Introduction To Gas Dynamics All Lecture Slides: Autumn 2009

Introduction to Gas Dynamics
All Lecture Slides
Teknillinen Korkeakoulu / Helsinki University of Technology
Autumn 2009
Gasdynamics Lecture Slides
1 Compressible flow
2 Zeroth law of thermodynamics
3 First law of thermodynamics
4 Equation of state ideal gas
5 Specific heats
6 The perfect gas
7 Second law of thermodynamics
8 Adiabatic, reversible process
9 The free energy and free enthalpy
10 Entropy and real gas flows
11 One-dimensional gas dynamics
12 Conservation of mass continuity equation
13 Conservation of energy energy equation
14 Reservoir conditions
15 On isentropic flows
16 Eulers equation
17 Momentum equation
18 A review of equations of conservation
19 Isentropic condition
20 Speed of sound Mach number
21 Results from the energy equation
22 The area-velocity relationship
23 On the equations of state
24 Bernoulli equation dynamic pressure
25 Constant area flows
26 Shock relations for perfect gas Part I
27 Shock relations for perfect gas Part II
28 Shock relations for perfect gas Part III
29 The area-velocity relationship revisited
30 Nozzle flow converging nozzle
31 Nozzle flow converging-diverging nozzle

32 Normal shock recovery diffuser
33 Flow with wall roughness Fanno flow
34 Flow with heat addition Rayleigh flow
35 Normal shock, Fanno flow and Rayleigh flow
36 Waves in supersonic flow
37 Multi-dimensional equations of the flow
38 Oblique shocks
39 Relationship between wedge angle and wave angle
40 Small angle approximation
41 Mach lines
42 Weak oblique shocks
43 Supersonic compression by turning
44 Supersonic expansion by turning
45 The Prandtl-Meyer function
46 Detached shocks
47 Shock-expansion theory
48 Reflection and intersection of oblique shocks
49 Cones in supersonic flow
50 Derivation of perturbation equation
51 Irrotational flow
52 Governing equations for small perturbation flows Part I
53 Governing equations for small perturbation flows Part II
54 Pressure coefficient
55 Boundary conditions
56 Flow past a wave-shaped wall an example
57 Flow past a wave-shaped wall subsonic case
58 Flow past a wave-shaped wall supersonic case
Compressible flow
In a nutshell, the term compressible flow refers to the fluids of

which there can be found significant variation of density in the flow
under consideration.
Compressibility is strongly related to the speed of the flow itself and
the thermodynamics of the fluid. A good grasp of thermodynamics
is imperative for the study of compressible flow.
For low-speed flow, the kinetic energy is often much smaller than
the heat content of the fluid, such that temperature remains more or
less constant.
On the other hand, the magnitude of the kinetic energy in a
high-speed flow can be very large, able to cause a large variation in
the temperature.
Some important phenomena strongly associated with compressibility
are the flow discontinuity and choking of the flow.

Compressible flow
To illustrate, consider a car at sea-level, 1 atm and 15 C, going at a

speed of 90 km/h. The density is 1.225 kg/m3 . At a stagnation
point, the density there is found to be 1.228 kg/m3 , a mere 0.27 %
difference. The temperature rises by 0.311 C and the pressure
changes by 0.38 %. Here, the incompressible assumption can be
applied.
Now, consider a typical air flow around a cruising jetliner at 10 km
altitude. The speed is now 810 km/h, while the ambient conditions
are 0.413 kg/m3 , 0.261 atm and 50 C. At the stagnation point the
temperature rises by over 25 C, while density and pressure changes
by more than 30 % and 45 %, respectively. It is clear that
compressibility must now be taken into account.
Compressible flow
An extreme example of
compressible flow in action is the
re-entry flow. Another is shown
here on the left as a jet fighter
seemingly punches through the
sound barrier. However, more
daily mundane applications can
also be found in flows through jet
Figure 1: Breaking the sound engines, or around a transport
barrier. . . ? aircraft.

Zeroth law of thermodynamics
The concept of equilibrium is fundamental to the study of

thermodynamics, yet it was not wholly appreciated until later. Thus,
it is deemed necessary to define the zeroth law of thermodynamics,
as a way to say that it precedes even the first law, as follows:
When two objects are separately in thermodynamic
equilibrium with a third object, they are in equilibrium
with each other.
A system is said to be in equilibrium if it is free of currents. The
term currents here refers to the flux of quantities such as mass,
momentum, or energy, which is caused by gradients in the system.
First law of thermodynamics
The first law is a statement of energy conservation. It states that the

increase in the internal energy level of a system E is equal to the
amount of heat Q flowing in from the surroundings and work W
done on the system by the surroundings. Mathematically, it can be
written as:
E = Q + W (1)
Note that E is a variable of state, while Q and W depend on the

process involved in the state change. And, for a small change of
state the law can be written in a differential form as:
dE = Q + W (2)

Consider this simple

idolized system of
cylinder-piston
arrangement,
assuming rigid walls.
Note that the only
avenue for work done
is through the
displacement of the
piston. Figure 2: Simple cylinder-piston arrangement
Thus, for the system considered here, the work term can be written
~ and displacement ~r as:
in terms of force vector F
Z Z Z
W = F ~ d~r = PA dr = P dV (3)
However, it can also be expressed in terms of pressure P and the

volume V , since the force is acting parallel to the displacement and
is equal to the pressure P multiplied by the piston surface area A.
Note that the negative sign indicates that the work is done by the
system if the system volume increases.

One can also distinguish between extensive and intensive variables

of state. An variable is extensive if its value depends on the mass of
the system. Such are the systems mass m, E and V .
On the other hand, an intensive variable is independent on the mass
of the system, temperature and P being typical examples.
Also, for every extensive quantity, such as E , one can introduce its
intensive counterpart e, the energy per unit mass, or specific energy.
Similarly, specific volume v = V /m can be defined.
Using specific variables, the first law can now be written in its
differential form as:
1
de = m Q P dv = q P dv (4)
Besides the energy E , one can also introduce the concept of

enthalpy H, which is defined as:
H = E + PV (5)
Thus, the first law can be expressed using enthalpy as:
dP
dh = de + P dv + v dP = q + v dP = q + (6)

where = 1/v is the mass density of the system.

Equation of state ideal gas
Measurements of thermal properties of gases have shown that for

low densities, the relationship among the variables approaches the
same form for all gases, which is referred to as the ideal gas
equation:
Pv = R ( + ) (7)
The characteristic temperature turns out to be the same for all
gases, = 273.15 C. Thus, it is much more convenient to adopt
Kelvin scale in favour of the Celcius scale, such that T = + .
Furthermore, the gas constant per unit mass R is a characteristic for
different gases. For air, this value is approximately 287.053 J/kg K.
Specific heats
Heat capacity of a substance C , an extensive property, is defined as

the heat amount required to change its temperature, and can be
expressed as:
Q
C= (8)
T
For gases, the heat transfer Q can occur by different processes.
Hence, the value of C is also dependent on the process path.
It is also more convenient to work with the notion of specific heat,
which is simply the heat capacity per unit mass.

Specific heats
In thermodynamics, it is empirically shown that, for a simple

system, any state variable can be expressed as a function of two
other variables. Hence, one only needs two values of specific heats
to adequately describe the values for all processes, typically the
values chosen are those for constant volume (isochoric) cv and
constant pressure (isobaric) cp processes:

q q
cv = and cp = (9)
dT v dT P
Specific heats
Assume that specific energy e is a funtion of v and T . Then:
e e
de = dv + dT = q P dv (10)
v T
or:
e e
q = dT + +P dv (11)
T v
Hence, the specific heats can be expressed as:

q e
cv = = (12)
dT v T

q e e v
cp = = + +P (13)
dT P T v T P

Specific heats
Likewise, using specific enthalpy h as a funtion of P and T :
h h
dh = dP + dT = q + v dP (14)
P T
or:
h h
q = dT + v dP (15)
T P
And now, the specific heats can be expressed as:

q h h P
cv = = + v (16)
dT v T P T v

q h
cp = = (17)
dT P T
The perfect gas
No gas is truly perfect. However, it is the simplest working

model. Further, experiences show that for a large range of
applications, the gases are under nearly perfect conditions to
justify the use of the model.
A thermally perfect gas is characterised by the fact that energy e is
solely a function of temperature:
e = e (T ) (18)
From the definition of enthalpy:
h = e + Pv = e (T ) + RT = h (T ) (19)

The perfect gas
Furthermore, it follows then that the specific heats are also functions
of only temperature:
de
cv = (20)
dT
de v
cp = +P = cv + R (21)
dT T P
Similarly:

dh P
cv = v = cp R (22)
dT T v
dh
cp = (23)
dT
The perfect gas
The followings then apply for perfect gases:
R = cp cv (24)
Z
e (T ) = cv dT (25)
Z
h (T ) = cp dT (26)
cp
= (27)
cv
where is the ratio of specific heats. For diatomic gases, e.g.,
nitrogen and oxygen, which constitute the bulk of the Earths
atmosphere at standard conditions, this value is found to be roughly
1.4.

The perfect gas
An important special case is the calorically perfect gas, where the

specific heats are constants, independent of temperature T . For this
special class, integrations can be carried out, resulting in:
Z
e (T ) = cv dT = cv T + constant (28)
Z
h (T ) = cp dT = cp T + constant (29)
Second law of thermodynamics
The first law of thermodynamics dictates that energy must be

conserved in a change of state for a system. It doesnt say in which
direction such change of state is allowed. Consider the following.
In this experiment, the weight is allowed to

drop, turning the paddles, which in turn
raises the temperature of the gas in the
adiabatic container. It is clear that one
cannot induce the wheel by itself to extract
the energy from the gas and lift the
weight, thus reversing the process.
Figure 3: Joules
paddle-wheel experiment

For a reversible change of state of an adiabatically enclosed system,

the following can be shown:
E
P= (30)
V
Now assume that similar relation can be found for T , using a

variable S, the entropy. Express E as function of S and V , such
that:
E E
dE = dS + dV = T dS P dV (31)
S v
Hence, using the first law of thermodynamics, the following is found
for a reversible process:
Q = T dS (32)
The concept of entropy can be generalized for any arbitrary process

between two states 1 and 2:
Z
Q
S2 S1 (33)
T
For thermally perfect gas, S, or the specific entropy s, can be given
explicitly, relative to some reference state:
R
Z Z Z
de + P dv cv
s= = dT + dv (34)
T T v
dh v dP R
Z Z Z
cp
= = dT dP (35)
T T P
And, for calorically perfect gas:
T v T P
s so = cv ln + R ln = cp ln R ln (36)
To vo To Po
Adiabatic, reversible process
The term adiabatic refers to the idea that no heat passes through the
boundary of the system, Q = 0.
The term reversible indicates that the system can be restored to its
initial state with no changes in both the system itself and its
surroundings, such that no energy is being dissipated. This is
signified as dS = Q/T .
In general, dS Q/T for natural processes. This means that
entropy increase can be attributed to either heat transfer, or to
irreversibilities.
A process that is both adiabatic and reversible is marked by the fact
that the entropy is kept constant, dS = 0.
Though it may sound very restrictive, plenty of important
applications in fluid flow can qualify. Consequently, isentropic
analysis is used rather extensively.
From the first law, the following can be written for isentropic
processes:
e e
dv + dT = P dv (37)
v T
h h
dP + dT = v dP (38)
P T
Which result in:

dT 1 e
= +P (39)
dv cv v

dT 1 h
= v (40)
dP cp P

For perfect gases, where R = cp cv and = cp /cv , these simplify

to:
v dT R
= = ( 1) (41)
T dv cv
P dT R 1
= = (42)
T dP cp
where generally depends on temperature T , except when

calorically perfect gas is assumed. In any case, these relationships
can be easily integrated.
For constant :
Z
v dT 1 T
ln = = ln (43)
vo ( 1) T 1 To
Z
P dT T
ln = = ln (44)
Po ( 1) T 1 To
And since = 1/v :
1 T
ln = ln (45)
o 1 To

The free energy and free enthalpy
In addition to energy E and enthalpy H, one can also define the

following two state variables: Helmholtz function (free energy) F
and Gibbs function (free enthalpy) G . These are related to E and H
as follows:
F = E TS (46)
G = H TS (47)
In differential forms, one then ends up with the followings:
dE = T dS P dV (48)
dF = S dT P dV (49)
dG = S dT + V dP (50)
dH = T dS + V dP (51)
Entropy and real gas flows
In reality, irreversible state changes increase the entropy. It can be

argued that during such process entropy are produced and left in the
system. In the case of simple one-dimensional flow the following
can be obtained. Let be defined as the rate of entropy change per
unit time and volume:
2
dT
T = 2 (52)
T dx
du 2

u = (53)
T dx
where is the coefficient of heat conduction, while denotes a

coefficient of viscosity.

Entropy and real gas flows
Experiments have shown that for large regions of the flow, these
gradients are small. Consequently, entropy production is very small
and may be neglected in these regions. This, however, do not apply
to regions where large gradients exist, such as across shock waves,
and flows inside the boundary layers, vortex cores and wakes. In
this regions, non-isentropic analysis must be employed.
One-dimensional gas dynamics
To begin the study of the motion of compressible fluids, it is

advisable to consider the case of one-dimensional flow, shown here
as a streamtube in the figure.
Figure 4: A streamtube

One-dimensional gas dynamics
Note that this definition still holds in the case of slowly varying
cross-sectional area along the axis x, provided that the variation is a
function of the axis A = A (x) only, unchanging in time; and that
flow properties, e.g., P and , are uniform across the cross-section,
including the velocity u = u (x), which is normal to the
cross-sectional area. These flow quantities can also be allowed to
vary with time t in case of non-stationary or unsteady flows.
In a one-dimensional incompressible flow, practically all information
is contained in the kinematic relation uA = constant, while pressure
is obtained independently using the Bernoulli equation.
On the other hand, compressible flows impose inter-dependency
between the conservation of mass and momentum, and the
area-velocity rule is no longer straightforward.
Conservation of mass continuity equation
Refer to the figure below.
Figure 5: Flow through a tube segment
Note that the mass flow rate across any cross-section is m = uA.
Also, the mass enclosed within the segment for small enough x
can be set equal to A x.
Conservation of mass continuity equation
The amount of fluid mass enclosed within the segment increases at

the rate of t (A x). Since mass is conserved, any sources or
sinks can be excluded, and any increases must be balanced by the
inflow through 1, minus the outflow through 2, such as:

t (A x) = x (uA) x (54)
Since x does not depend on time t, and may be divided through,

this results in the continuity equation:

t (A) + x (uA) = 0 (55)
And in the case of steady flow situations, the time dependency can
be dropped, and one is left with:
d
dx (uA) = 0 m = uA = constant (56)
Conservation of energy energy equation
Refer to the figure below.
Assume steady state

condition
Include the kinetic
energy contribution
1 2
2u
Assume small enough
displacements, such
that the pressure
variations at the pistons
are negligible
Figure 6: System for energy calculation

During some time increment t, the pistons moved from 12

positions to 10 20 positions. Also, some heat q is added to the
system. Furthermore, the work done by the pistons is:
w = P1 v1 P2 v2 (57)
According to the first law, the energy balance, which now includes
the kinetic energy contributions, is expressible as:
e2 + 21 u22 e1 + 12 u12 = q + w = q + (P1 v1 P2 v2 ) (58)

In terms of the enthalpy h = e + Pv = e + P/, this balance

equation becomes:
h2 + 12 u22 h1 + 12 u12 = q

(59)
Note that q as shown in the figure refers to heat from outside the
walls. Heat generated internally is included in h. Also, the energy
equation derived here relate the equilibrium states at sections 1 and
2 in the figure. It is valid even if there are regions on
non-equilibrium in-between, as long as the flow is at equilibrium at
the sections themselves.
In adiabatic flows, there is no heat transfer allowed, q = 0, and
the equation is refered as the adiabatic energy equation:
h2 + 12 u22 = h1 + 21 u12 (60)
If in addition, equilibrium exists all along the flow, then the

equilibrium equation is valid continously, and one ends up with:
h + 12 u 2 = constant (61)

In the case of equilibrium flows, one may obtain the differential

form:
dh + u du = 0 (62)
For a thermally perfect gas, this can be rewritten as:
cp dT + u du = 0 (63)
And, if the gas is calorically perfect, this can be integrated as:
cp T + 12 u 2 = constant (64)
Reservoir conditions
It is helpful to define the stagnation or reservoir conditions for

fluids in equilibrium, where the velocity is negligible and can be set
to zero. This allows one to obtain the stagnation / reservoir enthalpy
ho :
h + 12 u 2 = ho (65)
Figure 7: Flow between two reservoirs

Reservoir conditions
In the case of adiabatic flow, the reservoir enthalpy in the second

reservoir is also equal to that of the first reservoir, as well as along
the connecting channel:
ho0 = ho = h + 12 u 2 (66)
By assuming (calorically) perfect gas, h = cp T , one is left with:
cp To0 = cp To To0 = To (67)
Be aware that this is only for the perfect gas, otherwise only the
constancy of the stagnation enthalpy holds in the adiabatic flows.
Also note that often these stagnation or reservoir conditions are
referred to as the total conditions.
On isentropic flows
Refering again to the figure of the flow between two reservoirs, note
that the first law doesnt forbid the flow to be reversed, i.e., from
right to left. This illustrates the application of the second law:
so0 so 0 (68)
For the calorically perfect gas, one has:
To0 Po0 Po0

so0 so = cp ln R ln = R ln 0 (69)
To Po Po
This simply implies what one intuitively would have conclude,
which is that the downstream reservoir pressure cannot be greater
than upstream one, Po0 Po , in order for a spontaneous flow to take
place.

On isentropic flows
Looking back at the definition of entropy change:

1 1 1
ds = (de + P dv ) = dh dP (70)
T T
The following can be obtained:

s 1
= <0 (71)
P h T
This shows that an increase in entropy, at constant stagnation

enthalpy, must result in a decrease in stagnation pressure.
The irreversible increase in entropy, with the decrease in total
pressure, is due to the entropy production between the reservoirs.
On isentropic flows
This entropy production will only be absent if there are no

dissipative processes, i.e., the flow is in equilibrium throughout.
Only for such isentropic flow do so0 = so and Po0 = Po .
And now, the term total condition can now be broadened to include
conditions at any point of the flow, not just at the reservoirs as
follows:
The local total conditions at any point in the flow are the
conditions that would be attained if the flow there were
brought to rest isentropically.
In the light of this definition, the local stagnation, or total entropy is
equal to the local static entropy, so0 = s 0 .

On isentropic flows
For adiabatic flows of perfect gases, this allows for measurement of

flow entropy from the measurement of the local total pressure,
which under suitable conditions can be made using a simple pitot
probe, since the total temperature To is constant throughout:
0 Po0
s s = R ln (72)
Po
Be fully aware that for stagnation conditions to properly exist, it is
not sufficient that the velocity is zero. It is also necessary that
equilibrium conditions exist, and there are non-negligible gradients
to cause currents.
Eulers equation
Newtons second law states that the rate of change of momentum for
a body is proportional to the force acting on it:
~ = D (m~u ) = m D~u = m~a

F (73)
Dt Dt
where ~a is the acceleration suffered by the body.
It is customary to adopt Eulerian viewpoint when it comes to taking
derivatives of some field quantity associated with the fluid
particle, which is assumed to be a function of both time and its
position. That is to say that = (t, ~x ). Thus, in one-dimensional
flows, the following definition then applies:

d = dt + dx (74)
t x

Eulers equation
On the other hand, a particles position is also a time-dependent

variable. Hence, the rate of change of is expressible as:

D dx
= + = +u (75)
Dt t dt x t x
Be aware of the two parts involved: the local time derivative and the
convective derivative associated with the spatial changes to the
quantity as the particle travels with the fluid velocity u. Together,
they are called the total or material derivative.
Accordingly, the acceleration suffered by the fluid particle in a
one-dimensional flow can be written as:
Du u dx u u u
= + = +u (76)
Dt t dt x t x
Note that the velocity term occurs twice in the expression.
Eulers equation
At present, it is assumed
that the only forces acting
on the fluid are inviscid in
nature, no viscous
elements are included.
Furthermore, gravitational
effects are also neglected.
Thus, one ends up with
only pressure forces.
Figure 8: Pressure on a fluid particle
in diverging stream tube

Eulers equation
From the preceding figure, the pressure forces in the x-direction can
be summed up as:

X A P A
Fx = PA + Pm x P + x A+ x (77)
x x x
where Pm = P + 12 x Px is, for lack of better words, the mean

pressure between the two stations.
By expanding and neglecting the second order terms, one then ends
up with:
X P
Fx = A x (78)
x
Eulers equation
Recall that the mass inside the volume is m = A x. Putting terms

into Newtons second law, one ends up with:

u u P
A x +u = A x (79)
t x x
And by dividing out the term A x:
u u 1 P
+u = (80)
t x x
one ends up with the conservation of momentum for inviscid fluids,
often also called as the Eulers equation.

Eulers equation
For steady flows, t u = 0, t u = 0, and the Euler equation

becomes:
dP
u du + =0 (81)

When P is a known function of , one can integrate and obtain the
celebrated Bernoulli equation:
Z
1 2 dP
2 u + = constant (82)

For example, incompressible assumption = o gives:
1 2
2 o u + P = constant (83)
An isentropic process for perfect gas gives P/Po = (/o ) ,
resulting in:
1 2
2 u + 1 (P/) = constant (84)
Momentum equation
The continuity equation and the Eulers equation can be combined

to yield an integral form of momentum balance equation.
By multiplying the continuity equation by u and the Eulers
equation by A, one obtains:

u (A) + u (uA) = 0 (85)
t x

A (u) + uA (u) = A (P) (86)
t x x
One can then take the sum of the results and apply integration by
parts, which results in the following:
2
dA
(uA) + u A = (PA) + P (87)
t x x dx

Momentum equation
Figure 9: Control volume for the momentum equation
Refering to the figure above, the previous equation can be integrated

with respect to x from 1 to 2.
Momentum equation
Noting that the time differentiation can be taken out from inside the
integral sign, one now has:
Z 2
2 2

(uA) dx + 2 u2 A2 1 u1 A1
t 1
Z 2 (88)
= (P2 A2 P1 A1 ) + P dA
1
The first integral is the momentum of the fluid contained in the

control volume, while the last integral can be evaluated by defining
a mean pressure Pm , which can then be taken out of the integral
sign, and may include even contributions from viscous effects:
Z 2
P dA = Pm (A2 A1 ) (89)
1

Momentum equation
Thus, the left-hand-side is the rate of change of the fluids

momentum, consisting of the local contribution inside the control
volume, and the contribution due to the transport (flux) of the
momentum through the end sections.
The right-hand-side is the force in the direction of the flow, which
here is the x-direction, both from the end sections and also from the
walls.
This equation turns out to be more general than what the eye
catches. It is still valid even if there are regions of dissipation
(viscous effects) within the control volume, provided that they are
absent at the reference points 1 and 2.
Momentum equation
As shown here, the forces and fluxes on adjacent internal faces

cancel each other. Thus, any inequlibrium regions inside this
volume do not affect the end result.
Figure 10: Summing the forces and fluxes in the control volume
Finally, for steady flow in a duct of constant area:
2 u22 + P2 = 1 u12 + P1 (90)

A review of equations of conservation
So far, the followings have been derived for a one-dimensional flow:

the conservation of mass, of linear momentum, and of total energy
(eo = e + 12 u 2 ):

(A) + (uA) = 0
t x
dA
u 2 A =

(uA) + (PA) + P (91)
t x x dx

(eo A) + (ueo A) = (PuA)
t x x
Implicit in the expressions are the assumptions of negligible
transport coefficients, e.g., viscosity and heat conduction; adiabatic
condition at the walls; and that only pressure forces are present.
Isentropic condition
It is stated before that a flow which is both adiabatic and in

equilibrium is isentropic. This can be verified from the energy and
momentum equations recently derived.
For adiabatic and non-viscous flows, the followings are applicable:
dP
dh + u du = 0 and u du + =0 (92)

Eliminating u du between the two:
dP
dh =0 (93)

From the mathematical definition of entropy T ds = dh dP/,

this is simply saying that entropy is constant.

It is obvious that equilibrium cannot be strictly attained in real fluid,

since the fluid particle must continuously adjust itself upon
encountering new conditions. The rate of adjustments depends on
the existing gradients, and is a measure of the degree of
non-equilibrium in the flow. It has been mentioned before that
entropy production terms in one-dimensional flow are:
2 2
u T
and (94)
T x T2 x
Thus, entropy is always produced, since gradients are always

present, and coefficients of viscosity and conductivity are finite.
However, a large and useful part of fluid mechanics, can be studied
using idealized flows, in which the fluid is assumed to be inviscid
and non-conducting ( = 0 and = 0).
Yet, be fully aware that even if the actual fluid possesses extremely
low values of and , serious care must be taken in regions where
large gradients are present. Such exist as boundary layers, wakes,
vortex cores, and in supersonic flows, as shock waves.

Speed of sound Mach number
In principle, sound can be said to be waves of density and pressure

variations in space and in time.
Note that this is another hallmark of compressible flows, in that
small pressure changes can affect appreciable density changes, and
vice versa.
Furthermore, sound propagates at some finite speed through the
medium, while in incompressible flows, it is generally assumed to
be infinite.
And one can define the Mach number Ma as the ratio of flow
velocity u to that of the speed of sound a associated with the flow.
To learn more about the speed of sound in the fluid, it is instructive

to assume several things:
The disturbance is weak enough that only small changes are
produced, and the flow properties can be considered
more-or-less continuous
The disturbance is fast enough that the process is adiabatic, no
heat transfer is allowed to take place
The disturbance is slow enough that no significant deviation
from equilibrium is taking place during the process, a
quasi-static condition
Since the process is considered to be both adiabatic and in
equilibrium, it is then also isentropic.

Consider this thought experiment of a sound pulse caused by a

piston moving slowly into a still air in a duct of constant area.
(a) Laboratory frame (b) Moving frame
By considering the situation on the right figure as opposed to the

left, the problem becomes that of a steady state flow, since now the
wave front does not move in the frame of reference.
The continuity equation gives:
( + d) (a du) = a a d du = 0 (95)
And the momentum equation gives:
[( + d) (a du)] (a du) + (P + dP) = a2 + P

(96)
dP/a du = 0
By eliminating du from the two expressions, the following

expression for a is obtained:
a2 = dP/d (97)
But that still leaves out the question of how to relate the pressure
changes to the density changes.
At this point, it is illustrative to show how choosing correct

assumptions matter.
Isaac Newton had based his works on Boyles law, which stated that
for an isothermal process pressure is linearly dependent on density.
When combined with the ideal gas equation, it gives:

dP P
= = RT (98)
d T
Applying this to calculate the speed of sound:

a = RT (99)
Imagine his sheer disappointment when experiments showed that he

was short by more than 15%.
It shows that the fast adiabatic process of sound wave propagation

does not rule out temperature changes within the localized pressure
pulse. It is not a slow isothermal process, where the
temperature has had the chance to spread out evenly.
Now assume that pressure is a function of two state variables:
density and entropy. Earlier, it is stipulated that the sound motion is
an isentropic process. Hence, the following:

P P 2 P
dP = d + ds a = (100)
s
S s
For isentropic processes, the following relationships apply:

T P T
cp ln R ln = cv ln R ln = 0 (101)
To Po To o

Eliminating temperature from the equations gives the following

pressure-density relationship for isentropic process:

P dP d
= = (102)
Po o P
When combined with the ideal gas equation, the following correct
expression is obtained for the speed of sound in a calorically perfect
ideal gas: s
P p
a= = RT (103)

As mentioned earlier, one can now define Mach number as the ratio
of the flow velocity and the speed of sound Ma = u/a. It is clear
that its value depends on the local flow velocity and the flow
conditions. Generally, Ma is assumed to be a positive value.
Based on the value of Mach number, fluid flows can often be
divided into several (sometimes overlapping) regimes:
sub / super-sonic Mach number is less / greater than 1
transonic Roughly set around 0.8 Ma 1.25
hypersonic Usually said to be Ma > 5, but mostly defined by the
high degree of real gas effects, such as dissociation of
gas molecules

Results from the energy equation
Rewrite cp in terms of R and :
cp cp /cv (cp cv ) R
cp = (cp cv ) = = (104)
cp cv cp /cv 1 1
Use this result to rewrite the adiabatic energy equation for

calorically perfect gas:
1 2 1 2
2u + cp T = cp To 2u + 1 RT = 1 RTo (105)
By further employing the definition of speed of sound (a2 = RT ),

and Mach number (Ma = u/a), one ends up with:
To /T = (ao /a)2 = 1 + 1
2 Ma2 (106)
And the isentropic relations can also be recast in terms of the Mach
number:
1/(1)
1 2
o / = 1 + 2 Ma (107)
/(1)
1 2
Po /P = 1 + 2 Ma (108)
Be fully aware that in the case of density and pressure P, these

relationships are only for local reservoir values. They are not
necessarily the same throughout the flow field, unlike that of the
reservoir temperature To , and accordingly, the speed of sound ao ,
which are constant throughout.

These equations can have serious implications. For example, getting

a Mach 2 air flow in a blow-down wind tunnel with sea-level
conditions at the test section would require a reservoir condition of
7.8 atm and 540 K.
These requirements on the reservoir conditions only grow stronger
at higher Mach numbers. For air at Mach 7, To /T = 10.8 and
Po /P 4140. Getting sea-level conditions at the test section
requires the tank temperature to be hot enough to almost boil iron,
and the pressure to be almost 4 the pressure at the bottom of the
Mariana Trench.
One can also allow the test section to be very cold, without
liquefying the working fluid, and to operate at near-vacuum.
Furthermore, specialized gases may be used, such that 1, e.g.,
CF4 ( = 1.156) and C2 F6 ( = 1.085).
It is also convenient to define yet another reference condition, not

necessarily at the reservoir, which gives the stagnation condition.
The sonic condition provides an excellent choice, here marked by an
asterisk. And since Ma = 1, it follows that u = a . The point is
that the flow is to be brought to sonic point isentropically.
From the energy equation:
+1
u2 + 2
1 a2 = u2 + 2
1 a2 = 1 a2 (109)
And since T /To = 2/ ( + 1), the followings are computed for a
= 1.4 gas, which is an excellent approximation for air:
T a
= 0.83333 ; = 0.91287
To ao
(110)
P
= 0.63394 ; = 0.52828
o Po
Another useful measure can be also introduced, the ratio of the

actual flow speed to its would-be value if it were brought to sonic
point isentropically:
M = u/u = u/a (111)
Its relationship to the actual Mach number Ma is:
2 a2 a2 +1 2
2 u o 2 Ma ( + 1)
M = 2 2 2 = = (112)
a ao a 1 + 1
2 Ma
2 2/Ma2 + ( 1)
And the inverse:
2 M2 2
Ma = +1
= (113)
2 1 2
2 M
( + 1) /M2 ( 1)
Note that M behaves like Ma, in the sense that M S 1 when

+1
Ma S 1. And that M is bounded, limMa M = 1 .
The area-velocity relationship
Compressibility introduces new twist to the familiar area-velocity

rule, familiar in incompressible flows.
Begin by rearranging the steady one-dimensional continuity
equation uA = constant:
d du dA
+ + =0 (114)
u A
Also assuming an adiabatic inviscid flow, Eulers equation gives:

dP dP d d
u du = = = a2 (115)
d
d du
= Ma2 (116)
u
What this says is that at sub / super-sonic speed, the decrease in
density is slower / faster than the increase in velocity.
Substituting the result for d/ into the continuity equation gives

the following area-velocity relationship:
1 dA du dP
1 Ma2 = = 2 (117)
A u u
Ma 1 This is the incompressible limit: uA = constant

0 < Ma < 1 In the subsonic regime, the flow behaves similarly as
incompressible, in which the flow accelerates when
cross-sectional area decreases, and vice versa
Ma > 1 In the supersonic regime, the opposite behaviour takes
place, the flow decelerates with the decrease in
cross-sectional area, and accelerates when A increases
At Ma = 1, the flow reaches the sonic point. The preceding

equation demands, that in order for du to have a finite value, the
condition dA = 0 must be fulfilled.
At the sonic point, the flow can

cross over, from being subsonic
(supersonic) to being supersonic
(subsonic).
Figure 11: Throat area of a
duct
From previous discussion, the sonic point must occurs at the throat
area. This follows from the fact that the area must decreases to
accelerate a subsonic flow, and increases to accelerate it further in
the supersonic regime, and vice versa.
However, it does not mean that at the throat area the flow is sonic.
It only says that if sonic condition is reached, it is necessarily be at
the throat area.
If sonic condition is not attained, the throat area will simply be the
location for the maximum velocity for a subsonic flow, or the
minimum velocity for a supersonic flow, since du = 0 for Ma 6= 1.
Also, be aware that in the vicinity of the sonic point, Ma 1, the
flow is very sensitive to minute area changes, since the denominator
(1 Ma2 ) becomes very small. This is a major concern in
designing supersonic nozzles and wind tunnels.
On the equations of state
For a given cross-sectional area A = A (x), and defining

eo = e + 12 u 2 , the followings quasi 1-D flow equations are valid:

(A) + (uA) = 0
t x
dA
u 2 A =

(uA) + (PA) + P (118)
t x x dx

(eo A) + (ueo A) = (PuA)
t x x
An immediate problem can be seen in that there are 4 unknowns
(, u, P, e) and only 3 equations. To close the system, one needs a
state equation.

In practice, there are actually TWO equations of state used:

Thermal equation of state which expresses P = P (T , v )
Caloric equation of state which expresses e = e (T , v )
The first of these is perhaps the more familiar one. Examples are
the ideal gas law and the van der Waals gas equation:
RT
P = RT and P= a2 (119)
1 b
From the first law:

de = T ds P dv

e S P (120)
=T P =T P
v T v T T v
However, one also has the following:

e e
de = dT + dv (121)
T v v T
Thus, one can rewrite it and then integrate to obtain the caloric
equation of state e = e (T , v ):

P
de = cv dT + T P dv
T v
Z Z Z (122)
P
de = cv dT + T P dv
T v
For example, when applying ideal gas law P = RT , theR Tsecond 0

term on the right-hand-side vanishes, leaving e = eo + To cv dT .

Bernoulli equation dynamic pressure
In compressible flow, the dynamic pressure Q = u 2 /2 is no longer

simply the difference between the stagnation and static pressures. It
also depends on the Mach number. For a perfect gas:
1
2 u 2
= 12 a2 Ma2 = 2 PMa2 (123)
For adiabatic flows, applying the ideal gas law to the energy
equation gives the compressible Bernoulli equation:
1 2
2u + 1 (P/) = 1 (Po /o ) (124)
In fact, as shown before, this applies also for isentropic flows, which
would further allow to be eliminated:
h i
1 2 (1)/
2 u = 1 (Po /o ) 1 (P/Po ) (125)
Compressibility also introduces complexities in calculating the

pressure coefficient Cp with respect to some reference section 1:

P P 2 P
Cp = 1 2
= 2
1 (126)
2 u Ma P
For isentropic flows, this can be further recast as:

!
1
2 1 + 2 Ma2 1
Cp = 2 1 2
1 (127)
Ma 1 + 2 Ma
Note well that for compressible flows, a value of Cp > 1 is possible.

Surprisingly, there is a minimum value for Cp as well, found by
letting Ma , which simply results in:
2
Cp min = (128)
Ma2
There are two Cp values that are of special interest: the stagnation
value, and the critical value, defined at the location where the local
flow turns sonic.
The critical Cp is obtainable by setting Ma = 1 as:

P P 2 2 1 2 1
Cp crit = 1 2
= 2
+ Ma 1
2 u Ma +1 +1
(129)
Similarly, setting Ma = 0 gives the stagnation Cp :

Po P 2 1 2 1
Cp stag = 1 2
= 2
1 + Ma 1 (130)
2 u Ma 2
Recall that Cp = 1 (u/u )2 for incompressible flows. It remains

to be seen how to incorporate this into the compressible Cp .
The energy equation dictates that:
1 2 1 2 2 2
2u + 1 a = 12 u + 1
1 a (131)
By rearranging terms, one can easily obtain the following
relationship:
a2 u2 1 + 1 2

1 2 2 Ma
=1+ Ma 1 2 = (132)
2
a 2 u 1 + 1
2 Ma
2
Substituting that into the Cp expression, one finally ends up with:

u2

2 1 2 1
Cp = 2
1 + Ma 1 2
1 (133)
Ma 2 u

Constant area flows
Consider the control volumes of these steady flows, all enclosing

regions of non-equilibrium, in the following three figures.
(a) Finite region (b) Sharp jump (c) Localized volume
Note that in the third figure, the control volume has been shrank to
just enclose the portion of the shock wave normal to the flow, the
streamline ab, where the one-dimensional flow assumption is
locally valid.
Constant area flows
The figures show that a region of non-equilibrium can be idealized

to be a sharp jump discontinuity within the flow, a shock wave.
Be aware that now the gradients inside the vanishingly small control
volume are greatly increased, thus there will be an increase in the
entropy of the flow as it goes through.
Since the reference sections are outside the region of
non-equilibrium, the balance equations (mass, momentum and
energy) can be expressed as:
1 u1 = 2 u2 (134)
1 u12 + P1 = 2 u22 + P2 (135)
h1 + 12 u12 = h2 + 21 u22 (136)
These will form the basis for the derivation of the shock relations
for perfect gas in the following section.
Shock relations for perfect gas Part I
The following figure depicts similar situation encountered in

deriving the speed of sound wave. However, the assumption of a
weak disturbance has been discarded.
(a) Moving frame (b) Laboratory frame
The figure on the right describe the case of moving shock. It is

related to the stationary shock through the Galilean transformation.
Notably, ushock = u1 and upiston = u2 u1 .
From the continuity (1 u1 = 2 u2 ) and momentum equations:
21 u12 22 u22

1 1
P2 P1 = = 21 u12 (137)
1 2 1 2
This is the Rayleigh line equation in the P v diagram for a given
pre-shock condition.
Further solving for 1 u1 gives:
1
1 1 1 2
21 u12 = (P2 P1 ) = (P2 P1 ) (138)
1 2 2 1
Eliminate u2 term from the energy equation:
1 2 1 1 u1 2 1 2 2 1

1
h2 h1 = u1 = 1 u1 (139)
2 2 2 2 21 22

Substituting for 21 u12 and rearranging terms yields the Hugoniot

equation:
h2 h1 = 12 (P2 P1 ) (1/2 + 1/1 ) (140)
At this point, it is worth mentioning that both the Rayleigh line and
the Hugoniot curve have been derived without any mention of any
equation of state. Thus, it can be applied to the general case. The
role of the equation of state is then to define the relationship of the
enthalpy to the pressure and density.
For example, take the special case of a calorically perfect ideal gas:
P = RT and h = cp T
(141)

cp P P
h= =
R 1
Using it in the Hugoniot curve results in the following equation:

P2 P1 1 1 1
= (P2 P1 ) +
1 2 1 2 2 1
(142)
P2 +1 1 P1 +1 1
=
1 2 1 1 1 2
This is can be put into the Rayleigh line to yield:

1 1 1 1 + 1 1
2P1 = 21 u12 (143)
1 2 1 2 2 1
By discarding the trivial solution 2 = 1 , one finally obtains the
following density and velocity ratios across a shock wave:
1 1
2 u2 +1 2 P1
= = 1+ (144)
1 u1 1 1 1 u12

The shock strength is obtained from the momentum equation:
1 u12 1 u12

P2 u2 2
1= 1 = 1 (145)
P1 P1 u1 + 1 P1
Finally, temperature ratio can be obtained from ideal gas equation:

1 " +1 #
T2 P2 2 P2 1 + (P2 /P1 )
= = (146)
T1 P1 1 P1 1 + +11 (P2 /P1 )
Recall that the stagnation enthalpy is constant across the shock

wave. For perfect gas, that implies that the stagnation temperature is
also constant:
To2 = To1 = To (147)
The second law then states that the entropy change across the shock
wave can be found from the ratio of the stagnation pressures:

s2 s1 T2 P2 Po2
= ln ln = ln
R 1 T1 P1 Po1

Po2 P2 T2 1
= (148)
Po1 P1 T1
1 " +1
#
P2 1 1 + 1 (P2 /P1 ) 1
= +1
P1 1 + (P2 /P1 )
Note that for adiabatic flows, the stagnation pressure also allows one
to measure the available energy in the flow.

Shock relations for perfect gas Part II
The shock relationships derived previously can also be expressed in

terms of the upstream Mach number Ma1 . By dividing the
momentum equation by the continuity, and applying the ideal gas
equation, one ends up with:
P1 P2 a12 a22
u2 u1 = = (149)
1 u1 2 u2 u1 u2
Invoking the sonic condition, energy equation yields:
2 2 2 2 +1 2
u22 + a2 = u12 + a = a (150)
1 1 1 1
Combining these equations gives:

+1 2 1 1 1
u2 u1 = a (u1 u2 ) (151)
2 u1 u2 2
Algebraic simplification results in the Prandtl or Meyer relation:

u1 u2 = a2 M 2 = 1/M 1 (152)
This can be used to express the downstream Mach number Ma2 as a
function of the upstream Mach number Ma1 , and vice versa:
+1 2 1 2
2 Ma 2 1+ 2 Ma 1
=
1 + 1 2
2 Ma 2
+1
2 Ma 1
2
2 2 (153)
1 + Ma 1
Ma22 = 2 2
1 Ma 1 1
It is clear that there is a lower limit imposed on Ma2 , such that:

1
lim Ma2 = (154)
Ma1 2
Now, for the other relationships:

1 +1 2
2 u2 2 Ma 1
= =
1 u1 1 + 1 2
2 Ma 1
P2 2 2
2

2 1

(155)
= 1 + +1 Ma1 1 = +1 Ma1 2
P1
T2 h 2 2
1 i h 2 2
1 i
= +1 Ma1 +1 +1 1/Ma1 + +1
T1
Note that these are valid only for calorically perfect gases.
These shock relations, however, have not restricted the direction of
the shock wave process. They are applicable when Ma1 is either
subsonic (an expansion process), or supersonic (a compression
process). One then must refer to the second law for the answer.
This figure shows the entropy change s/R across the shock as a
function of the upstream Mach number Ma1 .
This figure clearly states that only compression shock wave is

physically allowed, such that Ma1 1 and Ma2 1.
Thus, as it was previously mentioned, downstream stagnation

pressure must be less than the upstream value, Po2 Po1 .
Additionally, static pressure P and temperature T , as well as density
, all increase across the shock wave, shown in this figure.
Furthermore, based on
these shock relationships,
an upper bound exists for
the density ratio:
2 +1
lim =
Ma1 1 1
(156)
Rewrite the pressure ratio equation using = Ma21 1:

2 2
Ma21 1 = 1 + +1

P2 /P1 = 1 + +1 (157)
Put this in the equation of entropy change across the shock wave:

1 " +1
#
s2 s1 P2 1 1 + 1 (P2 /P1 ) 1
= ln +1
R P1

1 + (P2 /P1 )
(158)

1
" #
1
2 1 1
= ln 1 + +1 1 + +1 (1 + )
For small 1, the Taylors expansion gives:
ln (1 + ) = 12 2 + 13 3 14 4 + (159)

Applying the expansion:

s2 s1
h
1 2

1
i
= 1 +1 + +1 1
R
2 2
/2 1 2
1 +1 + 1
+1 1 2
3 3
/3 1 2 1 3
+ 1 +1 + +1 1
(160)
4 4
/4 1 2
1 +1 + 1
+1 1 4 +
2 3 2 2 4
= 3(+1)2
(+1) 3 +
3
2
3(+1)2
2
Ma1 1 = 12 +1
2 (P2 /P1 1)3
Again, it shows that upstream flow must be supersonic. However, a
weak shock produces a nearly isentropic change of state.
Shock relations for perfect gas Part III
The previous relationships across a shock wave shown here have

been derived under the assumption of stationary shock wave.
These can also be applied to the case of moving shock waves
through Galilean transformation. The static property ratios stays the
same, but not stagnation property ratios.
(a) Moving frame (b) Laboratory frame
For example, note that the gas at rest, the left half of the right
figure, has its stagnation properties equal to its static properties, as
opposed to its counterpart on the left figure.
The area-velocity relationship revisited
For a steady one-dimensional flow, the mass flow rate m = uA is

constant. Suppose that this flow may be adjusted isentropically to
reach sonic condition (Ma = 1, A = A ), one can then write:
uA = u A
! +1
2 2 +1 1
A /o
= M2 = Ma 2 2
1 2
A /o 1+ 2 Ma
(161)
For flows that are entirely subsonic, the term A refers to the
fictitious area where the flow would reach Ma = 1, it does not
necessarily exist. However, if sonic or supersonic conditions are
attained, then it would correspond to the actual throat area.
Be aware that this relationship is valid only for isentropic flows due
to the use of /o relation.
In this figure, the area ratio A /A is plotted as function of the Mach

number Ma.
It is clear that for a

given area ratio,
there are two
corresponding
values of Ma:
subsonic, and
supersonic.
For example, a value of A /A = 0.6 would allow a flow with either

a Mach number of Ma = 0.3778, or Ma = 1.985.

Additionally, this area relation A /A can be expressed in terms of

the ratio of local static pressure and stagnation pressure P/Po :
+1 2 1

2
A 2 + 1 1 P P
= 1 (162)
A 1 2 Po Po
This relationship would

prove to become quite
useful for the discussion
on the nozzle flows in
the next sections.
Nozzle flow converging nozzle
As discussed previously, a subsonic flow cannot be accelerated

beyond sonic velocity in a converging nozzle. This is true
irrespective of pressure difference across the nozzle.
When the ambient pressure at the exit is equal to the reservoir
value, no flow will exist, m = 0.
As this back pressure at the exit is lowered, m steadily increases.
However, when the flow at the exit reaches Mae = 1, any further
decrease in the back pressure cannot be communicated to the
reservoir, since the signal can only travels at the local speed of
sound relative to the moving fluid particle. Thus, m stays constant,
and the flow is said to be choked.
The pressure ratio, beyond which no changes in m is obtained
without changing reservoir condition, or the nozzle geometry, is
called the critical pressure ratio.

Nozzle flow converging nozzle
For subsonic flows of a calorically perfect gas in a converging

nozzle, the mass flow rate m = uA can be expressed at the exit in
terms of pressure and temperature as:
v

u 1+
Pe Ae Mae Po Ae t 1 2 1
u
m = = Ma2e 1 + Mae
RTe RTo 2
(163)
This value increases as Mae increases, and it reaches its maximum
when Mae = 1:
v
u 1+
Po Ae t + 1 1
u
mmax = (164)
RTo 2
Nozzle flow converging-diverging nozzle
Now consider the following converging-diverging nozzle connected

to a reservoir. Define the back pressure Pb as the ambient (static)
pressure just outside the exit.
The diverging part will
allow the flow to either
decelerate back to
subsonic value, or if sonic
condition is reached at the
throat, to accelerate to
supersonic flow regime.

The figures show three different

nozzle flows for three different
exit pressures, deliberately
chosen to correspond to throat
Mach numbers of 0.25, 0.5 and
1. All three flows are entirely
subsonic in the diverging part.
The ratios of the mass flow rate
to its maximum value (m/mmax )
then are 0.4162, 0.7464 and
1.000, respectively.
Note that isentropic conditions
hold throughout the nozzle.
When the flow at the throat reaches sonic speed, the mass flow can
no longer be increased, and the flow is choked.
As the back pressure is lowered even further, a region of supersonic
flow begins to appear, starting from the throat down the diverging
part of the nozzle. This will terminate in a normal shock within the
nozzle, which is located farther downstream as the back pressure
further decreases, until it reaches the exit plane of the nozzle. Ergo,
no isentropic solution is possible for this range of back pressures.
This is due to the fact that the back pressure Pb is set higher than
what the exit pressure Pe would be if the entire nozzle beyond the
throat goes supersonic. Thus, a discontinous pressure increase, a
shock wave, is required to match the exit condition.

The figures show what can

happen if the exit pressure is low
enough to allow a supersonic
flow region past the nozzles
throat. Here, two of them
terminate at shock waves inside
the nozzle.
The dashed lines, on the other
hand, indicate the flow which is
supersonic throughout the nozzle.
Several different scenarios,
however, may happen at the exit,
depending on the back pressure
Pb .
A normal shock may exist just outside the exit plane.

At lower Pb , the flow may go through oblique shock waves at
the exit, which results in smaller pressure increase overall, as
opposed of having a single normal shock. The flow is said to
be over-expanded. Note however, that the flow may then go
through alternating series of expansion waves and more oblique
shock waves. This can be seen as a diamond pattern in a
rockets exhaust.
When Pb = Pe , isentropic condition prevails. This is the
design condition of the nozzle.
Finally, when Pb < Pe , the flow is said to be
under-expanded. Expansion waves will form at the exit.
Similar to the over-expanded case, alternating series of
expansion waves and oblique shock waves can then ensue.
Normal shock recovery diffuser
When the supersonic flow from the nozzle discharges directly into
the receiver, such that a normal shock wave stands at the exit, the
subsequent subsonic flow can be decelerated isentropically by the
use of a diffuser. Essentially, its a diverging extension of the nozzle.
The aim is to recover as much stagnation pressure as possible in the
receiving chamber, instead of simply obtaining the Pb at the exit.
Another configuration utilizes a long duct ahead of the exit. The
recompression then relies on the interaction of shocks and boundary
layers. For supersonic flows, this surprisingly results in a fairly
efficient and practical method.
Normal shock recovery diffuser
A second throat at the exit can also be utilized. Ideally, it would

decelerate the supersonic flow to sonic condition before subsequent
deceleration to subsonic regime. Since there is no losses, no
pressure difference is required, hence no power is needed. Of
course, real world would never allow this to happen.
However, a pressure difference still is required to start the flow. In
principle, this may also require the ability to rapidly adjust the area
of the second throat.
As the flow turns supersonic in the test section, the shock would
then travel downstream, until it is swallowed by the second throat,
to settle at the diverging part. The throat can then be adjusted to
bring the shock closer to the throat, where it would then be of
vanishing strength. Then the flow would be close to the ideal
condition.

Flow with wall roughness Fanno flow
Fanno flow refers to the flow through a constant area duct, where
the effect of wall friction w is considered.
However, it is further assumed that the wall friction does no work,
and does not change the total enthalpy level. Thus no heat transfer
from the friction effects.
This model is applicable to flow processes which are very fast
compared to heat transfer mechanisms. Thus, heat transfer effects
are negligible. This assumption can be used to model flows in
relatively short or insulated tubes.
Wall friction will cause both supersonic and subsonic Mach
numbers to approach Mach 1.
It can be shown that for calorically perfect flows the maximum
entropy occurs at Ma = 1.
Refer to the figure.

Continuity equation reads:
m
1 u1 = 2 u2 = (165)
A
Momentum equation reads:
X
2 2

Fx = A 2 u2 1 u1
= A (P2 P1 ) Aw w
Assume:
(166)
perfect gas Energy balance dictates:
steady flow
constant area h2 + 21 u22 = h1 + 12 u12 = ho (167)
no heat transfer
Define the following hydraulic diameter:

4A
Dh = (168)
perimeter
Additionally, define the (Darcy) friction factor f, that is dependent
on the flow Reynolds number Re and wall roughness:
4w
f= 1 2
(169)
2 u
By letting the state 2 be a differential change from state 1, the
momentum equation can be written as:
1 2
2 u f
u ( du + u d) + u du = dP dx (170)
Dh
Be aware that some places may use Fanning friction factor, which is
1
4 the value of Darcy friction factor used here.
Some manipulations of the flow equations give:

m d du
u = + =0 (171)
A u
dP d dT
P = RT = + (172)
P T
2

u 2 d Ma 2 du dT
Ma2 = = (173)
RT Ma2 u T
Applying these to the momentum equation:
1 2
2 u f
u du = dP dx
Dh
2
"
2
# (174)
f dx 1 Ma d Ma dT 2 dT
= +
Dh Ma2 Ma2 T Ma2 T

Due to adiabatic
assumptions,
the total temperature
To = T 1 + 1 2 Ma
2
is constant. This results in the following:
" #
1 2
d Ma2

dT 2 Ma
+ =0 (175)
T 1 + 1
2 Ma
2 Ma 2
Plugging this in, and simplifying gives:

!
d Ma2
2

f dx 1 1 Ma
= (176)
Dh Ma2 1 + 1
2 Ma
2 Ma2
As will be shown later, the wall friction causes the flow to either
speed up or slow down toward Mach 1. Thus, sonic condition is
chosen as the reference point, corresponding to the maximum length
of the conduit.
Define a mean friction factor f, assumed to be constant:

Z Lmax
1
f = f dx (177)
Lmax 0
Setting = Ma2 and integrating the previous equation:

Z 1 !
fLmax 1 1
= 2 1
d
Dh Ma 2 1 + 2
2 1
!
Z 1 +1
1 1 4
= 2
+ 1
2 d (178)
Ma 2 1+ 2
!
2 +1 2
1 Ma +1 2 Ma
= + ln
Ma2 2 1 + 1
2 Ma
2

Surprisingly, for the supersonic flows Ma > 1, there is an

asymptotic upper limit. Letting Ma goes to infinity gives:

fLmax +1 +1 1
lim = ln (179)
Ma Dh 2 1
Starting from the equation of the first law for perfect gas:
dT d dT dP
ds = cv R = cp R (180)
T T P
Assuming sonic condition as a reference point, and applying the
continuity and energy equations:
s s 1 T u
= ln + ln
R 1 T u

1 T To T
= ln + ln (181)
1 T To T

1 T 1 +1 T 1 1
= ln + ln ln
1 T 2 2 T 2 2
This is the Fanno line curve in the T s diagram.

The following figure shows an example of a Fanno line curve. Note

that the second law dictates that the flow is driven toward Mach 1.
Other relationships can be derived for the Fanno flow as well. From
the momentum equation:
!
1 1 2 2

dP + 2 Ma d Ma
= 2 1
P 1 + 2 Ma2 Ma2
1
! (182)
1 1 2 2

= 2
+ 1
d Ma
2 Ma 1 + 2 Ma2
Again, assuming sonic condition as a reference point:
1 2
P 1 1 1+ 2 Ma
ln = ln Ma2 ln +1
P 2 2 2
v (183)
u +1
P 1 u 2
= t
P Ma 1 + 12 Ma
2

Flow with heat addition Rayleigh flow
Rayleigh flow refers to diabatic flow through a constant area duct,

where the effect of heat addition or rejection is considered.
The heat addition causes a decrease in stagnation pressure, which is
known as the Rayleigh effect and is critical in the design of
combustion systems.
Heat addition will cause both supersonic and subsonic Mach
numbers to approach Mach 1.
Conversely, heat rejection decreases a subsonic Mach number and
increases a supersonic Mach number along the duct.
It can be shown that for calorically perfect flows the maximum
entropy occurs at Ma = 1.
Refer to the figure. Continuity equation reads:

m
1 u1 = 2 u2 = (184)
A
Momentum equation reads:
1 u12 + P1 = 2 u22 + P2
m
P2 P1 = (u2 u1 )
A
(185)
Assume:
Energy balance dictates:
perfect gas
steady flow h2 + 21 u22 = h1 + 12 u12 + q
constant area q = ho2 ho1 (186)
no wall friction = cp (To2 To1 )
To simplify expressions, it is customary to use sonic condition for

reference state 2, and letting the state 1 varies.
For perfect gas, the momentum equation can be rewritten as:
P 1+
P 1 + Ma2 = P (1 + )

= (187)
P 1 + Ma2
Plugging in the continuity equation in the perfect gas law gives:

P PMa T P 2 Ma2
T = = = (188)
R (m/A) R R (m/A)2
Combining:
T P 2 Ma2 (1 + )2 Ma2
= = (189)
T P2 2 2

1 + Ma
In addition:
u P /T (1 + ) Ma2
= = = (190)
u P/T 1 + Ma2
" #" #
2 2 1 2
To (1 + ) Ma 1 + 2 Ma
= 2 +1
(191)
To 1 + Ma2

2

" 1 2
#
1
Po 1+ 1+ 2 Ma
= (192)
Po 1 + Ma2 +1
2
It would be instructive to plot the flow on the T s diagram. To do
so, one would need to express P/P in terms of T /T , since:

s s T P
= ln ln (193)
R 1 T P

From the temperature and pressure ratios, one obtains:

2 " 1 #
T P 1+ P 1
=
T P P
" s # (194)
P +1 4 T
= 1 1
P 2 ( + 1)2 T
Note that there are two entropy values for each possible temperature
values, which can be shown to correspond to the subsonic and
supersonic branches.
As mentioned before during discussion on shock relations, this
Rayleigh curve is a line on the P v diagram:

v +1 1 P
= (195)
v P
The following figure shows a Rayleigh line curve. Note that the
second law dictates that the flow is driven toward Mach 1.

The maximum temperature point can be found from:
T (1 + )2 Ma2
= 2
T 1 + Ma2
2 2
(196)
d (T /T ) (1 + ) 1 Ma
2
= 3 =0
d Ma 2
1 + Ma
Thus, the point is located at Ma2 = 1/, which gives a maximum

temperature ratio of Tmax /T = (1 + )2 / (4). For example, a
= 1.4 gas results in Tmax /T 1.0286 at Ma 0.845.
Normal shock, Fanno flow and Rayleigh flow
Capping the discussions on flows in constant area conduits, recall

the following jump relations from the stationary normal shock:
mass: (2 u2 ) (1 u1 ) = 0
2 u22 + P2 1 u12 + P1 = 0

momentum: (197)
h2 + 12 u22 h1 + 12 u12 = 0

total enthalpy:
Finally, to complete the system, the ideal gas law P = RT has

been used to provide the link between pressure and temperature, as
perfect gas, e.g.,
well as the assumptions of calorically

h = cp T = 1 RT , and Ma = u/ RT .

Other than the momentum balance equation, Fanno flow uses

exactly the same equation set.
Similarly, Rayleigh flow employs the same set, except for the total
enthalpy balance equation.
What this entails is that for a given mass flow, the points,
corresponding to the states before and after the stationary normal
shock wave, must necessarily belong to the Fanno line curve AND
the Rayleigh line curve. In fact, the two points must be located at
the intersections of the two curves.
This will be shown on the next T s diagram, starting by putting the
continuity equation in the entropy change, resulting in:
s2 s1 1 T2 2 1 T2 u2
= ln ln = ln + ln (198)
R 1 T1 1 1 T1 u1
For Fanno line, the total enthalpy equation gives:

q q
u2 /u1 = 2cp (To T2 )/ 2cp (To T1 )
1 r
2 (199)
1 2 1 2
= 2 Ma 1 1+ 2 Ma 1 (T2 /T1 )
For Rayleigh line, the momentum equation, along with ideal gas
law, gives:
2 2

0 = 2 u2 + 2 RT2 1 u1 + 1 RT1
2 1+Ma21 1
0 = (u2 /u1 ) (u2 /u1 ) + Ma
Ma21 2 (T2 /T1 )
1
(200)
r
1+Ma 2 4Ma12
u2 /u1 = 2Ma21 1 1 2 (T2 /T1 )
1 (1+Ma21 )

Here is a T s diagram for a flow at Ma1 = 2. Note that State 1

refers to supersonic pre-shock condition, while State 2 refers to
subsonic post-shock condition.
Waves in supersonic flow
The following set of figures is commonly used to illustrate some

interesting aspects of wave propagation in supersonic flows:
(a) Ma < 1 (b) Ma 1 (c) Ma > 1

Ma < 1 The upstream air is smoothly adjusted for the passing

of the object, since the acoustic signals (pressure
waves) from it travels faster than the object itself.
Ma 1 The signals from the object is travelling at about the
same speed with the object itself, so the upstream air
receives no information about the event. There exists
a zone of silence just ahead of the object. The tight
clustering of the pressure waves also gives rise to a
thin layer of compression region, which is often visible
as a shock. In the past, this sudden pressure increase
has given birth to the idea of sound barrier.
Ma > 1 Even larger region in the flow is unaware of the

object, and some parts of it may even located
physically behind it, as the information carried by the
pressure front (the Mach wave) does not have enough
time to propagate. The angle of this Mach wave is
related to the speed as the object as:

1
= arcsin (201)
Ma
And in the case of the object being a sphere, this

pressure front forms what is often called Mach
cone.

Multi-dimensional equations of the flow
For the following topics, the following simplifying assumptions

about the flow is made:
Steady flow, /t = 0.
Two-dimensional flow, uz = 0 and /z = 0. Often, a 3-D
flow can be expressed as 2-D by using an appropriate
coordinate system.
No viscous stresses or heat conduction, the flow is isentropic
except across shocks.

Calorically perfect ideal gas, P = RT and h = 1 RT .
Multi-dimensional equations of the flow
1
ux2 uy2

Setting ho = h + 2 + , the governing equations for the
steady inviscid 2-D flows considered here are:

(ux ) + (uy ) = 0
x y

ux2 + P +

(uy ux ) = 0
x y
(202)

uy2 + P = 0

(ux uy ) +
x y

(ux ho ) + (uy ho ) = 0
x y

Oblique shocks
Consider a wedge in a supersonic flow, making an angle with the

flow. As a result, the flow is deflected by that much. Since this is a
supersonic flow, the deflection is accomplished through the means
of a shock wave.
For a small enough deflection, the shock wave is a straight line,

making an angle with respect to the incoming flow. The task now
is to relate the two angles for a given upstream Mach number.
Oblique shocks
Refer to the following schematic figure of a flow across a 2-D

oblique shock. Note how a coordinate system can be set up such
that the flow velocity can be resolved into its components, one that
is normal to the shock, and another that is tangential to it. The
equations governing the flow will be applied to this rotated system.
From geometry:
u1n = u1 sin
= u1t tan
(203)
u2n = u2 sin ( )
= u2t tan ( )

Oblique shocks
In the following analysis, another assumption is made, that is

/t = 0. This is to say that the shock will NOT influence the
tangential components of the flow, further reducing it to a 1-D
system:
d
(un ) = 0
dx
d
un2 + P = 0

dx (204)
d
(un ut ) = 0
dx
d
(un ho ) = 0
dx
Oblique shocks
Upon integration, and application of boundary conditions, those

equations result in the following:
1 u1n = 2 u2n
1 u12n + P1 = 2 u22n + P2
(205)
u1t = u2t
1 2
2 u1n + 1 (P1 /1 ) = 12 u22n + 1 (P2 /2 )
However, these are similar to the relations across a normal shock.

Thus, all the properties can be obtained in a similar fashion.

Oblique shocks
The tangential velocity stays constant across the oblique shock:

u1t = u1t = ut (206)
Jumps in the other flow properties are now dependent on the

normal component of the upstream Mach number:
p
Ma1n = u1n /a1 = u1n / P1 /1 = Ma1 sin (207)
From geometry, the wave angle is found to be:

tan = u1n /ut (208)
Similarly, the wedge angle can be calculated as:

tan ( ) = u2n /ut (209)
Oblique shocks
Thus, one has the following relationships for the oblique shocks:
u2n 1
+1 2
Ma21n 1

2 2 Ma 1n
= = =1+
1 u1n 1 + 1
2 Ma 2
1 n
1 + 1
2 Ma 1n
2
P2 2

2 1

2
Ma21n 1

= +1 Ma1n 2 = 1 + +1
P1
(210)
T2 h 2 2
1 i h 2 2
1 i
= +1 Ma1n +1 +1 1/Ma1n + +1
T1
1 1
Mat2 ut /a2 a2 T2 2
= = =
Mat1 ut /a1 a1 T1

Oblique shocks
Since it is required that Ma1n 1, there is a minimum value

imposed for the wave angle , such that min = arcsin (1/Ma1n ).
Recall that this expression is encountered earlier in the discussion
on Mach wave.
Naturally, the maximum value for corresponds to the case of
normal shock, max = /2.
In addition, the following relationships can also be established:
2 2 2
1 + Ma 1 sin
Ma22n = Ma22 sin2 ( ) = 2 2 2
1 Ma 1 sin 1
s2 s1 1
= 1 ln (P2 /P1 ) 1 ln (2 /1 ) = ln (Po2 /Po1 )
R
(211)
Relationship between wedge angle and wave angle
From the shock relationship and geometry, one obtains the

following:
u2n u2n /ut tan ( ) 2 + ( 1) Ma21 sin2

= = = = (212)
u1n u1n /ut tan ( + 1) Ma21 sin2
Expanding:
tan ( ) tan tan

= =
tan tan (1 + tan tan )
(213)
2 cos Ma21 sin2 1

(1 ) sin cos
tan = 2 2
= 2
cos + sin sin Ma 1 ( + cos 2) + 2
This equation can also be inverted to give a cubic equation for

sin2 , allowing an explicit solution for , for a given Ma1 and .
The following is the chart showing the relationship. Note that there
are a maximum wedge angle max allowed for attached oblique
shock. A greater value would result in detached shock. This
maximum wedge angle divides the solution for into two branches.
There is a maximum deflection angle max associated with each

Mach number, found by setting d/d = 0.
Smaller wave angles correspond to the weak shock.
lim0 = arcsin (1/Ma1 ), a Mach wave.
Applicable to most external flows, which must matches to
acoustic wave in far-field, but may exist in some internal flows.
Except for some small region close to max , the downstream
Mach number is supersonic, Ma2 1.
Larger wave angles correspond to the strong shock.
lim0 = /2, a normal shock wave.
Applicable mostly to internal flows, depending on the level of
the back pressure.
The downstream Mach number is subsonic, Ma2 1.
An important point is that if oblique shock waves are attached
to the leading edge of an object, the upper and lower surfaces
are independent of each other.
Small angle approximation
Rearranging the previous equation for = u2n /u1n :

1 +1 1
Ma21 sin2 = 2 cot tan ( ) 2
+1
(214)
Ma21 sin2 1= 2
2 Ma 1 sin sec ( ) sin
For small deflection angle , such that , this may be

approximated as:

2 2 +1 2
Ma1 sin 1 2 Ma 1 tan (215)
Furthermore, for large upstream Mach number, this can be further

simplified into a hypersonic approximation, noting that 1
while Ma1 1:
+1
2 (216)
Mach lines
From the earlier discussions, it has been mentioned that for the
weak oblique shock, the wave angle corresponding to an
ever-smaller deflection angle approaches a non-zero limit , such
that lim0 = = arcsin (1/Ma1 ).
At this point, the flow essentially experiences no changes
whatsoever. There is nothing unique about the point, it could be
anywhere in the flow. In fact, the angle is simply a characteristic
angle associated with the Mach number Ma1 . Thus, the name Mach
angle, which is the angle of the characteristic line emanating from
some point in the flow, relative to the direction of the flow.

Mach lines
In 2-D supersonic flows, there exist 2 families of characteristic lines,

the left-running, given the negative sign, and the positive
right-running. These are named relative to the streamline. And if
the flow is non-uniform, these lines are curved, since would vary
accordingly.
The name characteristics come from the mathematical theory of
partial differential equation. Supersonic flows are considered
hyperbolic, thus they have real characteristics. Subsonic flows are
considered elliptic, and the characteristics are imaginary.
This attribute manifests itself in the fact that Mach lines have a
sense of direction, which is toward increasing time. Hence, the
fact that there are no information going upstream.
Mach lines
This concept of Mach lines is depicted in the following figures.
(a) (b)
(c)

Weak oblique shocks
For small deflection angles, the following relationship has been

derived:
2 2 +1 2
Ma1 sin 1 = 2 Ma 1 tan (217)
And since 0, the right-hand-side can be approximated in terms

of Mach angle as:

( + 1) Ma21

2 2 +1 2
Ma1 sin 1 Ma1 tan = q (218)
2 2 Ma2 1 1
This equation shall be used as the starting point for the relationship
for weak oblique shock waves.
Weak oblique shocks
From the oblique shock relations, the shock strength is found to be

in the order of O ():

2
P 2 2 2
Ma 1
= Ma1 sin 1 q (219)
P1 +1 2
Ma 1 1
In addition, the change in entropy can be shown to be proportional

to the third order of the shock strength for weak oblique shock,
3

s = O :
3
s2 s1 +1 3 + 1 Ma21 3
(P/P1 ) (220)
12 2 12 2
q
R Ma2 1 1

Weak oblique shocks
The other properties can also be shown to be proportional to the

deflection angle as well. For example:
h i
1
T2 1 + 2 (P/P1 )
= [1 + (P/P1 )] h i
T1 +1
1 + 2 (P/P1 )
(221)

Ma 2
1 + 1 2
1

Ma21 Ma21 1
1+ 2
Ma1 1 2
1 + +1
2
Ma21
Ma1 1
Weak oblique shocks
For a closer look, one can define the wave angle as a slight
increase from the Mach angle , such that = + for 1.
With this approximation, one can obtain the following expressions:
q
1 + Ma21 1
sin sin + cos =
Ma1
q
Ma21 sin2 1 + 2 Ma21 1 (222)
" #
2
( + 1) Ma1
= 2

4 Ma1 1
Hence, for a small finite deflection angle , the direction of the shock
wave differs from the Mach wave by an amount = O ().

Weak oblique shocks
The expression relating the change of flow speed u across the

wave can also be developed as follows:
q
(u2n /ut )2 + 1
p
u2 tan2 ( ) + 1 cos
=q = p = (223)
u1 2 2
tan + 1 cos ( )
(u1n /ut ) + 1
Expanding the cosine terms, assuming small angles for and :
cos ( ) = cos ( + ) = cos ( ) sin

2 2
(224)
Ma1 1() Ma1 1
= Ma = Ma 1 2
1 1 Ma1 1
Weak oblique shocks
Thus, the expression for the speed ratio becomes:

1
u2 Ma21 1
=
u1 1 2
Ma1 1

= 1 1 + 2 + (225)
Ma21 1 Ma1 1

=1 + 2 +
Ma21 1 Ma1 1

1
Ma21 1
In terms of speed change, this becomes:

u
q (226)
u1 Ma2 1 1

Supersonic compression by turning
Supersonic flow may be compressed by turning it by an angle , and

forcing it to go through an oblique shock. However, this process of
turning may be done by a succession of small turns, each with a
magnitude of .
The series of these oblique shocks will eventually intersect one
another and coalesce into a single wave. However, near the wall,
each can be treated as independent from of the following one, and
the flow may be constructed step-by-step. This is as long as the
deflection does not become too great that the flow becomes
subsonic.
And as these turns become ever smaller, 0, one then obtains
a smooth turn. Along with this, the compression approaches
isentropic limit.
The situation can be depicted as follows.

For each wave associated with each , one has the followings:
P = O () and s = O 3 .

For the complete turn, = n, the overall pressure and entropy

changes are:
Pk P1 n
1 (227)
sk s1 n3 2 3
n
Hence, as the deflections become a smooth turn, the entropy rise
decreases by a factor of 1/n2 , to finally become vanishingly small,
approaching the limit of isentropic compression.
When the compression becomes isentropic, the change of speed

across a weak oblique shock becomes a differential expression:
du d
= p (228)
u Ma2 1
This equation applies continously through the isentropic turn, and
when integrated, will give a relation between Ma and .
Moving further away from the wall, the Mach lines will ultimately
converge to form a shock wave. This is due to the fact that an
intersection of two Mach lines would imply an infinitely high
gradient, since the Mach number is double-valued at this point.
Physically, however, as the lines approach each other, the gradients
would become large enough to invalidate isentropic assumption.

This notion of smoothness ultimately relies also on the scale of the

geometry. From far enough away, any smooth bend would still
appear as a sharp corner.
In the case of a channel flow, however, if the upper wall is located
close enough, then the gradients encountered during the turning may
still be small enough to warrant the use of isentropic compression.
And since the flow is isentropic, it may be reversed without
violating the second law of thermodynamics. In this case, the
reversing flow is an expansion flow.
The converging of the Mach waves can be seen in these figures.

Supersonic expansion by turning
Since a supersonic flow can be compressed by turning it, then it

also can be expanded by turning it the other way.
Unlike compression turning, there is no such thing as an expansion
oblique shock wave, since it requires a decrease in entropy. Instead,
the Mach lines are being directed away from each other, resulting in
a divergent pattern. This, in effect, would decrease the existing
gradients. Hence, an expansion flow is isentropic throughout.
For a sharp corner, this series would be centered at the corner, while
for a smooth turn, the waves would be spread out throughout the
turn.
Supersonic expansion by turning
The following figures illustrate this concept of supersonic expansion.

The Prandtl-Meyer function
The relationship between and Ma can be integrated as:

2

d u
Z Z p
1
d = 2 Ma2 1 (229)
u2
Here, > 0 indicates compression turn, and < 0 indicates
expansion turn. One can also define the Prandtl-Meyer function
(Ma) = o , such that = 0 correspond to Ma = 1.
The term d u 2 /u 2 can be rewritten as a function of Ma using the
following relations:
1 2
2 ao2 da 2 Ma dMa
a = 1
=
1+ 2 Ma
2 a 1 + 1
2 Ma
2 Ma
(230)
du 1 dMa
1 2
u = aMa = 1 + 2 Ma
u Ma
Now integrate the relationship to obtain explicit form of (Ma):

Z Ma " p 02 #
Ma 1 dMa0
(Ma) = 1 02 0 (231)
1 1 + 2 Ma Ma
Substituting 2 = Ma2 1:
Z "
Ma2 1 2 2
#
+1 d
(Ma) =
0 1 + 1
+1
2 1 + 2
Z Ma2 1 " #
1 1
= d (232)
0 1 + 1
+1
2 1 + 2
q q
+1 1 2

= 1 arctan +1 Ma 1
p
arctan Ma2 1

As the Mach number varies from 1 to , increases

monotonically, until it reaches a maximum value:
q
+1
max = 21 1 1 (233)
This figure illustrate the relation between and in simple

isentropic turns.
Detached shocks
Given a wedge at zero angle of attack, with a half-angle < max ,

the following phenomena can be described as the free stream flow
changes from subsonic to supersonic.

Detached shocks
Alternatively, one can increase the half-angle of the wedge to detach

the shock from it as follows.
The common theme between these two series of figures is that the
deflection angle for the detached shock is greater than the value
allowed in the --Ma equation.
Detached shocks
This angle max can be found by setting d/d = 0 for a given

Mach number. Allowing = cos 2, the --Ma equation becomes:
s " #
2

1 + 1 2/Ma
tan = = f ()
1 + 2/Ma2 +

f0

d f (234)
= = g ()
d 1+f2 1+f2
1 +1
g () =
1 2 1 2/Ma2 + 2/Ma2 +

Setting g () = 0 will yield a quadratic equation in . The correct

solution is chosen by noting that || 1. Plugging it back into the
--Ma equation will yield the value for max .

Detached shocks
In practice, there is no analytical treatment for the case of detached

shocks. Studies are conducted empirically, mainly through
experimental and numerical methods.
A detached shock exhibits the whole spectrum of oblique shock
behaviour, from normal shock, to strong and weak oblique shock, all
the way to Mach angle in the far-field. The shape and its
detachment / stand-off distance depend on the Mach number and the
geometry of the body.
During the early years of hypersonic flight research, it was found
that blunt nose is preferable to sharp nose, for survivability of a
re-entry vehicle, a finding deemed important enough to stay
classified for quite a long time. Thus, one sees that the Space
Shuttles have blunt leading edges, while supersonic fighters have
razor-sharp leading edges for lower drag.
Detached shocks
The complexities associated with detached shocks are mainly caused

by the presence of subsonic flow regimes. Hence, the shock now
also depends on the interaction with the downstream condition.
The following figures show sketches of what is usually observed.

Detached shocks
Some experimental results can be seen in this figure.
Detached shocks
As illustrated earlier, detached shocks are often encountered in

transonic flow regime, shown in the following series of figures.

Shock-expansion theory
The oblique shock wave and the simple isentropic wave provide the
basic tools for analysing many cases in supersonic flows, by patching
together the solutions, of which some are shown in this figure.
For example, the drag for the diamond airfoil is found as

D = (P2 P3 ) t, where t is the thickness at the shoulder.
Recall the dAlembert paradox, which states that a symmetrical body

immersed in an inviscid fluid experiences no drag. Yet, here the
airfoil clearly suffers from drag forces, while being put in an
inviscid fluid.
This is an example of supersonic wave drag. In supersonic flows,
drag force can exist, even in the idealized inviscid flows. It is
fundamentally different from the frictional drag and separation drag,
commonly associated with boundary layers in viscous flows.
It is worth mentioning that in real fluids, the shock wave and the
boundary layers do interact. In fact, the wave drag is ultimately
dissipated by viscosity within the shock wave, but it does not
depend on the value of the viscosity coefficient. Thus, the entropy
change across the shock wave is largely independent from the
detailed non-equilibrium process inside the shock wave itself.

Another example is the curved airfoil, which has continuous

expansion along its upper and lower surfaces. For the shock waves
to be attached, the leading edge and the trailing edge should be
wedge-shaped, with half-angle less than max .
For a flat plate at an angle of attack of o , the pressure on the upper
surface is independent from the lower surface. In fact, the entropy
rise experienced by flows along the two surfaces are not equal.
Hence, a slipstream emanates from the trailing edge, inclined at
some small angle relative to the free stream, across which the static
pressure is equal.
For this flat plate having a chord length of c, the lift and the drag
forces can be found as:
L = P20 P2 c cos o P20 P2 c

0
0
(235)
D = P2 P2 c sin o P2 P2 co
The shock waves and the expansion waves also interact with each
other. Here in the figure, it is shown that expansion waves intersect
the shock wave, curving and weakening it, and ultimately reducing
it to mere Mach wave in the far-field.
For most part, the reflected waves are very weak, and thus do not
affect the shock-expansion result for the pressure distribution.
Reflection and intersection of oblique shocks
An oblique shock wave can also be reflected by a wall. And since

the wall can alternatively be represented by a streamline, the
reflection can also be seen as an intersection of identical oblique
shocks of the opposing family. These are shown in the following
figures.
If the intersecting shocks are of

unequal strength, a slipstream is seen
to divide the flow into two portions,
across which the pressure and the
flow direction are identical.
However, the velocity magnitude can
be different across, thus it is also a
shear layer.
Additionally, density and temperature
This is due to different
may also differ, making the slipstream
entropy rise experienced by
also a contact surface.
the two portions of the
flow.

When the two intersecting shocks are from the same family, it
results in the merging of the two, shown in this figure. Note that a
wave of the opposing family is needed to equalize the pressure
across the slipstream.
The other figure shows what occurs when the expansion waves
intersect an oblique shock wave. Along with the shock weakening,
there are also weak reflected waves.
A more complex phenomenon in the shock reflection occurs when

the incident shock is strong enough such that the Mach number is
lower than the detachment value for the given deflection angle.
A nearly normal shock wave then appears near the wall, giving rise
a subsonic region. This is called the Mach reflection, and the
normal leg is often referred to as the Mach stem.
Note that since the flow now also depends on the downstream
conditions, no simple solutions can be obtained.
A slipstream dividing the
subsonic and the
supersonic regions
emanates from the triple
point of intersection,
shown in this figure.

Cones in supersonic flow
The flow field for a cone in supersonic flow at zero angle of attack is
not as simple as the one for wedge, where a uniform flow region is
found between the shock and the wedge. In this 3-D case, uniform
flow behind the shock would violate the continuity equation.
However, the flow possesses a property that can be exploited for the
analysis of the flow field, that is the cone can be considered to be
semi-infinite.
With this assumption, it

can be argued that the flow
conditions are constant
along each ray emanating
from the vertex of the cone.
An example of the this type of flow is shown in this figure. Note

that the flow possesses also azimuthal symmetry, thus one only need
to consider the flow along a meridian plane.

The solution has been given, first by Busemann, and later by Taylor
and MacColl. The procedure consists of fitting an isentropic conical
flow to the conical shock. The 3-D flow equations are rewritten in
terms of a single conical variable , which is then solved
numerically.
At the shock, the simple oblique shock jump relation can be used,
since locally it always applies to any shock surface. And the flow
behind the shock must be matched with the isentropic conical flow,
which then determines the solution.
Note that behind the shock, the streamlines are curved. And
compared to the flow over a wedge, the compression is lower due to
the 3-D relieving effect. Also, the detachment occurs at a much
lower Mach number.
The following figure shows the variation of shock wave angle as a

function of Mach number Ma and the cone angle .

Additionally, this figure shows the variation of pressure coefficient

Cp as a function of Mach number Ma and the cone angle .
Derivation of perturbation equation
In many problems in aerodynamics, the interest lies in knowing the

effect of small perturbation applied to a known fluid motion. This
may result in great simplification of the problems at hand.
The most common and obvious example is how a body placed in a
unform flow would change the flow field, as shown in this figure.
(a) Uniform flow (b) Perturbed flow

Derivation of perturbation equation
In this example, the coordinate system can be chosen to align with

the free stream flow velocity U . Furthermore, it is assumed for the
undisturbed flow, the properties are constant, e.g., , p and T .
This results in Ma = U /a . Correspondingly, the velocity field
can be expressed as ~u = hU , 0, 0i.
Now, assume that the flow field is perturbed by some small amount,
caused by the presence of a solid body in the flow, such that the
pertubed velocity field becomes:
~u = hU + ux , uy , uz i (236)
Note that these are small compared to free stream velocity:

ux uy uz
, , 1 (237)
U U U
Irrotational flow
Refer to the following figure, During a small time increment dt,

showing the rotations of two point P would have moved ux dt in
elements PR and PS on the the x-direction and uy dt in the
xy -plane, about the z-axis: y -direction.
Point R would have moved
(ux /x) dx dt in the x-direction
and (uy /x) dx dt in the
y -direction, relative to P.
Similarly, point S would have moved
(ux /y ) dy dt in the x-direction
and (uy /y ) dy dt in the
y -direction, relative to P.
The angles R and S can then be
described in terms of the velocity
increments, assuming right-hand-rule.
Irrotational flow
Define the angular velocity vector

~ as the average of the rotational
deformation rate of two mutually perpendicular elements, e.g., in
z-direction:

1 1 uy ux
z = 2 dt (R S ) = 2 x y (238)
Similar expressions can also be obtained for x-direction and

y -direction, which yields the following, recognizing that ijk is the
permutation symbol:
~ = 12 ~u
k = 21 ijk (uj /xi ) (239)
The vorticity vector is defined as twice the angular velocity ~ = 2~

.
For an irrotational flow, the vorticity is zero, which yields:
ui /xj = uj /xi (240)
Irrotational flow
The 3-D momentum equation for steady, inviscid flow is:

1
(~u ) ~u + P = 0 (241)
Furthermore, the following can be obtained from a vector identity:

1
2 (~
u ~u ) = (~u ) ~u +~u ( ~u ) = (~u ) ~u + 2 (~u
~ ) (242)
The second law of thermodynamics for an adiabatic flow, where ho

is constant, can be expressed as:
1
T s = h P = 12 (~u ~u ) 1
P
(243)
~ )] [ (~u ) ~u ] = ~ ~u
= [(~u ) ~u + 2 (~u
Hence, an irrotational, adiabatic flow implies isentropicity.

Governing equations for small perturbation flows Part I
Starting with the continuity equation and the momentum equation

for 3-D flows, here written in Cartesian tensor notation:

+ (ui ) = 0
t xi
(244)
uj uj 1 P
+ ui + =0
t xi xj
For this topic, only steady, adiabatic, irrotational ( ~u = 0) flows

are considered. Hence, according to Croccos theorem, the flows are
isentropic:
P P 2
= = a (245)
xj s xj xj
Plugging this back in the momentum equation, and taking scalar

product with the velocity vector, gives:
uj a2
ui uj + uj =0 (246)
xi xj
Combining this with the continuity equation results in the following
equation for steady inviscid flow:
uj uj
ui uj = a2 (247)
xi xj
And the speed of sound a can be obtained from the energy equation
for perfect gas as follows:
h i
1 2 1 2 1 2 2 2 1 2
2 U + 1 a = 2 (U + ux ) + uy + uz + 1 a
(248)
2 2 1 2 2 2

a = a 2 2U ux + ux + uy + uz
2 :
Expanding the equation for steady inviscid flow, and dividing by a
uy
Ma2 ux uz

1 x + y + z
uy2
h i
ux2 1 ux2 uz2
= Ma2 2
U
+ 2
2ux
U + 2
U
+ 2
U
+ 2
U
ux
x
uy2 uy2
h i
1 ux2 uz2 uy
+ Ma2 2
U
+ 2
2ux
U + 2
U
+ 2
U
+ 2
U y
uy2
h i
uz2 1 ux2 uz2
+ Ma2 2
U
+ 2 + 2ux
U + 2
U
+ 2
U
uz
z 2
U
h i h i
uy uz uy
Ma2 ux ux ux 2 uz

+ U x + x + Ma U 2 z + y

h i h i
uy uy
+ Ma2 U
ux
y + x
2
+ Ma U 2 uz ux uz ux
x + z

h i h i
ux uy uy
+ Ma2 uz
U
ux uz
z + x
2
+ Ma U 2 ux
y + x

(249)
This is the full, exact equation in terms of the perturbation

velocities.
Linear terms are collected in the left-hand-side, while the non-linear
terms are all on the right-hand-side.
For small perturbations, many terms can be neglected, e.g., the
second-order terms in perturbation velocities ux , uy and uz can be
dropped out, in comparison with the first order terms.
Depending on the flow regime of interest, whether it be subsonic,
transonic, supersonic or hypersonic, even further simplification is
possible.
For example, the completely linear equation, valid for subsonic and
supersonic flows, but not for transonic flows, can be obtained by
dropping the entire right-hand-side:
2 u
1 Ma xx + yy + u
u
z = 0
z
(250)

Governing equations for small perturbation flows Part II
Earlier, the full, exact equation has been obtained:

uy
Ma2 ux uz

1 x + y + z
uy2
h i
ux2 1 ux2 uz2
= Ma2 2
U
+ 2
2ux
U + 2
U
+ 2
U
+ 2
U
ux
x
uy2 uy2
h i
1 ux2 uz2 uy
+ Ma2 2
U
+ 2
2ux
U + 2
U
+ 2
U
+ 2
U y
uy2
h i
uz2 1 ux2 uz2
+ Ma2 2
U
+ 2 + 2ux
U+ + 2
U
uz
z 2
U 2
U
h i h i
uy uz uy
Ma2 ux ux ux 2 uz

+ U x + x + Ma U 2 z + y

h i h i
uy uy
+ Ma2 U
ux
y + x
2
+ Ma U 2 uz ux uz ux
x + z

h i h i
ux uy uy
+ Ma2 uz
U
ux uz
z + x
2
+ Ma U 2 ux
y + x

(251)
The first simplification that can be done involves neglecting

second-order perturbation terms, to obtain:
uy
Ma2 ux uz

1 x + y + z
h i
(1)ux
Ma2 ux ux ux ux

= U x + U x + x
h
(1)ux

uy

uy

uy
i (252)
+ Ma2 U y + U
ux
y + x
h i
(1)ux
Ma2 uz uz ux uz

+ U z + U z + x
Furthermore, the products of the the perturbation velocities and their

derivatives can also be dropped. However, one needs to first look at
the situation as Ma 1.

As Ma 1, the ux /x term on the left-hand-side becomes ever

smaller, such that it may not be justifiable to neglect the ux /x
term on the right-hand-side.
On the other hand, this does not affect the terms involving the
uy /y and uz /z, as well as the cross-differentiation terms, and
these can be dropped from the right-hand-side of the equation to
give this non-linear equation:
2 u 2 (+1)ux ux
1 Ma xx + yy + u
u
z = Ma U
z
x (253)
Note that this equation is valid for both subsonic and supersonic, as
well as transonic flow regimes, but not the hypersonic flow regime,
which requires different approximation to be used due to the large
value of Ma .
And if the transonic regime is excluded, a fully-linear equation then
can be used, obtained by neglecting the entire right-hand-side.
It has been assumed that the flow is inviscid. An additional

condition, that the flow is irrotational, can be assumed. This
imposes the following condition on the velocity field:
~u = 0 (254)
This condition allows the existence of a perturbation velocity

potential , such that:

ux = , uy = , uz = (255)
x y z
Hence, the small-perturbation equation, which includes the transonic
flow regime, can be rewritten as:
2 2 2 2
1 Ma x 2 + y 2 + z 2 = Ma2 +1
2
U x x 2 (256)

Pressure coefficient
The definition of the pressure coefficient is:

p p 2 p p
Cp = 1 = (257)

2 U 2 Ma2 p
In terms of local velocity ~u = hU + ux , uy , uz i:

h i/(1)
Cp = Ma 2
2 1 + 1
2 Ma2 1 U~u~2u 1

h i/(1)
2 2 2
ux +uy +uz
= Ma 2
2 1 1
2 Ma 2

2ux
U + U2
1

(258)
Using binomial theorem, the following is obtained:
2
2 2
u +u +u 2 2
2 2
u +u +u 2
Cp = U2ux + x U 2y z Ma4 U2ux + x U 2y z +

(259)
Pressure coefficient
Neglecting the cubic and higher-order terms, the previous binomial

expansion can be simplified to:
h u2 u 2 2
i
2ux 2 y u
Cp = U + 1 Ma U 2x + U 2 + U 2z (260)

For 2-D and planar flows, consistency with the first-order (linear)
perturbation equation can be maintained by retaining only the first
term:
Cp = U2ux (261)
On the other hand, for axi-symmetric flows, or flows over elongated
bodies, it is necessary to include also the transverse velocity terms
uy and uz :
uy2
h i
2ux uz2
Cp = U + 2
U
+ 2
U
(262)

Boundary conditions
The inviscid condition dictates that the flow velocity must be

tangential to the surface of a solid body. This is to say that the
velocity vector must be at the right-angle (90 ) with respect to the
normal vector of the solid surface.
If the solid surface is expressed functionally as f (x, y , z) = 0, this
flow tangency condition can be expressed mathematically in
Cartesian tensor notation as:
f f f f
~u f = ui = (U + ux ) + uy + uz =0 (263)
xi x y z
Since ux can be neglected relative to U , this condition can be
further simplified as:
f f f
U + uy + uz =0 (264)
x y z
Boundary conditions
As an illustration, a two-dimensional case shall be considered, for

f
which uz = 0, and = 0. This gives:
z
uy f /x dy
= = (265)
U f /y dx
Since dy /dx f is the slope of the body, which is perpendicular
to normal vector of the body surface, and uy /U approximates the
slope of the streamline, the boundary condition requires that the
velocity component uy at the body surface should be equal to U
multiplied by the slope of the body at that point.
To justify the use of small-perturbation assumption, the body needs
to be thin, |y | 1. Applying Taylor series expansion:
2 2

uy
y uy
uy (x, y ) = uy (x, 0) + y + + (266)
y (x,0) 2 y 2 (x,0)
Boundary conditions
For the 2-D case, it is only necessary to keep the first term, leading
to the following boundary condition:
uy (x, 0) = U (dy /dx)|body (267)
The previous expression is still valid for a planar 3-D case, since it
requires f /z 0. In this case, the boundary condition becomes:
uy (x, 0, z) = U (y /x)|body (268)
Note that this does not apply to axisymmetric cases, since the
transverse perturbation velocity cannot be expanded in power series
in the neighbourhood of the axis.
Finally, the far-field boundary condition can be set by requiring the
perturbation velocities to die out, or approaching some finite value,
depending on the nature of the problem.
Flow past a wave-shaped wall an example
A simple application of the small-perturbation theory in 2-D can be

served by the following example of a flow over a wave-shaped wall.
The surface of wall can be described as a sinusoidal shape of a
magnitude and a wavelength l = 2/:
y sin (x) = 0 (269)

Flow past a wave-shaped wall an example
Since this example only deals with subsonic and moderate

supersonic flows, the linear small-perturbation can be used, subject
to finite values of ux and uy at far-field:
2 2
1 Ma2

x 2
+ y 2
=0 (270)
At the wall, flow tangency is enforced:

dy
uy (x, 0) = U = U cos (x) (271)
dx body
Due to vastly different behaviours of the two flow regimes, the

subsonic case and the supersonic case shall be investigated
separately.
Flow past a wave-shaped wall subsonic case
Let m2 = 1 Ma2 , and rewrite:
2 1 2
+ 2 =0 (272)
x 2 m y 2
Using the separation of variable technique, (x, y ) = X (x) Y (y ),

the following is obtained:
X00 1 Y00
= 2 = k2 (273)
X m Y
This gives two ordinary differential equations, whose solutions are:
X = X1 cos (kx) + X2 sin (kx) (274)

Y = Y1 cosh (mky ) + Y2 sinh (mky ) (275)

Flow past a wave-shaped wall supersonic case
In the supersonic case, define 2 = Ma2 1, and rewrite:
2 1 2
2 =0 (276)
x 2 y 2
This equation is in the form of what is commonly referred to as the
second order wave equation. The solutions are simply the sum of
two arbitrary functions f and g :
(x, y ) = f (x y ) + g (x + y ) (277)
The lines x y = constant are recognizable as the characteristics

of the flow, carrying information toward downstream and upstream,
respectively.
But since this is a supersonic flow with limited upstream influence,
only one solution is valid, which is f .

Introduction To Gas Dynamics All Lecture Slides: Autumn 2009

Uploaded by

Copyright:

Available Formats

Introduction To Gas Dynamics All Lecture Slides: Autumn 2009

Uploaded by

Document Information

Original Description:

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

Introduction To Gas Dynamics All Lecture Slides: Autumn 2009

Uploaded by

Copyright:

Available Formats

Introduction to Gas Dynamics

All Lecture Slides

Teknillinen Korkeakoulu / Helsinki University of Technology

Gasdynamics Lecture Slides

31 Nozzle flow converging-diverging nozzle

In a nutshell, the term compressible flow refers to the fluids of

Gasdynamics Lecture Slides

To illustrate, consider a car at sea-level, 1 atm and 15 C, going at a

Gasdynamics Lecture Slides

Gasdynamics Lecture Slides

The concept of equilibrium is fundamental to the study of

Gasdynamics Lecture Slides

First law of thermodynamics

The first law is a statement of energy conservation. It states that the

Note that E is a variable of state, while Q and W depend on the

Gasdynamics Lecture Slides

Consider this simple

Gasdynamics Lecture Slides

First law of thermodynamics

However, it can also be expressed in terms of pressure P and the

Gasdynamics Lecture Slides

One can also distinguish between extensive and intensive variables

Gasdynamics Lecture Slides

First law of thermodynamics

Besides the energy E , one can also introduce the concept of

Thus, the first law can be expressed using enthalpy as:

where = 1/v is the mass density of the system.

Measurements of thermal properties of gases have shown that for

Gasdynamics Lecture Slides

Heat capacity of a substance C , an extensive property, is defined as

Gasdynamics Lecture Slides

In thermodynamics, it is empirically shown that, for a simple

Gasdynamics Lecture Slides

Assume that specific energy e is a funtion of v and T . Then:

Gasdynamics Lecture Slides

Likewise, using specific enthalpy h as a funtion of P and T :

Gasdynamics Lecture Slides

The perfect gas

No gas is truly perfect. However, it is the simplest working

From the definition of enthalpy:

Gasdynamics Lecture Slides

Gasdynamics Lecture Slides

The perfect gas

The followings then apply for perfect gases:

Gasdynamics Lecture Slides

An important special case is the calorically perfect gas, where the

Gasdynamics Lecture Slides

Second law of thermodynamics

The first law of thermodynamics dictates that energy must be

In this experiment, the weight is allowed to

Gasdynamics Lecture Slides

For a reversible change of state of an adiabatically enclosed system,

Now assume that similar relation can be found for T , using a

Gasdynamics Lecture Slides

Second law of thermodynamics

The concept of entropy can be generalized for any arbitrary process

Gasdynamics Lecture Slides

Adiabatic, reversible process