PII:SO895-7177(96)00144-6

Mathl. Comput. Modelling Vol. 24, No. 8, pp. 115-123, 1996 Copyright@1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0895-7177/Q6 $15.00 + 0.00

Linear Stability of a Detonation Wave with a Model Three-Step Chain-Branching Reaction

M. SHORT+ AND J. W. DOLD School of Mathematics, University of Bristol Bristol BS8 ITW, U.K. Abstract-The linearstabilityof a planar detonation wave with a three-step chain-branching reaction is studied by a normal mode approsch. The reaction model consists of a chain-initiation step and a chain-branching step governed by Arrhenius kinetics, with a chain-termination step which is independent of temperature. It mimics the essential reaction dynamics of a real chain-branching chemical system. The linear stability of the steady detonation wave to two-dimensional disturbances is studied with the chain-branching crossover temperature, i.e., the temperature at which chain-branching and chain-termination rates are equal, ss a bifurcation parameter. This parameter determines the ratio of the length of the chain-branching induction zone to the chain-termination zone within the steady detonation wave. The effect of linear transverse disturbances is considered for two values of in one the planar steady detonation wave is stable to the chain-branching crossover temperature: one-dimensional disturbances, while in the other it is unstable to such disturbances.
Keywords-Detonation, Stability, Chain-branching reactions.

P. INTRODUCTION
The problem of the one-dimensional linear instability of a detonation wave in a chemical mixture which was assumed to react via an idealised one-step Arrhenius reaction was first studied by Erpenbeck [1,2]. A Laplace transform technique was used to analyse the behaviour of small amplitude disturbances from the plane steady detonation wave. Later, Lee and Stewart [3] used a normal mode approach to address the linear instability problem. Their numerical shooting technique provided a straightforward way of calculating the stability spectra. Experimental studies on one-dimensional pulsating detonation instabilities such as [4] and on multidimensional transverse instabilities such as [5] demonstrate a distinct sensitivity to the type of the chemical mixture used. In particular, a majority of the mixtures used in these experiments involved the reaction of hydrogen and oxygen, which occurs through a chain-branching reaction mechanism. This involves a sequence of chain-initiation, chain-branching, and chain-termination stages. A small amount of reactant is converted into chain-carriers, which may be either free radicals or atoms, by means of the slow chain-initiation reactions. The chain-carriers are then rapidly multiplied through chain-branching reactions, while the rise in concentration of chainradicals is retarded by chain-termination steps which occur either through absorption at, the vessel walls or through three-body collisions in the interior. In the following, both the one- and twedimensional linear instabilities of a planar detonation wave with a model three-step chain-branching reaction are considered. The model consists of an Arrhenius type chain-initiation step with a large activation energy and an Arrhenius type
M. Short was supported in this work by an E.P.S.R.C. research grant and J. W. Dold was supported by an E.P.S.R.C. Advanced Fellowship. t Present address: Department of Theoretical and Applied Mechanics, University of Illinois, Urbana, IL 61801, U.S.A. Typeset 115 by A,#-T@

116

M. SHORTAND J. W. DOLD

chain-branching step with a moderate activation energy. These two stages are assumed to be thermally neutral. Finally, an exothermic temperature-independent chain-termination reaction converts chain-radicals into products. The reaction model thus retains the essential chemical dynamics of a real chain-branching reaction, unlike the idealised one-step Arrhenius kinetic model used previously [l-3]. The linear stability analysis is conducted for varying values of the chainbranching crossover temperature, i.e., the temperature at which the chain-branching and chaintermination rates are equal. This parameter has a major influence on determining the ratio of the length of the chain-branching induction zone to the chain-termination zone within the steady planar detonation.

2. MODEL
The thermodynamic behaviour of the fluid is modelled with the compressible reactive Euler equations written in nondimensional form as

~+pv.u=o,+p-lvp = g

0,

Dp- g+Pt =0,

where the variables p, u = (~1, uz), p, and e are the density, velocity, pressure, and specific internal energy, respectively. We assume a polytropic equation of state and an ideal thermal equation of state, such that T=?l e=&-Q (2) P where q represents the local chemical energy and T represents the temperature. The scales for the density, pressure, temperature, and velocity are the post-shock density, pressure, temperature, and sound speed, respectively, in a steady detonation wave, described in Section 3. The scalings for length and time are given below. Based on the early reaction model of Gray and Yang [6] in their study of the homogeneous explosion of a chain-branching reaction, we propose that the essential dynamics of a global chainbranching reaction can be represented by three main stages: chain-initiation, chain-branching, and chain-termination reactions with the rate constants Icr, kB, and kc, respectively. On this basis, a model three-step chain-branching reaction can be represented via the following three reaction stages: I: B: F + Y F + Y -+ 2Y Y + P (Initiation), (Chain-branching), (Chain-termination),
kB=eXp($(&-$)) kc = 1. (3)

c:

Kapila [7] has used a similar model to study the homogeneous explosion problem in the limit of high activation energies, except now the chain-initiation reaction is included. Based on the reaction dynamics of a number of chain-branching reactions (see [6] and, for example, [8]), the following assumptions are made regarding the model reaction mechanism. The chain-initiation and chain-branching rate constants kr and kB are taken to have Arrhenius temperature dependent form with inverse activation energies cl and eB, respectively. The chain-termination reaction is assumed to be first order, independent of temperature, and is modelled as having a fixed rate constant. The reaction model (3) is then the simplest realistic mechanism which reproduces the dynamics of a chain-branching reaction. Generalisations such as higher order chain-termination reactions, like those considered by Gray and Yang [6], could also be considered in order to further improve the model. The reference time t, is then scaled such that the chain-termination rate constant is unity, i.e., kc = 1, with the reference length set to t, times the sound speed

Detonation Wave Stability

117

behind initiation equations

the detonation

shock.

The remaining rates become then become

parameters

in (3) are TI and TB. These respectively, rate.

represent

chain-initiation

and chain-branching

crossover

temperatures,

in which the chainConsumption

and chain-branching for fuel and radical

ss fast as the chain-termination

Dt=
DY

Df

-rI

rg

(4

z=rI+TB-Tc,
where Tc

=y.

(5)

Finally

the chemical

energy

q is defined

[9] as

q=Q-Qf-(Q+D)y,

(6)

where Q > 0 represents the total chemical energy available in the unreacted mixture and D represents the amount of endothermic energy absorbed by the initiation and chain-branching reactions I and B in breaking down the reactant F into the energetic radical Y. In general, the chain-initiation post-shock detonation temperature reaction is energetically of unity, we assume TI >l in order to mimic this behaviour. and inhibited, so that given the steady

cI << 1 the chain-radicals can only grow initially

(7) at

On this basis,

the slow rate kr. The chain-branching reactions, on the other hand, have a lower activation energy and a rapid reaction rate once a sufficient concentration of chain-radicals is established. Therefore, we set TB < 1 and eB moderately small. (8) Finally, as noted in [7], a number of chain-branching reactions, e.g., that of a hydrogen-air reaction, liberate only a small amount of chemical heat energy during the initiation and branching stages. For the purposes of the present paper, it is assumed that the steps I and B in (3) are thermally neutral, i.e., D = 0. (9)

3. PLANAR

STEADY

DETONATION

STR?JCTURE
by

Behind a shock with Mach number M* in a steady one-dimensional detonation (indicated superscript *) pressure, velocity, and density satisfy the Rankine-Hugoniot relations u; =o, 8 where
Ms2= k~-W4*~+2 ,=vK2+l W2Wr 1)

p*=a+(l-a)dp,

u;=$$+M;,

,J - W * %

(10)

27M*2 The chemical energy

- (y - 1)

(Y + I)

and

b=

(l-@(y+l)

(11)

release q satisfies q* =Q-Qf*-Qy.

(12)

118

M. SHORT AND J.

W.DOLD

The spatial structure of the detonation is determined by numerical quadrature equations for fuel and radical concentration

of the first-order

(13) Immediately behind the shock, the following conditions apply:

p* = p* = T = 1,

u; = M,,

u; = 0,

= 1,

y* = 0.

(14)

The detonation

overdrive d is defined by

(15) where M& is the Chapman-Jouguet detonation velocity, determined by the velocity being exactly sonic uf = dm at the end of the wave where q* = Q. Figure 1 shows the steady detonation profiles for (a) fuel f and radical y and (b) temperature T, respectively, with distance behind the detonation shock for varying values of the chain-branching crossover temperature TB = 0.8, 0.85, 0.9, and 0.95 at fixed Q = 3, EI = l/20, &B= l/8, TI = 3, 7 = 1.2 and d = 1.2. For the lower values of TB, the steady detonation structure consists of a short chain-branching induction zone and much longer chain-termination region. As the chainbranching crossover temperature increases, the chain-branching induction zone becomes longer, with a lower peak concentration of chain-radicals being established. For the higher TB c 1, the detonation structure is dominated by a chain-branching induction zone being much greater in length than the chain-termination region, with only a very moderate rise in peak concentration of chain-radicals. For TB > 1, chain-branching is inhibited [9] and the reaction is, for all practical purposes, quenched. Only the extremely slow initiation reaction is able to bring about any changes at all. We will not consider TB 2 1 in this article.

4. LINEAR

STABILITY

ANALYSIS

We now present a normal mode linear stability analysis of the steady detonation structure (lo)-(13) with the model chain-branching reaction scheme (4)-(5) for a range of values of the chain-branching crossover temperature TB. This proceeds by defining a shock-attached coordinate system z=X+D,St-+(y,t), (16) where 0: is the detonation Mach number relative to the shocked gases and $(y,t) represents the perturbation to the steady shock location due to a small disturbance. Perturbations to the steady detonative wave structure are then sought in the form
a =

z*(2) + z(z)eateiky,

$=e

at e iky ,

(17)
(18)

where z = (V
v = l/p
211, 212,P,

f, YjT,

is the specific volume, Re(cy) is the disturbance growth rate, Im(cy) the disturbance frequency, and k the disturbance wavenumber. For brevity, we will omit the superscript * in what follows, the quantities without the superscript referring to the underlying steady-wave solution. Substituting expansions (17) into equations (l)-(6) results in a system of first-order linear differential equations for the vector of complex perturbation eigenfunctions z(z), namely cyz + AZ; + ikBz + Cz - & - ikBb = 0, b=z=, (19)

Detonation Wave Stability

119

0.8

0.6 G 0.4

5 x (a)

10

15

10 x (b)

15

Figure 1. Steady detonation profiles showing (a) f (solid lines) and y (dashed lines) and (b) T. The symbols mark corresponding profiles for TB = 0.80 (Cl), = 0.85 TB (O), TB = 0.90 (O), and TB = 0.95 (A).

where

0 A= 0 0 0 0

Ul

; 0
111

0 0 0
211

0 0 0 0
Ul 215

0 112

u2

0 0 0 0 0
0

-?I

Ul -yp 0 0 0 0 0

0 0
-I&

r
cc
0

I
0

B=

0 0 0 0

7.4

x Y
212

0 0 0
u2

0 0 0
IL.2 0

1
(20)

-yp 0 0
-

0 0

I
(21)

PZ

Ulx
u2z

0
0

0
0 YUls

0
0
0

0
0. --i; -A/+1_

-GJ
Pv

Px

fz

Pp
-Pp

Pf -Pf

_ -Pv

El2 0

120

M.SHORTAND

J. W. DOLD

The new quantities in C are

and (23) The perturbation shock conditions are determined from the Rankine-Hugoniot conditions ss

2 (1+ M2) a
v= (r+
$iM2&' 4 = (r+1)M2 7 4 =(D, -Mg)ik,

(24) f = 0, y = 0.

p = -;;yl;,

Finally, we apply a rear-boundary condition at z = 00 such that there should be no acoustic waves propagating upstream from infinity, leading to the acoustic radiation condition

(25) where Urb, Vb and cb are the unperturbed s-velocity, specific volume, and sound speed, respectively, in the burnt gases. For a given set of initial thermodynamic and chemical parameters characterising the steady detonation structure, the complex eigenvalues (Y and eigenfunctions Z(Z) are determined by a numerical shooting technique based on that of Lee and Stewart [3]. This involves taking an initial guess for (Y, integrating equations (19) from the perturbed shock conditions (24) into the burnt zone and iterating on the initial guess for a until the acoustic radiation condition (25) is satisfied in the burnt zone at a point where the reaction rate is exponentially small.

5. LINEAR STABILITY

RESULTS

Figure 2 shows the region of instability to one-dimensional disturbances of the steady detonation structure (lo)-(13) and the migration of nine unstable roots when the chain-branching crossover temperature varies in the range TB E [0.8,1.0] for Q = 3, er = l/20, Ed = l/8, TI = 3, 7 = 1.2, d = 1.2, and wavenumber k = 0. Figure 2a shows the behaviour of the growth rate Re(o) against TB while Figure 2b shows the corresponding frequency Im(o) of the unstable roots. For TB < 0.803, the steady planar detonation is found to be stable to one-dimensional disturbances. In this range, Figure 1 shows that the steady detonation structure is dominated by the temperature-independent chain-termination reaction, with only a short chain-branching induction zone. There are similarities here to the stability properties of a detonation wave with an ideal&d one-step chemical reaction model having zero activation energy as studied by Erpenbeck [lo]. The general conclusion, which is reinforced here, is that temperature-independent reactions render the detonation more stable than temperature-dependent chemical kinetics. At TB = 0.803, a neutrally stable mode (Y = or is present, having Re(crr) = 0. From Figure 2b, it is seen that the neutrally stable mode has a comparatively low frequency of oscillation with Im(crr) = 0.219. As TB increases through TB = 0.803, a Hopf bifurcation occurs rendering the detonation unstable to the low frequency disturbance. At TB = 0.821, a second unstable mode a2 appears having a much higher frequency than the mode ~1 with Re(as) = 0, Im(crz) = 1.11. As TB increases further, additional unstable roots appear, each having a higher frequency than the previous root. At TB = 0.85, there is the low frequency instability with Re(oi) = 4.62 x 10d2 and Im(crr) = 0.12 and two high frequency instabilities, op with Re(crz) = 4.01 x low2 and Im(crs) = 0.77, and (~3 with Re(as) = 9.97 x 10s3 and Im(cYs) = 1.49. Despite the fact that the

Detonation

Wave Stability

121

0.80

0.85

0.90 TB (a)

0.95

1.00

2.03

1.5 -

3 g

l.O-

0.5 -

0.0 0.80

, .
0.85

. . ,
0.90 Ts

. ,

,
0.95

, . .

,
1.00

(b)
Figure 2. The variation in (a) the growth rate Re(a) and (b) the frequency Im(a) of one-dimensional instabilities as TB is varied. Nine roots are plotted. Purely red roots are shown ss dashed lines.

detonation becomes more unstable the fastest growing mode at any into two purely real roots, shown a growth rate which continues to decays with TB.

as TB is increased, it is the low frequency mode than remains value of TB. At TB = 0.913, the low frequency mode splits by the dashed line in Figure 2a. One branch of the root has dominate with TB, while the other has a growth rate which

For the representative parameter values chosen above, the steady planar detonation is unstable to one-dimensional perturbations for su5ciently large chain-branching crossover temperatures, i.e., TB > 0.803. For smaller TB, the model chain-branching reaction scheme produces a detonation structure dominated by the chain-termination region. As the chain-termination reaction rate occurs independently of changes in temperature, there is no strong coupling between gasdynamic perturbations and the chemistry in this region. It is then seen that the detonation is stable to one-dimensional perturbations. As TB increases, the chain-branching induction zone, in which there is a stronger coupling between gasdynamic perturbations and chemistry, becomes longer.

122

M. SHOP AND J. W. DOLD

The detonation is then unstable initially to a low-frequency one-dimensionalperturbation, with additional high frequency i~tabiliti~ appearing as TB increasesfurther.

Figure 3. The variation in &(a) of transverse instabilities as the wavenumber k is varied for TB = 0.79. Two roots are shown.

k
Figure 4. The variation in RR(a) of transverse instabiliti- ss the wavenumber k is varied for ?& = 0.85. Three roots are shown.

Althou~ the stability of detonation waves to on~dimensional disturbances is relevant to the one-dimensional pulsating detonation waves seen in the experiments of Alpert and Toong [4], detonation waves travellingin shock tubes exhibit a complex transversewave structure [5] and not a one-dimensional pulsating type of instability. An important question which then arises is how the growth rates of transversedisturbances compare to those of one-dimensional perturbations. Figure 3 shows the change in the growth rate Re(cu) against transverse wavenumber k for the two modes cyi and cr2 in a detonation with TB = 0.79 and the other parametersfixed as above. For TB = 0.79, the planar detonation wave is stable to one-dimensional perturbations with R.e(ai) = -2.57 x 10e2 and Im(cur) = 0.25 and Re(a2) = -9.78 x 10m2and Im(crs) = 1.67 at ic:= 0. As the wavenumber k is increased from k = 0, both modes become unstable with or becoming unstable at k = 0.093 and 02 becoming unstable at k = 1.035. Cairns growth rates are IXe(crl) = 0.160 at k = 0.71 and Refc~) = 0.160 at k = 2.542. Thus, the second

DetonationWave Stability

123

mode (~2 possesses a greater maximum growth rate at a higher wavenumber, k = 2.542. Both modes become stable for sufficiently large wavenumbers, i.e., k = 1.56 for al and k = 3.78 for (~2. Figure 4 shows the variation of F&(a) against wavenumber k for the three modes ~1, 02, and ~3 with TB = 0.85. Here, the planar steady detonation is unstable to each of these modes for k = 0, possessing growth rates as given above. However, as k increases, the growth rates of each of these modes increases, demonstrating that the transverse instabilities are more unstable than the one-dimensional instabilities. The maximum growth rates of each of these modes are R,e(cwl) = 0.120 at k = 0.355, RR(W) = 0.137 at k = 1.13, and Re(a3) = 0.113 at k = 2.06. The maximum growth rates of the transverse disturbance are significantly larger than the corresponding growth rates of the one-dimensional instabilities. Again the second mode (~2 possesses the highest overall growth rate and all three modes are observed to stabilize at sufficiently high wavenumbers. Thus, in the former case where TB = 0.79, although the detonation is stable to one-dimensional disturbances, it is unstable to transverse disturbances. In the latter cases where TB = 0.85, the transverse disturbances would be predicted to grow much more rapidly than the one-dimensional disturbances ultimately giving rise to the complex transverse wave structure observed in experiments.

6. SUMMARY
The linear stability of a planar steady detonation wave has been studied where the chemical kinetics of the reactive material are modelled by a three-step chain-branching reaction. For sufficiently small chain-branching crossover temperatures, TB, the steady detonation structure is dominated by the temperature-independent chain-termination zone and the detonation is found to be stable to one-dimensional perturbations. As the length of the chain-branching induction zone is increased by increasing TB, the detonation becomes unstable to one-dimensional perturba,tions with the fastest growing instability having a low frequency of oscillation. For the case TB = 0.79, the steady detonation wave is stable to one-dimensional disturbances, but is found to be unstable to transverse disturbances. For the case TB = 0.85, the steady detonation wave is unstable to one-dimensional disturbances, but the maximum growth rate of transverse instabilities is found to dominate that of the one-dimensional instabilities.

REFERENCES
detonations,Phys. Fluids 5, 604-614 (1962). 1. J.J. Erpenbeck,Stabilityof steady-state equilibrium one-reactiondetonations,Phys. Fluids 7, 684-696 (1964). 2. J.J. Erpenbeck,Stabilityof idealized of instabilityof planedetonation, 3. H.I. Lee and D.S. Stewart, Calculation linearinstability:One-dimensional
J.F.M. 216, 103-132 (1990). 4. R.L. Alpert and T.Y. Toong, Periodicity in exothermic hypersonic flows about blunt projectiles, Acta A s&on. 17, 538-560 (1972). 5. R.A. Strehlow, Multi-dimensional detonation wave structure, Astro. Acta 15, 345-357 (1970). 6. B.F. Gray and C.H. Yang, On the unification of the thermal and chain theories of explosion limits, J. Phys. Chem. 68, 2747-2750 (1965). 7. A.K. Kapila, Homogeneous branched-chain reactions: Initiation to completion, J. Eng. Maths. 12, 2:!1-235 (1978). 8. F.A. Williams, Combustion Theory, Addison-Wesley, Reading, MA, (1985). 9. J.W. Dold and A.K. Kapila, Comparison between shock initiation of detonation using thermally-sensitive and chain-branching chemical models, Comb. Flame. 85, 185-194 (1991). Zero activation energy, Phys. Fluids 8, 10. J.J. Erpenbeck, Stability of idealized on&reaction detonations: 1192-1193 (1964).