Mathematical Models for High Speed Flows

Contents
1-

introduction ....................................................................................................................................... 2

2-

Governing Equations of Gas Dynamics ............................................................................................ 2

2.1 - Introduction ................................................................................................................................... 2

2.1.1 - The mass conservation law ................................................................................................... 4
2.1.2 -

Conservation of Momentum................................................................................................. 5

2.1.3 -

Conservation of Energy ......................................................................................................... 7

2.1.4 -

The Integral and Differential Form of the Governing Equations .......................................... 9

2.2 -

The Euler Equations .................................................................................................................... 10

2.3 - Thermodynamic Relations .......................................................................................................... 13

2.3.1 - Equilibrium Real Gas ........................................................................................................... 13
2.3.2 -

The specific heat capacities and the speed of sound ......................................................... 15

2.3.3 -

Non-Equilibrium Real Gas ................................................................................................... 17

2.4 - Simplified Euler equations .......................................................................................................... 18

2.4.1 - 1D Compressible Flows with Area Variation ....................................................................... 18
2.4.2 -

Cylindrical and Spherical Symmetry.................................................................................... 19

2.4.3 -

Axi-Symmetric and Plane (2D) Flows .................................................................................. 20

2.5 - Mathematical properties of the Euler equations.......................................................................... 21

2.5.1 - The 1D Euler equations ....................................................................................................... 21

3-

2.5.2 -

Two-Dimensional Euler Equations ...................................................................................... 26

2.5.3 -

The 3D Euler Equations ....................................................................................................... 29

2.5.4 -

Conclusion ........................................................................................................................... 31

References ....................................................................................................................................... 33

1-

INTRODUCTION

The role of this work is to review systematically the well-known Euler equations
of gas dynamics in the frame of actual aerothermodynamics. In the frame of the present work,
the supersonic/hypersonic flows under consideration are flows dominated by the occurrence of
(strong) shock waves and, for hypersonic flows, significant local high temperature effects.

2-

GOVERNING EQUATIONS OF GAS DYNAMICS

2.1 -

Introduction

In this part we recall the standard form of the classical fluid dynamics equations written
in eulerian form. The fundamental assumption is that the fluid is a continuum (see the
considerations about the low-density case before). For the evolution of a fluid in N 1 spatial

dimensions, the description involves ( N + 2 ) fields. These are the mass density, the velocity

field and the energy. The fundamental equations of fluid dynamics are derived following the
next three universal conservation laws:
1. Conservation of Mass.
2. Conservation of Momentum or the Newtons Second Law.
3. Conservation of Energy or the First Law of Thermodynamics.
The resulting equations are the continuity equation, the momentum equations and the
energy equation, respectively. The system of governing equations is closed by:
1. Statements characterizing the thermodynamic behavior of the fluid, which are called
state equations. The state equations do not depend on the fluid motion and represent
relationships between the thermodynamic variable: pressure (p), density ( ),
temperature (T), specific internal energy per unit mass (e) and total enthalpy per unit
mass (h). The microscopic transport properties like the dynamic viscosity ( ) and the
thermal conductivity (k) are also related to the thermodynamic variables (see for
instance Sutherlands relations.
2. Constitutive relations, which are postulated relationships between stress and rate of
strain and heat flux and temperature gradient. The previously mentioned microscopic

transport properties appear in the constitutive relations as proportionality factors

(between stress and rate of strain and heat flux and temperature gradient) and must be
determined experimentally.
The flow is considered as known if, at any instant in time, the velocity field and a precise
number of thermodynamic variables are known at any point. This number equals two for a real
compressible fluid in thermodynamic equilibrium (for example, the pressure and the density).
There is some freedom in choosing a set of variables to describe the fluid flow. A possible
choice is the set of primitive or physical variables, namely the density, pressure and velocity
components along the three directions of a reference frame, i.e. the set ( , p, u, v, w ) . Another

choice is the set of the so-called conservative variables which are the density, the three

momentum components and the total energy per unit mass, i.e. the set ( , u , v, w, E ) . The

set of conservative variables results naturally from the application of the above fundamental
laws of conservation. From a computational point of view, there are some advantages in
expressing the governing equations in terms of the conservative variables. We will call
conservative methods the class of numerical methods based on the formulation of the
governing equations in conservative variable. The primitive and conservative variables are
dependent variables. They are depending on time and space coordinates, i.e. on the
independent variables. For example, the density is = ( t , x, z , y ) . Recall that the number of

the spatial coordinates is N. The reader is invited here to notice an important difference
between the governing equations and the closure conditions. The governing equations are
exact mathematical conditions that must be satisfied by the flow variables. The state equations
and the constitutive relations are only practical models and they depend on the degree of
knowledge the user has about the fluid and flow under consideration. Thus, they represent
sources of uncertainties that are introduced in the mathematical model of the flow.
Let us now consider the new scalar field

(t ) = ( x, y, z, t )dV

(2-1)

where the volume of integration V is enclosed by a piece-wise smooth boundary surface

A = V that moves with the material under consideration. It can be shown, [1, 2] that the
material derivative of is given by:
D
=
Dt

dV + A (n V ) dA
t

(2-2)

where n = (n1 , n2 , n3 ) is the outward pointing unit vector normal to the surface A. This equality
can be generalized to vectors ( x, y, z , t ) as follows:
D
=
Dt

dV + A (n V )dA
t

(2-3)

The surface integrals in the last two equalities may be transformed to a volume integral by
virtue of Gausss theorem, [1]. This states that for any differentiable vector field
= ( 1 , 2 , 3 ) and a volume V with smooth bounding surface A the following identity holds:

(n
)dA =
div dV .

(2-4)

The above identity also applies to differentiable scalar and tensor fields. If we use it, for
example, to transform the second term of the right hand side of equation (1-5) into a volume
integral, one obtains:
D
=
Dt

t + div ( V ) dV .

(2-5)

In the above relations the volume V is arbitrary but moving with the fluid particles. Let J be the
Jacobian of the transform that links two successive positions of the volume V. There is a
theorem due to Euler which states that the time derivative of the Jacobian is, [2]:
DJ
= J divV .
Dt

(2-6)

2.1.1 - The mass conservation law

The law conservation of mass can now be stated in integral form by identifying the scalar
in (1-2) as the fluid density . By assuming that no mass is generated or annihilated within the
material control volume V, we have:

dV + A n ( V )dA =
0.
t

This is the integral form of the law of the conservation of mass. This integral conservation law
may be generalised to include sources of mass, which will then appear as additional integral

(2-7)

terms. A useful reinterpretation of the integral form results if now V is a fixed control volume
and we rewrite it as:

dV = dV =
n ( V )dA .
A
dt V
t

(2-8)

Thus, the time-rate of change of the mass enclosed by the volume V, in the absence of sources
or skins, is due to the mass flow through the boundary of the control volume V. One obtains
from (2-7) that:

+ div( V ) dV =
0
V t

(2-9)

As V is arbitrary it follows that the integrand must vanish, that is:

0
t + ( u ) x + ( v) y + ( w) z =

(2-10)

which is the differential form of the continuity equation.

2.1.2 - Conservation of Momentum

As done for the mass equation, we now provide the foundations for the law of
conservation of momentum, derive its integral form in quite general terms and show that under
appropriate smoothness assumptions the differential form is implied by the integral form. A
control volume V with bounding surface A is chosen and the total momentum in V is given by:

(t ) = VdV

(2-11)

The law of conservation of momentum results from the direct application of Newtons law: the
time rate of change of the momentum in V is equal to the total force acting on the volume V.
The total force is divided into surface forces fs and volume fv given by:

fs =
S dA, fv =
g dV
A

Here g is the specific volume-force vector and may account for inertial forces, gravitational
forces, electromagnetic forces and so on. S is the stress vector, which is given in terms of a
stress tensor as S= n . The stress tensor can be split into a spherical symmetric part
due to pressure p, and a viscous part :

(2-12)


= pI

(2-13)

Application of Newtons Law for the momentum of the fluid in the fixed volume V (independent
of time) gives:
(2-14)
d

V (n V ) dA + f s + f v
( V ) dV =
( V )dV =

V t
A
dt V
We interpret the above as saying that the time rate of change of momentum within the fixed
control volume V is due to the net momentum inflow over momentum outflow, given by the
first term in (2-14), plus the effects of surface and volume forces. Substituting the stress tensor
into (2-12) and writing all surface terms into a single integral we have:
(2-15)

( V )dV =
V (n V )g+ pn n ] dA + dV
[

V t
A
V
This is a general or integral statement of the momentum conservation law and it is valid even
for the case of discontinuous solutions.

The differential conservation law can now be derived from (2-15) under assumption
that the integrand in the surface integral is sufficiently smooth so that Gausss theorem may be
invoked. The first term of the integrand of the surface integral can be rewritten as:
V (n V ) =
n V V,

where the dyadic (tensor) product V V is a tensor. The three columns of the left hand side
are:
T

u (n V ) =n u 2 , uv, uw ,
T

v (n V ) =n uv, v 2 , vw ,
T

w (n V ) =n uw, vw, w2 .

Using the Gausss theorem for each of the surface terms:

[ div( V
g V ) + gradp div ] dV + dV
( V) =
V t
V
V
As this is valid for any arbitrary volume V the intregrand must vanish, i.e.:

[ V
g V + pI
( V ) + div
t

]=

(2-16)

(2-17)

This is the differential conservative form of the momentum equation, including a

source term due to volume forces. In order to use it for a given fluid flow the components of
the viscous stress tensor must be provided through a constitutive relation.

2.1.3 - Conservation of Energy

As done for mass and momentum we now consider the total energy in a control volume V, that
is:

(t ) = EdV

(2-18)

where E is the total energy per unit volume:

1
1

E= e + V 2 = e + ( u 2 + v 2 + w2 ) .
2
2

(2-19)

Here e is the specific internal energy and the second term represents the specific kinetic
energy.
Before we discuss the equations in more detail, let us make a physical comment: the
derivation of the energy equation relies upon the assumption that, in a fluid in motion, the
fluctuation around thermodynamic equilibrium are sufficiently weak so that the classical
thermodynamics results hold at every point and at all times. In particular, the thermodynamic
state of the fluid is determined by the same state variables as in classical thermodynamics, and
these variables are determined by the same state equations (see the devoted paragraph in this
chapter).
The energy conservation law states that the time rate of change of total energy (t ) is
equal to the work done, per unit time, by all forces acting on the volume plus the influx of
energy per unit time into the volume. The surface and volume forces in (2-12) give rise to the
following two terms:

Esurf =
p (V n)dA
+ V (n )dA
A

(2-20)

and

=
Evol

(V g)dV .

(2-21)

In (2-20) Esurf is the work done by the surface forces: the first term corresponds to the work
done by the pressure while the second term corresponds to the work done by the viscous
stresses. In (2-21) Evol is the work done by the volume force g.
To account for the influx of energy into the volume we denote the energy flux vector by
Q = (q1 , q2 , q3 ) . The flow of energy per unit time across a surface dA is given by the flux
(n Q)dA . This gives the total influx of energy to be included in the equation of balance of

energy:

Einf =
(n Q) dA

(2-22)

The total internal energy variation, i.e. the energy conservation law in integral form for a fluid
volume V fixed in space and independent of time, is therefore:

[n ( pV +V
Q)g+ V (n
EdV + (n EV )dA =
V t
A
A

] dA + V (

)dV

(2-23)

The differential form of the conservation of energy law can now be derived by assuming
sufficient smoothness and applying Gausss theorem to all surface integrals of (2-23). Direct
application of Gausss theorem gives:

n ( pV )dA =
div( pV )dV
A

n QdA =
divQdV
A

The work done by the viscous stresses can be transformed by first observing that
V (
n n
) =V ( ). This follows from the symmetry of the viscous stress tensor . Hence:

V (n =
)dA

div
V ( )dV

Substitution of these volume integrals into the integral form of the law of conservation of total
internal energy gives:

div
E
p
V

V
Q
V

g
(
)
0
+
+

(
)

dV =

V t

Since the volume V is arbitrary the integrand must vanish identically, that is:

(2-24)

E
+ div ( E + p ) V
0
V Q+ V (g ) =
t

This is the differential conservative form of the law of conservation of energy with a
source term accounting for effect of body forces; if these are neglected we obtain the
homogeneous energy equation corresponding to the Navier-Stokes equation. The energy flux
vector Q must be specified through a constitutive relation. Further, when viscous and heat
conduction effects are neglected one obtains the energy equation corresponding to the
compressible Euler equations.

2.1.4 - The Integral and Differential Form of the Governing Equations

The previous derivation of the governing equations of compressible flows, such as the
compressible Euler and Navier-Stokes equations stated earlier, is based on integral relations on
control volume and their boundaries. The differential form of the equations results from
further assumptions on the flow variables (the smoothness of the flow field variables is
assumed). In the case of the Navier-Stokes equations, the smoothness of the solution is
naturally assured by the diffusion due to the viscosity and heat conductivity. On the contrary, in
the absence of viscous diffusion and heat conduction one obtains the Euler equations. These
admit discontinuous solutions and the smoothness assumption that leads to the differential
form no longer holds true. Thus, one must return to the more fundamental integral form
involving integrals over control volumes and their boundaries. However, for most of the
theoretical developments the differential form of the equations is preferred. From a
computational point of view there is another good reason for returning to the integral form of
the equations.
We have determined previously the integral form of the governing equations, as well as
the conservative differential form of the equations. One notices that the name Euler and/or
Navier-Stokes originally given to the momentum equations (or equations of motion) transfers
to the entire system of equations. We summarize below the general laws of conservation of
mass, momentum and total energy, written both in differential and integral form:
a) the continuity equation:

t + div( V ) =
0 or
d
0
dV + n ( V )dA =
A
dt V

(2-25)

b) the momentum equation:

( V )t + div
[ Vg V + pI

, or
]=

[ div( V gV ) + gradp div

( V ) +
V t
V
c) the energy equation

] dV =
V

dV .

Et + div [ ( E + p )V
V Q+ =] V (g ) or
d
EdV + [n ( EV + p
V + Q) V (nV
=
g] dA
A
dt V
where g = ( g1 , g 2 , g 3 ) is a body force vector.

( )dV

At this point, all the terms, such as the viscous stress tensor, the heat flux, the pressure,
the temperature and the specific internal energy are expressed as functions of the conservative
variables.

2.2 -

The Euler Equations

In this section we consider the time-dependent Euler equations. The continuity,

momentum and total internal energy equations are forming a system that takes the name of
the inviscid momentum equations, [1]. This system of non-linear hyperbolic conservation laws
governs the dynamics of a compressible fluid, such as gases (or liquids at high pressure). This is
possible only within the framework of the hypothesis that the fluid is an ideal one, without
viscosity and without heat conduction. This means that both microscopic transport properties
of the fluid (i.e., the dynamic viscosity and the heat conduction coefficients) are completely
neglected, by the previous assumption.
When body forces are included via a source term vector but viscous and heat conduction
effects are neglected the mass, momentum and total energy conservation laws become the so
called Euler equations that may be written in conservative form as:
U t + Fx (U) + G y (U) + H z (U) =
S(U)

where the vector of conservative variables is U and the convective fluxes are F, G, H in the x, y
and z directions, respectively. Their explicit forms are:

(2-26)


u

=
U =
v , F

w
E

u
u 2 + p

uv=
, G

puw

u ( E + p )

v
uv

v 2 +=
p , H

vw
v( E + p )

w
uw

vw
2

w + p

w( E + p ).

First, it is important to note that the fluxes are nonlinear functions of the conserved variable
vector. Any set of partial differential equations written in the form (2-26) is called a system of
conservation law (in differential formulation). Recall that the differential form assumes smooth
solutions, that is, partial derivatives are assumed to exist. It is also clear now that the integral
form of the equations is an alternative way of expressing the conservation laws in which the
smoothness assumption is relaxed so that to include discontinuous solutions.
The term S = S(U) is a source or forcing term. Due to the presence of the source term,
the equations are said to be inhomogeneous. There are several physical effects that can be
included in the forcing term, like: body forces such as gravity, injection of mass, momentum
and/or energy. Usually, S(U) is a prescribed algebraic function of the flow variables and does
not involve derivatives of these, but there are exceptions. If S(U) = 0 one speaks of
homogeneous Euler equations. This is the case of aerodynamic applications, when the body
forces due to gravity and other sources are negligible. We also mention here that there are
situations in which source terms arise as a consequence of approximating the homogeneous
equations to model situations with particular geometric features (axy-symmetric flows, for
instance). In this case the source term is of geometric character, but we shall still call it a source
term.
Sometimes it is convenient to express the equations in term of the primitive or physical
variables , u, v, w and p. By expanding derivatives in the conservation law form and using the
mass equation into the momentum equations and in turn using these into the energy equation
one can re-write the thee-dimensional Euler equations for ideal gases with a body-force source
term as:

(2-27)

t + u x + v y + w z + (u x + v y + wz ) =
0
ut + uu x + vu y + wu z +
vt + uvx + vv y + wvz +

wt + uwx + vwy + wwz +

px =
g1 ,
py =
g2 ,

pz =
g3 ,

0
pt + upx + vp y + wpz + p (u x + v y + wz ) =
For computational purposes it is the conservation law form (2-26) that is most useful. The
formulation in primitive variables (2-28) is more convenient to theoretical developments.

We have presented in this section the time-dependent Euler (and Navier) equations of
fluid dynamics. The equations are accompanied by equations of state and by constitutive
relations and also by initial and boundary . Both models can be used for homogeneous gases
and/or liquids at high pressure. We appreciate that for supersonic/hypersonic flows the Euler
equations are a representative and useful model for both theoretical developments, [3], and
computational purposes, [1, 2]. From a mathematical point of view, these models are systems
of partial differential equations. Since with a few exceptions one cannot find exact solutions of
these equations for practical problems, the only way is to solve them is to use numerical
techniques. It is generally accepted that the discretization techniques must be based not only
on the underlying physics bust also on the mathematical properties of the partial differential
equations. It is therefore useful to begin with the analysis of the mathematical nature of the
governing equations of the flow before trying to solve them numerically. Further, the equations
of compressible fluid flow reduce to hyperbolic conservation laws (i.e. the Euler system) when
the effects of viscosity and heat conduction are neglected. The hyperbolic part represents the
convection and pressure gradient effects and it can be identified in the equations even the
above physical effects are not negligible. This is the reason why we allocate further a special
section to study the hyperbolic partial differential equations.
We now conclude this section with a brief discussion of initial and boundary conditions.
Usually, we solve the systems of equations for t 0 prescribing the values of all the unknowns
at t = 0 . The question of the boundary condition is much more delicate. Simply said, one
impose physical boundary condition at the solid boundaries for some of the variables, and
these can be of Dirichlet or Neumann type (depending on the variable type). Since the
boundary conditions are replacing the governing equations on the boundaries, the rest of the
variables are determined via numerical techniques. These are strongly related to the numerical

(2-28)

solvers and thus we will discuss the imposition of the boundary conditions in the context of
numerical solvers.
A short comment on the Navier-Stokes model is also necessary. The equations
presented in this chapter can be used for the numerical simulation of laminar flows only. At
practical Reynolds numbers, the effects of the turbulence must be taken into account. The
Reynolds Averaged Navier-Stokes (RANS) equations and the Large Eddy Simulation (LES)
equations have formally the same mathematical aspect, with minor differences. The huge
difference comes from the necessity of modeling the turbulence effects, mainly in the vicinity
of solid boundaries. As we already mentioned, it is beyond the purposes of this book to discuss
about turbulence modeling and turbulence models. More details about the derivation of these
and other equations can be found in [1, 2, 4, 5].

2.3 -

Thermodynamic Relations

2.3.1 - Equilibrium Real Gas

In thermodynamic equilibrium there are two independent variables, [3, 4, 7]. Thus, two
equations are necessary combining the four variables, which are the pressure (p), the density
(), the absolute temperature (T) and the internal energy per unit mass (e). These equations are
the so-called thermal and the caloric equations of state.
The thermal equation of state (EOS) for a real gas (like the air at usual and high
temperatures is) can be written as
p
= RT Z ( , T ) ,

(2-29)

where R is the specific gas constant per unit mass and Z is the compressibility factor. For nondissociating and non-ionizing gases and if the pressure is low-enough so that the van de Waals
effects are negligible, the compressibility factor Z 1 and thus the thermal ideal gas equation
can be applied:
p
= RT .

(2-30)

The caloric equation of state is a relation between the internal energy per unit mass e
and two dependent variables, i.e. the density and the temperature and can be written in
general as:
e= e ( , T ) .

(2-31)

In a non-dissociating and non-ionizing gas the specific internal energy does not depend on the
density and thus the caloric equation of state of a real gas is
e = e (T ) .

(2-32)

(2-33)

It can be shown that (2-32) is, [2, 3, 4]:

=
e e=
(T )

c (T ) dT ,
V

where cV (T ) is the specific heat at constant volume.

Below 775 K (about 5000C) the specific heat in air is independent of the temperature
and the air behaves as a calorically perfect/ideal gas, i.e. the air is a thermally and calorically
ideal gas and:
=
e e (=
T ) cV T , =
cV

5
=
R
R.
1
2

(2-34)

Above 775 K molecular vibrations are excited and become important and the specific
heats at constant pressure and volume become dependent on the temperature. The thermal
and caloric equations of state can be combined to give the pressure in terms of the density and
internal energy per unit mass similar to (2-27):
(2-35)

p
= RT ( , e ) Z ( , e ) .

The implementation of the state equation (2-35) of real gases in equilibrium in a code requires
p
p
and
,
e
e

p p ( ,e ) and of its partial derivatives,

the knowledge of pressure function =

if the speed of sound is needed by the algorithm. For air the real gas description is given by
Mollier, [10] or from tables/routines approximating the pressure as function of density and
internal energy.

2.3.2 - The specific heat capacities and the speed of sound

The heat capacity at constant pressure cp and the heat capacity at constant volume cv
(specific heat capacities) are now introduced. In general, when an addition of heat dQ changes
the temperature by dT the ratio c = dQ / dT is called the heat capacity of the system. For a
thermodynamic process at constant pressure one obtains dQ =+
de d ( pv) =
dh, where the
definition of the specific enthalpy is
h= e +

(2-36)

Assuming h = h(T , p ) , the heat capacity cp at constant pressure becomes:

h
cp =

T p

(2-37)

The heat capacity cv at constant volume may be written, following a similar argument, as:

e
cv =

T v

(2-38)

The speed of sound is a variable of fundamental interest. For flows in which particles
undergo unconstrained thermodynamic equilibrium one defines a new state variable a, called
the equilibrium speed of sound or just speed of sound. Given a caloric equation of state such
as p = p ( , s ), one defines the speed of sound a as:

p
a=
s
where s = the specific entropy. This basic definition can be transformed in various ways using
established thermodynamic relations. For instance, given a caloric EOS in the form h = h( p, ),
we can also write, [2]:

h
h
1 p
p
dp + d = Tds + d + ds
s
s
p
p p

Settings ds=0 (isentropic conditions) and using the definition (2-39) we obtain successively:

(2-39)

 h

p
2
,
a =
h
1

p

and for a thermally ideal gas (perfect gas) characterized by the EOS of the form h= h(T):

a2 =

h

p p
1
h

p

h
From = c p (by definition) and for a thermally ideal gas one obtains the widely
p
used relation for the speed of sound:
=
a

() R=

()

(2-40)

Remark: One notices the dependence of the ratio of the specific heats on the temperature.
For a general material (gas, but not only) the caloric EOS is a functional relationship
involving the variables p--e. The derived expression for the speed of sound a, depends on the
choice of dependent variables. Two possible choices and their respective expression for a are:
=
p p ( ,=
e), a
=
e e( , p=
), a

pe + p

e
p

2 e p e p
p

(2-41)

where subscripts denote partial derivates.

Various isentropic exponents can be defined, depending on the assumption of the state
of the gas, [4]. For instance, for a thermally ideal gas:

h
c p T d h
=
= =
,
e d e
cv
T

(2-42)

or for a perfect gas:

=

cp h
.
=
cv e

In the case of perfect gas all the possible definitions are identical, [4]. However, the
implementation of the state equation of real gases in equilibrium in an Euler model requires the
p
p
and
, if the speed of sound is
e
e

p p ( , e ) and the derivatives

knowledge of =

needed by the numerical algorithm. For air , the real gas description can be taken from [10, 9].

2.3.3 - Non-Equilibrium Real Gas

The composition of the dry air atmosphere is given in Table 2.1. The high temperature
occurring in hypersonic flows cause the excitation of vibrations and dissociation of N2 and O2
and the formation of N, O and NO. At even higher temperatures it occurs also ionization. The
argon an carbon dioxide are considered non-reacting components.
Table 2.1. Composition of atmospheric dry air, [6].

From [4], for instance, on a typical shuttle re-entry trajectory, the real gas effects have to
be taken into account for Mach numbers greater than 4. At altitudes below 40 km the air is
approximately in local thermal and chemical equilibrium. At altitudes above 40 km, due to the
decreasing free stream density the chemical relaxation time increases to the same order of
magnitude as the characteristic flight time leading to a non-equilibrium state of the gas. Finite
chemical reaction rates lead to chemical non-equilibrium and finite energy exchange rates of
vibrational degrees of freedom lead to thermal non-equilibrium, [4]. The bow shock layer
consists of a thin shock, a nearly inviscid region and a thin boundary layer, [3, 4]. However, the

(2-43)

influence of thermal non-equilibrium is expected to be rather small since the thickness of the
thermal non-equilibrium layer is only a few millimeters, [4].
We consider that the reacting air neglecting ionization can be modeled using five species
(i.e. nitric oxide NO, molecular N2 , O2 , and atomic N, O). Each species behaves as a perfect gas
and the Dalton law allows the calculation of the mixture pressure equation of state as a sum of
species partial pressures. Then, a chemical reactions model that includes a number of reactions
must be added, [4, 6]. There are models with 17 equations, [4] or less, see [6]. The
computational efficiency is of primary importance for the use of such a model in a coupled
manner with an Euler solver.

2.4 -

Simplified Euler equations

In this section we wish to consider successively simplified versions, or submodels, of the

3D Euler governing equations and their closure conditions. There are many reasons and many
ways to use and derive simplified models. On one hand, it is clear that any simplification leads
to the reduction of the generality of the equations. On the other hand, a rational simplification
may lead to a significant reduction of the computational effort without penalties on the quality
of the solution. There are some possibilities that can be exploited to derive simplified models,
and perhaps the most common of them are the reduction of the dimensionality of the problem
and the introduction of some new assumptions/hypothesis. One can also augment the basic
equations by adding source terms to account for additional physics. This is not exactly a
simplification of the basic model; on the contrary, adding some semi-empirical source terms
may enlarge the area of practical applications that can be solved with that model.

2.4.1 - 1D Compressible Flows with Area Variation

Compressible inviscid flows with area variation arise naturally in the study of fluid flow
phenomena in shock tubes and nozzles. One may start from the three dimensional
homogeneous version of Euler equations to produce, under the assumption of smooth area
variations, a quasi-one dimensional system with a geometric source term S(U) , namely:

U t + Fx (U) =
S(U) ,

(2-44)

where

u
uAx
1

2
pu 2 Ax .
U=
u , F =+
u p , S =
A
E
u ( E + p )
u ( E + p ) Ax

(2-45)

Here x denotes distance along the tube, nozzle, etc.; A is the cross-sectional area and in general
is a function of only space and time, that is A = A( x) . The flow variables are averaged
quantities on the transversal cross-section area. In this common case one can re-write the
equations in the more convenient form:
U t + Fx (U) =
S (U) ,

(2-46)

where

A
U = A u , F =
AE

A u
A( u 2 + p ) , S =

Au ( E + p )

0
pA .
x
0

(2-47)

The simplified version of (2-44) is the case without the geometric source term, namely:

0
U t + F(U) x =
These equations are useful for solving shock-tube type problems. Further, under suitable
physical assumptions they produce even simpler mathematical models. In all the submodels
studied so far we have assumed some thermodynamic closure condition given by equations of
state.

2.4.2 - Cylindrical and Spherical Symmetry

Cylindrical and spherical symmetric wave motion arises naturally in the theory of
explosion waves in air and other compressible media. In these situations: a) the microscopic
transport properties may be neglected due to the very fast physical processes and b) the
multidimensional Euler equation may be reduced to essentially one-dimensional equations with
a geometrical source term vector S(U) to account for the second and third spatial dimensions.
One can write therefore:

(2-48)

U t + Fr (U) =
S(U)

(2-49)

where

u
u

2
U=
u 2
u , F =+
u p , S =
r
E
u ( E + p )
u ( E + p )

(2-50)

Here r is the radial distance from the origin and u is the radial velocity. When = 0 one deals
with plane one-dimensional flow. When = 1 one deals with a simplified cylindrically
symmetric flow, which is an approximation to two-dimensional axy-symmetric equations when
no axial variations are present (v = 0) . For = 2 we have spherically symmetric flow, an
approximation to three-dimensional flow. Approximations of this kind can easily be solved
numerically to a high degree of accuracy by a good one-dimensional numerical method. These
accurate one-dimensional solutions can then be very useful in partially validating two and three
dimensional numerical solutions of the full models.

2.4.3 - Axi-Symmetric and Plane (2D) Flows

Here we consider domains that are symmetric around a coordinate direction. We

choose this coordinate to be the x-axis and is called the axial direction. The second coordinate is
r, which measures distance from the axis of symmetry z and is called the radial direction. There
are two component of velocity, namely the axial u (t , x, r ) and the radial v(t , x, r ) velocities.
Then the three dimensional (homogeneous) conservation laws are approximated by a two
dimensional (inhomogeneous) problem with geometric source terms S(U) , namely:

U t + Fr (U) + G z (U) =
S(U)

(2-51)

where

,F =
U=
v

E

u
2

u + p , G =
uv

u ( E + p )

v
u
uv

, S = u
u 2 + p
r uv

u ( E + p )
v( E + p)

When = 0 one deals with plane, two-dimensional flow. When = 1 one deals with an axysymmetric flow.

(2-52)

2.5 -

Mathematical properties of the Euler equations

In this section we apply some mathematical tools to find the basic properties of the
time-dependent Euler equations. As seen in previous sections, the Euler equations result from
neglecting the effects of viscosity, heat conduction and body forces on a compressible medium.
Here we show that these equations form a system of hyperbolic conservations laws. In
particular, we study those properties that are essential for the numerical solution of the
equations: the eigenstructure of the equations (that is, the eigenvalues and eigenvectors), the
properties of the characteristic fields and the basic relations across rarefactions, contacts and
shock waves and finally the solution of the Riemann problem.
We assume in what follows that the working gas is thermally and calorically perfect.
This only simplifies the analysis but do not represent a particular case. The use of more
complicated state equations only changes the way the sound speed is determined but does not
change the conclusions obtained within the perfect gas assumption [1, 2, 4, 5].

2.5.1 - The 1D Euler equations

The conservative differential form of the 1D Euler equations was derived before and is:

U t + F(U) x =
0,

(2-53)

where U and F(U) are the vectors of conserved variables and fluxes, given respectively by:

u1
f1 u

U =u2 u , F = f 2 u 2 + p
u3 E
f3 u ( E + p )
Here is density , p is the pressure, u is particle velocity and E is total energy per unit volume,
1
E = ( u 2 + e) , where e is the specific internal energy given by the caloric equation of state
2
e = e( , p ) . For ideal gases one has:

(2-54)

(2-55)

p
,
( 1)

e = e( , p ) =

with = c p cv denoting the radio of specific heats. From (2-55) and using equation (2-40) one
has the speed of sound a as:
a = ( p 2 e ) e p =

p
.

(2-56)

The 1D conservation laws may also be written in quasi-linear form:

U t + A(U)U x =
0,

(2-57)

where the coefficient matrix A(U) is the Jacobian matrix:

f1 u2
def f1 u1
F
A(U) =
= f 2 u1 f 2 u2
U
f3 u1 f3 u2

f1 u3
f 2 u3 .
f3 u3

(2-58)

After some manipulations, one obtains using (2-58) the Jacobian A in the form:

u2
1

A(U) =
( 3)

2
u1

3
u u
u2
2 3
2 + ( 1)
u1
u1

1
u
(3 ) 2
u1
u
3
( 1) 2
u1 2
u1

u3

1 ,

u2

u1

(2-59)

or using the speed of sound and the fluid velocity:

0
1
0

A(U) =
( 3)u 2
(3 )u
1 .

2
2
1 2 u 3 a u 3 2 u 2 + a
u
)
2 (
2
1
1

The Euler equations with the ideal-gas satisfy the homogeneity property:

(2-60)

F ( U ) = A(U)U .

(2-61)

The homogeneity property can be proof by direct calculations and is useful in deriving the
family of numerical schemes called Flux Vector Splitting, [1, 2, 4, 5].
In what is concerning the eigenstructure of the 1D Euler equations, the eigenvalues of
the Jacobian matrix A are:

1= u a,
2 = u, .
3= u + a.

(2-62)

=
The corresponding right eigenvectors are, using the total enthalpy H

R (1)

( E + p) :

1
1

.
(2)
(3)
=
u , R =
u a , R =
u+a
1 2
H ua
H + ua
u
2

(2-63)

The eigenvalues are all real and the eigenvectors R (1) , R (2) , R (3) form a complete set of
linearly independent eigenvectors. Following the theory of systems of first order partial
differential equations, the time-dependent, one-dimensional Euler equations for ideal gases are
strictly hyperbolic, because the eigenvalues are all real and distinct. It is worth to remark that
hyperbolicity remains a property of the Euler equations for more general equations of state.
The Euler equations in general and the 1D Euler equations in particular may be written
in terms of variables other than the conserved variables used in (2-53) and (2-54). These new
formulations are called non-conservative formulations and one can notice that for smooth
solutions all formulations are equivalent. For theoretical developments the non-conservative
formulations have some advantages over their conservative counterpart. For the onedimensional case one possibility is to choose a vector V = ( , u , p ) of primitive or physical
T

variables, with p given by the equation of state. Expanding derivatives in the mass equation, we
obtain:

Vt + A(V)Vx = 0 ,
where

(2-64)


V =
u , A(V)
=
p

0
u
0 u 1 .

0 a 2 u

(2-65)

The change of variable does not change the (hyperbolic) nature of the 1D Euler equations. Thus
the matrix A(V) has three real eigenvalues and three linearly independent eigenvectors, that is:

1= u a,
2 = u,
3= u + a,

(2-66)

and

(1)

1
1
1

a .
(2)
(3)
=
a , K =
0 , K =

a 2
0
a 2

(2-67)

By changing the variables, the eigenvalues remain the same but the eigenvectors have other
structure. The left eigenvectors of the matrix A(V) are:

1 a ) ,

=
L(2) ( 2 , 0, 2 a 2 ) ,

L(3) = ( 0, 3 , 3 a ) .

=
L(1)

( 0, 1 ,

(2-68)

where 1 , 2 , 3 are scaling factors or normalisation parameters.

For 1 =
(such that L(i)K(i) = 1, a normalization criteria) one
, 2 =
1, 3 =
2a
2a
obtains:

( 0,
L
=
(1,
L = ( 0,
L(1)
=
(2)
(3)

1 2a 2 ) ,

2
0,
1 a ) ,

2a , 1 2a 2 ) .

2a ,

The theory of systems of first order partial differential equations uses the concepts of
characteristics and characteristics equations, respectively. Thus, the characteristic directions
in the space-time domain are defined by:

(2-69)

dx dt = i for i = 1,2,3 ,

(2-70)

and the corresponding characteristic equations are defined by:

L(i ) =
dV 0, along dx=
dt i , =
for i 1, 2,3 .

(2-71)

In the case of the Euler equations, one obtains the following three characteristic equations (in
fact, they are differential relations that hold true along characteristic directions):
0 along dx dt =
1 =
u a,
dp adu =

2
0 along dx dt =
2 =
u, .
dp a du =
dp + adu =
0 along dx dt =
3 =
u+a

(2-72)

The characteristic directions and equations are useful and necessary in the numerical solution
of the Euler and Navier-Stokes equations, being used for the correct implementation of the
boundary conditions as well as for the solution of local Riemann problems.
The Riemann problem for the 1D, time dependent Euler equations (2-53) with initial
data (U L , U R ) is the Initial Value Problem (IVP) defined by the governing equations and the
constant initial conditions:

0, < x <
U t + F(U)
=
x

(2-73)

U if x < 0, ,
(t 0,=
( x) L
U
=
x) U (0)=
U R if x > 0.
The physical analogue of the Riemann problem is the shock-tube problem, in which the left and
right side velocities are zero. The Riemann problem is a key problem in the numerical solution
of compressible flows, [1, 2, 5]. Its exact solution is complicated, but a linearized one can be
obtained and solved using acceptable approximations, [1, 2]. This is key problem in the socalled Flux Difference Splitting family of solvers used in numerical solution of compressible
flows.
Remark on the entropy. The entropy is defined by:
p
s s0 =
cV ln

= =
u the
where s0 is a constant. It can be shown that along the characteristic line dx dt
2
0 so long as the flow remains smooth.
entropy is constant, i.e. st + u sx =

(2-74)

2.5.2 - Two-Dimensional Euler Equations

The two-dimensional version of the Euler equations in differential conservative form is:
0,
Ut + F ( U ) x + G ( U ) y =

(2-75)

with

u
, F
U =
=
v

E

u
u 2 + p

, G
=
uv

u ( E + p )

v
uv

.
v2 + p

v ( E + p )

(2-76)

The Jacobian matrix A ( U ) corresponding to the flux F ( U ) is given by (the Jacobian of

the other flux may be calculated similarly):

0
1

u 2 + 1 ( 1) V 2
(3 ) u
2
dF
A (=
U) =
v
uv
dU

2
2
u 2 ( 1) V H H ( 1) u

( 1) v 1

.
u
0

( 1) uv u

(2-77)

The eigenvalues of A and the corresponding right eigenvectors are:

1= u a,
=
=
u,
2
3
4= u + a

and

(2-78)

 1
ua
1
, R ( 2)
=
R( )
=
v

H au

1
u

( 3)
v=
,R

1V 2
2

0
0
=
, R ( 4)
1

v

1
u+a

H + ua

(2-79)

Thus, the Jacobians of the fluxes have real eigenvalues and linearly independent eigenvectors.
In contradistinction with the 1D case, one of the eigenvalue is multiple.
The Jacobian matrix A ( U ) is diagonalizable, because there is a similar transform so
that it can be written as:

A ( U ) =RU
( U )R

( U)

( ),

(2-80)

where R ( U ) is the non-singular matrix the columns of which are the right eigenvectors
i
R ( ) ( U ) , R 1 ( U ) is its inverse and ( U ) is the diagonal matrix with the eigenvalues i ( U )

given by (2-78) as the diagonal entries.

Remark on the Rotation Invariance. The fluxes and the vector of conservative variable
given in (2-80) satisfy an important property, called the rotational invariance. Thus, given an
unit vector n = ( cos ,sin ) normal to a line, where is the angle formed by x-axis and n,

0 2 , the normal flux Fn ( U ) to the line is:

F=
n

n
( F, G ) =

cos F ( U ) + sin G ( U ) ,

(2-81)

The rotational invariance means that the 2D Eulers equations satisfy for all angles and
vectors U the property:

Fn = cos F ( U ) + sin G ( U ) = T-1F ( TU ) .

The matrix T = T ( ) is the rotation matrix and T-1 ( ) is its inverse:

(2-82)

0
0
0
1
0 cos sin 0
, T1
=
T =
0 sin cos 0

0
0
1
0

0
1
0 cos

0 sin

0
0

0
sin
cos
0

0
0
.
0

(2-83)

The rotational invariance property of the two-dimensional time dependent Euler equations is
fundamental in numerical calculation but also allows us to proof that the equations are
hyperbolic in time.
By definition, the 2D time dependent Euler system (2-75) is hyperbolic in time if for all

admissible states U and angles the Jacobian of the normal flux A n ( U, =

) cos A + sin B

is diagonalizable ( A and B are respectively the Jacobian matrices of the fluxes F ( U ) and

G ( U ) ), [1, 2, 5]. That is, there exist a diagonal matrix ( U, ) and a non-singular matrix
R ( U, ) such that:
A n ( U,) =UR ( UR
, ) U ( , )

( , ) .

(2-84)

where R ( U ) is the non-singular matrix the columns of which are the right eigenvectors
R (i ) ( U ) , R 1 ( U ) is its inverse and ( U ) is the diagonal matrix with the eigenvalues i ( U )

given by (2-78) as the diagonal entries. By taking the derivative with respect to U of the normal
flux given by (2-82) one obtains:
cos A + sin B
= T-1
A=
n

d F ( TU )
= T-1 A ( TU ) T .
T
d ( TU )

(2-85)

Further, using (2-80) it results that the equality (2-88) may be obtained:
An

-1
T=

( TU )R ( TU) T1 ( T ) )R TU( -1 (TU ) ) R ( TU) ( T1 (

( R TU

),

(2-86)

provided that:

R ( U, ) = T-1 R ( T ( ) U ) ,
( U, ) = ( T ( ) U ) .
Thus, by direct matrix manipulations we have proofed that the condition (2-84) holds and
therefore the two dimensional Euler equations are hyperbolic time, [1. 2].

(2-87)

2.5.3 - The 3D Euler Equations

The previous results proved for the one- and two-dimensional Euler equations can be
extended to the time-dependent three dimensional Euler equations, (2-26) and (2-27). One
may recall them here:
U t + Fx (U) + G y (U) + H z (U) =
S(U) ,

where:

u

=
U =
v , F

w
E

u
u 2 + p

uv=
, G

puw

u ( E + p )

v
uv

v 2 +=
p , H

vw
v( E + p )

w
uw

vw .
2

w + p
w( E + p ).

The Jacobian matrix A corresponding to the flux F ( U ) is given by:

H u 2 a 2

F
uv
=
A =
U
uw

1
2
u ( 3) H a
2

0

(3 ) u
v
u
0
,
w
0
u
0

H u 2 uv uw u

0
v

0
w
0

(2-88)

1
a2
where H =( E + p ) = V 2 +
, V 2 =u 2 + v 2 + w2 , = 1 . This matrix has five real
2
( 1)

eigenvalues:

1= u a,
=
=
=
u,
2
3
4
5= u + a
The matrix of corresponding right eigenvectors is

(2-89)

 1
ua

v
R=
w

H ua

1
u
v
w
1 2
V
2

0
0
1
0
v

1
0 u + a
v
0

w
1

w H + ua

(2-90)

while its inverse is

a
a
H + (u a ) u +

4 2
2u
2 H + a

( 1)
2va 2
R 1 =

2a 2

2 wa 2

0

a
a
H ( u a ) u +

2v

2w

2a 2

2a 2

(2-91)

The Jacobian matrix A ( U ) given by (2-88) is diagonalizable, because there is a similar

transform so that it can be written as:

A ( U ) =RU
( U )R

( U)

( ),

(2-92)

where R ( U ) is given by (2-90), R 1 ( U ) is given by (2-91) and ( U ) is the diagonal matrix

with the eigenvalues i ( U ) given by (2-89) as the diagonal entries.
The time-dependent three dimensional Euler equations are also rotationally invariant,
that is for the normal flux through a surface they satisfy, for all angles ( y ) , ( z ) and vectors U,
the equality:
y
z
y
z
y
=
Fn cos ( ) cos ( ) F ( U ) + cos ( ) sin ( )G ( U ) + sin ( ) H=
( U ) T-1F ( TU ) .

Here the rotation matrix T = T ( y ) , ( z ) is the product of two simple rotations matrices, [2]:

(2-93)

=
T T=
( y ) , ( z ) T( y ) T( z ) .

(2-94)

where

T ( y)

0
0
0
0
1

y
( y)
0 sin ( ) 0
0 cos
z
=
0
0
1
0
0 , T ( )

0 sin ( y ) 0 cos ( y ) 0

0
0
0
1
0

0
1

( z)
0 cos
0 sin ( z )

0
0

0
0

0
sin (

z)

cos (
0
0

z)

0 0

0 0
0 0
1 0

0 1

(2-95)

and

0
1
0 cos ( y ) cos ( z )

T = 0
sin ( z )

(y)
(z )
0 sin cos
0
0

0
0
(z )
cos sin
sin ( y )
(z )
cos
0
(y)
(z )
sin sin
cos ( y )
0
0
(y)

0
0
0

0
1

Similar to the two-dimensional case, the three-dimensional Euler equations are proofed
to be hyperbolic in time, [1, 2].

2.5.4 - Conclusion

The Euler equations are hyperbolic in time irrespective the space dimension is. The
inviscid fluxes are homogeneous and the variables and multi-dimensional fluxes satisfy the
rotational invariance condition. These mathematical properties are useful for the development
of the numerical algorithms dedicated to the solution of the Euler equations for
supersonic/hypersonic flows.

(2-96)

3-

REFERENCES

1. Danaila, S., Berbente, C., Metode Numerice in Dinamica Fluidelor, Editura Academiei
Romane, Bucuresti , 2003.
2. Toro, E.F., Riemann Solvers and Numerical Methods for Fluid Dynamics, Springer-Verlag,
Berlin, Heidelberg, 1997.
3. Anderson, J.D., Jr., Hypersonic and High Temperature Gas Dynamics, McGraw-Hill Series in
Aeronautical and Aerospace Engineering, 1998.
4. Weiland C. and all, Numerical Methods for Aerodynamic Design I, Space Course 1991,
RWTH Aachen, 1991.
5. Feistauer, M., Felcman, C., Straskraba, I., Mathematical and Computational Methods for
Compressible Flow, Clarendon Press, Oxford, 2003.
6. Iannelli, J., Characteristics Finite Element Methods in Computational Fluid Dynamics,
Computational Fluid and Solid Mechanics Series, Springer, 2006.
7. Masatsuka, K., I do like CFD, Vol. 1, Yorktown, 2009.
8. Park, C., Hypersonic Aerothermodynamics: Past, Present and Future, Intl J. of Aeronautical
and Space Sci. 14(1), pp. 1-10, 2013.
9. Barbante, P.F., Magin, T.E., Fundamentals of hypersonic flight Properties of high
temperature gases, RTO-EN-AVT-116, Paper presented at RTO AVT Lecture Series on
Critical Technologies for Hypersonic Vehicle Development, held at the von Krmn
Institute, Rhode-St-Gense, Belgium, 10-14 May, 2004.
10. Srinivasan, S., Tannehill, J.C. and Weilmuenster, K.J. Simplified Curve Fits for the
Thermodynamic Properties of Equilibrum Air, by S., ISU-ERI-Ames-86401, Iowa State
University, 1986.