORS. J. F. Charts for the Determination of Suoer-
sonic ~ i lo; r Against ~nclinkdPlanes and Axially Symmetric Cones, NACA
Tech. Note, No. 1373 (1947).
18. MOECKEL, W. E. Interaction of Oblique Shock Waves with Regions of Variable
Pressure. E n t r o ~ v and
. Energy. NACA Tech. Note. No. 2725 (1952).
19. EGGERS,A. J ., ~~.,'a"n'd SYVERTSG, C. A. Inviscid FI'OW About Airfoils a t High
Supersonic Speeds, NACA Tech. Note, No. 2646 (1952).
20. ~OHANNESEK, N. H. Experiments on Two-Dimensional Supersonic Flow in Cor-
ners and over Concave Surfaces, Aero. Research Crmncil, F.M. 1669 (Jan.
1952).
PROBLEMS
16.1. Derive the oblique shock relations of Art. 16.2 by a transformation
of the normal shock rehtions a s suggested in Art. 16.1.
16.2. Derive t h e relation between temperature ratio and pressure ratio for
oblique shocks.
16.3. Demonstrate that the pressure ratio and the density ratio across an
oblique shock depend only on k and on MI sin a. Note that as a consequence
the normal shock tables give directly the pressure and density ratios for oblique
shocks if the inlet Mach Number M, of the normal shock is interpreted as the
value of MI sin u for the oblique shock.
16.4. Investigate the branches of the hodograph shock polar for values of
uz/c* greater than ul/c*. What is the physical significance of these branches?
Can they be used for the graphical solution of problems, and, if so, in what way?
What are the maximum values of u ~ / c *and v ~ / c *on the right-hand branches,
in terms of ul/c*?
16.5. Demonstrate that the hodograph shock polar is a circle for an initial
Mach Number of infinity.
16.6. (a) Investigate the branches of the 6-p shock polars for p2/p1 less than
unity, and discuss the physical significance of these branches.
(b) Derive the equations of the 6-p shock polars.
16.7. Compare incompressible flow in a concave corner with supersonic
flow in a concave corner.
16.8. By expanding p2/p1 as a function of (p, - pl)/pl for both the Rankine-
Hugoniot relation and for the isentropic relation, demonstrate that the two
relations give identical results up to and including terms in the second order of
shock strength.
16.9. Derive the oblique shock relations with the assumption that there is
sufficient longitudinal heat conduction to produce equal temperatures on the
two sides of the shock.
Compare the results for such an oblique shock in a general way with the
corresponding results for an adiabatic oblique shock.
16.10. Demonstrate the following limiting relations for oblique shocks as
the initial Mach Number M1 becomes very large:
tan u k+l
+-
tan (u - 6) k-1
16.11. (a) Show that the shock angle urn, corresponding to the maximum
turning angle 6, is given explicitly in terms of MI by the following expression:
(b) Show that the shock angle u* corresponding to the turning angle 6* for
which the downstream Mach Number is unity is given explicitly in terms of MI
Ch. 16 PROBLEMS
by the following expression:
16.12. Demonstrate the following series relation for the stagnation-pressure

ratio in terms of the initial Mach Number and the shock angle, a relation espe-
cially useful for weak shocks where the shock angle is slightly larger than the
Mach angle:
16.13. (a) Demonstrate that if two points in the flow have the same pressure
and same stagnation temperature, but different stagnation pressures, the veloci-
ties are related as follows:
(b) From the above result show with the help of momentum considerations
that the pressure drag of a two-dimensional supersonic profile in an infinite
medium is directly related to the entropy increases across the shock waves in
the flow field.
(c) From the result in part (a) show that when a slip stream appears in an
adiabatic flow, the Mach Number is higher on the side of the slip stream having
the lower entropy.
(d) Show that if the oblique shocks produced by a supersonic, two-dimen-
sional profile did not interact with rarefaction waves in such a way that the
oblique shocks were ultimately weakened to become Mach waves, the drag of
such a profile would be infinite.
16.14. A parallel stream with a pressure of 10 psia and a Mach Number of
2.0 approaches a sharp l o 0 corner (see sketch).
(a) Assuming that the shock is of the weak variety, calculate the following
quantities: the final Mach Number, the final pressure, the change of entropy,
and the ratio of the final flow per unit area to the initial flow per unit area.
(b) Make a sketch to scale showing the streamlines, the shock line, and the
Mach lines upstream and downstream of the shock.
(c) Compare the results of part (a) for final Mach Number and final pressure
with the analogous result for a 10"-turn based on linearized theory, a 10"-turn
based on the Prandtl-Meyer corner-type flow, and a lo 0 -turn based on the
second-order theory.
16.15. Compare the linearized theory with the oblique-shock theory and the
second-order theory by plotting curves of final Mach Number and of the frac-
tional change in pressure (Ap/p) against the deflection angle, basing each curve
on a constant initial Mach Number.
Use constant initial Mach Numbers of 1.5 and 2.5, and carry each curve from
zero deflection to the maximum deflection angle.
16.15. Air flows in the passage shown in the sketch with an initial Mach
Number of 2.0. Determine the maximum turning angle 6 for which three regular
reflections of the original shock arc= possible.
16.17. An initially uniform parallel flow undergoes a 10"-turn across shock A

and a subsequent 10"-turn across shock B (see Fig. 16.29a). Assuming that
MI = 3.0 and pl = 100, determine the values of M, 8, and p in fields 2, 3,4, and 5.
16.18. The sketch shows the inlet of a n unsymmetrical oblique-shock
diffuser. Calculate the pressure and flow direction in fields 4 and 5 by
(a) The first-order theory
(b) The second-order theory
(c) The exact shock theory (in this case find also the value of M4*-M5*)
16.19. Consider a triple-shock configuration A-B-C as shown in the sketch.

Find the values of M* and flow direction 0 in fields 2, 3, tt~d4 for M ~ =* 1.50
and 81 = 0, if
(a) 82 = +4" (Answer: M2* = 1.421; Ma* = 0.759; M4*= 0.679; =
64' = -6')
(b) O2 = +12O (Answer: M2* = 1.231; Ma* = 0.523; Ma* = 0.696; O3 =
O4 = +9.2O) '
Sketch the shock directions to scale.

Ch. 16 PROBLEMS 589
16.20. Consider a shock configuration A-B-C as shown in the sketch, where

A is a normal shock. Find the dimensionless velocity ratio, M*, and flow direc-
tion, 8, in fields 2, 3, and 4 if MI* = 1.50 and 81 = 0. Sketch the shock directions
to scale.
(Answer: O2 = 7.5";O3 = 04 = 0; M2* = 1.350; M3* = 0.788; M4* = 0.666)
16.21. Consider an oblique shock A incident on a slip plane dividing regions

1 and 2 (see sketch). These regions have different Mach Numbers but the same
pressures and stagnation temperatures.
(a) Assunling that wave A is very weak and that the velocity difference
Setween zones 1 and 2 is very small, show that the condition that there be no
reflected wave C is that M1 s 4 2 .
(b) Assuming that MI = 2.0, M2 = 1.9, 6~ = loo, determine 68, 6c, Os, M4,
and Ms.
16.22. A wedge-shaped probe (see sketch) may be used for estimating the
Mach Number of a supersonic wind tunnel.
(a) In a particular instance a wedge with a 10" in-
cluded angle is used. Schlieren photographs show
that the totd included angle of the shock front is /
- c
98". Estimate the Mach Number of the tunnel.
(b) What is the minimum Mach Number for
which a 10"-probe can be used in this way as a
Mach Number indirator?
\
FROB.16.22.
16.23. Discuss the suitability of conventional subsonic profiles for supersonic

speeds. Discuss the suitability of conventional supersonic profiles for subsonic
speeds. Consider the design or invention of a profile which is to operate mainly
a t supersonic speeds but must be capable also of developing high lift a t subsonic
speeds.
16.24. Using the results of the second-order theory for thin profiles, deter-
mine an expression for the angle of zero lift, ao, in terms of shape integrals and
Mach Number. Show from this result that a. must be positive for positive
camber, and'compare this conclusion with subsonic flow past thin profiles.
Show that for a given profile the value of a(] increases with M,.
16.25. A flat-plate profile of chord I is to be tested a t Mach Number 2.0 in
a wind tunnel of rectangular cross section, the tunnel height being H. Deter-
mine the minimum ratio HI1 required to avoic! interference, for
(a) Zero angle of attack
(b) An angle of attack of 10"
16.26. Consider the flow past a symmetrical double-wedge profile of 10%
thickness ratio a t Mach Number 2.0, and determine the maximum angle of
attack for which the leading-edge shock will be attached.
16.27. Consider the flow past a flat-plate profile at M = d 2 . Assuming

a skin-friction coefficient of 0.005, determine the maximum lift-drag ratio and
the corresponding angle of attack, using
(a) The linearized theory
(c) The exact theory
16.28. Consider the flow past a symmetrical double-wedge profile of 10%
thickness ratio, a t Mach Number 2, and with an angle of attack of 5". Calculate
the lift, drag and moment coefficients for this profile by
(a) The exact theory
(c) The linearized theory
16.29. Given the curved shock shown in the sketch with MI = 2.0. At
points a and b, which are separated by the distance h = 0.1 ft, the shock angles
are a,,= 35" and m, = 55". The stagnation temperature is 530°R.
Determine approximately the rotation of the fluid downstream of the shock

on a streamline midway between a and b, in seconds-'.
16.30. Design a practical supersonic biplane to operate at Mach Number 2.5
and zero angle of attack, taking note of starting problems, and with the aim of
obtaining as large
L~
a cargo volume within the biplane wings as possible.
16.31. Consider a parallel, two-dimensional jet leaving a nozzle at e with
Me= 2.0. (a) Plot the deflection angle of the jet, 6, as a function of the premm
Ch. 16 PROBLEMS 591
ratio p,/p,, for values of the latter between unity and the value for which a
normal shock stands in the exit plane. (b) Determine the range of pressure
ratios, p,/p,, for which regular reflection of the oblique shock wave originating
a t the nozzle exit is not possible.
16.32. A supersonic stream leaves a nozzle in parallel flow (region a) with a

Mach Number of 2.0 and a pressure of 10 psia. The pressure of the atmosphere
into which the jet discharges is 15 psia.
(a) Calculate the pressures in regions b and c.
(b) Make a sketch to scale showing streamlines and shock lines.
(c) Assuming the pressure a t the nozzle entrance to be maintained constant,
what is the maximum atmospheric pressure for which this general type of flow
configuration is possible? Describe the nature of the flow pattern when the
exhaust-region pressure is raised above this limiting value.
(d) Compare the results of part (a) with the results of calculations based on
the first-order and second-order theories.
16.33. A two-dimensional supersonic diffuser is to be designed as shown in

the sketch for a Mach Number of 2.5. The ratio hS/hl is to be chosen so that the
diffuser will barely swallow the initial shock, and the ratio i/hl is to be selected
so as to obtain the wave pattern shown.
(a) Determine h2/hl and L/hl.
(b) Neglecting friction, compare the over-all stagnation-pressure ratio of
this diffuser with the stagnation-pressure ratio of a diiuser in which a normal
shock occurs a t Mach Number 2.5.
(c) Compare the results of part (a) with the linearized theory.
Symmetrical
t Porabola
........ ....................................
..~ :..:..:
............... ~ . .~,
:.i
~ . . :....... ...............:
... ..........
.......;c..
:>:.:::::
:<.<x...:.:.,.,.
?.................... ...., ..................
.:.::::..?:,~5*
............................................
I.."
16.34. Using second-order theory, find Cr.and Cn for the profile shown in
the sketch in terms of &, t/1, and a.
16.35. A two-dimensional double-wedge profile (see sketch) is at zero angle
of attack in an air stream of Mach Kumber 2.016. Ignoring friction, and using
exact methods,
(a) Calculate the lift coefficient.
(b) Calculate the drag coefficient.
(c) Calculate the ratio Lc.,,/l where I,., is distance aft of the leading edge
to the center of pressure, and I is the chord.
(d) Estimate the downwash angle, E , immediately downstream of the trailing
edge.
. Sketch the wave system s t great distances from the profile.
it: :ompare the results obtained above with similar results based on the sec-
ond-order theory.
16.36. (a) Demonstrate the following relation between turning angle, initial
Mach Number, and shock strength exp~essedin terms of the fractional pressure
k+lpz-pi
- ( 1 +-Ui )
tan 6 = rt -1 P 2 - PI
k PI k+lpz-pi
Pl
(b) From this show that, if an obliyw -hock crosses a slip line which divides
parallel flows having Mach Numbers Mia and MI=, the condition that the shock
be transmitted without any reflection whatsoever is that
~1.4' (1 + k + 1 7 ) ~PI Pi )
-
-
P2 - Pi ( 2
PI
+ P2 - Pl)
pl
MIB2 =
(c) Show from part (a) that for a fixed pressure ratio across the shock, there
is a maximum turning angle, 6, (,I, which occurs a t a Mach Number given by
and that the corresponding value of maximum turning angle is given by
(d) From part (c) show that if an oblique shock crosses a weak slip line divid-
ing parallel flows of nearly the same Mach Number, the condition that there
shall be no reflection a t the slip line is approximately
APPENDIX
Appendix A
THEORY OF CHARACTERISTICS
A S . Introductory Remarks
The differential equations describing the motion of compressible
fluids in two or three dimensions are, with the exceptions of a few simple
cases or under greatly simplifying assumptions, not easily soluble by
analytical methods. Fortunately, there is a class of problems for which
solutions may be rather easily obtained by means of numerical or
graphical procedures. Such is the case when the motion is described by
a partial differential equation of hyperbolic type. We shall see that
the possibility of constructing solutions arises from the special properties
of certain characteristic curves which satisfy second-order, partial differ-
ential equations of hyperbolic type with two independent variables.
The purpose of this Appendix is to outline, for those readers not al-
ready familiar with the subject, the mathematical theory underlying
the method of characteristics. Particular attention will be given to
those aspects of the theory which are relevant to problems of compress-
ible flow (readers are referred to Reference 1 for a more complete treat-
ment than is given here).
NOMENCLATURE
functions of x, y, q,, and velocity potential
Pu
speed of sound signifies conditions along
ratio of specific heats characteristic curve
radius in cylindrical and signifies conditions along
polar coordinates characteristic of family
radius in spherical coor- 1.
dinates signifies conditions along
time characteristic of fam-
velocity components ily I I .
Cartesian coordinates
Some of the important physical problems which are easily handled

by the method of characteristics may be summarized as follows (it is
assumed throughout that we are dealing with a perfect gas having
595
596 THEORY OF CHARACTERISTICS APP. A
neither viscosity nor thermal conductivity, and that there are no dis-
continuities in fluid properties such as occur in shock waves):
Two-Dimensional, Plane, Supersonic, Irrotational, Steady Flow
(Chapter 16). Application of Euler's equation, the continuity equation,
and the condition of irrotationality yields
(c2 -
2
9z )9m + (c2 - '~Y2)~yy= 0
- 29z(~y9xy (A.1)
where cp is the velocity potential, x and y are the Cartesian space coordi-
nates, c is the local sound velocity, and fixdenotes d29/axz, etc. The
velocity components u and v are given by
The local sound velocity depends on 9=and cpu according to the energy
equation,
k-1
c2 = co2 - -(9x2
2
vU2)+ (A.2b)
where co is the sound velocity a t the stagnation temperature.

Axially Symmetric, Supersonic, Irrotational, Steady Flow (Volume
11, Chapter 17). Application of Euler's equation, the continuity equa-
tion and the condition of irrotationality yields
c2
+ +
(c2 - (pZ2)~zz- 29z~r'~zr (c2 - vr2)9rr - ( ~ t .= 0
r
(A.3)
where 9 is the velocity potential, x and r are cylindrical space coordinates

(x being along the axis of symmetry and T normal to the axis of sym-
metry), and c is the local sound velocity. The velocity components u
and v in the x- and r-directions, respectively, are given by
u = (o,; v = $0, (A.4a)
The local sound velocity is, according to the energy equation, given by
where co is the sound velocity at the stagnation temperature.

Non-steady, One-Dimensional Motion with Constant Cross-Sec-
tional Area (Volume 11, Chapter 24). Application of the Euler and
continuity equations, together with the introduction of a velocity
potential, yields
(c2 - (4z2)%z - 29zPzt - 9 t t = 0 @a
where x is the one-dimensional space coordinate measured along the
direction of motion, t is the time coordinate, cp is the velocity potential,
APP. A INTRODUCTORY REMARKS 597
and c is the local sound velocity. The fluid velocity in the x-direction
is given by
u = cox (A. 6a)
The local sound velocity is related to P
(, and cot through the equation
(A.6b)
where co is the sound velocity a t any arbitrary location and time.

Non-steady, Plane, Cylindrically Symmetrical Motion. Application
of the Euler and continuity equations, together with the introduction
of a velocity potential, yields
2
c2
(c2 - -2
~r ) ~ r r ~ ~ -r ~ t t -t - ~r = 0 64.7)
r
where r is the radius in polar coordinates, t is the time coordinate, (P is
the velocity potential, and c is the local sound velocity. The fluid
velocity in the r-direction is given by
The local sound velocity is related to p , and cot by an expression

similar to Eq. A.6b, except that pz is replaced by (P,.
Non-steady, Spherically Symmetrical Motion. Application of the
Euler and continuity equations, together with the introduction of a
velocity potential, yields
c2
2 2
(c - V R )VRR - 2'~R'f'~t - Ptt - 2-(PR
R
=0 (-4.9)
where R is the radius in spherical coordinates, t is the time coordinate,

q is the velocity potential, and c is the local sound velocity. The fluid
velocity in the R-direction is given by
U = (PR (A.10)
The local sound velocity is related to c p and
~ ( ~ by
t an expression
similar to Eq. AXb, except that rp, is replaced by p ~ .
Quasi-hear Differential Equation of Second Order in Two Inde-
pendent Variables. Eqs. A.1, A.3, A.5, A.7, and A.9 are all nonlinear
partial differential equations of second order with two independent
variables. They are sometimes called quasi-linear differential equations,
as they are linear in the derivatives of highest order. All of them have
where the coefficients A, B, C, and D are, in general, functions of x, y,

cp,, and VU.
598 THEORY O F CHARACTERISTICS APP. A
The remainder of this Appendix is devoted to a study of certain p r o p

erties of this differential equation, a study which leads to a convenient
procedure for solving the practical problems mentioned previously.
Occasionally it will be helpful to think in terms of a physical analogue
to the mathematical analysis. For this purpose we shall consider two-
dimensional, supersonic flow (Eq. A.l) as a typical illustrative example.
Then rp will be thought of as the velocity potential, rp, will be the x-com-
ponent of velocity (u), and rp, will be the y-component of velocity (v).
Since the rp,, cp,-plane is, for this case, usually called the hodograph plane,
it will be convenient terminology to employ the same phrase as a
generalization for the cp,, cpv-plane corresponding to Eq. A.ll.
A.2. The Characteristic Curves

We consider here only regions of the flow where the fluid properties
are continuous. This means that the derivatives cp, and rp, (the velocity
components) are continuous functions of x and y. Hence, we may
write, for arbitrary increments dx and dy,
Integral Surfaces. Each solution to the differential equation (Eq.

A.ll) may be thought of as a three-dimensional surface in x, y, rp space,
called an integral surface, and defined by a function p = cp(x, y) which
satisfies Eq. A.ll. We may imagine the integral surface to have drawn
on it various space curves, each of which has a certain projection on the
x, y-plane. A small step along one of these projected curves a t a certain
point determines corresponding values of drp, and dcp, according to
Eqs. A.12 and A.13.
Definition of Characteristics. We now ask: Is it possible that, on
certain curves lying on an integral surface, the derivatives of (p, and rp,
(e.g., the derivatives of the velocity components, and hence of all fluid
properties) might be discontinuous? Such curves, when they exist, are
called the characteristic curves of the solution, and their projections on
the x, y-plane are called the physical characteristics.
Eq. A.11 is applicable a t any point on the integral surface. More-
over, Eqs. A.12 and A.13 are applicable to the increments correspond-
ing to an infinitesimal length of any curve lying in that surface. Since
we are interested in possible indeterminacies of the derivatives cp,,, pup,,,
APP. A THE CHARACTERISTIC CURVES 599
and p,,, we write these equations in the following manner:
These may be regarded as simultaneous, linear, algebraic equations in

the variables p,,, rp,, and a,. Solving for rp,, we obtain, in determi-
nant notation
I 0 dcp, dy I-
-
+
A dy dcp, - D dy dx C dx dp,
(A.14)
A dy2 - 2B dx dy f C dx2
At any point (x, y, cp) on the integral surface, the values of rp, and p,
are known from the shape of the surface. Consequently the coefficients
A , B, C, and D are known a t all points of the surface. I n general, there-
fore, the value of rp,, (and, by similar reasoning, the values of p,, and
p,,) is determined by Eq. A.14, except when both the numerator and
denominator of Eq. A.14 are equal t o zero. The latter condition gives
the answer to our original query, for when pz,, p,,, and p,, are indeter-
minate, they may be discontinuous.
Physical Characteristic Curves. By setting the denominator of Eq.
A.14 equal to zero, we obtain the differential equation of the projections
of the characteristic curves in the physical plane, namely,
(A. 15a)
or, after solving the quadratic,
Hodograph Characteristic Curves. By means of Eq. A.15b the slopes

of the characteristics in the physical plane may be found. However, the
coefficients A , B, and C, in terms of which the slopes (dy/dx),hnr are ex-
pressed, depend in part on p, and p,. I n order to construct the charac-
teristics in the physical plane, it is therefore necessak to determine how
rp, and p, vary along the characteristics. This is most easily done by
finding the equation of the characteristics in the co,, p,-plane.
Since rp,, is, in general, finite, i t is necessary that the numerator as

well as the denominator of Eq. A.14 vanish. Therefore we write
Solving for d9,/dqX, we obtain
which is further simplified by introducing Eq. A.l5b, thus yielding
B f d B 2 . - AC
(A. 16)
c
Eq. A.16 gives the slopes of the characteristics in the pX,9,-plane in
terms of q,, 9,,x, and y. It may be recalled that Eq. A.15b gives the slopes
of the characteristics in the x, y-plane also in terms of qz, 9,, x, and y.
Given suitable initial data, therefore, the characteristic curves in the
physical and hodograph planes may simultaneously be constructed nu-
merically or graphically in stepwise fashion. Since the physical data of
interest (such as pressure, velocity, etc.) may be found from the values
of qz and 9, a t each point, a solution of the problem is thereby deter-
mined. Before going into details, we shall discuss certain general fea-
tures of the method.
Classification of Differential Equations. Examination of Eqs. A .15
and A.16 shows that the differential equation A.ll may be of three
types, depending on the sign of (B2 - AC). These types are defined as
follows:
Hyperbolic type. The value of (B2 - AC) is positive, so that Eqs.
A.15 and A.16 each have two real roots. Two characteristic curves pass
through every point of the physical and hodograph planes.
Parabolic type. The value of (B2 - AC) is zero, and there is one real
root to Eqs. A.15 and A.16. This type is not of great practical signif-
cance.
Elliptic type. The value of (B2 - AC) is negative, so that Eqs. A.15
snd A.16 have no real roots, and the characteristic curves are imaginary.
Applying these criteria to Eqs. A.l, A.3, A.5, A.7, and A.9, it is found
Lhat (i) for the two cases of steady motion the differential equation is
hyperbolic for supersonic flow and elliptic for subsonic flow, whereas
(ii) for the three cases of non-steady motion the differential equation is
nlways of hyperbolic type. Hence real characteristics always exist for
the non-steady motions, but exist for the steady motions only when the
flow is supersonic.
APP. A THE CHARACTERISTIC CURVES 601
Properties of Characteristic Curves. CHARACTERISTIC LINES VERSUS

S HOCK LINES. The physical characteristics for two-dimensional, super-
sonic flow are the Mach lines (Chapter 15), and are inherently different
from lines of shock discontinuities (Chapter 16). On a shock line the
fluid properties (velocity, pressure, etc.) are discontinuous. On a char-
acteristic curve, however, the fluid properties themselves are continu-
ous, but their derivatives may be discontinuous.
CHARACTERISTIC CURVES AS SOLUTIONS TO THE DIFFERENTIAL EQUA-
TION. Eqs. A.15b and A.16 satisfy the differential Eq. A.11. The latter
may accordingly be solved by simultaneous solution of Eqs. A.15b and
A.16. Although this simultaneous solution does not directly yield
cp = cp(x, y), it yields a result even more useful for practical purposes,
namely, the velocity components a t each point of the x, y-plane. In-
deed, the possibility of arriving a t solutions of Eq. A.ll merely by con-
structing the characteristic curves in the physical and hodograph planes
constitutes the most important practical aspect of the theory of charac-
teristics. The original task of solving a difficult nonlinear partial dif-
ferential equation of second order (Eq. A.ll) is thereby replaced by the
comparatively easy task of simultaneousIy solving two pairs of ordinary
differential equations of first order.
Thinking again of the solution to Eq. A.ll as a surface in cp, x, y-
space, the shape of this surface may be approximated by a skeleton
formed by a finite number of wires lying in the surface. If these wires
follow the characteristic curves of the surface, then the physical charac-
teristics (Eq. A.15b) are the projections of these wires in the x, y-plane,
and the hodograph characteristics (Eq. A.16) are the representation of
the slopes (p, and cp,) of the wires. The physical and hodograph charac-
teristics taken together may be used to construct the wire skeleton and
thus to represent the solution of Eq. A.ll.
CHARACTERISTIC CURVES AS LOCI OF POSSIBLE DISCONTINUITIES IN
DERIVATIVES OF FLUID PROPERTIES. The characteristic curves are
possible branch lines on an integral surface, in the sense that the second
and higher derivatives of cp are indeterminate and may xndergo discon-
tinuities a t points on these curves. Thus it follows that regions of flow
of different types (i.e., described by expressions which are analytically
different) may be joined, or ''patched," on the characteristics. For this
reason the characteristic curves are sometimes called patching curves.
In two-dimensional, supersonic flow, for example, a region of uniform
parallel flow can be joined to a region of nonuniform flow along a Mach
line.
An extension of the foregoing reasoning leads to the conclusion that
supersonic flows with prescribed initial conditions are inherently non-
analytic in charact,er. For example, if we are dealing with the super-
sonic flow past an arbitrary body or through an arbitrary duct, it is

almost certain that no analytic expression exists which describes the
entire flow pattern; a t best analytical expressions may be found for
describing certain limited regions of the flow (e.g., regions of simple-
wave sow).
When t.he flows in.two adjoining regions are described by expressions
which are analytically different, the arguments given above lead to the
conclusion that the two regions are necessarily adjoining along a char-
acteristic curve. In two-dimensional, supersonic flow, for example, a
region of uniform parallel flow can adjoin a region of nonuniform flow
only on a characteristic curve.
Equations of the Characteristics for Certain Simplifying Conditions.
There are two families of characteristics, corresponding to the plus and
minus signs in Eqs. A.15b and A.16. We shall refer to them respectively
as family I and family II. We may then summarize the equations of
the characteristics as follows:
Substantial simplifications of procedure are possible under the three

special sets of circumstances described below.
COEFFICIENTS A , B, C D EPEND ONLY O N x A N D y. Under these con-
ditions, i t is evideut from Eqs. A.17 that the characteristics in the physi-
cal plane are independent of the hodograph coordinates. That is, they
may be found by direct integration of Eqs. A.17 without the need for
simultaneously constructing the characteristics in the hodograph plane.
Important applications include (i) two-dimensional, supersonic, steady
motion with small perturbations (Chapter 14), and (ii) one-dimensional,
non-steady motion with small perturbations in a constant-area duct
(Volume 11, Chapter 23).
HOMOGENEOUS EQUATION. When the coefficient D is zero, the dif-
ferential equation A.11 is homogeneous. It may then be shown from
Eqs. A.15b and A.16 that
APP A CONSTRUCTING CHARACTERISTIC CURVES GO3
These relations show (i) that the physical characteristic I is normal to

the hodograph characteristic 11, and (ii) that the physical characteristic
11 is normal to the hodograph characteristic I . This has obvious
practical implications for the graphical construction of the charac-
teristic nets.
HOMOGENEOUS EQUATION WITH A, B, C AS FUNCTIONS ONLY OF rp,

A N D rp.,Under these conditions, Eqs. A.18 indicate that the hodograph
characteristics are completely independent of the physical coordinates.
That is, they may be found by direct integration of Eqs. A.18 without
the need for simultaneously constructing the physical characteristics.
The hodograph characteristics may be found once for all from Eqs. A.18
and then may be used for all problems, irrespective of the nature of the
flow in the physical plane. The physical characteristics are then found
for each particular problem, starting with the given initial conditions,
from Eqs. A.17. The construction is facilitated by the orthogonality
relations of Eqs. A.19, especially inasmuch as the hodograph character-
istic curves are known beforehand.
Important applications include (i) two-dimensional, steady, super-
sonic flow (Chapter 15), and (ii) non-steady motion in constant-arca
ducts (Volume 11, Chapter 24).
A.3. Method of Constructing Characteristic Curves

Initial-Value Theorem. We shall now prove a theorem which is basic
both to an understanding of the physical significance of the characteris-
tics and to the method of employing the characteristics for practical
computations. This theorem is related to the basic initial-value prob-
lem of hyperbolic differential equations. It states that if the values of
rp, and rp, are given at all points of a noncharacteristic curve 1-2' in the
x, y-plane (Fig. A . l a ) , then the entirejlow pattern i s determined (and may
be computed) within the quadriluteral formed by the characteristic curves
passing through points 1 and 2'. By "noncharacteristic" is meant that
the curve 1-2 is nowhere tangent to a characteristic curve. The demon-
stration of this theorem will a t the same time illustrate a general method
of stepwise computation which may be used for solving numerically any
problems of the types under discussion.
General Numerical Solution. Along the noncharacteristic curve 1-2

in Fig. A.la we suppose that we know the values of qz and rp, (e.g., the
velocity components) a t each point. The image curve 1'-2' may there-
fore be constructed in the hodograph plane (Fig. A.lb).
The method of solution involves stepwise construction of a finite
n w h e r of characteristic curves forming a characteristics net or mesh.
Accordingly, we select a number of discrete points along 1-2 as points

through which these characteristics pass. For illustrative purposes we
take four points, namely, 1, 2, 3, and 4. These in turn have four corre-
sponding hodograph points, l', 2', 3', and 4'.
The characteristics will form a network of the general form shown in
Figs. A.la and A.lb, where solid lines indicate characteristics of family I
and dashed lines characteristics of family II. The problem is to deter-
mine this network, for, when this is done, the values of qz and qg are
X
(4
FIG.A.1. Characteristics net.
(a) Physical plane.
known a t each point of the physical plane and consequently the entire
flow pattern in terms of the velocity distribution is established. In the
stepwise construction the curvilinear characteristics are replaced by
straight-line segments connecting the lattice points of the network.
To establish a point such as 7 and its hodograph image 7', we proceed
as follows:
(i) Using Eq. A.l7a, the value of (dyldx)~a t point 4 is computed,
using the values of A, B, and C corresponding to point 4. A line with
the slope (dyldx)~is tentatively passed through point 4.
(ii) Using Eq. A.l7b, the value of ( d y l d z ) ~a ~
t point 1 is computed,
using the values of A, B and C corresponding to point 1. A line with the
slope ( d y l d x ) ~is~tentatively passed through point 1. Its intersection
with the line drawn in step (i) establishes, a t least approximately, the
location of point 7.
(iii) The value of (dqg/dcpz)r at point 4' is computed from Eq. A.18a,
and a line is tentatively drawn through 4' with the slope (dvu/dqz)r.
APP. A CONSTRUCTING CHARACTERISTIC CURVES 605
In computing the value of ( d y / d i , ~ in
~)~Eq. A.18a1 we use the approxi-
mate finite-difference form
The value of y, - y4 is known approximately from step (ii), and the

value of (rp2)7rmust, for the present, be estimated roughly and subse-
quently be improved as indicated later.
(iv) After a computation similar to that of step (iii), a line through
point 1' is tentatively drawn with the value of the slope (drpy/dcp,)II
corresponding to 1'. The intersection of this line with that of step (iii)
tentatively establishes point 7'.
(v) An iteration process is used to establish points 7 and 7' more
accurately. Starting with the tentative locations found previously, the
location of point 7 is found more accurately by drawing the lines through
points 1 and 4 with slopes corresponding to the mean conditions between
points 1 and 7, and between points 4 and 7, respectively. Likewise,
point 7' is found more accurately by drawing the lines 1'-7' and 4'-7'
with their respective mean slopes. The iteration is continued until
satisfactory convergence has been attained.
To summarize, points 7 and 7' may be found from information con-
cerning s, y, qz, and i , ~ , a t points 1 and 4. Points 6 and 6' may be
found in like manner from information a t 3 and 4; and points 5 and 5'
from 2 and 3. Proceeding in this way, points 8 and 9 are then found
from 5, 6, and 7, and finally point 10 is found from 8 and 9. The en-
tire flow pattern in the region 1-2-10 is thus established. It is evident
that the flow pattern in the region 1-2-11 may also be calculated by a
similar procedure.
We have therefore proved that, given the values of qz and 9, along
the noncharacteristic curve 1-2, the entire flow pattern may be deter-
mined within the characteristic quadrilateral 1-10-2-11.
Other Types of Initial Data. If the initial-value curve 1-2 is a t any
point tangent to a characteristic, the construction described above will
not yield results near the point of tangency. Hence the curve 1-2 must
be noncharacteristic in order that the solution in the characteristic
quadrilateral be c~mplet~ely determined by the data along 1-2.
I n addition to the one shown in Fig. A.1 there are several other initial-
value problems of practical importance for which the method of charac-
teristics yields solutions in certain regions:
Fig. A.2a. If the values of (p, and (p, are given along two intersecting
characteristics, 1--2 and 1-3, then the type of computation described
previously leads to a solution within the characteristic quadrilateral
1-2-34.
Fig. A.2b. The values of cpz and (p, are given along the noncharac-
teristic curve 1-2, and either the magnitude or the direction of the veloc-
ity vector is given along the curves 1-4 and 2-3 (note that if curves 1-4
or 2-3 were solid boundaries, then the direction of the velocity would be
known along these curves; or if

Y the curves 1 4 or 2-3 were the
boundaries of a free jet exposed
to a known pressure, then the
magnitude of the velocity would
be known along these curves).
' With these initial data, a solution
may be obtained in the region
where the characteristics net may
be constructed, namely, the fig-
ure 14-6-3-2-5.
1
Fig. A.2c. The values of (p,
P
x and cp, are given along the char-
(4 acteristic curve 3-5, and either
Fra. A.2. Various types of initial-value the magnitude or direction of the
problems.
velocity vector is given along the
noncharacteristic boundary 1-2. By constructing the characteristics
net, beginning with these initial data, the solution may be uniquely
determined in the region 3 - 5 4 .
A.4. Simple Waves

A degenerate type of flow known as a simple-wave $ow plays a
curiously important role in practical problems and occurs far more
APP. A SIMPLE WAVES 607
frequently than one would superficially expect for a flow so special in
nature. Simple waves occur only when the coefficient D in Eq. A . l l is
zero and when A , B, and C depend only on rp2 and (p,. These conditions
will henceforth be assumed.
Definition of Simple-Wave Flow. We shall define a region of simple-

wave flow as a region where cp, and rp, are not independent but are
rather unique functions of each other. The implications of this are
found by examining the differential equation A . l l . I n the light of the
assumption that pz and rp, depend only on each other, the derivatives
in Eq. A . l l may be expressed as
Substituting these into Eq. A . l l , with D = 0, we find that
Multiplying through by drpu/drpzand then solving the quadratic equa-

tion for the latter, we obtain two possible solutions:
Hodograph of Simple Wave. Comparing this with Eqs. A.18, we see

that the relation between p, and py in a simple-wave flow is the same as
that on one of the fixed hodograph characteristics. The term "fixed"
hodograph characteristics is used here because, as shown previously,
the hodograph characteristics are immediately integrable when D = 0,
irrespective of the particular physical problem. Consequently, a physi-
cal region of simple-wave flow is mapped in the hodograph plane as a
single characteristic curve. Such a flow is singly degenerate inasmuch as
a line on the physical plane maps as a point on the hodograph plane,
and an area on the physical plane maps as a line on the hodograph plane.
Simple Wave in Physical Plane. Let us now investigate the nature

of a line of constant cp, in the physical plane. Since cp, is related uniquely
to rp, in a simple-wave flow, the line is one of constant r p , as well as of

constant rp,, and therefore all fluid properties will be constant on such
a line. The increment in rp, may be written in the usual manner in terms
of dx and dy; employing Eqs. A.20 in addition, we obtain
On a line where cp, is constant, drp, = 0. On such a line, therefore, the

foregoing equation yields
Keeping in mind that drpv/drp, is the slope of a hodograph charac-

teristic, a comparison of this equation with Eqs. A.19 leads to the con-
clusion that the lines of constant fluid properties in the physical plane
are identical with one family of physical characteristics. Examination of
Eqs. A.17 shows that the physical characteristics under consideration are
straight lines; this follows because A, B, and C depend only on pz and
P ~and
, the latter in turn are constant on the physical characteristics.
To sum up, a special type of flow is possible when D = 0 and when
A, B, and C depend only on q, and rp,. In this simple-wave flow one
family of physical characteristics consists of straight lines, with uniform
fluid properties on each such characteristic. Each of the physical charac-
teristics maps as a point on the hodograph plane. The locus of these
points is the hodograph characteristic of opposite family.
Auxiliary Theorem. An important theorem (I) which follows from the
foregoing results states that if a region of constant state (i.e., unifmm
properties throughout) adjoins a region of nonuniform $ow, then the latter
must be a simple-wave $ow. This may be proved by noting that the two
regions must be joined on a physical characteristic, and that a physical
characteristic in a uniform flow must be a straight line. The significance
of this theorem lies in the fact that regions of uniform flow are almost
always present in practical applications.
I n two-dimensional, supersonic, simple-wave flow one family of Mach
lines is straight and the fluid properties are uniform on each such Mach
line. If the Mach lines of family 11 are straight, then the entire flow
is mapped on a hodograph characteristic of family I, and vice versa.
Such a flow is called a Prandtl-Meyer flow. A region of uniform, paral-
lel flow can only be adjoined to a Prandtl-Meyer flow.
APP. A REFERENCES AND SELECTED BIBLIOGRAPHY
R E F E R E N C E S AND SELECTED BIBLIOGRAPHY

1. C OURANT, R., and FRIEDRICHS. K. 0. Supersonic Flow and Shock Waves. New
2. D A L I ~R., H. Some Jfalhemntical Aspects of Compressible Flow. Report ACA-20
of the Australian Council for Aeronautics (1946).
3. I SENBERG, J. S., and L IN, C. C. The Method of Characteristics in Compressible
Flow, Part I: Steady, Supersonic Flow, Air Materiel Command Tech. Report,
NO. F-TR-1173A-ND (1947).
4. ROBERTS,R. C., and LIN, C. C. The Method of Characteristics in Compressible
Flow, Part 11: Unsteady Flow, Air Materiel Command Tech. Report, No. F-
TR-1173-ND (1947).
Appendix B
TABLES OF COMPRESSIBLE-FLOW FUNCTIONS
Collected in this Appendix are Tables B.l to B.8 for facilitating numerical
calculations.
Listed below are (i) an index to the tables, (ii) nomenclature for the tables,
(iii) the sources of the tables, and (iv) a bibliography of additional tables and
charts available in the literature.
INDEX TO TABLES
TABLE PAGE
B.1. Properties of the Standard Atmosphere . . . . . . . . . . . . . 612
B.2. Isentropic Flow, k = 1.4 . . . . . . . . . . . . . . . . . . . 614
B.3. Normal Shock, k = 1.4 . . . . . . . . . . . . . . . . . . . . 621
B.4. Frictional, Adiabatic, Constant-Area Flow (Fanno Line), k = 1.4 . 626
B.5. Frictionless, Constant-Area Flow with TO-Change (Rayleigh Line),
k=1.4.. . . . . . . . . . . . . . . . . . . . . . . . . 628
B.6. Influence Coefficients, k = 1.4 . . . . . . . . . . . . . . . . . 630
B.7. Hodograph Characteristic Functions for Two-Dimensional, Isen-
tropic, Supersonic Flow, k = 1.4 . . . . . . . . . . . . . . . 632
B.8. Coefficients for Second-Order Theory of Two-Dimensional, Supersonic
Flow, k = 1.4 . . . . . . . . . . . . . . . . . . . . . . . 633
NOMENCLATURE FOR TABLES

B.1. See footnote to Table B.1. B.5. See Chapter 7.
B.2. See Chapter 4. B.6. See Chapter 8.
SOURCES OF TABLES
B.1. T HE STAFF O F THE AMES 1- B Y %FOOTSUPERSONIC WIND T UNNEL SECTION.
Notes and Tables for Use in the Analysis of Supersonic Flow, NACA Tech.
Note, No. 1428 (1947).
B.2 to B.6. SHAPIRO, A. H., HAWTHORNE, W. R., and EDELMAN, G. M. The Mechan-
ics and Thermodynamics of Steady One-Dimensional Gas Flow with Tables
for Numerical Solutions, Meteor Report, No. 14, Massachusetts Institute
of Technology Guided Missiles Program, Cambridge, Mass. (1947).
B.7. STAFF OF THE G AS TURBINE LABORATORY, Massachuset.ts Institute of Tech-
nology, Cambridge, Mass. (1948).
B.8. Same source as for Table B.1.
610
APP. B BIBLIOGRAPHY OF TABLES AND CHARTS 61 I
BlBLIOGRAPHY OF PUBLISHED TABLES AND CHARTS

1. DAILEY, C. L., and WOOD, F. C. Computation Curves for Compressible Fluid
Problems. New York: John Wilev & Sons. Inc.. 1949.
2. EMMONS, H. W. Gas Dynamics l'abl&for ~ i r .N&V York: Dover Publications,
1947.
-.
- - ~
3. KEENAN, J. H., and KAYE, J. Gr1.9 T a b h . New York: John Wiley & Sons, Inc..
1948.
4. SHAPIRO, A. H., HAWTHORNE, W. R., and EDELMAN, G. The Mechanics and
Thermodynamics of Steady One-Dimensional Gas Flow, in Handbook of
Supersonic Aerodynamics, NAVORD Report 1488 (Vol. I), Superintendent of
Documents, U. S. Govt. Printing Office (1950). Seealso EDELMAN, G . M., and
SHAPIRO, A. H. Tables for Numerical Solution of Problems in the Mechanics
and Thermodynamics of Steady, One-Dimensional Gas Flow Without D i s
continuities, Jour. App. Mech., Vol. 14, No. 4 (1947), p. A-344.
5. KOPAL, Z. Table of Supersonic Flow Around Cones, Tech. Rep., No. 1, M.I.T.
Center of Analysis, Mass. Inst. of Tech., Cambridge (1947).
6. KOPAL,2. Tables of Supersonic Flow Around Yawing Cones, Tech. Rep.,
No. 3, M.I.T. Center of Analysis, Mass. Inst. of Tech., Cambridge (1947).
7. KOPAL,Z. Supersonic Flow Around Cones of Large Yaw, M.I.T. Center of
Analysis, Mass. Inst. of Tech., Cambridge (1947).
8. ISENBERG, J. S. Monograph 11: The Method of Characteristics in Compressible
Flow; Part IA: Tables and Charts, Air Maten'el Command Report, No. F-TR-
1173-ND (Dec., 1947).
9. R O C K E J.~ ,A,, and HAYES, W. D. Monograph 11. The Method of Character-
istics in Compressible Flow; Part IC: Two-dimensional flow with large entropy
changes, Air Materiel Command Report, No. 102-AC49/6-100 (Dec., 1949).
10. NEICE, M. M. Tables and Charts of Flow Parameters Across Oblique Shocks,
NACA Tech. Note. No. 1673 (1948).
11. HUCKEL,V. Tables df ~ y ~ e r g e o h e t rFunctions
ic for Use in Compressible-Flow
Theory, NACA Tech. Note, No. 1716 (1948).
12. BURCHER. M. A. Compressible Flow Tables for Air, NACA Tech. Note, No.
1592 (1948).
13. T URNER , L. R., ADDIE,A. N., and ZIMMERMAN, R. H. Charts for the Analysis
of One-Dimensional Steady Compressible Flow, NACA Tech. Note, No. 1419
(1948).
14. AMES AERONAUTICAL LABORATORY. Notes and Tables for Use in the Analysis of
Supersonic Flow, NACA Tech. Note, No. 1428 (1947).
15. IVEY, H. R., STICKLE, G. W., and S C H U E ~ L EA. R , Charts for Determining the
Characteristics of SharpNose Airfoils in Two-Dimensional Flow a t Super-
sonic Speeds, NACA Tech. Note, No. 1143 (1947).
16. MOECKEL, W. E., and CONNORS, J. F. Charts for the Determination of Super-
sonic Air Flow Against Inclined Planes and Axially Symmetric Cones, NACA
Tech. Note, NO. 1373 (1947).
17. ORDNANCE AEROPHYSICS LABORATORY. Compressible Flow Tables and Graphs, in
Handbook of Supersonic Aerodynamics, NAVORD Report 1488 (Vol. 2),
Superintendent of Documents, U. S. Govt. Printing Office (1950).
18. ROSENHEAD, L. (ed.). Selection of TabEm for TJse in Calculations of Compressibk
Airflow. New York: Oxford Uiiversity Press, 1952.
TABLES OF COMPRESSIBLE-FLOW FUNCTIONS APP. B
TABLE B.l
PROPERTIES OF THE STANDARD ATMOSPHERE
-
Svmbols: See end of Table B.1.
App. B PROPERTIES OF THlE STANDARD ATMOSPHERE 613
TABLE B.1. PROPERTIES OF THE STANDARD ATMOSPHERE (Continued)
Symbols: h = height above sea level, f t

t = temperature, degrees F
c = speed of sound, ft/sec
p = pressure, ibf/ft2
3
p = mass density, slug/ft
p = coefficient of viscosity, slug/ft see
v = @lp,kinematic viscosity, ft2/sec
TABLES OF COMPKESSIBLEFLOW FUNCTIONS APP. B
TABLE B.2
ISENTROPIC FLOW
Perfect Gas, k = 1.4
-
Notes: (1) For values of M from 0 to 5.00, all digits to the left of the comma are
valid for linear interpolation. Where no comma is indicated in this
region, all digits are valid for linear interpolation.
(2) The notation .Ox429 signifies ,000429. The notation 53704 signifies
5,370.000.
APP. B ISENTROPIC FLOW 615
TABLE B.2. ISENTROPIC FLOW (Continued)

See Notes at beginning of this table.

616 TABLES OF COMPRESSIBLE-FLOW FUNCTIONS



4.t M* PlPo P/PO TITO A/AS F/F* A P

AX'%
-
618 TABLES OF COMPRESSIBLE-FLOW FUNCTIONS APP. B
TABLE B.2. ISENTROPIC FLOW (Contznued)

---
See Notes at beginning of this table. ,
TABLE B.2. ISENTROPIC FWW (Cmtinued)

Perfect Gas. k = 1.4

620 TABLES OF COMPRESSIBLEFLOW FUNCTIONS APP- B
TABLE B.2. ISENTROPIC FLOW (Concluded)

M M* P/PO P /PO T/To A/A' F/F* p

A .P
See Notes a t beginning of this table.

NORMAL SHOCK
TABLE B.3
NORMAL SHOCK
Notes: (1) For values of M from 1.00 to 3.00, all digits to the left of the comma are
valid for linear interpolation. Where no comma is indicated in this
region, all digits are valid for linear interpolation.
622 TABLES OF COMPRESSIBLEFLOW FUNCTIONS APP. B
TABLE B.3. NORMAL SHOCK (Continued)

--
See Notes at beginning of this table. >

NORMAL SHOCK
TABLE B.3. N ORMAL SHOCK (Continued)

See Notes at beginnine: of this table. -2

624 TABLES OF COMPRESSIBLEFLOW FUNCTIONS APP. B
TABLE B.3. N ORMAL S HOCK (Continued)

-~

NORMAL SHOCK
TABLE R.3. N ORMAL SHOCK (Concluded)


TABLES OF COMPRESSIBLEFLOW FUNCTIONS APP. B
TABLE B.4
FRICTIONAL, ADIABATIC, CONSTANT-AREA FLOW (FANNOLINE)
~-
v/v*
h4 T/T* P/P* PO/PO* and F/F* 4.fLmdD
P*/P
APP. B FANNO LINE 627
FLOW (FANNO
TABLE B.4. FRICTIONAL, ADIABATIC, CONSTANFAREA L INE)
(Concluded)
,
TIT* PIP* PO/PO* and FIF* 4fLrnanlD
P*/P
TABLE B.5
FRICTIONLESS, CONSTANT-AREA FWWWITH C HANGE IN STAGNATION TEMPERATURE
(RAYLEIQH LINE)
P*/P
and
v/v*
APP- B RAYLEIGH LINE 629
TABLE B.5. FRICTIONLESS, CONSTANT-AREA F LOW WITII C HANGE IN STAGNATION

T EMPERATURE (RAYLEIGIII L INE) (Cmcluded)
630 TABLES OF COMPRESSIBLEFLOW FUNCTIONS APP. 13
TABLE B.6
-"
and an2
Fw/2 -Fw
The notation .03429signifies 0.000429. The notation 63704 signifies 5,370,000.

APP. INFLUENCE COEFFICIENTS
TABLE B.6. INFLUENCE COEFFICIENTS (Concluded)

k = 1.4
M M~ 1 - M~ FA Ff and an'd
Fw/2 -Fw
5.0625
5.2900
5.5225
5.7600
6.0025
6.2500
6.5025
6.7600
7.0225
7.2900
7.5625
7.8400
8.1225
8.4100
8.7025
9.00
12.25
16.00
20.25
25.00
36.00
49.00
64.00
81.00
LOO. 00
m
TABLE B.7
(aand a are in degrees)
w a a - o a a - w M
APP. B COEFFICIENTS FOR SECOND-ORDER THEORY 633
TABLE B.8
COEFFICIENTS FOR SECOND-ORDER THEORY OF TWO-DIMENSIONAL, SUPERSONIC
FLOW, k = 1.4
INDEX
(References are to pages of Vol. I only.)
Absolute temperature scale, 32 thickness drag of, 443
Accelerating flows, 507 trailing-edge flow of, 584
Acoustics, 48 two-dimensional, 567-80
Action. zone of. 500 types of, 575
~ d i a b a t i cellipse, 58-59 wake of, 578
Adiabatic flow wave drag of, 443
critical velocity for, 79-81 thin, supersonic flow for, 427
dimensionless velocitv M* of. 81-82 triangular profile, 574
energy equation of, 80-81 ' two-dimensional, 567-80
flow per unit area of, 82 Analogue techniques, 289
maximum velocity for, 79-81 Angular velocity, 269
one-dimensional, 160-1 78 Area change, flow with, 219 55
of perfect gas, 78-82 Area change and friction, flow with,
a t constant area, 162-73 24142
stagnation-temperature ratio for, 80 Area ratio, for isentropic flow, 85-86
Adiabatic Drocess. definition of. 29 Aspect ratio, effect on compressibility,
entropy chan g e' for, 34 400, 411-15
Adiabatic wall temperature, 211-12 generalized correlation of. 422
Affine transformation. 319 ~ r m o s ~ h e r&operti%
e, of, 612
Air, properties of, 612 Atomic explosion, 211
Airfoils; see also Wings Axially symmetric flow, 291,294-95,290,
biconvex, 568 596
completely symmetrical, 574
double-wedge, 568 Ballistics, 49
lift-drag polars of, 571 Barotropic fluid, 278
shock-characteristic theory of, 566 Bend. concave. 583
subsonic ~ e r n o u l lconstant,
i 276
pressure distributions for, 379-81 Bernoulli's equation, 276, 278
a t supersonic speeds, 571 Biconvex airfoil. 568
thickness of. 384-87 ~ i p l a n e supe&nic,
, 455, 579
in wind tunhel, 378 Body, free, 11
without stagnation points; see Bump Body, isolated, 11
supersonic Body forces, 18
aerodvnamicallv " ootimum.
* 575 Rodv of revolution: see Revolution.
biplane, 579 body of
center of pressure of, 444, 446, 574 Bomb, drag of, 408
downwash of. 576 Borda re-entrant orifice. 111
drag of, 442, ,446; 572 Boundary layer, 10, 26546, 282, 379
experimental data for, 4 4 7 4 8 displacement thickness of, 266
flow past, 501 and interaction with shocks, 135-37,
induced drag of, 442 266, 381-82, 580-84
lift of, 441, 445, 472 and pressure distribution, 308
lift-drag polar of, 446 separation of, 379, 380, 415
lift-drag ratio of, 443 Boundary of system, 11
linearized theory of, 4 4 0 4 9 Bump, flow past, 350, 372-74
of minimum thickness drag, 575 Burning, slow, 207-11
moment of, 444, 572
performance of, 445 Camber, in subsonic flow, 384-87
pressure drag of, 442 Cancellation of waves, 450, 452, 493
second-order theory of, 571-78 Cauchy-Riemann equations, 351
shape of, 571 Cavitation. 475
skin-friction drag of, 443 Center of ;ressure, of supersonic airfoils,
theory vs. experiment, 575 446, 572
INDEX
Centered waves, 474; see also Corner Compressibility burble, 326, 384
flow; Prandtl-Meyer wave effect. of aspect ratio on, 400
streamlines of, 525 Compressibility of liquid, 69
Chapman-Jouguet rule, 210-11 Compressible flow
Characteristic coordinates, 478 classification of, 58-59
Characteristic curves, 482-86, 598; see hypersonic, 59
also Method of characteristics subsonic, 59
eauations of. 602 supersonic. 59
hodograph, 599 transonic, 59
initial-value theorem of, 603-6 Compression shock; see Shocks
method of constructine. 603-6 Compression wave, steepening of, 127-3 1
orthogonality relationyor, 485, 603 Compressor, interference in, 315
physical, 599 Compressor cascades, 427, 462
properties of, 601 supersonic, 505
and shocks. 601 Concave bend, 583
as solut,ions, 601 Condensation
Characteristic function flow with, 219-55
series exoression of. 561 moisture. flow with: see S i r n ~ l eTo-
table of,'632 Change
Characteristic quadrilateral, 498, 499 Condensation shock, moisture, 203-5
Characteristics; see also Hodograph char- Conduct,ion of heat, Fourier law of, 36
acteristic: Method of characteristics Conductivitv. thermal. definition of. 36
hodograph, '467-68, 478, 485 ~onservatioh'of ma&; see also 'Con-
table of, 468 tinuity equation
working chart of, 467 principle of, 12
method of, 462-523; see also Char- for control volume, 13
acteristic curves; Method of char- Continuity equation, 14, 283-84; see also
acteristics Conservation of mass
illustrative example of, 479, 486-89 in cylindrical coordinates, 290
rules for using, 478 in spherical coordinates, 299
unit processes for, 492-94 in steady flow, 15
physical, 484 Continuum
physical significance of, 499 concept of, 4, 23
theory of, 482-91, 595-608 domain of, 56-58
Chemical reaction, flow with, 206-11, Euler's description of, 11
219-55; see also Simple To-Change Lagrange's description of, 11
Choking, 234-36, 257 limitations of, 56-58
due to heating, 192-93, 201-3, 205 mathematical description of, 10
due to Simple To-Change, 192-93, properties of, 5, 35
201-3, 205 Control surface, 12
in explosion wave, 210-11 Control volume, 12
frictional, 162, 170-78 First Law of thermodynamics for, 39
for isentropic flow, 89-94 Second Law of thermodynamics for, 40
for isothermal flow. 183 Converging nozzle
rules of, 235-36 discharge coefficient of, 93
thermal, 192-93, 201-3, 205 as flow meter, 93
Circular-arc profile, in subsonic flow, 375 as flow regulator, 93
Circular cvlinder with varying pressure ratio, 91-93
in subsonic flow, 367 Converging-diverging nozzle
in free jet, 368 operat,ing characteristics of, 1 3 9 4 3
in wind tunnel, 357-58, 367 with varying pressure ratio, 93-94
Circulation, 26749, 278-80, 284 Cooling, flow with; see Simple To-Change
Clausius, inequality of, 32 Corner flow, supersonic, 46344; see also
Coefficient Prandtl-Meyer wave; Simple-wave;
of discharge; see Discharge coefficient Centered waves
of friction; see Friction coefficient Critical Mach Number; see Lower critical
of heat transfer, 212-13 Mach Number
influence; see Influence coefficients Critical point,, 258
pressure; see Pressure coefficient Critical pressure ratio, 76
Combustion Crit,ical speed, 79-81
choking in, 198,201-2 Critical stsate,83
example of, 197-98 Critical velocity, in adishatic flow, 79-81
flow with, 206-11; see also Simple Crocco's thcorcm, 281-82
7'0-Change Crossing of waves; see Intersection of
Combustion chamber, 219 waves
INDEX 637
Curved shocks. 549. 55941: see also of spheres, 410

Shocks for subsonic flow, 386
Curved wall, supersonic flow past, 464 of supersonic airfoils, 442, 446, 572
Curvcs, characteristic, 598; see also Char- thickness, 443
acteristic curves: Method of charac- wave, 443
teristics Drag coefficient, 163, 379; see also Drag
Cycle, 24 Drag divergence, Mach Number of,
First Law for a, 28 384-87
Duct
d'Alembert Paradox, 449 flow in
d1Arcy's formula, 189 fed by converging nozzle, 173-75
Decelerating flows, 507 fed by converging-diverging nozzle,
Deflagration, 207-11 175-78
Dependence, domain of, 499 with friction, 159-86
Detached shocks, 548 isothermal, 178-83; see also Iso-
Detonation, 207-11 thermal flow in ducts
Detonation wave, 219 with varying pressure ratio, 173-
Diffuser 78
isentropic flow in, 73
oblique-shock, 566 Efficiency
supersonic; see also Inlet, supersonic of nozzle, 98-99, 110
of air-breathing engines, 147-51 of supersonic diffuser, 151-52
allowable contraction for, 144-51 thermal, of heat engines, 29
devices for starting of, 147 Ejector. 219
efficiency of, 151-52 Elbow
of fixed geometry, 143-52 for accelerated flow, 3 5 9 4 0
friction in, 147 supersonic, 504
limiting contraction ratio of, 149 Ellipse, adiabatic, 58-59
one-dimensional, 143-52 Ellipsoid, subsonic flow past, 398-401
operating conditions of, 14-7 Ellipsoid of revolution
overspeeding of, 150 lower critical Mach Number of, 407
special problems of, 1 4 3 4 4 subsonic flow past, 405
starting of, 144-51 Elliotic cvlinder
of variable geometry, 147-52 lirt of, $76
for wind tunnels, 144-47 lower critical Mach Number of, 407
Dimensional analysis, 315 subsonic flow past, 354-56, 358, 369
of inviscid flow, 308-9 Elliptic differential equation, 420, 600
Dimensions, of dynamical quantities, 20 Energy
Direction index, 450 internal, 27
Discharge coefficient for perfect gas, 41
of Bordn orifice, 111 U , for pure substance, 35
of nozzle, 90-100 kinetic, 36
of orifice, 359 flux of, 39
of sharp-edged orifice, 100 potential, 36
Discontinuity, surface of; see Slip line; Encrgy effects, flow with; see Simple
Vortex sheets 7'0-Change
Displacement thickness, 266 Energy equation
Distorted models, 315 in kinematic form, 80-81
Disturbances, propagation of, 499 for steadv flow. 3 9 4 0
Domain of dependence, 499 Engines. heat. 29'
Domain of influence, 499 Enchalpy
Double wedge definition of. 36 .-
lift-drag ratio of, 446 flux of, 39 '
~
supersonic flow past, 444, 568 of perfect gas, 41, 43

Drag, 449; see also Airfoils; Drag coeffi- stagnation, 75
cient Entropy
and aspect ratio, 401 definition of, 33
of bomb, 408 evaluation of, 34
of flat plate normal to stream, 360 increase of
induced. 442 principle of, 34
of &iAgs, 412 in shock, 119-20
internal, flow with, 219-55 in weak shock, 125
and Mach Number. 383 of perfect gas, 43
Equality of temperature, 24
Equal-temperature process, 25
INDEX
Equation of continuity; see Continuity Flow with waves of one family; see
equation Corner flow; Prandtl-Meyer flow;
Equations of motion, 265-97 Simple waves
Euler's, 273-78; see also Euler's equa- Fluid
tions barotropic, 278
linearized. 428 definition of, 3
Equipotenti'al lines homogeneous, 278
for flow past circular cylinder, 297 inviscid, 10
and streamlines. 295-97 Newtonian. 10
Euler. method of, '11 perfect, 10'
Euler's equations Fluid curve, 269
in Cartesian coordinates, 273-75 Fluid line. 269
integration of Fluid rot$tion. 269-73
for irrotational flow, 277-78 Flux
along a streamline, 275-77 of enthalpy, 39
in streamline coordinates, 281 of kinetic energy, 39
Evaporation, flow with, 219-55; see also of mass. 14
Simple To-Change of moment of momentum, 20
Exchanger, heat; see Heat exchanger of momentum, 17
Expansion wave, spreading of, 128-31 Force, units of, 20-21
Explosion Force break, 383-84
atomic, 211 Force-divergence Mach Number, 384-87
hvdrocarbon. 206-7 Forces
surf%e, 18
Fanno line, 114-15, 160-61, 192, 240 types of, 18
eouation of. 155 Fourier law of heat conduction, 36
tible of. 626 Friction
Field method, 491 one-dimensional flow with
Fields in su~ersonicflow. 477 adiabatic, 159-86, 2 1 S 5 5
Film coeffiiient of heat: transfer. 212- choking effects in, 170-78
curves for, 169
~ i r s t L a wof thermodynamics, 26 example of, 169-70
for control volume. 37-39 formulas for, 16649
for a cycle, 28 isothermal, 178-83; see &o Iso-
Flame front, 219 thermal flow in ducts
propagation speed of, 207, 209-11 laminar, 189
Flat late limiting duct length for, 170-71
noimal to subsonic stream, 360 in lone ducts. 173-78
subsonic flow past, 448 of perrect gas. 162-73
supersonic flow past, 444, 448 tables for, -169, 626
lift-drag ratio of, 446 simple, 159-78; see also Friction, one-
Fliegner's formula, 85 dimensional flow with
Flow; see also One-dimensional flow Friction and area change, flow with, 241-
accelerating, 507 42
adiabatic; see Adiabatic flow ~ r i c t & and heat transfer, flow with,
decelerating, 507 242-55
irrotational; see Irrotational motion Friction coefficient, definition of, 163
isentropic; see Isentropic flow experimental, 1 8 4 8 6
mass rate of, 14 for subsonic flow, 184-85
onedimensional for supersonic flow, 185-86
adiabatic, 160-78; see also Friction, Friction factor; see Friction coefficient
one-dimensional flow with, adi- Frictional effects on gas flow; see Fric-
abatic tion, one-dimensional flow with
with friction. 159-86; see also Fric- Frictional flow, table of, 626
tion, one-dimensional flow with Frictional heating, 211-12
potential; see Potential flow Frictionless flow, dimensional analysis of,
rotational; see Rotation; Rotational 308-9
flow
Flow coefficient, of nozzle, 99-100 Gas, perfect; see Perfect gas
Flow per unit area Gas constant, definition of, 41
isentropic, 84-85 and molecular weight, 41
maximum, 85 universal, 41
Flow with friction; see Friction, one- Gas constants, relations among, 42
dimensional flow witk Gas dynamics, 49
INDEX
Generation of wave, 449, 463, 493 Incompressible flow

Gladstone-Dale equation, 63 Bernoulli's equation for, 278
Gothert's rule, 316-21, 336 similarit,~law for, 321-22
for bodies of revolution, 402-5 square network for, 297
for one-dimensional How, 331-32 velocity potential for, 289
for spheres, 409 Index of refraction, 63
for sweptback wings, 416 Induced drag, 442
for three-dimensional flow, 394-99 of wings, 412
for two-dimensional passages, 330- Influence
32 domain of, 499
fo; wind tunnels. 327-30 range of, 500
for wings, 410 ' Influence coefficients, 226-31
Green's theorem, 268-69 table of, 228, 231, 237-38, 630
Initial-value theorem
Heat, definition of, 27 for hyperbolic equations, 603-6
mechanical equivalent of, 28 for supersonic flow, 497-500, 603-6
in reversible processes, 33 Injector, 219
specific, 36 Inlet, supersonic, 506; see also Diffuser,
of perfect gas, 42 supersonic
Heat conduction, Fourier law of, 36 of fixed geometry, 148-51
Heat engines, 29 limiting contraction.ratio for, 149
thermal efficiency of, 29 overspeeding of, 150
- . 214-55: see also Heat
Heat exchanger. starting of, 147-51
transfer of variable geometry, 150-51
approximate method for low Mach Integral surface, 598
Numbers. 249-51 Interference in wind tunnel, 315,335,403
with constant heat flux. 24647. 251- Interference pressure coefficient, 329
52 Interferogram
-.
with constant wall temperature, 243-
01
evaluation of, 62-65
fringe-displacement, 63-65
graphical integration for, 2 4 4 4 8 infinite-fringe, 61-63
in nuclear reactor, 252-55 Interferometer, 60-65
numerical integration for, 2 4 4 4 8 Internal energy, 27
Heat transfer; see also Heat exchanger of perfect gas, 41, 43
coefficient of, 212-13 U , for pure substance, 35
experimental correlation of, 213 Intersection of oblique shocks, 553-59
flow with, 219-55; see also Simple of same family, 558
TO-Change Intersection of waves, 450, 453, 494
Heat transfer and friction, flow with, Inviscid flow, dimensional analysis of,
242-55 308-9
Heating Inviscid fluid, definition of, 10
flow with; see Simple TO-Change Irreversibility, 30
frictional, 211-12 and perpetual motion, 31
Hodograph characteristic, 468, 478, 485 and Second Law, 31
related to shock polar, 552 Irrotational How; see Irrotational motion
Hodograph characteristic curves, 599 Irrotational motion, 266, 267-273, 280,
Hodograph characteristic function, 467 282, 284, 286
chart, of. 467
- - - - in axially symmetric flow, 294
table of: 468, 632 in cylindrical coordinates, 290
Hodograph coordinates, 338 example of, 271-73
Hodograph equations, 3 3 8 4 1 in spherical coordinates, 299
Hodograph method, 288, 336 Isentropic flow
Hodograph plane, 336, 465 applications of, 73
equation of stream function in, 341 charts for, 87
Hodograph shock polars, 542, 552 choking in, 89-94
Hodograph streamlines, 336-37 of compressible liquid, 11
and physical streamlines, 336-37, 341 with deviations from perfect gas l a m
Homogeneous fluid, 278 -- -.
9.%97
Hyperbolic differentialequation, 429,600 in diffuser, 73
Hypersonic How, 59 formulas in powers of Mach Nu,
--
QFi
Impulse function general features of, 74-78
definition of, 86 illustrative example of, 88
for isentropic flow, 86-87 for low Mach Numbers, 94-95
Impulse turbine blade, 504 Mach Number relations for, 83-8
INDEX
Isentropic flow-Continued and Mach Number, 315, 326, 376, 383

in nozzle, 73 Pranrltl-Glauert rule for, 322-23
one-dimensional, 74-78, 239 for subsonic flow, 385
of perfect gas, 83-88 of supersonic airfoils, 441
in stream tube, 73 Lift-curve slope, effect of Mach Number
tables for, 88, 614 on, 326
with van der Waals gas, 96-97 Lift-divergence Mach Number, 384-87
with variable specific heat, 9 6 9 7 Lift-drag polar
Isentropic process of subsonic profiles in supersonic flow,
enthalpy change in, 43 571
of perfect gas, 43 of supersonic profile, 446
Isentropic stagnation pressure, 75 Lift-drag ratio, of supersonic airfoil, 443
Isentropic stagnation state, 83 Limit line, 540
Isothermal flow in ducts. 178-83 Linear differential equation, 288, 289
charts for, 181 Linearized theory
choking effects for, 183 assumptions of, 308
formulas for, 180-81 pressure coefficient in, 309-10
a t low Mach Number, 182-83 and second-order theory, 561-63
variation of fluid properties for, 179-80 subsonic, 303
supersonic, 42656
Jet compared with experiment, 44748
contraction ratio of, 358-59 general solution for, 429
overexpanded, 454, 565 geometrical interpretation of, 434-
turbine nozzle. 566 37
underexpanded, 454, 479-81, 501 modified, 463, 477
Joukowski profile and shock theory, 553
lift of, 370 for three-dimensional subsonic flow,
lower critical Mach Number of, 370 394-99
in subsonic flow, 369 Liquid
Joule's constant, 28 change in wave shape for, 131
shock wave in, 155
Lower critical Mach Number, 348, 349
Karman-Tsien coordinate correction for- ~ -
of bump, 372-74
mula, 353 of circular-arc profile, 375
Karman-Tsien method, 344-51 of circular cylinder, 367
for flow with circulation, 357 of circular cylinder in wind tunnel,
with polygonal approximation, 357 358, 368
profile correction for, 351-56 of ellipsoid, 398400
Karman-Tsien pressure correction for- of ellipsoid of revolut.ion, 407
mula, 323, 327, 341; see also Kar- of elliptic cylinder, 358, 369, 407
man-Tsien rule of Joukowski profile, 370
Karman-Tsien rule, 370, 374, 375 of profile, 387
accuracy of, 349-50 of sphere, 409
chart for, 348 and sweepback, 418
comparison with experiment, 350
comparison with Prandtl-Glauert rule, Mach angle, 54, 432
349 Mach cone, 51-54
Kelvin temperature scale, 32 Mach line, 463, 484, 499; see also Mach
Kelvin's theorem, 278-80, 282 wave
Kinetic energy, 36 Mach Number
flux of, 39 constant, flow with, 241-42
definition of, 54
Lagrange, method of, 11 influence of independent variables on,
Laitone's modification of Prandtl- 233
Glauert rule, 324-27 lower critical; see h w e r critical Mach
Lambda shock, 381 Number
Laplace's equation, 289, 294, 297 relation to Mf, 81-82
Lattice-point method, 491 as similarity parameter, 56
Layer, boundary; see Boundary layer Mach quadrilateral, 477, 498
Length, units of, 20-21 Mach reflection, 555-57
Lift, of supersonic airfoils, 445, 572; see Mach wave, 601; see also Mach line;
also Airfoils; Lift coefficient Pressure wave: Wave
Lift coefficient; see also Airfoils; Lift compared with oblique shock, 531
of circular-arc profile, 375 four basic types of, 436
Gothert's rule for, 321 interaction with oblique shock, 541
INDEX
Mach wave direction of, 120-21
left-running, 432 in ducts, 135-37, 583
as limit of oblique shock, 551 with energy effects, 203-11
pressure-deflection relation for, 435 entropy increase in, 119-20
right-running, 433 example of, 123
Mass formation of, 125-31
conservation of; see Conservation of formulas for, 118-20
mass; Continuity equation governing equations of, 114-16
flux of, 14 impossibility of rarefaction, 120-21
units of, 20-21 interaction with boundary layer, 135-
Mass flow parameter, for adiabatic flow, 37. 583
82 in liquid, 155
Mass rate of flow, 14 moving, 137-39
Maximum velocity, for adiabatic flow, in perfect gas, 116-22
79-8 1
.. - - Prandtl's equation of, 118
Mechanical equivalent of heat, 28 Rankine-Hugoniot equation of, 121-22
Method of characteristics, 462-528; see stagnation-pressure loss in, 119
also Characteristics, method of; strength of, 123-24
Characteristic curves tables for, 123, 621
compared with experiment, 514 thickness of, 131-34
graphical, 494 transformed from oblique shock, 530
grid size vs. accuracy of, 497 of vanishing strength, 124
illustrative example of, 479, 486-89 velocity of, 124
initial-value theorem of, 497-500, weak, i23z25
603-6 Normal shock wave; see Normal shock;
numerical, 494-97, 603 Wave
for rotational flow. 51623 Normal stresses, 18
rules for using, 478 work done by, 38
and shock theory, 553 Nozzle; see also Isentropic flow
theory of, 595-608 converging
unit processes for, 492-94, 495 discharge coefficient of, 93
Mixing as flow meter, 93
h , see also Simple gas in-
f l .~ ~.w i t239; as flow regulator, 93
jection with varying pressure ratio, 91-93
one-dimensional flow with, 219-55 converging-diverging
Models, distorted, 315 operating characteristics of, 1 3 9 4 3
Moisture condensation, flow with, 219- with varying pressure ratio, 93-94
55; see also Simple To-Change discharge coefficient of, 99-100
Moisture condensation shock, 203-5 efficiency of, 98-99, 110
Moment of supersonic airfoils, 572; see isentropic flow in, 73
also Airfoils: Moment coefficient overexpanded, 454, 565
Moment coefficient performance of, 98-100
effect of Mach Number on, 315, 383 rocket, 219
Prandtl-Glauert rule for. 322-23 sharp-cornered, 512
of supersonic airfoils. 44k su~ersonicwind tunnel. 427. 507-16
Moment of momentum tuibine, 566
flux of, 20 underexpanded, 454, 479-81, 501
theorem of, 19 velocitv coefficient of. 99
Momentum, flux of, 17 wind tinnel, 502
Momentum theorem, 16 Nuclear reactor, 219
for moving reference frames, 18 heat transfer in, 252-55
working
Motion
- form of. 17 Nusselt Number, 213
equations of; see Equations of motion Oblique shocks, 52944; see also Shocks
irrotational; see Irrotational motion approximate treatment of, 552
perpetual, of second kind, 29 charts for, 536-38
compared with Mach waves. 531
Newtonian fluid, definition of, 10 crossing of, 555
Newton's second law; see Euler's equa- crossing slip line, 561, 592
tions curved. 549. 559-61
Nonlinear differential equation, 288 detached, 548
Normal shock, 112-54, 199-200; see also equations of, 532-39
Shocks; Wave formation of. 540. 563
curves for, 122 formulas for,' 535'
development of, 12532 generation a t concave bend, 583
INDEX
Oblique shocks-Continued definition of, 41
geometry of, 53941 enthalpy of, 41, 42
impossibility of rarefaction, 539 entropy of, 43
interaction with shocks, 541 equation of state of, 41
interaction with waves, 541, 559, 563 flow of, adiabatic constant-area, 162-
intersection of, 553-59 73
limit of as Mach wave, 551 internal energy of, 41, 43
maximum deflection for, 548, 586 isentropic flow of, 83-88
minimum Mach Number for, 548 isentropic process in, 43
nearly normal, 550 normal shock in, 116-22
Prandtl relation for, 534 ratio of specific heats of, 42
Rankine-Hugoniot law for, 533 shock thickness in, 133-34
reflection of, 553-59 specific heats of, 42
from boundary layer, 582 velocity of sound in, 47
from free boundary, 554 Perpetual motion
from wall, 554 and irreversibility, 31
and Second Law of thermodynamics, of second kind, 29
539 Perturbation potential, 305, 428
series solution of, 562 Perturbation velocities, 305
sonic flow after, 550, 586 Perturbations, small, method of, 303; see
streamline turning in, 539 also Linearized theory
strong, 544-48 Physical characteristic curves, 484, 599
tables for, 536 Pipes, flow in; see Ducts, flow in
transformed from normal shock. 530 Pitot tube
and wave drag, 569 subsonic, 94-95, 154
weak, 54448, 551-53, 587 supersonic, 153-54
One-dimensional ao~roximation.73-74 Plate, flat; see Flat plate
One-dimensional flow; see also low
A.
Polars, shock; see Shock polars

with area change, 219-55 Potential, velocity: - . see also Velocity
with chemical reaction, 219-55 potential
with evaporation or condensation, perturbation, 305, 426, 428
--- --
31 9-55
with friction, 159-86; see also Friction,
Potential energy, 36
Potential flow, 265-66
flow with Potential motion, 286
with friction and area change, 24142 Potential vortex. 271-73
with friction and heat transfer, 242-55; Pound of force, definition of, 20-21
see ako Heat exchanger Pound of mass, definition of, 20-21
general method of solution, 236-37 Prandtl-Glauert method, 370-72
Gothert's rule for, 331-32 Prandtl-Glauert rule, 321-24, 336, 370,
with heat transfer, 219-55; see also 375, 376, 383, 400
Simple TrChange comparison with ex eriment, 326-27
influence coefficients of, 237-38
influence of inde~endentvariables on.
2
comparison with arman-Tsien rule,
349
232-36 and drag, 383
with internal drag, 219-55 Laitone's modification of, 324-26
with mixing, 219-55 for wind tunnels, 327-30
simple types of, 23841 Prandtl-Meyer flow; see Centered waves;
table for, 239 Corner flow: Prandtl-Meyer wave;
transition through sonic speed, 235-36 Simple waves
unsteady, 596 Prandtl-Meyer wave, 46344, 539
with wall friction, 219-55 impossibility of, in concave corner, 540
Optical methods, 5 9 4 8 Prandtl number, as similarity parameter,
comparison of, 5 9 4 0 56
Orifice, discharge coefficient of, 100, 359 ~randtl'sequation
Orthogonal network of cquipotential and for normal shock, 118
streamlines, 296-97 for oblique shock, 534
Overexpanded nozzle, 454, 565 Pressure
center of; see Center of pressure
Parabolic differential equation, 600 stagnat;on, 75
Patching curves, 482, 601 total, 75
Perfect fluid. definition of. 10 Pressure coefficient, 107, 309
Perfect gas dimensional analysis of, 308-9
adiabatic flow of, 78-82 effect of Mach Number on, 309, 3i4-
change in wave shape for, 131 15, 327
constants of, 42 for elliptic cylinder, 369
INDEX 643
Gothert's rule for, 320 Relaxation method, 297, 336, 377-79
interference, 329 Relief effect, three-dimensional, 399,
Karman-Tsien rule for, 347 400, 414, 416
lateral decay of, 314 Reversibility, 30
linearized, 309-10, 428 Revolution
Prandtl-Glauert rule for. 322-23 body of
for small perturbations, 428 disturbance of. 402
Pressure distribution drag of, 408 '
and boundary layer, 308 experimental results for, 404, 407
and Mach Number, 327 Gothert's rule for. 402-5
on wing profiles, 379-81 pressure coefficien't for, 402
Pressure drag, 442 subsonic flow past, 401-10
and oblique shocks, 569 thin. 405-6
~- -
Pressure index, 450 ellipsoid of, subsonic flow past, 405

Pressure ratio, critical, 76 Reynolds' analogy, 213, 243
Pressure stresses, work done by, 38 Reynolds Number
Pressure wave, 433, 463; see also Mach of shock wave, 132-34
wave; Shock wave; Wave as similarity parameter, 56
Process, adiabatic; see Adiabatic process of supersonic wind tunnel, 1 0 3 4
definition of, 24 Rocket motor, thrust of, 100-3
equal-temperature, 25 Rocket nozzle, 219
irreversible, 30 Rotation, 267, 269-71
reversible, 30 flow with; see Rotational flow
Profiles; see Airfoils in streamline coordinates, 280-81
Prolate spheroid, subsonic flow past, 406 supersonic flow with, 51623
Propagation of disturbances, 499 and thermodynamic properties, 280-82
Properties, thermodynamic, units of, 37 Rotational flow, 266, 271, 282, 51623
Property
specific value of, 35 Saddle point; see Critical point
of a system, 24 Schlieren method, 5940, 65-67
per unit mass, 35 analysis of, 66-67
Pure substance, definition of, 34 Second Law of thermodynamics, 30
internal energy of, 35-36 in analytical form, 34
for control volume, 40
Quasi-linear equation, 597 corollaries of. 31
and irreversibility, 31
Ram jet, 219 Second-order theory, 56143, 571-78
combustion in, 201-2 coefficients of. 562
Rankine-Hugoniot curve, 125 table of, 63'3
Rankine-Hugoniot equation, 121-22, compared with experiment, 575
5237 ---
552 Separation of boundary layer, 379, 380,
Rank& temperature scale, 32 A1 5
Rarefaction wave, spreading of, 128-31 shadowgraph, 5940, 67-68
Rayleigh line, 115-16, 191-93, 240 Shaft work, 39
equation of, 155 Sharpcornered nozzle, 512
table of, 628 experiments on, 514
Rayleigh-Janzen method, 336, 36547, Shear deformation, rate of, 270-71
271 Shear stress, 18
fo;spheres, 409 definition of. 9
Rayleigh pitot-tube formula, 153-54 Shear stresses.'work done by,
- . 38
Reactor, nuclear, 219 Shear work, 38-39
Recovery factor, 211-13 Shock-boundary layer interaction,
Reflection 576, 580-84
Mach, 555-57 Shock-characteristic theory, 575
of oblique shock, 553-59 Shock polars, 54244
from boundary layer, 582 hodograph, 542, 552
from free boundam. " ,493. 554 and characteristic curve, 552
from wall, 554 pressure-deflection, 544
regular, 555 Shock wave; see Normal shock; Oblique
of wave shocks; Shocks; Wave
from free boundary, 450, 452 normal ; see Normal shock
from surface of discontinuity, 461 oblique; see Oblique shocks
from wall, 450, 451, 493 Shocks, 266; see also Normal shock;
Refraction, index of, 63 Oblique shocks; Wave
Regular reflection, 555 and characteristics, 601
Shocks-Continued 606-8; see also Centered waves;
curved. 282. 549. 5 5 9 4 1 Corner flow; Method of character-
detached, 548 ' istics; Prandtl-Meyer wave
development of, 125-32 calculation rules of, 467
direction of. 116 zavitation in, 475
as discontirhities, 112 centered; see Cent,ered waves
dissipative effects in, 131-32 complete, 474, 475
entropy increase of, 119-20 exact solution for, 465-77
for weak shock, 124-25 example of, 468, 470
explosion, 206-11 features of, 471
formation of, 125-31 hodograph of, 607
impossibility of rarefaction, 120-21 left-running, 472
interaction of, with boundary layer, patching of, 608
13537, 266, 381-82, 58&84 in physical plane, 607
lambda, 381 right-running, 473
moisture condensation, 203-5; see also streamlines of. 475-77
Simple To-Change tables for, 468
moving, 137-39 Skin-friction coefficient, 379
normal; see Normal shock Slender bodies, flow past; see Linearized
oblique. 529-84; see also Oblique theory
shbcks Slip line, 519, 544, 555
Reynolds Number of, 132-34 crossed by oblique shocks, 561, 592
strength of, 123-24, 551 Slow burning, 207-11
swallowing of, 145-51 Slue. definition of. 20-21
thickness of, 112, 131-34 azd pound force, 21
in oerfect ws. 133-34 and pound mass, 21
ta6le for, r33134 Small perturbations; see Linearized the-
for weak shocks, 133 ory
velocity of, 48
. A .
variation of fluid ~ronertieswithin. 134 Sonic point
passage through, 258
and wave drag, 569 stability of, 260
Silence, zone of, 500 special conditions at, 255
Similarity laws, 303, 315-17; see also Sound, velocity of, 4549; see also Ve-
Gothert's rule; Prandtl-Glauert rule locity of sound
for flow far from profile, 324 in air, 48
for incompressible flow, 321-22 in incompressible fluid, 48
subsonic, 303-32 in liquid, 69
compared with experiment, 326-27 in perfect gas, 47
for one-dimensional flow, 331-32 transition through, 235-36
for two-dimensional passages, 330-32 Sources and sinks, method of, 401
for wind tunnels, 327-30, 335 Specific heat
for twodimensional subsonic flow, definition of, 36
Gothert's, 315-21 of perfect gas, 42
Laitone's, 324-26 Speed, flame propagation, 207, 209-11
Prandtl-Glauert, 321-24 Speed of sound; see Velocity of sound;
summary of, 323 Sound, velocity of
for wings, 410-15 Spheres
Similarity parameters, 54-56 drag of, 410
Simple area change; see Isentropic flow subsonic flow past, 409-10
Simple friction; see Friction, onedimen- Spherical coordinate system, 299
sional flow with Stagnation enthalpy, 75
Simple gas injection, 2 3 9 4 1 Stagnation pressure, 75
Simple heating; see Simple To-Change loss of, in normal shock, 119
Simple To-Change, 190-211, 239 Stagnation state, 83
chart for, 196 Stagnation temperature, 79
choking due to, 192-93, 201-3 Stagnation-temperature ratio, 80
example of, 197-98 Stagnation velocity of sound, 79-81
formulas for. 195-96 State, change of, 24
of perfect g&, 193-203 equation of, for perfect gas, 41
tables for, 196, 628 equilibrium, 40
variations of fluid properties in, 198-99 nonequilibrium, 40
Simple types of flow, 2 3 8 4 1 of a system, 24
table of, 239 Steady flow
Simple wave flow; see Simple waves continuity equation for, 15
Simple waves, 431, 464, 489-91, 602, definition of, 11
INDEX 645
energy equation of, 3 9 4 0 for sweptback wings; see Sweptback

one-dimensional, 15 wings
Stokes' theorem, 269 for thin bodies of revolution, 4 0 5 4
Stream function. for axially - symmetric
- three-dimensional, 394-99
flow, 294 wind tunnel corrections for, 327-30
definition of, 291-92 for wings, 410-15
differential eauations of. 291-92 generalized correlation of, 422
equation of Subsonic pitot tube, 154
'for inconiprcssible flow, 293 Substance, pure
i n C:~rtcs~:tn
- - ~ ~ o o r t l i ~ n292
t w 94 definition of, 34
in cvlindrical coordinates. 294-95 internal energy of, 35-36
in &dograph plane, 341 ' Suction pressure, a t nose of airfoil, 449
physical meaning of, 292 Superposition of solutions, 289
of plane source, 299 Supersaturation, 203
and velocity potential, 2 9 5 9 7 Supersonic biplane, 579
Stream tube, isentropic flow in, 73 Supersonic diffuser; see Diffuser, super-
Streamline coordinates, 280 sonic
Streamline curvature method, 387-91 Supersonic flow, 59
Streamlines compared with subsonic, 76-78, 429,
and equipotential lines, 295-97 448, 500
for flow past circular cylinder, 297 in duct with friction, 104-5
hodograph, 336-37, 341 Karman's rules of, 51
Strength of shock, 123-24, 551 linearized, 427-56
Stress. Mohr's circle of, 9 general solution for, 429
normal, 18 method of characteristics for, 482-91
work done by, 38 with rotation, 516-23
shear, 9, 18 with small perturbations, 427-56
work done by, 38 Supersonic inlet; see Diffuser, supersonic;
state of, 9 Inlet, supersonic
hydrostatic, 9 Supersonic nozzle, 426, 507-16; see also
in inviscid fluid, 9 Supersonic flow
in viscous fluid, 9 Supersonic pitot tube, 153-54
surface. 18 Supersonic wind tunnel, 459, 502
Surface, control, 12
Subsonic Row, 59 integral, 598
in accelerating elbow, 3 5 9 4 0 Surface forces, definition of, 18
for airfoil in wind tunnel, 378 Surface of discontinuity; see Slip line;
angle of attack in, 384 Vortex sheets
for bodies of revolution, 401-10 Surface stresses, 18
drag of, 408 Surroundings, definition of, 11
for bump, 372-74 Sweepback; see Sweptback wings
for circular arc profile, 375 Sweptback wings
for circular cylinder, 367 compressibility effects for, 417
in free jet, 368 drag of, 421
in wind tunnel, 357-58, 367 experimental results for, 419-22
compared with supersonic, 76-78, 429, generalized correlation of, 422
448, 500 Gothert's rule for, 416
effect of camber in, 384-87 lift of, 420, 421
effect of Mach Number on flow far lower critical Mach Number of, 418
from profile, 324 subsonic flow past, 415-22
effect of thickness in, 384-87 System
for ellipsoids, 398401 boundary of, 11
for elliptic cylinder, 354-56, 369 definition of, 11
a t angle of attack, 358, 376 property of, 24
experimental results for, 326-27, 350, state of, 24
379-87 Tangent gas, 336
for flat d a t e normal to stream. 360 definition of, 342
influenceof compressibility in, 460 limitations of, 3 4 3 4
for Joukowski profile, 369 polygonal extension of, 357
Karman-Tsien method for, 34440 properties of, 3 4 2 4 3
lift coefficient for, 376 Tangent-gas approximation, 341-44
pressure distributions in, 3 7 9 4 1 Temaeratme
similarity laws for, 303-32 adiabatic wall, 211-12
for spheres, 409-10 equality of, 24
646 INDEX
Temperaturdonlinued flame propagation, 207, 209-1 1
ratio of stagnation t o static, 80 fluid. 6
scale of, 24 line integral of, 267
stagnation, 79 maximum, for adiabatic flow, 79-81
total, 79 perturbation, 305
Temnerature scale of shock wave, 48
ab$olut,e,32 of sound. 4 5 4 9
Kelvin, 32 in air, '48
Rankine. 32 in incompressible fluid, 48
~ e m ~ e r a t & e - e n t rdiagram,
o~~ 34 in liquid. 69
Thermal conductivity, definition of, 36 passage throu g h, 258
Thermal efficiency of heat engines, 29 stability of, 260
Thermodynamic properties, units of, 37 in perfect gas, 47
Thermodynamics special conditions at, 255
First Law of, 26 stagnation, 79-81
for a cycle, 28 transition through, 235-36
laws of Velocity plane, 465
for irrotational frictionless flow, 284 Velocity potent~al,definition of, 285
of nonequilibrium states, 40 equation of, 284-91
Second Law of. 30 for axially symmetric flow, 291
in analytical'form, 34 in Cartesian coordinates, 286-89
for control volume, 40 in cylindrical coordinates, 289-0;
corollaries of, 31 for incompressible flow, 289
Zeroth law of, 25 linearized, 304-8
Thickness mathematical features of, 288
displacement in polar coordinates, 291
of boundary layer, 266 solution of, 288
effects of, in subsonic flow, 384-87 for two-dimensional flow, 289
Thickness drag, 443 of line vortex, 299
Thickness of normal shock, 112 lines of; see Equipotential lines
Thin airfoil, supersonic flow past, 434; perturbation, 305, 426, 428
see also Linearized theory and stream function, 295-97
Thin bodies, flow past; see Linearized Viscosity
theory coefficient of, 9
Thomson's theorem, 278-80, 282 effects of, in design, 506
Three-dimensional relief effect, 399, 400, Viscous stresses, 18
414. 416 Volume, control, 12
Thrust of rocket motor, 100-3 Vortex, potential, 271-73
Time, units of, 20-21 Vortex sheets, 519, 544, 555
Total pressure, 75 Vorticity; see Rotation
Total temperature, 79
Transformation, affine, 319 Wake, wave system in, 578
Transition
""
through speed of sound, 235- Wake losses, effect of Mach Number on,
60 383
Transonic flow, 59, 378 Wall
Turbine, blade cascades of, 427 curved, supersonic flow past, 464
impulse blade of, 504 wave-shaped, flow past, 310-15
interference in, 315 Wave; see also Mach wave; Pressure
nozzle jet of, 566 wave; Shocks
Two-dimensional passages, subsonic sim- cancellation of, 450, 452, 493
ilarity rule for, 330-32 centered; see Centered waves
compression, steepening of, 127-31
Underexpanded jet, 454, 479-81, 501 detonation, 219
Underexpahded nozzle, 454, 47941, 501 development of, 158
Unit processes, for waves, 450, 492-94 analysis of, 128-31
Units of dynamic quantities, 20-21 expansion, spreading of, 128-31
Units of thermodynamic quantities, 37 explosion, 206-11
Unsteady one-dimensional flow, 596 generation of, 449, 463, 493
Upper critical Mach Number interaction with oblique shocks, 559
for bump, 372-74 intersection with wave, 450, 453, 494
definition of, 372 left-running, 465
lumped, 449
Velocity Mach; see Mach wave
critical, 79-81 normal shock, 112-54; see also Normal
dimensionless, M*, 81-82 shock
INDEX
of one family, 431 Wind tunnel
rarefaction. soreading - of.. 128-31 airfoil in, 378
reflection of corrections for, 327-30
from free boundary, 450, 452, 493 interference in, 315, 335, 403
from surface of discontinuitv. " ,461 supersonic, 459, 462, 502
from wall, 450, 451, 493 nozzles for, 507-16
right-running, 467 performance of, 152
shock Wings
with energy effects, 203-11 effect of aspect ratio on, 411-15
obliaue. 529-84; see also Oblique experimental results for, 412-15
shbcks of finite span
thickness of, 131-34 lifecurve slope for, 411-12
velocity of, 48 subsonic flow past, 410-15
steepness of, rate of change of, 158 generalized correlation of, 422
of two families, 477-89 induced drag of, 412
unit processes for, 450, 492-94
Wave drag, 443 422
-
lift of.. generalized correlation of.
and oblique shocks, 569 lift-curve slope of, 425

Wave equation, 429 supersonic; see Airfoils, supersonic
Wave shape, change in, 130-31 swe~tback:see Sweatback wings
Wave-shaaed wall work'
subsonic flow past, 310-15 definition of, 25
supersonic flow past, 438-40 done bv normal stresses. 38
~~ -
Weak shock, 551-53 done b; shear stresses, 38

entropy change in, 124-25, 552 measurement of, 26
Weak shock waves, 123-25, 551-53 recognition of, 25
Wedge shaft, 38-39
double shear, 38-39
liftdrag ratio of, 446
supersonic flow past, 437, 444, 568 Zeroth law of thermodynamics, 25
single, 448 Zone of action, 500
Wilson line, 204 Zone of silence, 500

Dynamics of Compressive Fluid Flow PDF

Uploaded by

Copyright:

Available Formats

Dynamics of Compressive Fluid 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

Dynamics of Compressive Fluid Flow PDF

Uploaded by

Copyright:

Available Formats

The Dynamics

JOHN WlllFY S SONS

JOHN WILEY & SONS

All Rights Reserved

Library of Congress Catalog Card Number: 53-8869

The Velocity of Sound. Physical Differences Between Incompres-

Part II. One-Dzmensional Flow

5 NORMAL SHOCK WAVES 112

Simple-Heating Relations for a Perfect Gas. Choking Effects in

8 GENERALIZED ONE-DIMENSIONAL CONTINUOUS FLOW 219

Part 111. Introduction to Flow in Two and Three Dimensions

The Physical Significance of Irrotational Motion. Euler's Equa-

Part N. Subsonic Flow

Part V . Supersonic Flow

18 SUPERSONIC FLOW PAST WINGS OF FINITE SPAN

Part VI. Mixed Flow

22 DRAG AND LIIT AT TRANSONIC SPEEDS

Part VIZ. Unsteady Motion in One Dimension

Part VZZZ. Flow of Real Gases with Viscosity and

1.1. Introductory Remarks

anomalously as to defy being neatly placed in a pigeonhole. For our

The great simplification afforded by the concept of a continuum is

1.2. Properties of the Continuum

imagine the smallest volume which can be regarded as continuous t o

circular area 6A in this plane surrounding P. It is a matter of experience

where a, is the acceleration in the x-direction. From the geometry of

and then, noting that dx dy is negligible compared with dy, we obtair~

By similarly applying Newton's second law in the y-direction, we

and, since the direction -y was chosen without prejudice, we conclude

The constant of proportionality is called the coeficient of viscosity,, p

In Chapter 9 a study of the kinematics of two-dimensional motion

where u and v are the respective velocity components in the x- and

of Lagrange, usually used in the dynamics of discrete particles, wherein

Then for arbitrary increments in x, y, z,and t, the corresponding incre-

1.3. Systems and Control Volumes

1.4. Conservation of Mass

system moves in the general direction of the streamlines. The two

where 6mII,+,, represents the amount of mass crossing the elementary

and so the conservation law may now be expressed as

where & is an element of control volume, p is the local mass density

where p is the local instantaneous mass density in the neighborhood of

a form which is usually called the epwction of continuity, since it is based

It is interesting to observe that the same result applies to incompress-

tional area a t section 1, measured in a plane normal to the axis of the

FIG.1.7. Flow through a control volume.

The combination of terms

represents the time rate of change of x-momentum within space I. In

In the limit, as dt vanishes, we note that Z F , becomes the algebraic

which is the fundamental principle in the dynamics of fluid motion.

1.6. Theorem of Moment of Momentum

to the element of mass m, Ft is the x

I n Cartesian form, and considerirg moments about, say, the z-axis,

1.7. Units and Dimensions

where go is an experimentally determined constant whose magnitude

REFERENCES AND SELECTED BIBLIOGRAPHY

1.2. An incompressible fluid flows in a pipe of radius R. At the inlet, sec-

(a) Demonstrate that V/V,, = s.

2.1. Introductory Remarks

2.2. The First Law of Thermodynamics