A new concept for the modeling of detonation waves in

multiphase mixtures

Ashwin Chinnayya *, Eric Daniel*, Richard Saurel* , Grard Baudin
+
,C. Le Gallic
+

* PolytechMarseille, 5 rue E. Fermi, 13453 Marseille Cedex 13, France
also INRIA Projet SMASH, 2004 route des Lucioles, 06902 Sophia Antipolis
+
DGA/DCE/Centre dEtudes de Gramat, 46500 Gramat, France

This paper addresses two key issues of detonation wave modelling in hydrocodes. The first
issues deals with the computation of the multiphase mixture variables. We develop a new
model that deals accurately with non-equilibrium multiphase mixtures. The second issue lies
in the computation of thermodynamic variables in so called mixture cells when using Eulerian
hydrocodes. We show that the new multiphase model is able to deal with mixture cells
resulting of numerical smearing of interfaces as well as physical mixtures, which result of
explosive decomposition.

INTRODUCTION
A new numerical concept is presented to
deal with two basic problems encountered
in the computation of detonation waves in
condensed energetic materials.
The first problem is due to the
multiphase feature of the condensed
energetic material involving several
compressible materials. The Euler
equations (plus several species
conservation equations) do not provide
enough thermodynamic information to
determine the mixture pressure and sound
speed. The mixture EOS is most times very
inaccurate
1,2,3
.
The second problem is related to the
numerical treatment of material interfaces.
Our numerical strategy adopts full eulerian
methods without front tracking, Level Set
or interface reconstruction. But it is well
known that Eulerian methods create
artificial mixtures due to numerical
diffusion at the interfaces
1
.
In a recent paper
4
, some difficulties
related to multiphase approaches have
been cleared. The key idea is to discretize
the multiphase mixture at the microscopic
level and then to average the discrete
equations. It provides a new discrete model
as well as the numerical method. This is
done in the opposite way of what was done
before. Indeed, it was conventional to
obtain a system of PDEs on the basis of
averaging procedures, and then to
discretize the corresponding PDE system.
Nevertheless this method needs a lot of
extensions to deal with the applications of
confined multiphase detonations:
* The materials are reactive.
* The equations of state are complex.
*The mixture behaves essentially with a
single velocity and pressure
2
. It is then
necessary to adopt a description of the
mixture that relaxes intrinsically towards
pressure and velocity equilibrium between
all the phases.
* The mixture involves an arbitrary
number of phases.
* The geometries are 2D axisymmetric.

In this paper, the reactive multiphase
model is first written in a conventional
continuous form. Then the basis of the new
method is developed. It is then extended to
account for the various relaxation effects.
It is then validated upon a set of illustrative
computations.

CONTINUOUS MODEL
In this section we recall a BN type
1,2,3,5

continuous multiphase model used for
detonation waves computations.
2
Each pure phase k is assumed to obey the
reactive Euler equations. The mixture is
composed of N phases ( | | N k , 1 ). Each
phase k is composed of
k
N chemical
species and the m
th
species is defined by its
mass fraction
k m
Y
,
.
A conventional averaging method provides
the following system.

= +
+ +
= + +
+ +
= + +
= +
+ = +
mk
m
k
u
m
Y div
t
k
m
Y
i
E
k
m
k i
p
k
F
i
u
k i
u
i
p
k
u p E div
t
k
E
i
u
k
m
k
F
k i
p
k
p
k
u u div
t
k
u
k
m
k
u div
t
k
i
k
m
k k i
u
t
k
&
r
& &
r
r r
r
r
&
r
r r
r
&
r
&
&
r
) (
) (
) ) ( (
) (
) ( ) (
) (
) (
) (

Standard notations of the multiphase
literature are used. The index i refers to
interfacial quantities which need to be
modeled in order to close the system. The
modeling of these interfacial quantities
remains a difficult fundamental issue. The
determination of the drag force (
k
F
r
),
pressure relaxation rate (
k
& ) is another
difficulty. The numerical approximation of
these equations is also difficult due to the
presence of non-conservative terms and
fluxes involving many elementary waves
that render the system difficult to upwind.
The new approach, developed in the next
section solves these difficulties. A
consequence is that numerical mixtures,
resulting of the explosive decomposition
and artificial smearing of interfaces will be
computed accurately.

THE DISCRETE EQUATIONS
METHOD (DEM)
A numerical model for multiphase flows
that relaxes intrinsically towards
mechanical equilibrium (velocities and
pressures equilibrium), is developed in this
section.
The DEM considers the multiphase
mixture at the microscopic/grain scale. At
this scale, the interactions between phases
are given by the solution of Riemann
problems between two fluids governed by
the reactive Euler equations. Thus,
microscale interface problems are solved
with the help of the Godunov scheme. The
solution of these interface problems are
averaged over the control volume and
provide macroscopic discrete equations of
the mixture.
For the sake of simplicity we first consider
that mass transfer is absent. The
multiphase control volume corresponds to
the macroscopic computational cell
i
C . It
contains N phases.
The pure fluid equations are multiplied by
an indicator function
k
X to select the
appropriate fluid. In the computational cell
i
C , between the time 0 and t , the
evolution of the variables of the phase
k
F
obey the following equation:
0
0
=
(

|
.
|

\
|

t
C
k
i
dxdt
x
F
t
W
X (1)
where ) , , , , 1 (
m
Y E u W = and
) , , , , 0 ( ) (
2
m
uY pu Eu p u u W F + + = . The
first equation is a trivial identity. The
indicator function obeys the topologic
equation: 0 =

x
X
t
X
k k

where represents the local interface
velocity
6
.
After some algebraic manipulations
equation (1) becomes:
(

=
(

|
.
|

\
|

t
i
C
k
t
i
C
k k
dxdt
x
X
W F
dxdt
x
F X
t
W X
0
0
) (
) ( ) (

We have to evaluate these three integrals.
i lag i conv i temp
I I I
, , ,
= +
where
3
(

=
t
C
k
i temp
i
dxdt
t
W X
I
0
,
) (
(temporal term)
(

=
t
C
k
i conv
i
dxdt
x
F X
I
0
,
) (
(convective flux)
dxdt
x
X
W F I
t
C
k
i lag
i
(

=

0
,
) ( (Lagrangian flux)

This equation represents the solution of
interface problems in the cell and its
resolution is the key point of the present
multiphase approach.
Each control volume
i
C contains initially
N phases. Each phase is discernable and
the mixture is assumed immiscible. The
mixture consists in a succession of fluids
layers separated by interfaces as depicted
in Figure 1.
0
t
o
T B
1
T
2
T
M
T C
1 M
T
o
D
1
D
2
D
M
D
1 + M
D
0
2 / 1
x x
i

M
i
x x
+ 2 / 1
1 3

M
D D
1
x
2
x
1 M
x
A D
Figure 1: Description of the 1D topology in the
cell
i
C
During time evolution, 2 + M lagrangian
domains have to be considered as depicted
on the previous figure. Each domain may
contain any fluid
k
F . The trajectories paths
j
T M j .. 0 = separate the different zones
and represent the different interface
trajectories inside the fixed volume control
i
C during the time step t . Each trajectory
path
j
T , M j .. 0 = satisfy the equation of
motion t x t x
j j j
+ = ) ( . Note that for
1 .. 1 = M j the
th
j interface always lie into
the computational cell
i
C because the time
step is assumed sufficiently small. They
are named internal interfaces.
0
T lie into
i
C only if the sign
0
is positive.
M
T lie
into
i
C only if the sign
M
is negative.
The domain
0
k
D randomly contains the
fluid
0
k
F for ] , 1 [
0
N k ; the domain
1
k
D
randomly contains the fluid
1
k
F for
] , 1 [
1
N k and so forth. The fluids in the
domains
kj
D and
1 + kj
D lie on both sides of
the
th
j interface, 1 .. 0 + = M j . In the domain
0
k
D , the fluid
0
k
F may come from the
1 i
C
computational cell. In the domain
1 + M
k
D ,
the fluid
1 + M
k
F may come from the
1 + i
C
computational cell. This defines a random
configuration ) ,..., (
1 0 + M
k k P of the
considered topology in cell
i
C .
For computational purposes, we define a
numerical projection of the microscopic
variables W X
k
. This numerical projection is
valid for any random configuration (event):
( ) ( )
(

=
+ +
i
C
n n
i k
dx t x event XW
x
event XW ) , ( ) (
1
) ( ) (
1 1
,

The macroscopic variables at the end of
time step 1 + n are given by:
( )

=
+
event
n
i k
event XW event P W ) ( ) ( * ) ( ) (
1
,

where ) (event P is the probability of event .
For any configuration, a shock-capturing
scheme is considered for the variables
( ) )) ,..., ( ( ) (
1 0
1
, +
+
M
n
i k
k k P XW obtained by an
average projection:
(

=
+
+
+
+
i
C
n
M
M
n
i k
dx t x k k P XW
x
k k P XW
) , ))( ,..., ( )( (
1
)) ,..., ( )( ) ((
1
1 0
1 0
1
,

The solution for the discrete value of
k
XW ) ( requires the calculation of the three
integrals defined previously:
i temp
I
,
,
i conv
I
,
,
and
i lag
I
,
.
Approximation of the temporal terms
( )
( )

=
+
+

=
1
0
1
, ,
1
,
) ( ) (
M
j
j j n
i k k i k
n
i k i temp
x x W X
W X x I
j

where
kj j i k
X = ) (
,
, with
kj
the Kronecker
symbol.
Approximation of the convective flux
4

| |

{ }
A
i k
D
i k
t i
i k i conv
F X F X t dt F X I
2 / 1 2 / 1 0
2 / 1
2 / 1 ,
) ( ) (
+
+

= =
where
) , ( ) ( ) (
1
2 / 1 2 / 1
+
+ +
=
M
k
M
k
D
i k
D
i k
F X F X F F
) , ( ) ( ) (
1 0
2 / 1 2 / 1 k k
A
i k
A
i k
F X F X F F

= .
and
) , ( ) ( ) , ( ) (
) , ( ) (
2 / 1 1 , 2 / 1 ,
2 / 1
q p i q i k q p i p i k
q p
D
i k
F X F X
F X
F F F F
F F

+ +
+
+
+
+
=

where

>
=
+
+
otherwise
if
*
F F
F F
0
0 ) , (
) , (
*
2 / 1
pq q p
q p i
u F
F
and

>
=

+
otherwise
if
F F
F F
) , (
0 0
) , (
*
*
2 / 1
q p
pq
q p i
F
u
F
and
kp p i k
X = ) (
,
.
The convective flux
) , ( ) ( ) (
1 0
2 / 1 2 / 1 k k
A
i k
A
i k
F X F X F F

= is computed
in the same way.
) , ( ) ( ) , ( ) (
) , ( ) (
2 / 1 , 2 / 1 1 ,
2 / 1
q p i q i k q p i p i k
q p
A
i k
F X F X
F X
F F F F
F F

+
=

Approximation of the Lagrangian flux
( ) dxdt
x
X
W F I
t
i
C
k
i lag (

=

0
,

The expression of the Lagrangian flux is
( ) pu mE p mu m W F + + = , , ,
where ) ( = u m .
In absence of mass transfer = u . Then the
Lagrangian flux is given by
( ) pu p u W F , , 0 , =
W F is constant across the interface and
the gradient of the indicator function is
zero except on interfaces.
{ | |
| | | | }

=
+ +

+
+ =
1
1
2 / 1 2 / 1
2 / 1 2 / 1 ,
) ( ) (
) (
M
j
j
i k
j
i
C
i k
C
i
B
i k
B
i i lag
X W F X W F
X W F t I

| |
j
i k
X denotes the jump of the indicator
function of the
th
k phase across the
th
j
interface.
B
i
W F
2 / 1
) (

is function of
1
k
F
and
0
k
F and of the sign of the velocity of
the interface B .
| |
| |
0
1 ,
1
,
0 1
2 / 1
2 / 1 2 / 1
) ( ) ( ) , ( ) (
) (
k i k k i k k k i
B
i k
B
i
X X W F
X W F

=
F F

where

>
=
+

otherwise
if
*
F F
F F
0
0 ) , ( ) (
) , ( ) (
*
2 / 1
pq q p
q p i
u W F
W F

<
=

otherwise
if F F
F F
0
0 ) , ( ) (
) , ( ) (
* *
2 / 1
pq q p
q p i
u W F
W F

The other Lagrangian fluxes are computed

in the same way.
Thus, for a specific event ) ,.. (
1 0 + M
k k P , the
numerical scheme reads
( )
( )
{ }
{ | |
| | | | }

=
+ +

+

=
+
+
+
+
+ +

=
1
1
2 / 1 2 / 1
2 / 1 2 / 1
2 / 1 2 / 1
1
0
1
, ,
1 0
1
) ( ) (
) (
) ( ) (
) ( ) (
)) ,..., ( (
M
j
j
i k
j
i
C
i k
C
i
B
i k
B
i
A
i k
D
i k
M
j
j j n
i k
j
k i k
M
n
i k
X W F X W F
X W F t
F X F X t
x x W X
k k P W X x






The numerical scheme for the microscopic
variables of a particular event has been
developed. It must be averaged in a proper
manner in order to get the numerical
scheme for the macroscopic variables.
Each numerical scheme is multiplied by
the discrete probability ) ,.., (
1 0 + M
k k P of the
particular event:
( )
n
i
M
k
n
i
M
k
n
i k
n
i k M
k k P
1 ,
1
, ,
1
1 ,
0
1 0
... ) ,.., (
+
+
+
=
In this definition, the notation of the event
and its probability are the same for the
sake of simplicity in the notations.
The numerical multiphase model is finally
obtained by summing over all the different
configurations.
( )

=
+
=
+
+
+
+
=
N
M
k N k
M
n
i k M
n
i k
k k P XW k k P
W
.. 1
1
;...; .. 1
0
1 0
1
, 1 0
1
,
)) ,.., ( ( ) ( * ) ,.., (
) (
In the following, we note:

=
=
+
= N
M
k N k .. 1
1
;...; .. 1
0

and
j
k
i
k j i
j
k
i
k
k k P P ... ) ,..., (
,
= = .
Note that
1
,
1
,
+ +
=
j
k
j
k
i
k
j
k
i
k
P P and
j
k
l
k
l
k
i
k
j
k
i
k
P P P
, , ,
= for all
j l i
k k k . Note also
that 1
,
=

j
k
i
k
P .
After some algebraic manipulations of the
temporal term we get:
5
=
|
|
.
|

\
|

=
+
+
1
0
1
, , 1 0
)) ( ) (( ) ,.., (
M
j
j j n
i k
j
k i k M
x x W X k k P
n
i k
n
i k
W x
, ,

That yields the compact formula
{ }
n
i k
n
i k i temp
W W x I
,
1
, ,
) ( ) ( =
+


Let us examine the averaged convective
flux at
2 / 1 + i
x :
D
i conv
I
2 / 1 , +
.
The numerical convective
) , ( ) (
1
2 / 1
+
+
M
k
M
k
D
i k
F X F F depends only on the
variables of the different fluids
M
k
F and
1 + M
k
F . Therefore
( )

|
|
.
|

\
|
=
+
+
+
+
+
) , ( ) (
) , ( ) ( * ) ,.., (
1
1 ,
1
,
1
,
0
1
1 0
M
k
M
k
D
k
n
i
M
k
n
i
M
k
M
k k
M
k
M
k
D
k M
F X P
F X k k P
F F
F F



+
+
+ |
|
.
|

\
|
=
1
,
0 1
1 ,
1
,
) , ( ) (
M
k k
M
k
M
k
D
k
n
i
M
k
n
i
M
k
P F X F F
with 1
1
,
0
=

M
k k
P . So

+
+ +
+
+
= ) , ( ) (
1
2 / 1 1 ,
1
, 2 / 1 ,
M
k
M
k
D
i k
n
i
M
k
n
i
M
k
D
i conv
F X I F F
In the same way

= ) , ( ) (
1 0
2 / 1 ,
1
1 ,
0
2 / 1 , k k
A
i k
n
i k
n
i k
A
i conv
F X I F F
That means that
{ }
A
i conv
D
i conv i conv
I I t I
2 / 1 , 2 / 1 , , +
=

The averaged Lagrangian flux is computed
in the same way.
{ }
int
, 2 / 1 , 2 / 1 , , i lag
B
i lag
C
i lag i lag
I t I I t I + + =
+

where
| |

+
+
+
= ) ) , ( ) ((
1
1 ,
1
,
C
k
M
k
M
k
C n
i
M
k
n
i
M
k
C
lag
X W F I F F
( | | )

=

B
k k k
B n
i k
n
i k
B
lag
X W F I ) , ( ) (
1 0
,
1
1 ,
0
F F
| |

=
+
=
1
1
,
1
,
int
,
) (
M
j
j
i k
j
i
n
i
j
k
n
i
j
k i lag
X W F I

The numerical scheme thus reads:
{ }
{ }
int
,
2 / 1 , 2 / 1 , ,
1
,
) ( ) (
i lag
B
lag
C
lag
A
i conv
D
i conv
n
i k
n
i k
I
x
t
I I
x
t
I I
x
t
W W

+ +

=
+
+


Because
M
k
F belongs to
i
C and
1 + M
k
F
belongs to
1 + i
C , the convective flux
D
i conv
I
2 / 1 , +
can be written in a more
convenient way:
( )

= =
+ +
=
N r N q
i r i q
D
k
n
i r
n
i q
D
conv
F X I
.. 1 ; .. 1
1 1 , ,
) ( , ) ( ) ( F F
For similar reasons,
A
i conv
I
2 / 1 ,
,
C
i lag
I
2 / 1 , +
,
B
i lag
I
2 / 1 ,
can be written in the following
form:
( )

= =

=
N r N q
i r i q
A
k
n
i r
n
i q
A
conv
F X I
.. 1 ; .. 1
1 , 1 ,
) ( , ) ( ) ( F F
| | ) ) ( , ) )(( ) ( (
.. 1
.. 1
1 1 , ,
=
=
+ +
=
N r
N q
i r i q
C
k
C n
i r
n
i q
C
lag
X W F I F F
| |

=
=

=
N r
N q
i r i q
B
k
B n
i r
n
i q
B
lag
X W F I
.. 1
.. 1
1 , 1 ,
) ) ( , ) )(( ) (( F F
Because the different fluids
j
k
F , 1 .. 1 = M j
belong only to
i
C , the average Lagrangian
flux reads:
| |

= = =
=
1
1 .. 1 ; .. 1
, ,
int
,
) (
M
j N r N q
j
i k
j
i
n
i r
n
i q i lag
X W F I
Each convective or lagrangian fluxes is
computed with the help of a Riemann
solver. Note that the convective and
Lagrangian fluxes evaluated at the
boundary cell do not depend on the
number of internal interfaces that lie into
the computational cell.
The opposite remark stands for the
Lagrangian flux associated to internal
interfaces. Since the number of internal
interface is large, some simplifications are
possible and are detailed in the next
section. Before doing these simplifications,
let us mention that the general multiphase
discrete model is summarized by the
various formulas of the present column.

RELAXATION PROCESSES

It can be easily shown that the sum over all
the internal interfaces of the lagrangian
fluxes is, in this specific case, equal to:
|
|
.
|

\
|
=

= =
+
+

1
1 1
1
,
,
1
,
int
,
) ( lim
M
j
N
j
k
j
k
j
i k
j
i
n
i
j
k
n
i
j
k
M
i lag
X W F I

It means that the total amount of internal
interfaces tends to infinity. Infinity means
a large value, which is not useful to
determine. It is important to note that this
large number of interface is very small
compared to the number of molecules: the
media is still a separate flow mixture out of
equilibrium and not a molecular mixture.
6
Let us split the contribution of these
Lagrangian fluxes from the convective and
Lagrangian fluxes of the boundaries. Their
contribution to the evolution of
conservative variables is given by the
algebraic equation:
{ }
int
, ,
1
,
) ( ) (
i lag
n
i k
n
i k
I t W W x =
+

For a two-phase flow, the corresponding
semi-discrete equations associated to fluid
1 are:
) (
1 2 1
p p
t
=

0 ) (
1
=

t

) ( ) (
1 2 1
u u u
t
=

) ( ) ( ) (
1 2 1 2 1
p p P u u V E
t
I I
=

When the number of internal interfaces
tends to infinity ( M ), it can be easily
shown that the various differences of
Lagrangian fluxes of the Riemann problem
for the Euler equations provide the
preceding system where and tend to
infinity. It means that the pressures and the
velocities relax instantaneously. This result
can be generalized for N phases.

NUMERICAL STRATEGY

The numerical model stiffens as the
number of internal interfaces increases.
Indeed, the direct resolution of the
numerical model would require infinitely
small time step. We propose a numerical
strategy to get free from this constraint.
The scheme is be written in the following
form

t
W
i k

,
) (
{ } { }
4 4 4 4 4 4 4 4 4 4 3 4 4 4 4 4 4 4 4 4 4 2 1
Terms Transport
I I
x
I I
x
B
i lag
C
i lag
A
i conv
D
i conv 2 / 1 , 2 / 1 , 2 / 1 , 2 / 1 ,
1 1
+ +
+

+
0
1
int
,
=

4 43 4 42 1
Terms Internal
I
x
i lag

To alleviate the CFL constraint associated
with internal interfaces the solver is split
into two parts:
We first solve the Transport step:
0 ) (
) (
= +

W Transport
t
W

It is followed by the Internal step
1,3
:
0 ) (
) (
= +

i
i
W Internal
t
W

The scheme is extended to second order by
using a classical MUSCL strategy. A
special care is brought for the extension to
second order of the various non
conservative terms.
The extension to 2D configurations is
obtained by using a dimensional splitting.
The source terms involving chemical
reactions, mass transfer or other
phenomena that are not included in the
Lagrangian fluxes, are integrated after a
complete time step by using a well adapted
ODE solver for stiff equations
7
.

RESULTS
Detonation reaction zone
In this test, the solution of the multiphase
model is compared to the exact solution for
a steady detonation wave. A shock wave is
transmitted to an explosive by a piston
with a given velocity.
We compute the shock to detonation
transition and the stabilisation of the
detonation wave. The solid and the
gaseous phase (detonation products) obey
the same equation of state : e p ) 1 ( = ,
with a polytropic coefficient equal to 3.
This calculation corresponds to the ZND
model solution and the conditions used
here are those given in Fickett and Davis
8
.
In this specific case, the source terms
representing the mass transfer terms for the
gaseous products are:
)) ( , , , (
explo explo explo explo explo explo explo
PROD GAS
react
q E m u m m m
Mass
+
=
& & & &
The term
react
q corresponds to the heat of
combustion ( kg MJ . q
react
/ 5156 4 = ). The
mass transfer is given by
) (
explo explo explo

m
k m = & with
1 6
10 . 2

= s k
and

=
k k m
is the mixture density.
7
Initially, the pressure is set at 1 bar and the
densities of the both phases are equal to
3
/ 1600 m kg . The initial volume fraction of
the gaseous product is
6
min
10

= . The
piston velocity is 1000 m/s.
The total length of the domain is 0.3 m.
1000 cells are used for the mesh, with a
constant length (3.3 cells/mm).
In Table 1, relevant analytical values are
reported. They are compared with the
numerical results.


Pressure
(10
8
Pa)
Density
(kg/m
3
)
Velocity
(m/s)
CJ Point 289 2133 2125
Neumann
spike
578 3200 4250
Table 1: Characteristic values of the ZND model
for detonation wave

Analytical Numerical
Velocity of the
detonation front
8500 m/s 8486 m/s
Length of the
reaction zone
5.31 10
-3
m 5.2 10
-3
m
Table 2: Comparison between ZND characteristic
values and numerical characteristics values

The numerical velocity of the detonation
zone and the reaction zone thickness are in
very good agreement with the analytical
solution (Table 2).
In Figure 2, the numerical pressure is
plotted with solid lines while the analytical
solution in the reaction zone is plotted with
symbols. The Neumann spike and the
Chapman-Jouguet point are well
reproduced by the numerical model. The
numerical solution converges to the exact
one in the reaction zone. The density and
of course the temperature of the gaseous
product is not equal to the ones of the
explosive. This obvious remark is however
a proof that models based on the Euler
equations and mixture equations of state
based on thermal equilibrium are wrong
for this type of application.

0
100
200
300
400
500
600
0 0.05 0.1 0.15 0.2 0.25 0.3
x (m)
Pressure (10
8
Pa)
1500
2000
2500
3000
3500
4000
0 0.05 0.1 0.15 0.2 0.25 0.3
x (m)
Densities (kg/m
3
)
Figure 2: Numerical solution of the ZND
problem, Neumann spike and CJ point value
Detonation of a condensed CHNO-PA-
Al non ideal energetic material
The multiphase model is used to calculate
the detonation properties of a condensed
energetic material. For a highly sub-
oxygenated material, the chemical
decomposition is given by:
(1) C
a
H
b
N
c
O
d
dH
2
O + (b/2-d)H
2
+ c/2N
2
+ aC
(2) Al + G Al
2
O
3
+ C + G'
(3) NH
4
CLO
4
1.5 H
2
O +0.5 N
2
+ 1.25 O
2
+ HCl
(4) C + O
2
CO
2

(5) H
2
+ 0.5 O
2
H
2
O
A piston impacts the reactive material with
the velocity of
1
. 400

s m . The overall
computational domain (one meter long) is
at the initial pressure of 1 bar. Both phases
are at rest. We have chosen the initial
gaseous composition: 1
2
=
O H
Y , 0 =
i
Y for the
other gases with an initial volume fraction
of 10
-3
.
We compute the shock to detonation
transition and the stabilisation of the
detonation wave. The spatial resolution is
one cell per millimetre. The reactive
material is governed by a multi-component
Mie-Gruneisen equation of state whereas
the detonation gaseous products are
8
governed by a multi-component H9
equation of state.
All the chemical decomposition are based
on a Vieilles law. The different rates of
change of the species concentrations | |
i
X
are modelled by:
| | | |

=
+
=
|
|
.
|

\
|

=
react
s g
j
l
N
j
N N
l
l
j
ij ij
i
X X
1 1

&

j
is a function of the relaxed pressure:
n
o
j
j
p
p
A
|
|
.
|

\
|
=

1
, 2 = n .
o
p is a reference
pressure. The various rates of change of
the mass fractions
i
Y are modelled by:
( ) ( )

= = =
|
|
.
|

\
|
=
react
N
j
g
N
kg
j
kg
kg g
s
N
ks
j
ks
ks s
j
ij ij
i
Y Y m
1 1 1
) ( ) (

&
After integrating the whole system of
equations in a 1D code, the 1D numerical
simulation represents a detonation wave in
a section of infinite area. The numerical
results are shown hereafter. A magnified
view of the variables inside the reaction
zone is shown on Figure 3: Two
characteristic time scales are visible for the
chemical decomposition condensed
material (3a). The decomposition of the
organic molecule CHNO is the fastest. Its
decomposition releases part of the energy
contained into the solid. Without this
initial released energy, the other highly
energetic components of the solid
(Ammonium Perchlorate PA and
Aluminium Al ) cannot decompose. The
PA decomposes totally but there is not
enough oxygen in the reactive mixture to
allow the Al to burn totally. There is still
energy that can be released in the system.
a)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0.85 0.86 0.87 0.88 0.89 0.9 0.91 0.92
Yk solid mass fractions

b)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.85 0.86 0.87 0.88 0.89 0.9 0.91 0.92
Yk solid mass fractions

c)
0
0.2
0.4
0.6
0.8
1
0.85 0.86 0.87 0.88 0.89 0.9 0.91 0.92
Yk gas mass fractions

d)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.85 0.86 0.87 0.88 0.89 0.9 0.91 0.92
Yk gas mass fractions

Figure 3: Solid components mass fractions:
a): CHNO (circles), PA (triangles), Al (squares)
b): Al
2
O
3
(circles), C (squares)
Gas components mass fractions:
c): CO
2
(circles), H
2
O (triangles), O
2
(squares)
d): N
2
(circles), HCl (triangles), H
2
(squares)

The combustion of both PA and Al has
thickened the reaction zone because the
rates of their decomposition are lower.
The condensed material volume is about
30 % (cf. solid and gas volume fractions in
Figure 4) even far from the reaction zone.
0
50
100
150
200
250
300
350
400
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Pressure (10
8
Pa)

9
1600
1800
2000
2200
2400
2600
2800
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Mixture Density (kg/m
3
)

0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Solid Volume Fraction
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Gas Volume Fraction

Figure 4: Pressure, mixture density, solid and gas
volume fraction

Two-dimensional detonation simulation
In this section a two-dimensional
simulation is presented. We simulate a
projectile high-velocity impact (1930
1
.

s m ) onto a tank filled with a reactive
liquid (Nitromethane). The transmitted
shock wave elevates the liquid temperature
and pressure. This pressure and
temperature elevation initiates the
chemical decomposition of the liquid
explosive. The shock transits to a
detonation. The increase in pressure will
put into motion the different walls of the
tank. A schematic view of the system is
presented in Figure 5.
The chemical decomposition is based on
the theory of the state transition. The
reactive material and the detonation
products obey a Mie-Grunesen EOS
(Cochran-Chan and JWL). The copper
obeys a modified stiffened gas EOS. The
ideal gas law is used for the surrounding
air.
Explosive
Projectile
(Cu)
Confining Material
(Cu)
Surrounding
Air
x
y
Symmetry axis

Figure 5: Presentation of the system simulated by
a two-dimensional calculation.
The total amount of equations is large:
twenty equations five equations per
phase- are solved at each grid point. But
the method is able to solve all flow
features (physical mixtures inside the
reaction zone) as well as material
interfaces under a unique mathematical
formulation and numerical scheme.
The mixture density is depicted on Figure
6. In Figure 6a the initial mixture density
is plotted and one can observe the
envelope as well as the projectile. In the
next view, the projectile has penetrated the
envelope, which is deforming. The initial
shock wave is now transmitted to the
explosive (6b) In Figure6c, a curved
detonation wave is observed which
propagates into the liquid. One can also see
in this figure, that a very low-pressure zone
is setting in the projectile. This is due to
the interaction of two rarefactions wave.
The first one comes from the
depressurisation behind the projectile
because of its motion. The second one is a
reflected rarefaction wave induced by the
interaction of the shock and the liquid.
In Figure 6d, the detonation wave has
reached the upper wall. It is noticeable that
the left wall above the projectile has a
motion toward the left: it is due to the
pressure difference with the atmosphere.
In the last view (6e), the entire explosive is
burnt. The upper wall has now a motion
that will turn to a complete destruction of
the tank.
10
6a
0.02 0.04 0.06
X
0.01
0.02
0.03
0.04
0.05
Y

6b
6c
6d
6e
Figure 6 : Mixture density evolution for the 2D
case
The new approach proposed in the present
paper is able to solve accurately very
different problems like physical multiphase
mixtures as well as material interfaces and
artificial mixtures, produced by numerical
diffusion zones of contact discontinuities.
It has been extended to the resolution of
turbulence problems in plasma physics
9

and phase transition fronts propagation in
superheated liquids. We believe that this
new approach may have important
applications in several problems of
physical, mathematical and industrial
importance.

REFERENCES
1
Saurel R. and Abgrall R., 1999, A multiphase
Godunov method for compressible multifluid
and multiphase flows, J. Comp. Phys. 150, pp
425-467.
2
Kapila A.K., Menikoff R., Bdzil J.B., Son S.F. and
Stewart D.S., 2000, Two-phase Modelling of
DDT in granular materials: Reduced Equations,
Phys. Fluids, 13, pp 3002-3024.
3
Saurel R. and Le Metayer O., 2001, A multiphase
model for compressible flows with interfaces,
shocks, detonation waves and cavitation, J.
Fluid Mech., Vol. 431, pp 239-271.
4
Abgrall R. and Saurel R., 2002, Discrete equations
for physical and numerical compressible
multiphase mixtures , J. Comp. Phys. , in
revision.
5
Baer M.R. and Nunziato J.W., 1986, A two-phase
mixture theory for the deflagration-to-
detonation transition in reactive granular
materials, Int. J. Mult. Flows, 12, pp 861-889.
6
Drew D. and Passman S.L., 1998, Theory of
Multicomponent Fluids, Springer-Verlag.
7
Mott D.R., Oran E.S. and Van Leer B., New quasi-
steady-state and partial equilibrium methods for
integrating chemically reacting systems, J.
Comp. Phys., 164, pp 407-428.
8
Fickett W. and Davis W.C., 1979, Detonation,
Univ. of California Press.
9
Saurel R., Gavrilyuk S. and Renaud F. (2002) A
multiphase model with internal degree of
freedom. Application to shock bubble
interaction. J. Fluid Mech., submitted
10
Chinnayya A., Daniel E. and Saurel R. (2002)
Computation of detonation waves in
heterogeneous energetic materials. J. Comp.
Phys., in revision