An Introduction to Compressible Flow for Propulsion Applications

Antonio L. Sánchez

MAE, UCSD

January 2017
 Preface
The following notes are intended as a short introduction to Compressible Flow to be used in
connection with 113 (Fundamentals of Propulsion)

2
Contents

1 Ideal Gas Flow 1

The Navier-Stokes equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
The speed of sound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
The Mach number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Steady gas flow around a sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
The Euler equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Quasi-steady flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Steady ideal flow of a gas: stagnation flow properties . . . . . . . . . . . . . . . . . . . 6

2 Shock and Expansion Waves 9

Jump conditions across discontinuity surfaces . . . . . . . . . . . . . . . . . . . . . . . 9
Normal discontinuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Jump conditions across shock waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Oblique shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Prandtl-Meyer expansions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3 Nozzle Flow 25
Quasisteady ideal gas flow in pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Subsonic and supersonic flow in pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Critical magnitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Flow in a convergent nozzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Flow in convergent-divergent nozzles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

i
ii
List of figures

1.1 Perturbation regions for subsonic and supersonic flows. . . . . . . . . . . . . . . . 2

1.2 Steady gas motion around a sphere. . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1 Jumps across a discontinuity surface. . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.2 Hugoniot curve and Rayleigh line. . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Velocity jumps across a normal discontinuity. . . . . . . . . . . . . . . . . . . . . 14
2.4 Piston-supported shock wave. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 The variation of β with δ for constant values of M1 . . . . . . . . . . . . . . . . . 17
2.6 The variation of p2 /p1 with M1 for constant values of M2 and δ. . . . . . . . . . 18
2.7 Supersonic flow over a wedge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.8 Supersonic flow over a blunt body. . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.9 Supersonic flow over a concave wall. . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.10 Supersonic flow over a convex wall. . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.11 Differential evolution across a Mach wave. . . . . . . . . . . . . . . . . . . . . . . 24
2.12 Supersonic stream discharging to the vacuum. . . . . . . . . . . . . . . . . . . . . 24

3.1 The variation of the exit Mach number M (L) and mass flow rate G with pc /pa as
obtained from (3.16) and (3.17) for γ = 1.4. . . . . . . . . . . . . . . . . . . . . . 28
3.2 Pressure and Mach number distributions along a nozzle for different values of pa /pc . 29
3.3 Flow in a convergent-divergent nozzle. . . . . . . . . . . . . . . . . . . . . . . . . 31

iii
Chapter 1

Ideal Gas Flow

The Navier-Stokes equations

To study the motion of a perfect gas, the conservation equations of continuity
∂ρ
+ ∇ · (ρv̄) = 0, (1.1)
∂t
momentum
Dv̄
ρ = −∇p + ∇ · τ̄¯′ + ρf¯m , (1.2)
Dt
and energy
De
= −p∇ · v̄ + τ̄¯′ : ∇v̄ − ∇ · q̄,
ρ (1.3)
Dt
must be supplemented with the equations of state
p
ρ= and e = eo + cv T (1.4)
Rg T

together with the constitutive equations

2
τ̄¯′ = µ(∇v̄ + ∇v̄ T ) + (µB − µ)(∇ · v̄)¯Ī and q̄ = −k∇T (1.5)
3
and associated state functions

µ = µ(T ), µB = µB (T ) and k = k(T ). (1.6)

The energy equation (1.3), which is written for a non-radiative chemically frozen gas, can be
alternatively expressed in terms of the enthalpy h = e + p/ρ to give

Dh Dp
ρ − = τ̄¯′ : ∇v̄ − ∇ · q̄, (1.7)
Dt Dt
and also in terms of the entropy, defined from T ds = dh − dp/ρ, yielding

Ds
ρT = τ̄¯′ : ∇v̄ + ∇ · (k∇T ). (1.8)
Dt
 Ideal Gas Flow

The thermal equations of state can be written for the enthalpy and entropy as
 
s − so p
h = eo + cp T and = ln , (1.9)
cv ργ
where cv and cp are the specific heat at constant volume and at constant pressure for the gas,
with γ = cp /cv . According to (1.9), a flow is effectively isentropic when the entropy variations
that appear ∆s are much smaller than the corresponding specific heat.

The speed of sound

Besides the enthalpy and the entropy, another thermodynamic quantity of interest in fluid-flow
analysis is the speed of sound, a, which is the velocity at which the small pressure perturbations
propagate in a compressible medium at rest, to be computed according to
 
2 ∂p
a = . (1.10)
∂ρ s

For a perfect gas, straightforward differentiation of the second equation in (1.9) leads to

a = (γp/ρ)1/2 = (γRg T )1/2 , (1.11)

indicating that the speed of sound is a function of the temperature. The reader can verify that
for air at T = 300 K the above expression yields a ≃ 340 m/s.

The Mach number

at at
v<a (a−v)t v>a (v−a)t
vt vt
(a+v)t (a+v)t

Figure 1.1: Perturbation regions for subsonic and supersonic flows.

Compressibility effects associated with the fluid motion are characterized by the Mach number
v
M= , (1.12)
a
defined as the ratio of the velocity magnitude v = |v̄| to the speed of sound a. Flows can be
subsonic, sonic or supersonic if the value of the Mach number is M < 1, M = 1, or M > 1,

2
 Ideal Gas Flow

respectively. Since the perturbations move with respect to the flow with the speed of sound,
subsonic and supersonic flow show markedly different behaviors, as sketched in Fig. 1. As can be
seen, a perturbation in a subsonic flow ends up affecting all of the flow field, because it propagates
upstream sufficiently fast to overcome convection. In supersonic flow, however, convection sweeps
downstream the perturbation, which affects only a conical region (termed the “Mach cone”) of
semi-angle α such that sin(α) = 1/M .

Steady gas flow around a sphere

To compute a given flow field one needs to integrate (1.1) and (1.2) together with one of the
alternative forms of the energy equation - i.e., (1.3), (1.7), or (1.8) - with appropriate initial and
boundary conditions. To begin to introduce some needed concepts, it is of interest to consider,
as an illustrative example, the steady buoyancy-free motion of a gas around a sphere of radius
R with negligible radiation and chemical reaction. The ambient conditions in this case are given
by v̄ = U∞ ēx , T = T∞ , and p = p∞ whereas on the sphere surface we find v̄ = 0 and T = Tw ,
where Tw is assumed to be constant.

T
8

p
Tw z y
8

U
8

R x

Figure 1.2: Steady gas motion around a sphere.

In this case, the problem reduces to that of integrating

∇ · (ρv̄) = 0 (1.13)
ρv̄ · ∇v̄ = −∇p + |∇{z
· τ̄} ¯′ (1.14)
| {z } | {z }
2 /R
ρ∞ U∞ ∆p/R µU∞ /R2

ρT v̄ · ∇s ¯′
= τ̄| {z
: ∇v̄} + ∇ · (k∇T ) (1.15)
| {z } | {z }
2 /R2
µU∞
ρ∞ T∞ U∞ ∆s/R k∆T /R2

supplemented with the constitutive law for τ̄¯′ given in (1.5) and with the equations of state
 
p s − s∞ p/p∞
ρ= and = ln (1.16)
Rg T cv (ρ/ρ∞ )γ

and subject to the boundary conditions


|x̄| = R : v̄ = 0, T = Tw
(1.17)
|x̄| ≫ R : v̄ = U∞ ēx , T = T∞ , p = p∞ .

3
 Ideal Gas Flow

Note that, for convenience, the thermal equation of state in (1.16) has been written using the
ambient conditions for the arbitrary reference state.
The estimated orders of magnitude are given below each term in the momentum and energy
equations. The character of the solution depends on the value of the Reynolds number
O(ρv̄ · ∇v̄) ρ∞ U∞ R
Re = = . (1.18)
O(τ̄¯′ ) µ
When Re ≫ 1, the viscous forces are found to be negligible in most of the flow field. The
simplified balance equation that appears can be used to estimate the characteristic value of the
spatial pressure differences appearing across the flow field according to
2
ρv̄ · ∇v̄ ∼ −∇p → ∆p = ρ∞ U∞ . (1.19)
| {z } | {z }
2 /R
ρ∞ U∞ ∆p/R

It is of interest to compare these differences with the ambient value p∞ according to

 
∆p ρ∞ U∞2 U∞ 2 2
= =γ = γM∞ . (1.20)
p∞ p∞ a∞
As can be seen, the compressibility ratio ∆p/p∞ is proportional to the square of the Mach
number
M∞ = U∞ /a∞ . (1.21)
Appreciable compressibility effects appear in flows with M∞ > ∼ 1, whereas in flows with M∞ ≪ 1
the resulting spatial pressure differences are much smaller that the ambient pressure. In the latter
case, one may rewrite the equation of state given in (1.16) in the simplified form p∞ = ρRg T .
According to the energy equation (1.15) the entropy of a fluid particle crossing the flow field
changes due to the effects of viscous dissipation and thermal heat conduction. The corresponding
characteristic entropy increase due to each separate effect, (s − s∞ )φ and (s − s∞ )k , can be
estimated by order-of-magnitude comparisons according to
(s − s∞ )φ M2
ρT v̄ · ∇s ∼ τ̄|¯′ {z
: ∇v̄} → = γ(γ − 1) ∞ (1.22)
| {z } cv Re
2 /R2
µU∞
ρ∞ T∞ U∞ (s−s∞ )φ /R

and
(s − s∞ )k γ ∆T 1
ρT v̄ · ∇s ∼ ∇ · (k∇T ) → = . (1.23)
| {z } | {z } cv P r T∞ Re
ρ∞ T∞ U∞ (s−s∞ )k /R k∆T /R2

where P r = µcp /k is the Prandtl number (P r = 0.7 for air). The above expressions clearly
indicate that when the Reynolds number is large the resulting flow is effectively isentropic,
thereby reducing (1.15) to s = s∞ .

The Euler equations

According to the previous discussion, the effects of molecular transport, including viscous forces,
heat conduction and also viscous dissipation, can be neglected in the first approximation when-
ever the condition Re ≫ 1 is satisfied, where Re = ρc vc Lc /µc is the Reynolds number of the

4
 Ideal Gas Flow

problem, defined from the characteristic values of the density and viscosity, ρc and µc , velocity
vc , and length Lc 1 . The simplified set of conservation equations, called the Euler equations, is
given by
∂ρ
+ ∇ · (ρv̄) = 0, (1.24)
 ∂t 
∂v̄ Dv̄
ρ + v̄ · ∇v̄ = ρ = −∇p + ρf¯m , (1.25)
∂t Dt
 
∂s Ds
ρT + v̄ · ∇s = ρT = 0. (1.26)
∂t Dt
This last equation indicates that the entropy of each fluid particle remains constant in the flow
evolution. We say that the flow is isentropic. If at the initial instant all fluid particles have the
same value of s, then in the subsequent evolution the entropy will remain uniform in the flow
field, so that ∇s = 0; when this happens, the flow is said to be homentropic.
For the integration of (1.24)–(1.26), one needs to provide initial conditions, given by the initial
values of the velocity and thermodynamic state at every point of the integration domain. Because
of the absence of the second-order spatial derivatives, the Euler equations cannot however satisty
all of the boundary conditions. For the aerodynamic problem of flow around a body the subset
of boundary conditions to be satisfied by the Euler equations (1.24)–(1.26) include those far from
the body, i.e., v̄ = U∞ ēx , p = p∞ , and T = T∞ at |x̄| ≫ R. On the body surface, however,
we cannot impose either the temperature or the velocity, so that integration of (1.24)–(1.26)
provides in general a solution with T 6= Tw and v̄ 6= 0 on the body surface |x̄| = R. For the
velocity, the presence of the pressure forces in the momentum equation prevents the fluid from
penetrating the solid surface, so that the resulting velocity component normal to the wall is
strictly zero, that is, for integrating (1.24)–(1.26) we must impose v̄ · n̄ = 0 on the body surface.

Quasi-steady flow
For a flow with characteristic length Lc , characteristic velocity vc , and characteristic time of
flow variation tc , the importance of the unsteady terms in the conservation equations is mea-
sured through the so-called Strouhal number St, which compares the orders of magnitude of the
unsteady and convective terms according to
O(∂ρ/∂t) O(∂v̄/∂t) O(∂s/∂t) Lc /vc
St = = = = . (1.27)
O(∇ · (ρv̄)) O(v̄ · ∇v̄) O(v̄ · ∇s) tc
For a given problem, the values of Lc , vc , and tc can be estimated from the boundary conditions.
For instance, for flow of a fluid stream of variable velocity U∞ (t) (e.g., U∞ = Uo sin Ωt) around
a body (e.g., a sphere), the value of Lc will be given by the characteristic body size (e.g., the
1
Note that, regardless of the value of the Reynolds number, molecular transport is also negligible in unsteady
flows when the condition
ρc L2c
≫1
µc tc
is satisfied, where tc is the characteristic time of flow variation. Under those conditions, the molecular-transport
terms are found to be negligible compared with the unsteady terms, so that, for instance, ρ∂v̄/∂t ≫ ∇ · τ̄¯′ in the
momentum conservation equation.

5
 Ideal Gas Flow

sphere radius R), whereas the values of vc and tc will be the characteristic free-stream velocity
(e.g., its mean value vc = Uo ) and the characteristic time of free-stream velocity variation (e.g.,
tc = Ω−1 ).
Flows with St ≪ 1 are said to be quasi-steady, in that, for their description, one may in the
first approximation neglect the time derivatives in the conservation equations, which therefore
become

∇ · (ρv̄) = 0, (1.28)
ρv̄ · ∇v̄ = −∇p + ρf¯m , (1.29)
v̄ · ∇s = 0. (1.30)

The flow is termed quasi-steady, as opposed to strictly steady, because the resulting solution
depends on the time through the slowly-varying boundary conditions. For instance, for the
sphere example with variable free-stream velocity, when St = RΩ/Uo ≪ 1 the instantaneous
velocity field around the sphere v̄(x̄, t) will be that corresponding to steady motion with free-
stream velocity U∞ = Uo sin Ωt. Note that the condition St ≪ 1 implies that the fluid-particle
residence time Lc /vc is much smaller than the time required to change appreciably the boundary
conditions, so that the boundary conditions remain almost constant during the time that it takes
each fluid particle to cross the flow field. As a result, each individual fluid particle behaves as
though the flow were steady, with the values of the boundary conditions corresponding to the
instant of time at which they cross the flow field.

Steady ideal flow of a gas: stagnation flow properties

When the mass forces are conservative, i.e., f¯m = −∇U , we may rewrite (1.29) in the form

∇p
v̄ · ∇v̄ = − − ∇U. (1.31)
ρ

The projection of this vector equation along a given stream line, obtained by scalar multiplication
by v̄/v with use of the identity v̄ · ∇v̄ = ∇(v 2 /2) − v̄ ∧ (∇ ∧ v̄), yields

∂ 2
1 ∂p
v /2 + U + = 0, (1.32)
∂l ρ ∂l

where
∂ v̄
= ·∇ (1.33)
∂l v
denotes the derivative along stream lines.
According to (1.30),
∂s
= 0, (1.34)
∂l
i.e. the entropy is constant along stream lines, with a value

s = s0 , (1.35)

6
 Ideal Gas Flow

that may be different for different stream lines and may also vary slowly with time. This last
result can be used to simplify the description of steady gas flow at large Reynolds numbers.
Using the definition of entropy T ds = dh − dp/ρ along with (1.34) yields
∂h 1 ∂p
= , (1.36)
∂l ρ ∂l
which can be substituted into (1.32) to give
 
∂ v2
h+ + U = 0. (1.37)
∂l 2
In many gas-flow applications mass forces have a negligible effect on the gas motion, because the
velocity is sufficiently large for the associated Froude number to satisfy
O(ρv̄ · ∇v̄) vc2
Fr = = ≫ 1. (1.38)
O(ρf¯m ) Lc fmc
When this condition is satisfied, then (1.37) indicates that
v2
h+ = h0 , (1.39)
2
is constant along a given stream line. For known values of s0 and h0 , the equations (1.35)
and (1.39) together with the equations of state

T = T (s, h), p = p(s, h), ρ = ρ(s, h), and a = a(s, h) (1.40)

enable the computation of the values of s, h, T , p, ρ, and a in terms of the local value of the
flow velocity v. The constants s0 and h0 together with the corresponding values T0 , p0 , ρ0 , and
a0 , obtained by substituting the values s0 and h0 in the equations of state (1.40), are called
stagnation flow properties. The explicit form of the equations of state for a perfect gas, given
above in (1.4), (1.9), and (1.11), can be used to express the different stagnation flow properties
in terms of the local Mach number according to
 (γ−1)/γ  γ−1
h0 T0  a0 2 p0 ρ0 γ−1 2
= = = = =1+ M . (1.41)
h T a p ρ 2
For steady (or quasi-steady) ideal gas flow with negligible body forces, the values of s0 ,
h0 , T0 , a0 , p0 , and ρ0 remain constant along a given stream line and (1.41) can be employed to
compute h, T , a, p, and ρ as a function of the local value of the Mach number M . For M ≪ 1,
equation (1.41) indicates that the gas moves with small variations of the temperature, density,
and pressure. The latter can be computed by expanding
 
p0 γ − 1 2 γ/(γ−1)
= 1+ M (1.42)
p 2
for M ≪ 1 to give the familiar Bernoulli’s equation

p + ρ0 v 2 /2 = p0 , (1.43)

thereby illustrating the similarities between gas motion at low Mach numbers and liquid motion.

7
 Ideal Gas Flow

8
Chapter 2

Shock and Expansion Waves

For the solution of the Euler equations to represent adequately a given large-Reynolds-number
flow, we need to consider in general the existence of discontinuity surfaces, across which the
fluid properties or their derivatives exhibit finite jumps. Mach cones, for instance, represent an
example of weak discontinuities with jumps in derivatives. Boundary layers (and also vortex
sheets and mixing layers) are examples of tangential discontinuities. We shall focus below on
the analysis of normal discontinuities, showing that they are always compression waves, also
called shock waves, with specific results given below for the jump conditions across normal
and oblique shocks. Conversely, expansion waves cannot be discontinuities; rather, they are
continuous isentropic waves, to be analyzed below in the specific case of steady flow around a
sharp corner (the so-called Prandtl-Meyer expansion flow).

Jump conditions across discontinuity surfaces

To determine the conditions that relate the different flow properties that exist on the sides of a
given discontinuity surface we shall apply the conservation equations to the infinitesimally small
control volume shown in figure 2.1. As can be seen, the control volume includes faces of area
dσ, lying on the two sides of the surface, separated by a small distance ε. In a reference frame
moving with the discontinuity, mass conservation can be expressed in the form
Z Z
d
ρdV + ρv̄ · n̄dσ = 0. (2.1)
dt Vc Σc

The first integral is proportional to the volume of the control volume, of order εdσ, and therefore
becomes negligibly small when a infinitesimally thin control volume satisfying ε ≪ (dσ)1/2 is
considered. In that limit, only fluxes across the two faces of area dσ need to be computed when
evaluating the second integral, which therefore provides the simplified continuity equation

ρ1 v1n = ρ2 v2n = m, (2.2)

where m denotes the local value of the mass flux across the discontinuity and the subscript
n is used to denote normal velocity components, with the subscript t employed below for the
tangential components. Volume integrals can also be neglected when evaluating the momentum
 Shock and Expansion Waves

and energy conservation equations, which can be seen to provide

2 2
ρ1 v1n + p1 = ρ2 v2n + p2 (2.3)
ρ1 v1n v1t = ρ2 v2n v2t (2.4)

and    
p1 v12 p2 v22
m e1 + + = m e2 + + . (2.5)
ρ1 2 ρ2 2

v1 t v1 n n
n
dσ
v1 v2
v2 n v2 t
ε

Figure 2.1: Jumps across a discontinuity surface.

Equations (2.2)–(2.5) apply to all discontinuities. Tangential discontinuities, such as boundary

layers and vortex sheets, are defined by the condition that the mass flux is identically zero, that
is, m = 0. From (2.2) it then follows that v1n = v2n = 0 and, from (2.3), it is seen that p1 = p2 ,
indicating that the jump in pressure across tangential discontinuities is zero, a condition used
before when analyzing jets and boundary layers. With m = 0, the conservation of tangential
momentum and energy, expressed in (2.4) and (2.5), are identically satisfied, regardless of the
values of v1t 6= v2t , ρ1 6= ρ2 , and T1 6= T2 .

Normal discontinuities
When the mass flux across the discontinuity is nonzero, then it follows from (2.4) that

v1t = v2t , (2.6)

whereas (2.5) reduces to

v12 v2
h1 + = h2 + 2 , (2.7)
2 2
indicating that, in a reference frame moving with the discontinuity, the stagnation enthalpy (and,
therefore, the stagnation temperature) is conserved.

10
 Shock and Expansion Waves

For given values of v1n , ρ1 , and p1 , equations (2.2), (2.3), and

γ p1 v1n 2 γ p2 v2n 2
+ = + (2.8)
γ − 1 ρ1 2 γ − 1 ρ2 2
determine v2n , ρ2 , and p2 , with (2.8) corresponding to the energy conservation balance (2.7)
written with use made of (2.6) and of the thermodynamic identity h = [γ/(γ − 1)]p/ρ.
Straightforward manipulation of (2.2) and (2.3) leads to the equation of the so-called Rayleigh
line  
p2 m2 ρ1
−1= 1− . (2.9)
p1 ρ1 p1 ρ2
On the other hand, writing (2.8) in the form
    
γ p2 p1 1 2 2 1 2 1 1 1 1
− = (v1n − v2n ) = m − + (2.10)
γ − 1 ρ2 ρ1 2 2 ρ1 ρ2 ρ1 ρ2
and using (2.9) to eliminate m2 yields
γ+1 ρ1
p2 γ−1 − ρ2
= γ+1 ρ1 . (2.11)
p1 γ−1 ρ2 − 1

This last equation, relating the jumps in pressure and density across the discontinuity, is known
as Hugoniot curve. Observation of the numerator and denominator in (2.11) indicates that the
solution for the density ratio across the discontinuity must lie in the range
γ−1 ρ1 γ+1
< < (2.12)
γ+1 ρ2 γ−1
while the range of possible pressure ratios extends in principle for 0 < p2 /p1 < ∞, giving the
solution shown in Fig. 2.2 for γ = 1.4.
For a given mass flux m, the values of p2 /p1 and ρ1 /ρ2 are determined by the point where
the Rayleigh line (2.9) and the Hugoniot curve (2.11) cross in Fig. 2.2 (of course, a second
solution p2 /p1 = ρ1 /ρ2 = 1 always exists, but that is just the trivial solution corresponding to a
continuous flow with no jumps).
To investigate the different possible solutions, we begin by noting that the slope of the Rayleigh
line can be written in terms of the Mach number associated with the normal component of the
incident velocity M1n = v1n /a1 according to
m2 v2 2
− = − 1n = −γM1n . (2.13)
ρ1 p 1 p1 /ρ1
This value is to be compared with the slope of the Hugoniot curve at ρ1 /ρ2 = 1, which can be
computed by straightforward differentiation of (2.11) to give
 
d(p2 /p1 )
= −γ. (2.14)
d(ρ1 /ρ2 ) ρ1 /ρ2 =1

For M1n = 1 the Rayleigh line is tangent to the Hugoniot curve at p2 /p1 = ρ1 /ρ2 = 1, which is
therefore the only solution of (2.9) and (2.11), whereas a nontrivial crossing point appears for
M1n 6= 1 according to

11
 Shock and Expansion Waves

10

Hugoniot curve
8

6
p 2 /p 1

Rayleigh line (M1n > 1)

3

Rayleigh line (M1n < 1) p2/p1 = (ρ2/ρ1)−γ

1

0
0 1 2 3 4 5 6
(γ−1)/(γ+1) ρ1/ρ2 (γ+1)/(γ−1)

Figure 2.2: Hugoniot curve and Rayleigh line.

• If the normal component of the incident velocity is supersonic (M1n > 1), then the Rayleigh
line crosses the Hugoniot curve along the upper stretch, so that the solution corresponds
to a compression with
p2 γ−1 ρ1
∞> > 1 and < <1
p1 γ+1 ρ2

• If the normal component of the incident velocity is subsonic (M1n < 1), then the Rayleigh
line crosses the Hugoniot curve along the lower stretch, so that the solution corresponds to
an expansion with
p2 ρ1 γ +1
1> > 0 and 1 < <
p1 ρ2 γ −1

Although both compressions and expansions are apparently valid solutions of (2.9) and (2.11),
only compression waves, termed shock waves, may exist in reality. To prove this point, one
may use thermodynamic arguments based on the second principle, which indicates that, in an
adiabatic process, the entropy change s2 − s1 can never be negative, i.e., the entropy can never

12
 Shock and Expansion Waves

be destroyed. For a perfect gas,

 
s2 − s1 p2 /p1
= ln , (2.15)
cv (ρ2 /ρ1 )γ
and one may use (2.11) to compute the entropy variation. As seen in Fig. 2.2, for ρ1 /ρ2 < 1 the
Hugoniot curve (2.11) lies above the isentrope p2 /p1 = (ρ2 /ρ1 )γ . Compression waves, therefore,
are associated with positive entropy increments
   
γ+1 ρ1 γ+1 ρ1
s2 − s1 γ−1 − ρ2 / γ−1 ρ2 − 1
= ln   > 0, (2.16)
cv (ρ2 /ρ1 )γ

as can be calculated from (2.15). Conversely, in expansion waves with ρ1 /ρ2 < 1 the Hugoniot
curve (2.11) lies below the isentrope p2 /p1 = (ρ2 /ρ1 )γ , so that the entropy change computed
from (2.15) would be negative (s2 − s1 < 0), thereby violating the second principle of thermo-
dynamics.
Hence, normal discontinuity surfaces are always shock waves, with jumps of thermodynamic
properties in the ranges
p2 γ−1 ρ1 T2
∞> > 1, < < 1, and ∞ > > 1, (2.17)
p1 γ+1 ρ2 T1
the latter to be computed from the equation of state T2 /T1 = (p2 /p1 )/(ρ2 /ρ1 ). The normal
component of the incident velocity v1n must be supersonic, so that the inequalities
M1 > M1n > 1 (2.18)
are always satisfied. Use of (2.3) and (2.6) leads to the additional conditions
γ−1 v2n v2
< < 1 and < 1. (2.19)
γ+1 v1n v1
It can also be seen that
M2n < 1. (2.20)
It should be noted that, despite this last result, in oblique shock waves the downstream velocity
can be however supersonic, i.e., M2 > 1.

Jump conditions across shock waves

To resolve for the jumps across a generic shock wave, it is convenient to consider first the change
in the normal component of the velocity along with the jumps in thermodynamic properties,
and address subsequently the deflection of the flow δ as a function of the incident angle β (see
Fig. 2.3). The development begins by using (2.2), (2.3), and (2.8), which are rewritten below for
convenience
ρ1 v1n = ρ2 v2n , (2.21)
2 2
ρ1 v1n
+ p1 = ρ2 v2n+ p2 , (2.22)
γ p1 v1n 2 γ p2 v2n 2
+ = + . (2.23)
γ − 1 ρ1 2 γ − 1 ρ2 2

13
 Shock and Expansion Waves

v1 v2 t v2 n
v2
β
v1 n v1 t δ

Figure 2.3: Velocity jumps across a normal discontinuity.

For a gas of known specific-heat ratio γ, these three equations determine the downstream values
v2n , p2 , and ρ2 in terms of v1n , p1 , and ρ1 . The dependence can be simplified by using the Π
theorem of dimensional analysis to give
 
v2n = f1 (v1n , p1 , ρ1 , γ)   v2n /v1n = ϕ1 (M1n , γ)
p2 = f2 (v1n , p1 , ρ1 , γ) ⇒ p2 /p1 = ϕ2 (M1n , γ) (2.24)
 
ρ2 = f3 (v1n , p1 , ρ1 , γ) ρ2 /ρ1 = ϕ3 (M1n , γ)

The functions ϕ1 , ϕ2 , and ϕ3 can be determined by manipulation of (2.21)–(2.23) to give (we

leave the needed development as an exercise for the reader)
 −1 2 2 +1−γ
v2n ρ2 2 + (γ − 1)M1n p2 2γM1n
= = 2 and = , (2.25)
v1n ρ1 (γ + 1)M1n p1 γ+1

which can be combined with T2 /T1 = (p2 /p1 )/(ρ2 /ρ1 ) and M2n2 = M 2 (ρ /ρ )−1 (p /p )−1 to
1n 2 1 2 1
give
T2 2 + 1 − γ)[2 + (γ − 1)M 2 ]
(2γM1n 2
2 + (γ − 1)M1n
1n 2
= 2 and M2n = 2 +1−γ. (2.26)
T1 (γ + 1)2 M1n 2γM1n
The above formulae take simplified forms in limiting cases of interest, including very strong
shocks with M1n ≫ 1, for which
 −1 2 2
v2n ρ2 γ−1 p2 2γM1n T2 2γ(γ − 1)M1n 2 γ−1
= = , = , = and M2n = , (2.27)
v1n ρ1 γ+1 p1 γ+1 T1 (γ + 1)2 2γ

14
 Shock and Expansion Waves

and weak shocks with ε = M1n − 1 ≪ 1, for which

 −1
v2n ρ2 2ε p2 2γε T2 2(γ − 1)ε
= =1− , =1+ , =1+ (2.28)
v1n ρ1 γ+1 p1 γ+1 T1 γ+1
and 1 − M2n = ε. The associated entropy increase can be computed from (2.15) to give (s2 −
s1 )/cv ∼ ε3 , indicating that these weak shocks are nearly isentropic.
For normal shock waves, such that the incident flow forms a angle β = π/2 with the shock,
no flow deflection occurs, so that M1 = M1n and M2 = M2n , and the above equations (2.25)
and (2.26) are all we need to determine the jumps across the shock. As an illustrative example,
we can use them to determine the flow induced by a piston moving with constant velocity in a
long cylindrical tube containing a gas. The piston, initially at rest, is suddenly accelerated to
reach a constant velocity VP , generating a shock wave that moves with velocity VSW along the
tube. The gas ahead of the shock is at rest, with initial thermodynamic properties characterized
for instance by the sound velocity a1 and pressure p1 . The shock wave sets the fluid in motion,
so that the gas between the shock and the piston moves with uniform constant velocity, equal to
that of the piston.
FIXED REFERENCE FRAME MOVING REFERENCE FRAME

VP p2 VSW p1 v2 =VSW −VP v1 =VSW

Figure 2.4: Piston-supported shock wave.

The velocities ahead and behind the shock v1 = v1n and v2 = v2n and the corresponding Mach
numbers M1n and M2n appearing in (2.25) and (2.26) are measured relative to the shock, so in
analyzing the problem we need to use a reference frame moving to the right with velocity VSW . In
that reference frame, the incident and downstream velocities are v1 = VSW and v2 = VSW − VP ,
whereas the incident Mach number is M1n = M1 = VSW /a1 . In terms of these quantities, the
first equation in (2.25) yields
v2 VP 2 + (γ − 1)(VSW /a1 )2
=1− = , (2.29)
v1 VSW (γ + 1)(VSW /a1 )2
which can be rewritten in the form
   
VSW 2 γ + 1 VP VSW
− − 1 = 0. (2.30)
a1 2 a1 a1
Solving now for VSW /a1 gives
"     #1/2
VSW γ + 1 VP γ + 1 2 VP 2
M1n = = + 1+ , (2.31)
a1 4 a1 4 a1

which can be used in (2.25) and (2.26) to determine the pressure, density and temperature in
front of the piston as a function of VP /a1 .

15
 Shock and Expansion Waves

The above piston problem can be used as a simplified representation of the flow induced by a
high-speed train entering a tunnel, the main simplification being that our piston occupies the
whole transverse section of the tube, whereas in reality the train occupies only a fraction of the
tunnel section. The results indicate that, as the train enters the tunnel, a sudden overpressure
develops immediately ahead. The shock is weak for VP ≪ a1 , when

γ + 1 VP p2 γ VP
M1n − 1 = ≪ 1 and −1= ≪ 1,
4 a1 p1 2 a1

which explain why, in conventional trains, the problems encountered at the entrance of a tunnel
are minor. Relative pressure differences of order unity appear however as VP approaches values
of the order of the sound velocity a1 . That is the case of high-speed trains, for which special
measures need to be taken to remedy the overpressure problem at the tunnel entrance.

Oblique shocks
For known values of M1 and β one may compute

M1n = M1 sin β, (2.32)

which in turn determines from (2.25) and (2.26) the jumps of thermodynamic properties and the
values of
2 2 + (γ − 1)M12 sin2 β
M2n = (2.33)
2γM12 sin2 β + 1 − γ
and
v2n 2 + (γ − 1)M12 sin2 β
= . (2.34)
v1n (γ + 1)M12 sin2 β
To determine the flow deflection, we use (2.6) written in the form

vt = v1 cos(β) = v2 cos(β − δ). (2.35)

Using the equations v1n = v1 sin(β) and v2n = v2 sin(β − δ) to give

tan(β − δ) v2n
= (2.36)
tan(β) v1n

and solving for δ with use made of (2.34) yields

M12 sin2 (β) − 1

tan(δ) = 2 cot(β) . (2.37)
2 + M12 [γ + cos(2β)]

Once δ is known one may easily compute

M2n
M2 = , (2.38)
sin(β − δ)

with M2n determined from (2.33).

16
 90

80
Figure 2.5: The variation of β with δ for constant values of M1 .

1.1

1.2
1.3
70 1.4
6 8 10 20
1.5 3.8 4.0 4.5 5
1.6 1.7 3.0 3.2 3.4 3.6
1.8 1.9 2.0 2.2 2.4 2.6 2.8
M <1
2

M2 > 1
60

50
!
17

40

30
γ = 1.40

20

Shock and Expansion Waves

M1 δ

10
β

0
0 5 10 15 20 25 30 35 40 45 50
 30
28
26
Figure 2.6: The variation of p2 /p1 with M1 for constant values of M2 and δ.

24 γ = 1.40
22
20 , p2 40
18 M2
16 M1 , p 1 δ
14

12

10
.0 35
9 =1
M2
.5
8 =1
M2
7 .0
=2
M2
p 2 /p 1

6
30
.5
18

5 =2
M2

4 25 .0
=3
M2
k
oc

3 3.5
sh

=
al

20
m

M2
or
N

0
4.

Shock and Expansion Waves

15 =
2 M2

10 5
4.
=
M2

1
1 1.5 2 2.5 3 3.5 4 4.5 5
M1
 Shock and Expansion Waves

The results are summarized in Figs. 2.5 and 2.6, where the angles are given in degrees. In
particular, Fig. 2.5 indicates that for given values of M1 and δ two different solutions are found if
δ < δmax and no solution exists for δ > δmax , where the value of δmax increases for increasing M1 .
For δ < δmax , the solution with larger β corresponds to a stronger shock wave (i.e., larger value of
M1n = M1 sin β). The corresponding value of M2n is sufficiently small that the flow downstream
is subsonic, that is, M2 < 1. On the other hand, the solution lying along the lower part of
the curve of constant M1 in Fig. 2.5 corresponds to a lower value of β, and therefore a weaker
solution including a values of M2 that are generally supersonic (except for a small region near
δ = δmax , as indicated in the figure). Because the associated downstream flow is supersonic, the
weak-shock solution is more stable, in that perturbations can never reach it from downstream.
Because of their stable character, when both strong or weak solutions may exist, weak solutions
tend to prevail in realistic configurations.
For each value of M1 , there exist two solutions with zero deflection angle δ = 0. One is the
normal shock wave, corresponding to β = π/2 and M1 = M1n , which is the strongest possible
wave for that value of M1 . The other, corresponding to β = µ with
 
−1 1
µ = sin , (2.39)
M1

is the weakest shock wave for that value of M1 ; the incident angle is such that M1n = M1 sin β =
1. These limiting solutions are termed Mach waves. According to what we mentioned be-
low (2.28), these weak shocks are effectively isentropic compressions. A Mach wave is seen to
bound, for instance, the Mach cone sketched in Fig. 1.
Figure 2.5 can be used, for instance, to solve the supersonic flow over a wedge, characterized by
the appearance of an oblique shock, that forms an angle θ with the wall, deflecting the incoming
stream to make it parallel to the wedge, as sketched in Fig. 2.7. If the incident Mach number
is M1 = 2 and the wedge semi angle is δ = 15o , then two solutions are possible, as can be seen
in Fig. 2.5. Following the arguments given above, we select the√weaker shock, corresponding to
β ≃ 45o , thereby giving θ = β − δ = 30o and M1n = M1 sin β = 2. Using now (2.25) and (2.26)
yields p2 /p1 ≃ 2.17 and M2n ≃ 0.734. This last value can be substituted into (2.38) to give
M2 = 1.47. The problem could also be solved by using Fig. 2.6 instead, an exercise left for the
reader to attempt. Also of interest is that, for the flow over a wedge, no solution involving an
oblique shock exists for δ > δmax ≃ 23o . In that case, the shock detaches to form a bow shock
wave, which is locally planar at the center line and becomes increasing inclined farther away.
These bow shock waves are found in general in front of blunt bodies, as seen in the experimental
visualization shown below in Fig. 2.8.
Regarding the flow over a wedge (or rather the equivalent case of the flow around a concave
corner), it is of interest to consider what happens when, instead of a sharp corner, one encounters
a smoothly curved wall, as shown in Fig. 2.9. The perturbations associated with the continuous
increase of slope are necessarily weak near the wall, and therefore give rise to a system of Mach
waves that propagate into the stream with a local angle of incidence β = µ such that, at each
position, Mn = M sin β = 1. For the concave wall, the value of M decreases and the pressure
increases as the flow crosses the system of Mach waves, giving an increasing value of µ and
causing the Mach waves to approach one another. The compression reinforces as a result of the
interaction between the different Mach waves, so that a shock wave of finite strength eventually
emerges. This complex interaction process is at the origin of the oblique shock that is seen to

19
 Shock and Expansion Waves

M2
M1
β p2
p1
θ
δ

Figure 2.7: Supersonic flow over a wedge.

emerge from the corner when the observation distance is sufficiently large for the curved wall to
appear as a sharp corner (see Fig. 2.9).

Prandtl-Meyer expansions
Let us now consider the case of supersonic flow over a convex wall, represented in Fig. 2.10. A
system of Mach waves originates from the wall as it curves downwards. In this case, however,
the value of M increases and the pressure decreases as the flow crosses the system of Mach
waves, giving an decreasing value of µ and causing the Mach waves to open up. As a result, the
expansion wave generated as the supersonic stream turns around the convex corner is formed by
a fan of Mach waves that cause the stream lines to evolve smoothly. Since the resulting flow is
steady and isentropic, all stagnation magnitudes are preserved across the expansion wave.
Besides the condition of conservation of all stagnation properties,
 (γ−1)/γ  γ−1  a 2
T0 p0 ρ0 0 γ−1 2
= = = =1+ M , (2.40)
T p ρ a 2

the solution for the convex-corner expansion, commonly known as Prandtl-Meyer expansion,
requires knowledge of the relationship between the flow deflection and the flow acceleration,
which comes from consideration of the infinitesimal evolution due to a given Mach wave. As
shown schematically in Fig. 2.11, the condition that the tangential component of the velocity

20
 Shock and Expansion Waves

Figure 2.8: Supersonic flow over a blunt body.

must be conserved as the flow is deflected can be used to relate the velocity increment and the
deflection according to
dv/ tan µ
dθ = . (2.41)
v
From (2.39) it follows that
1
tan µ = √ . (2.42)
M2 − 1
To obtain dv/v we begin by differentiating v = M a to give

dv dM da
= + . (2.43)
v M a
On the other hand, differentiating the last equation in (2.40) for a constant value of a0 yields

da 1 (γ − 1)M dM
=− , (2.44)
a 2 1 + (γ − 1)M 2 /2

which can be combined with (2.43) to provide

dv 1 dM
= 2
. (2.45)
v 1 + (γ − 1)M /2 M

Substituting (2.42) and (2.45) into (2.41) finally gives

√
M2 − 1 dM
dθ = 2
(2.46)
1 + (γ − 1)M /2 M

as a relationship between the deflection and the Mach number increment.

This last equation can be integrated to relate the total deflection θ with the change in Mach
number as the supersonic stream in Fig. 2.10 turns around the corner according to
Z M2 √
M2 − 1 dM
θ= 2
, (2.47)
M1 1 + (γ − 1)M /2 M

21
 Shock and Expansion Waves

p2 >p1
M2<M1
p1
M1

M1
M1n=1

MACH WAVES

Figure 2.9: Supersonic flow over a concave wall.

which can be alternatively written as

θ = ν(M2 ) − ν(M1 ) (2.48)

in terms of the so-called Prandtl-Meyer function

Z M √
M2 − 1 dM
ν(M ) = 2 /2 M
(2.49)
1 1 + (γ − 1)M
given in explicit form by
r r 
γ+1 γ−1 p
ν(M ) = tan−1 2
(M − 1) − tan−1 ( M 2 − 1) (2.50)
γ−1 γ+1

with tan−1 representing the arctangent (the inverse of the tangent function).
For the problem represented in Fig. 2.10, if the corner angle θ and the Mach number of the
incoming stream M1 are known, then one may use (2.48) to determine the value of M2 and then
use (2.40) written in the form
 (γ−1)/γ  γ−1  2
T2 p2 ρ2 a2 1 + (γ − 1)M12 /2
= = = = (2.51)
T1 p1 ρ1 a1 1 + (γ − 1)M22 /2
to obtain the ratios of thermodynamic properties across the expansion. For instance, if θ = 15o
and M1 = 2, then ν(M1 ) = 26.38o . Using this value in (2.48) gives ν(M2 ) = 41.38o , which can
be employed to obtain from (2.50) the value M2 ≃ 2.6.

22
 Shock and Expansion Waves

p1
M1
µ1
µ2 M2n=1
M1 M2>M1
p2 <p1 θ
M1n=1

MACH WAVES
M2>M1

Figure 2.10: Supersonic flow over a convex wall.

In other occasions, what determines the flow deflection is the pressure that exists downstream.
An example of interest occurs at the exit of nozzles of thrust rockets used for satellite attitude
control. Near the nozzle edge the flow is locally planar and corresponds in the first approximation
to a Prandtl-Meyer expansion, as indicated in Fig. 2.12. The surrounding pressure is very small,
so that the expanding gas reaches very high Mach numbers M2 . If we assume for simplicity
complete vacuum by using p2 = 0, then for any finite values of M1 and p1 one obtains M2 = ∞,
as can be seen from (2.51). The associated flow deflection must be calculated from (2.48) with
ν(M2 ) = ν(∞) = 130.45o . If the value of M1 is sufficiently small, then the deflection may exceed
90o , giving rise to a reverse jet flow that may impinge on the satellite wall, possibly damaging it
(for instance, for M1 = 2 one obtains ν(M1 ) = 26.38o and θ = 104.07o ). This could be avoided
by ensuring that the value of M1 remains above M1 ≃ 2.56, so that θ < 90o .

23
 Shock and Expansion Waves

v v
µ
dθ µ

dv
tan µ { µ

{ dv

Figure 2.11: Differential evolution across a Mach wave.

M 1 >1

p2 =0
M2

Figure 2.12: Supersonic stream discharging to the vacuum.

24
Chapter 3

Nozzle Flow

Quasisteady ideal gas flow in pipes

For the large values of the Reynolds number typically found in nozzles, the flow is ideal. For
steady operation with negligible body forces the energy and momentum equations can be pro-
jected along a given streamline to give

∂s ∂ v2
= 0 and (h + ) = 0, (3.1)
∂l ∂l 2

indicating that the entropy s and the sum h + v 2 /2 remain constant, i.e.,

s = s0 , (3.2)

and
v2
h+ = h0 . (3.3)
2
If the nozzle is fed from a container with uniform thermodynamic properties sc and hc , we may
evaluate the above expressions by following the streamline upstream into the container to yield
s0 = sc and h0 = hc .
Since the nozzles are slender, in that the ratio of their length L and characteristic diameter D
satisfies L/D ≫ 1, the resulting streamlines are aligned, with longitudinal and transverse velocity
components vT /vL ∼ D/L ≪ 1, as follows from the continuity equation. Correspondingly, the
characteristic changes of pressure across the nozzle section δT p ∼ ρc vT2 are much smaller than the
characteristic pressure variation along the nozzle δL p ∼ ρc vL2 according to δT p/δL p ∼ (D/L)2 ≪
1, so that one may assumed that the pressure remains uniform across each pipe cross section.
With s = sc everywhere in the pipe and with p = p(x), where x is the distance measured along
the nozzle, it is clear that all other thermodynamic variables (e.g., h = h(s, p), T = T (s, p),
ρ = ρ(s, p),. . . ) are also only a function of x. For instance, h(s, p) = h(sc , p(x)) = h(x). This
last result can be used in (3.3) to demonstrate that v = v(x).
In view of the above considerations, the solution can be determined from (3.2) and (3.3) together
with
G = ρ(x)v(x)A(x), (3.4)
 Ideal Nozzle Flow

where G is the constant mass flow rate circulating in the pipe. Equations (3.2) and (3.3) can be
written with use made of the equations of state in terms of the different stagnation flow variables,
according to
 2  (γ−1)/γ  γ−1
h0 T0 a0 p0 ρ0 γ−1
= = = = =1+ M (x)2 , (3.5)
h(x) T (x) a(x) p(x) ρ(x) 2

where h0 = hc , T0 = Tc , ao = ac , p0 = pc , and ρ0 = ρc can be evaluated from the container

properties. Similarly, it is convenient to rewrite (3.4) in term of the local Mach number M (x) to
give
 − γ+1
γ−1 2
2(γ−1)
G = ρ0 a0 A(x)M (x) 1 + M (x) . (3.6)
2

Subsonic and supersonic flow in pipes

It is of interest to comment on the effect of the Mach number on the character of the fluid motion.
For liquid flow in a pipe of varying cross section, our intuition tells us that as the cross section
decreases the velocity must increase to accommodate the constant volume flux Q = v(x)A(x),
while the corresponding pressure p(x) must decrease to provide the needed acceleration, as follows
from Bernoulli’s equation p + ρo v 2 /2 =constant. Things are not always quite so intuitive for gas
flow, for which the resulting behavior depends on whether the flow is subsonic or supersonic. To
see how changes in cross-section area affect the acceleration, we begin by differentiating (3.4) to
give
1 dρ 1 dv 1 dA
+ =− . (3.7)
ρ dx v dx A dx
On the other hand, for a perfect gas, the condition of isentropic flow (3.2) gives
p p0
γ
= γ, (3.8)
ρ ρ0

providing
1 dρ 1 1 dp
= 2 (3.9)
ρ dx a ρ dx
upon differentitation, with a2 = γp/ρ. Since

1 dp dh
= (3.10)
ρ dx dx

as follows from the condition of isentropic flow T ds = dh − dp/ρ = 0 and with

dh dv
= −v , (3.11)
dx dx
obtained by differentiating (3.3), (3.7) finally yields

1 dv 1 dA
(1 − M 2 ) =− . (3.12)
v dx A dx

26
 Ideal Nozzle Flow

For subsonic flow (M < 1) along a convergent nozzle (dA/dx < 0) the velocity increases, as
can be seen from (3.12), and the enthalpy, temperature, pressure, density, and velocity of sound
decrease, as seen in (3.5) and (3.11). With v increasing and a decreasing the associated Mach
number M = v/a continuously increases as the cross section decreases. Because of the presence
of the factor (1 − M 2 ) in (3.12) the opposite behavior is found for supersonic flow in a convergent
nozzle, that is, the flow decelerates and the Mach number decreases, while h, T , p, ρ, and a
increase. It is therefore not possible to accelerate a subsonic stream beyond sonic conditions
with use made of a convergent nozzle.
Observation of (3.12) reveals that a divergent nozzle (dA/dx > 0) is required to accelerate a
supersonic flow. It also indicates that, if sonic conditions M = 1 are reached at a given section
x, at that location the cross-section area must satisfy dA/dx = 0. These findings suggest that
supersonic flow can be therefore achieved by utilizing a convergent-divergent nozzle so that the
flow accelerates through the converging stretch to reach sonic conditions at the throat (the section
of minimum area), further accelerating downstream into the supersonic regime in the diverging
stretch. We shall see that, depending on the conditions, the flow in convergent-divergent nozzles
may include a number of complicating features, including shock and expansion waves, to be
treated in a later chapter.

Critical magnitudes
For steady gas flow in a pipe (3.5) provides the variation of the different thermodynamic variables
with M . It can be used to determine, for instance, their critical values

h0 T0  a 2  p (γ−1)/γ  ρ γ−1 γ + 1
0 0 0
= ∗ = = = = . (3.13)
h∗ T a∗ p ∗ ρ∗ 2

found at a section where the flow is sonic. At that critical section, (3.6) provides
 − γ+1
∗ ∗ γ+1 2(γ−1)
G = G = ρ0 a0 A , (3.14)
2
which can be written in the form
   γ+1
A 1 2 γ−1 2 2(γ−1)
= 1+ M (3.15)
A∗ M γ+1 2
thereby relating the values of A and M found at a given section with the area A∗ of the section
where sonic conditions would be attained.

Flow in a convergent nozzle

We begin by analyzing the discharge of a gas container with known gas properties (pc , ρc ,
Tc , ac , . . . ) through a convergent nozzle of minimum cross section A(L) at the outlet x = L
into a surrounding gas atmosphere at ambient pressure pa . The solution, depending on the
thermodynamic properties in the container and on the ambient pressure pa , can be determined
with use made of (3.5) and (3.6) with T0 = Tc , ao = ac , p0 = pc , and ρ0 = ρc . An outflow

27
 Ideal Nozzle Flow

is established whenever pa < pc . The flow is strongly subsonic for small pressure differences
pc − pa ≪ pc , when the gas discharges into the outer atmosphere as a low-Mach-number jet with
pressure p(L) = pa at the exit section x = L. This last condition enables the Mach number
M (L) to be computed from (3.5) according to
 (γ−1)/γ  1/2 " (γ−1)/γ #1/2
pc γ−1 2 pc
=1+ M (L)2 → M (L) = −1 , (3.16)
pa 2 γ−1 pa

to be substituted into (3.6) to determine the gas flow rate through the nozzle

 1/2  − γ+1 "  γ−1 #1/2

2 pc 2γ pc γ
G = ρc ac A(L) −1 . (3.17)
γ−1 pa pa

As the value of pc /pa increases so does the value of M (L) given by (3.16) and also the value of G
given by (3.17), with variations given in Fig. 3.1 for γ = 1.4. Using the value of G, computed for
a given value of pc /pa from (3.17), in (3.6) enables the Mach number distribution to be obtained
along the nozzle, whereas (3.5) provides the accompanying distributions of pressure, density, and
temperature. Sample schematic profiles are given in Fig. 3.2.
γ = 1.4
1.2

1 M (L) 1

0.8

0.6
0.579
0.4
G
ρ c a c A(L)
0.2

0
1 1.5 1.893 2 2.5
p c /p a

Figure 3.1: The variation of the exit Mach number M (L) and mass flow rate G with pc /pa as
obtained from (3.16) and (3.17) for γ = 1.4.

As discussed above, the Mach number at the exit section increases for increasing values of pc /pa .
The jet solution with p(L) = pa and the computation procedure described above continue to
hold as long as pa remains above a critical choking value pCH at which the flow reaches sonic
conditions at the outlet section. This choking value is determined from (3.16) with M (L) = 1 as
 γ/(γ−1)
pc γ+1
= (3.18)
pCH 2

giving pc /pCH ≃ 1.893 for γ = 1.4 (pCH ≃ 0.53pc ). For these choking conditions the mass flow

28
 Ideal Nozzle Flow

rate reaches the critical value G∗

 − γ+1
∗ γ+1 2(γ−1)
G = ρc ac A(L) , (3.19)
2
given in (3.14). The flow cannot accelerate past sonic conditions in the convergent nozzle by
decreasing the ambient pressure. Instead, the solution becomes choked, that is, for values of
pa < pCH the flow is sonic at the exit, and becomes independent of the outer pressure 1 . In
particular, the mass flow rate, which for pa > pCH increases for increasing values of pc /pa ,
remains equal to G∗ , independent of pa , for pa < pCH . In the choked flow that emerges for
pc /pa > [(γ + 1)/2]γ/(γ−1) , the condition M (L) = 1 therefore replaces p(L) = pa . The pressure
at the exit section is p(L) = pCH > pa , so that the jet flow with parallel streamlines found
downstream from the pipe outlet for pa > pCH is replaced for pa < pCH by a complicated flow
pattern including expansion waves that deflect the streamlines outwards.

L
pa
pc
x
ac
ρc

1 p1 /pc
p/pc p2 /pc
pCH /pc = 0.53

x
1 M= 1
M M2
M1

Figure 3.2: Pressure and Mach number distributions along a nozzle for different values of pa /pc .

In summary, the discharge of a container of pressure pc into the atmosphere, at pressure pa ,

through a convergent nozzle of minimum cross section A(L) depends on the value of pc /pa , so
that

1
Note that the decrease of the ambient pressure below pCH cannot be “felt” within the nozzle, because the flow
at the exit is already sonic, and therefore the perturbations cannot travel upstream, so that for pa < pCH the flow
in the nozzle becomes independent of pa .

29
 Ideal Nozzle Flow

p(L) = pa
 1/2  (γ−1)/γ 1/2
2 pc
M (L) = γ−1 pa −1 ≤1
 1/2  − γ+1   γ−1 1/2
2 pc 2γ pc γ
 γ/(γ−1) G = ρc ac A(L) γ−1 pa pa − 1
pc γ+1
If pa ≤ 2 : L
pa
pc x
ac
ρc

M (L) = 1
 −γ/(γ−1)
γ+1
p(L) = pCH = 2 pc > pa
 − γ+1
γ+1 2(γ−1)
G = G∗ = ρc ac A(L) 2
 γ/(γ−1) L
pc γ+1
If pa > 2 : pa
pc
x
ac
ρc

Flow in convergent-divergent nozzles

The flow in convergent-divergent nozzles is more complicated. The nozzle geometry is defined
by the values of the exit area Ae and throat area At (the section of minimum cross-section area).
The nozzle is assumed to be connected to a reservoir, where the gas pressure is p0 , and discharges
to the atmosphere, where the pressure is pa . Different solutions emerge depending on the value
of pa /p0 , as indicated in Fig. 3.3.
Let un begin by considering the case where the ambient pressure is only slightly smaller than p0
(case a in Fig. 3.3). As a result of the pressure difference, there appears a slow isentropic steady
outflow. For the resulting subsonic flow, the pressure decreases as the Mach number increases in
the convergent part of the nozzle, and opposite behaviors are observed in the divergent stretch.
The minimum pressure and the maximum Mach number are therefore reached at the throat.
The flow discharges to the ambient as a subsonic jet. The condition that the pressure at the exit
equals the ambient value can be used to compute the Mach at the exit section from

 (γ−1)/γ
p0 γ −1 2
=1+ Me , (3.20)
pa 2

30
 Ideal Nozzle Flow

a Mt <1 Me<1

p0
b Mt =1 Me<1

1.0
p
a
c Mt =1 Me<1
b
p0 c

0.528
d
e d Mt =1
f
g

e Mt =1 Me> 1

M
f

f Mt =1 Me> 1
1.0
d
c
b

a
g Mt =1 Me> 1

Figure 3.3: Flow in a convergent-divergent nozzle.

31
 Ideal Nozzle Flow

which in turn determines the mass flow rate

  γ+1
γ − 1 2 − 2(γ−1)
G = ρ0 a0 Ae Me 1 + Me
2
 1/2  − γ+1 "  γ−1 #1/2
2 p0 2γ p0 γ
= ρ0 a0 Ae −1 , (3.21)
γ−1 pa pa

with the stagnation properties for the flow being those found in the reservoir. The Mach number
at the throat Mt > Me can be obtained from
 − γ+1  − γ+1
γ−1 2 2(γ−1) γ−1 2 2(γ−1)
Ae Me 1+ Me = At Mt 1+ Mt (3.22)
2 2

obtained by equating the values of G evaluated at the throat

 − γ+1
γ−1 2 2(γ−1)
G = ρ0 a0 At Mt 1+ Mt (3.23)
2

and at the exit.

The value of Mt continues to increase for decreasing values of pa /p0 , reaching sonic conditions
when pa = pCH (case b in Fig. 3.3). The value of pCH /p0 can be determined as a function of
At /Ae by equating the mass flow rate at the throat,
 − γ+1
γ+1 2(γ−1)
G = ρ0 a0 At (3.24)
2

evaluated from (3.23) with Mt = 1, to that crossing the exit section, evaluated from (3.21),
yielding
  γ+1  1/2  − γ+1 "  γ−1 #1/2
At γ + 1 2(γ−1) 2 p0 2γ p0 γ
= −1 . (3.25)
Ae 2 γ−1 pa pa

It can be seen that, for a given value of At /Ae , this last equation is satisfied by two different
values of pa /p0 . One of the two solutions, corresponding to the choking value pa = pCH , is
associated with an exit Mach number Me < 1, to be evaluated from (3.20) with pa = pCH . There
exists, however, a second isentropic solution with pa < pCH in which the flow discharges as a
supersonic jet with pe = pa = pSJ and Me > 1. For instance, for At /Ae = 0.477 the two solutions
of (3.25) and (3.20) are pa /p0 = pCH /p0 = 0.943 with Me = 0.29 and pa /p0 = pSJ /p0 = 0.0865
with Me = 2.25.
The results indicate that the convergent-divergent nozzle can be choked more easily than the
convergent nozzle (note that for At /Ae = 0.477 a value pa /p0 = 0.943 is sufficient to choke the
flow, to be compared with the value pa /p0 = 0.528 required in convergent nozzles). The flow
remains choked for pa < pCH , which implies that, regardless of the value of pa , the flow in the
convergent part of the nozzle is identical, with the mass flow rate taking always the constant
value (3.24). In the isentropic solution discharging as a supersonic jet (case f in Fig. 3.3) the
flow is therefore subsonic in the convergent stretch, sonic at the throat, and supersonic in the

32
 Ideal Nozzle Flow

divergent stretch, leaving the nozzle with pe = pa , so that neither expansions nor compressions
are needed to adapt the pressure to the ambient value.
It is of interest to consider what happens when the ambient pressure is below pSJ and also when
it takes intermediate values in the range pSJ < pa < pCH . For pa < pSJ (case g in Fig. 3.3),
the flow conditions at the exit are exactly those found for pa = pSJ , so that pe = pSJ > pa
(underexpanded flow). An expansion is therefore needed outside the nozzle to decrease the
pressure to the ambient value. The expansion wave that appears is locally planar near the nozzle
rim and therefore corresponds there to a Prandtl-Meyer expansion, which deflects the stream
outwards as it decreases its pressure to pa while increasing its Mach number beyond that found
at the exit section.
Similarly, when the ambient pressure is slightly above pSJ (case e in Fig. 3.3), the flow condi-
tions at the exit are also exactly those found for pa = pSJ . In this case, however, the flow is
overexpanded, and a shock wave is needed to increase the pressure to the ambient value. When
pa is only slightly above pSJ , an oblique weak shock is sufficient. As the value of pa further
increases, the needed shock wave becomes stronger, and the associated incidence angle increases
towards β = π/2. There is a critical value of pa for which the compression occurs through a
normal shock wave standing at the exit section (case d in Fig. 3.3). For even larger values of pa
the shock wave migrates into the divergent stretch of the nozzle, so that downstream from the
throat there is a region of supersonic flow that ends at the inner shock, which is followed by a
region of subsonic flow eventually discharging to the ambient as a jet, as indicated in Fig. 3.3.
These steady solutions including planar shock waves within the divergent part of the nozzle are
hard to observe in reality, because they are often affected by boundary-layer separation immedi-
ately downstream from the shock, which may lead to significant changes in the downstream flow
structure, including at times unsteady flow variations and asymmetric flow patterns.

33