Vous êtes sur la page 1sur 74

PERGAMON Progress in Energy and Combustion Science 27 (2001) 1–74

www.elsevier.com/locate/pecs

Fundamental aspects of the heterogeneous flame in the


self-propagating high-temperature synthesis (SHS) process
A. Makino*
Department of Mechanical Engineering, Faculty of Engineering, Shizuoka University, Hamamatsu 432-8561, Japan
Received 30 April 1999; received in revised form 7 February 2000; accepted 7 February 2000

Abstract
Recent progress on understanding fundamental mechanisms governing the Self-propagating High-temperature Synthesis
(SHS) process, which is characterized by the flame propagation through a matrix of compacted reactive particles and is
recognized to hold the practical significance in producing novel solid materials, is reviewed. Here the focus is not only on
the theoretical description of the heterogeneous nature in the combustion wave, which has not been captured by the conven-
tional premixed-flame theory for a homogeneous medium, but also on the extensive comparisons between the predicted and
experimental results in the literature. Topics included are the statistical counting procedure used for deriving governing
equations of the heterogeneous theory, flame propagation in the adiabatic condition, flame propagation and extinction under
heat loss conditions, effects of bimodal particle dispersion on the combustion behavior, those of external heating by electric
current, the transition boundary from steady to pulsating combustion, and the initiation of the combustion wave by use of the
external heating source. The importance of heterogeneity in the combustion wave, that is the particle size of the nonmetal or the
higher melting-point metal, has been emphasized for fundamental understanding of such combustion behavior as flame
propagation, extinction, and initiation. Potentially promising research topics are also suggested. 䉷 2000 Elsevier Science
Ltd. All rights reserved.
Keywords: Combustion synthesis; Self-propagating high-temperature synthesis; Flame propagation; Burning velocity; Range of flammability;
Extinction; Ignition; Ignition delay time

Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. The SHS process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1. General characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2. Scope of the present survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3. Phenomenological description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3. Formulation of the heterogeneous theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1. Surface regression rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2. Statistical counting procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.1. Change of distribution function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.2. Conservation equations in general form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.2.1. Overall continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.2.2. Species conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.2.3. Momentum conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.2.4. Energy conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

* Tel.: ⫹ 81-53-478-1050.
E-mail address: tmaki@eng.shizuoka.ac.jp (A. Makino).
0360-1285/00/$ - see front matter 䉷 2000 Elsevier Science Ltd. All rights reserved.
PII: S0360-128 5(00)00004-6
2 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

3.3. Heterogeneous theory for the SHS process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13


3.3.1. Size distribution function in quasi-one-dimensional form . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.2. Conservation equations for steady, one-dimensional form . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.2.1. Overall continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.2.2. Species conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3.2.3. Momentum conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3.2.4. Energy conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3.3. Further simplification of the governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4. Flame propagation in the adiabatic condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.1. Dominant parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.2. Burning velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.3. Range of flammability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.4. Theoretical and experimental results for Ti–C system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.5. Applicability of the heterogeneous theory for other systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.5.1. Available experimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.5.2. Effects of mixture ratio, m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.5.3. Effects of degree of dilution, k . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.5.4. Effects of initial temperature, T0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.5.5. Effects of particle radius, R0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.6. Approximate expression for the burning velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.6.1. Derivation of the approximate expression for m ⱕ 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.6.2. General behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.6.3. Comparisons with numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.6.4. Approximate expression for m ⱖ 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.7. Some remarks on the flame propagation in the adiabatic condition . . . . . . . . . . . . . . . . . . . . . . 30
5. Flame propagation in the nonadiabatic condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.1. Governing equations and boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2. Characteristics in the nonadiabatic condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.3. Turning-point determined by the thermal theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.4. Experimental comparisons for Ti–C system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.4.1. Range of flammabilty and extinction limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.4.2. Burning velocity under heat loss condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.5. Experimental comparisons for other systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.5.1. Boride synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.5.2. Synthesis of intermetallic compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.6. Approximate expression of the heat loss parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.6.1. Asymptotic expansion analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.6.2. Correction term for the heat loss parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.6.3. General behavior of the heat loss parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.6.4. Applicability of the analytical expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.7. Some remarks on the flame propagation under heat loss condition . . . . . . . . . . . . . . . . . . . . . . . 41
6. Other aspects of the nonadiabatic flame propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.1. Bimodal particle dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.1.1. Equivalent particle radius . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.1.2. Experimental comparisons for spherical carbon particles . . . . . . . . . . . . . . . . . . . . . . . . 43
6.1.3. Experimental comparisons for diamond particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.2. Representative length of the cross-sectional area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.3. Correspondence between the heterogeneous theory and the homogeneous theory . . . . . . . . . . . . 47
6.4. Field activated SHS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.4.1. Heat-input parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.4.2. Temperature profiles outside the combustion wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.4.3. Effective range of electric field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.4.4. Experimental comparisons for Si–C system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.5. Some remarks on the several, other, important factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7. Boundary between steady and pulsating combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7.1. Criterion for the appearance of cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 3

7.2. Linear stability analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52


7.2.1. Basic solution for planar propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.2.2. Asymptotic solution and linear stability boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.2.3. Experimental comparisons for the boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.2.4. Boundary at the adiabatic condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.3. Transverse effects on the range for steady combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.4. Some remarks on the boundary of steady combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
8. Initiation of the combustion wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
8.1. Flame initiation induced by igniter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.1.1. Inert stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.1.2. Transition stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
8.1.3. Limiting solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
8.1.4. Experimental comparisons for the ignition delay time . . . . . . . . . . . . . . . . . . . . . . . . . . 61
8.1.5. Expression for the ignition energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
8.1.6. Results for the ignition energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
8.2. Flame initiation by use of radiative heat flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
8.2.1. Model definition and inert stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
8.2.2. Transition stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
8.2.3. Results and experimental comparisons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
8.2.3.1. Situation without heat loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
8.2.3.2. Situation with heat loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
8.3. Some remarks on the flame initiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
9. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
9.1. Summary of the present survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
9.2. Area for further research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

1. Introduction combustion synthesis in exploring and/or explaining its


combustion phenomena. This may be attributed to the fact
There has been increasing interest in the combustion that the main concern of the combustion has been on produ-
synthesis of new materials which would play important cing heat, for the sake of heating and/or heat-supply to heat
roles in future technologies in various fields. Many of the engines. As for the combustion synthesis, its main concern
materials available today, such as ceramics (inorganic is not the heat, but the materials produced as combustion
compounds) and intermetallic compounds, fall into a cat- products.
egory of chemical compounds, so that we can expect that However, it should be recognized that combustion synth-
combustion would make a great contribution in producing esis is one of the research subjects of combustion science
new materials and/or enhancing productivity of these ma- and technology because of the existence of combustion
terials, because of the existence of strong chemical reactions phenomena, although objects to be produced by combustion
accompanied by the combustion. The combustion synthesis, might be very different from each other. Furthermore, since
defined as “materials synthesis” by combustion, is therefore it is a new research field in combustion, combustion syn-
an interdisciplinary research subject in the fields of combus- thesis should be studied systematically, from the viewpoint
tion and material science, for producing ceramics and/or of combustion science and technology. This trend appeared
intermetallic compounds in lumps, thin films of oxides in the late 1980s and has been reflected by the establishment
and/or diamond, ultra-fine powder of inorganic compounds, of a session on Materials Synthesis in the International
etc. as combustion products. Symposium on Combustion, since 1990.
Research for the combustion synthesis has been very Generally speaking, methods of combustion synthesis can
active in the field of material science. On the other hand, be divided into two main classes: one is burner flames which
research from the viewpoint of combustion has not been so produce thin films of oxides and/or diamond, and ultra-fine
active, in spite of the important fact that synthesis of ma- powder of inorganic compounds; the other method is to
terials by combustion can only be accomplished by virtue of produce ceramics (inorganic compounds) and/or intermetal-
combustion. Then, it remains unsolved whether the knowl- lic compounds in lumps, by use of a kind of powder metal-
edge of combustion hitherto obtained is applicable to the lurgy called the Self-Propagating High-Temperature
4 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Nomenclature
A Reduced surface Damköhler number ‰ˆ Da exp…⫺Ta =T†Š
a Parameter …a3 ˆ f ⫺ 1† in Section 4; length of the rectangular cross section in Section 6; thermometric
conductivity in Section 8.1; radius of heating area by radiative heat flux in Section 8.2
a0 Leading term of the normalized mass burning rate …ˆ m=ma † in Sections 5 and 7
B Frequency factor for the surface reaction in Section 3; constant in Section 7
B~ Frequency factor for the “reaction” term in the conventional homogeneous theory
b Parameter …b3 ˆ f ⫹ 2g ⫹ 1† in Section 4; width of the rectangular cross section in Section 6
C Constant; maximum temperature in Section 7
c Specific heat
D Mass diffusivity in Sections 3–7; constant in Section 8.
Da Surface Damköhler number; …ˆ BR=D†
E Electric field in Section 6.4; energy for establishing combustion wave in Section 8
F Thermal runaway criterion in Section 8.2
f Distribution function
fst Stoichiometric mass ratio
G Size distribution function
H Heat-input parameter in Section 6.4
h Enthalpy
hrad Radiative heat-transfer coefficient
I Integrated quantity
Jn Bessel function of the first kind of order n in Section 8.2.2
K Constant in Section 7; derivative of temperature in nondimensional form in Section 8.1.2
k Inverse of the arithmetic mean of melting point and adiabatic combustion temperature in nondimensional form
k~ Over-all transverse wave-number in Section 7.3; nondimensional parameter related to normalized mass burning
rate in Section 8.1
L Volumetric heat loss
Le Lewis number
M Melting parameter
m Mass burning rate for deflagration
n Number of N particles
p Static pressure in Section 3; perimeter of rectangular cross section in Section 6.3
P Size ratio defined in Eq. (143) in Section 6.1.1
Q Heat flux
q0 Heat of combustion per unit mass of N species
R Average radius of N particles
R_ Rate of change of particle size R …ˆ dR=dt†
r Radius of compacted specimen
S Cross-sectional area of compacted specimen
s Variable which represents a change of particle size in Section 3; stretched variable of time in reactive/diffusive
region in Section 8.2
s~ Stretched variable of time in transient/diffusive region in Section 8.2
T Temperature
T0 Standard temperature
Ta Activation temperature for the reaction
Td Activation temperature for the condensed phase mass diffusivity
t Stretched variable of temperature in Section 5.6.1; time in Sections 7 and 8.
U Diffusion velocity of species
u Heterogeneous burning velocity
v Velocity of a particle
W Molecular weight
X Normalized heat loss, defined as X ˆ exp…⫺2bC=a 02 † ˆ exp…⫺C1 † in Sections 5 and 7; stretched variable of
location in reactive/diffusive region in Section 8.2
x Physical spatial coordinate
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 5

Y Mass fraction
y Stretched variable of mass ratio of fluid in Section 5.6.1
Z Mass ratio of fluid

Vectors and tensors


E Unit tensor
F Force vector
f Body force vector
P Total pressure tensor
q Heat flux vector
T Viscous stress tensor
U Diffusion velocity vector
u Veclocity vector for fluid
v Veclocity vector for a particle

Greek symbols
a Normalized temperature rise
b Spalding transfer number in Section 3; Zeldovich number in Sections 4–8
G Mass ratio of small particles to large particles in Section 6
g Mass ratio of fluid to solid
D Damköhler number for flame initiation in Section 8
d( ) Delta function
d Thickness of combustion wave
e Emissivity
z Normalized mass fraction
h Variable which represents a change of particle size in Section 3; inner variable of length in Section 5.6.1
u Normalized temperature
k Degree of dilution defined as the initial mass fraction of the diluent
L0 Mass burning rate eigenvalue
L Heat penetration coefficient in Section 8
l Thermal conductivity
l u, l d Exponents which represent temperature distribution in Section 6.4 and Section 7
m Mixture ratio defined as the initial molar ratio of nonmetal, N, to metal, M, divided by the corresponding
stoichiometric molar ratio
n Stoichiometric coefficient
j Normalized mass ratio of fluid
1⫺j Normalized mass ratio of solid
r Density
s Nondimensional coordinate
sc Electric conductivity in Section 6.4
s ST Stefan–Boltzmann constant
s Stretched coordinate in the downstream in Section 5.6.1.
t Nondimensional time
t xx xx component of the viscous stress tensor in Section 3.
f Inverse of the mixture ratio m in Section 4.6 and 5.6; heat loss in Section 8.2.
w Function which represents dependence of the “reaction” term on reactants
w~ Nondimensional parameter related to heat loss in Section 8.1
x Surface regression rate
C Heat-loss parameter
c Variable for particle size distribution in Section 3; stretched variable for temperature in Section 8
V Constant
v Mass rate of production by homogeneous chemical reaction in Section 3; stretched variable of location in
transient/diffusive region in Section 8.2
6 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Subscripts
a Adiabatic condition
c Initiation of the combustion wave
cr State of extinction
d Downstream
E Equivalent
e Edge of boundary
f Fluid
I Inert heating
ig Igniter
in Material flow to the fluid from particles
L Large particles
M Metal or lower melting-point metal
m Melting point
max Maximum
N Non metal or higher melting-point metal
P Combustion product
p Perturbed term
r Representative value in Section 6.4; radial direction in Section 8.2
rel Relative
S Small particles
s Solid or surface
sb Neutral stability boundary
sp Specimen, that is, medium to be ignited
TMD Theoretical maximum density
t Differentiation with respect to time
u Upstream
z Differentiation with respect to location in the axial direction
j With respect to mass fraction of fluid Z
u With respect to temperature
t Ignition delay
0 Initial or unburned state
∞ Final or burned state in the adiabatic condition
Superscripts
t Nondimensional
— Average over all velocities in Section 3.

Synthesis (SHS) process. In the present review, emphasis is compacted reactive particles (cf. Figs. 1 and 2). From the view-
put on the latter, because of its various advantages, to be point of combustion science and/or technology, as shown in
examined. Fig. 2 photographically, the SHS process falls into a category
of the flame propagation, because a reaction initiated at one
end of a compacted medium self-propagates through an
2. The SHS process unburned medium, in the form of a combustion wave.
The velocity of the combustion wave, depending on
2.1. General characteristics systems, varies from 1 to 250 mm/s. The high temperature
needed for synthesis (perhaps more than 2000 K for inor-
Among various types of combustion synthesis, wide ganic compounds and more than 1000 K for intermetallic
attention has been given to the SHS process, proposed by compounds) can be supplied by the self-sustained exother-
Merzhanov and co-workers [1–3] in Russia in the late mic chemical reactions. The potential advantages of this
1960s, because not only bulk processing of materials, but process are as follows [4–9]: rapid synthesis; self-heating;
also formation of some elemental combinations not energy savings; self-purification due to enhanced impurity
previously synthesized, can be performed [4–9] by initiat- outgassing; near-net-shape fabrication, etc. In addition,
ing and passing a combustion wave through a matrix of more than 500 kinds of materials, including carbides,
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 7

borides, silicides, nitrides, sulfides, hydrides, intermetallics,


and complex composites, are reported [10] to be synthesized
by applying the SHS process not only for solid–solid
systems but also for solid–gas and/or solid–liquid systems.
Table 1 shows only a few of examples of inorganic
compounds synthesized by the SHS process. Compounds
synthesized are being considered for use as electronic mate-
rials, materials resistant to wear, corrosion, and heat.
Furthermore, it can even be applied to the synthesis of
shape-memory alloys, hydrogen-storage alloys, and high-
temperature superconductors [7–9]. Thus, it is now well
recognized that the SHS process can be of practical signifi-
cance in producing novel solid materials.
However, if restricted to synthesizing homogeneous
materials in composition and/or texture, as is the case
for conventional materials, the SHS process would not
be attractive to researchers in the field of combustion. It
was in the late 1980s when various attempts began to
discover possibilities for improving the specific nature
of the conventional, homogeneous materials for the
purpose of producing advanced materials with useful,
multifunctional characteristics generated by heterogen-
eity in composition and/or texture. Among these
advanced materials, Functionally Graded Materials
(FGMs), which are composed of different material
components such as ceramics and metals with continu-
ous profiles in composition, structure, texture, material
Fig. 1. Schematic drawing of combustion synthesis by the SHS strength, and thermophysical properties, have attracted
process [5]. special interest as advanced heat-shielding/structural
materials in future space applications. At that time,

Fig. 2. A sequence of the advance of a combustion wave in the SHS process for Ti–C system; stoichiometric mixture without dilution; diameter
is 18 mm and length is about 50 mm; the initial temperature T0 is 300 K and radius R0 of the carbon particle is 5 mm; the maximum temperature
is about 3200 K and the burning velocity u is 17 mm/s. (a) Just after the ignition; (b) 1 s; (c) 2 s; and (d) 3 s.
8 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

production of FGMs by use of the SHS process

TiB2 –Al2O3, etc.


(especially for producing ceramics in FGMs with gradual

FeSi2, CoSi2
Fe4N, Fe8N

NiS, Ni3S2,
profiles in composition) was considered because of its rapid-

CoS, CoS2
FeB, NiB
ity in production. Although it has proven effective, it was
VIII
also recognized that in order to control the manufacturing
process, dependence of flame propagation speed on various
dominant parameters is indispensable. In addition, range of

MoB–Al2O3,
flammability with respect to various parameters is urgently
MnB, MnB2

required because the combustion wave cannot be main-


Mn7C3

tained outside the range of flammability. Of course, depen-

MnSi

MnS
VII

dence on various parameters is also required in preventing


flame extinction. From an academic point of view, these
areas are within the parameters combustion researches

MoSe2, WSe2
CrSi2, MoSi2,
Mo3Si, WSi2
Cr3C2, MoC,

MoS2, WS2, have pursued. That is, researchers in the field of combustion
Mo2B5, WB

WC–Al2O3,
CrB2, MoB,

Mo2C, WC

are strongly encouraged to clarify these problems through


knowledge of combustion.
CrN
VI

In addition to its practical utility, the SHS process


commands fundamental interest because of its interesting
and diverse phenomena, such as steady planar propagation,
NbSe2, TaSe2
NbSi2, TaSi2

B4C–Al2O3,

pulsation, spinning, and repeated combustion, which can be


VSi2, V3Si,
NbB2, TaB

TaN, Ta2N
VC, Nb2C,

VN, NbN,
NbC, TaC
VB, NbB,

observed during the flame propagation in the solid compacted


TaS2,

medium [1–3,11,12]. In pulsating combustion which is also


V

called self-oscillation combustion, the combustion wave


travels in a planar, but pulsating manner, which frequently
results in materials with a laminated structure (cf. Fig. 3). In
SiC–Al2O3, SiC–MgO,

spinning combustion the combustion wave is nonplanar, and


one or more hot spots are observed to move along a spiral path
over the surface of the sample specimen, inside of which reac-
Si3N4, TiN,

TiSi, TiSi2,
ZrB2, HfB2

TiH2, ZrH2
ZrSi, ZrSi2
TiS2, ZrS2,
PbS, TiSe2

tants experience only heating, without reaction. In repeated


TiB, TiB2,

ZrN, HfN
ZrC, HfC
SiC, TiC,

combustion, after the passage of the combustion wave, another


combustion wave is initiated and propagates through the
IV

burned medium, yielding the final combustion products from


the intermediates. Fundamental understanding of flame propa-
gation, extinction, and/or initiation in the SHS process is not
CeSi2, DySi2
YSi2, LaSi2,
AlB2, LaB6,

PrH2, NdH2
B4C, Al4C3

TiC–ZrO2,
BN, AlN,

only of interest for its own sake, but also of great use for
potential improvement in the manufacturing process.
CeB2

NdN

CeS

Various experimental works have been conducted, and


Examples of inorganic compounds synthesized by the SHS process

III

many of the accomplishment have been summarized in


several good review papers [5–9]. In order to elucidate
TiC–SiC, TiC–Al2O3,

effects of various system parameters on the flame behavior,


theoretical studies have also been conducted keeping up
with experiments. The basic theory, originally derived by
MgB2, MgB4,

Mg2Si, CaSi2
MgB6, CaB6

Mg3N2, SrN

ZnSe, CdTe

Novozhilov [13] for the combustion of solid propellant, and


ZnS, CdS,

extended by Merzhanov [14] for the SHS process, has been


Group in periodic table

CaC2

BaN

applied to examining stability [15] and unsteady behaviors,


II

such as pulsating [16] and spinning [17] combustion. Effects


of competing reactions [18], particle size distribution [19],
TiC–TiB2,

heat loss [20,21], Arrhenius mass diffusion [22], etc. have


CuInSe2

also been studied. Subsequent accomplishments by theor-


Cu2Si

Cu2S,

CsH2
CuN

etical works are well summarized in recent review papers


I

[23,24]. Note here, however, that theoretical description in


those works has primarily been based on the premise that
Chalcogenide

synthesis is accomplished through the passage of a premixed


Compound

flame in a homogeneous medium in which reactants are well


Complex

Hydride
Carbide
Table 1

Silicide
Nitride
Boride

mixed at the molecular level before arrival of the flame front


of the combustion wave.
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 9

Fig. 3. Laminated structure in the combustion products; Ti–C system for stoichiometric mixture without dilution; the initial temperature T0 is
300 K. (a) Radius of the carbon particle R0 is 0.5 mm and the relative density r rel is 0.56. (b) R0 ˆ 0:5 mm and rrelˆ 0:63: (c) R0 ˆ 0:5 mm and
rrel ˆ 0:72: (d) R0 ˆ 2:5 mm and rrel ˆ 0:47: (e) R0 ˆ 2:5 mm and rrel ˆ 0:64:

2.2. Scope of the present survey medium of compacted reactive particles. Again we are
required to construct an appropriate model which can prop-
Although much work has been conducted both experi- erly represent the combustion situation in the SHS process,
mentally and theoretically, a general lack of consistency by taking its heterogeneity into account in an essential
has been noted [25] in the combustion data and no firm manner.
understanding of the combustion mechanism has been The present paper, restricted to theoretical accomplish-
achieved. This may be attributed to the fact that compari- ment, aims to complement the extensive reviews by Fran-
sons between experimental and theoretical results have been khouser et al. [5], Munir and Anselmi-Tamburini [7],
rare, regardless of their importance for comprehensive Merzhanov [8] and Varma and Lebrat [9] regarding the
understanding of various combustion characteristics. experimental accomplishment, and those by Merzhanov
Through agreement between theoretical and experimental [4], Merzhanov and Khaikin [23] and Margolis [24] regard-
results over wide range of parametric quantities for various ing the theoretical accomplishment. Indeed, because of the
combinations of metal and nonmetal (or another metal of a broad coverage of some of these reviews, a narrow range of
higher melting point) systems, appropriateness of theories topics will be discussed here, focused on the heterogeneity
should therefore be examined. in the SHS process, to which little attention has been paid,
When such comparisons are conducted with conventional although it is indispensable for realistic descriptions of
models in the homogeneous premixed-flame theory, there combustion behavior. The present survey also emphasizes
exists an inherent deficiency that cannot account for size understanding of the fundamental mechanisms govern-
effect of particles, while experiments have firmly estab- ing the flame propagation, extinction, and/or initiation in
lished that the flame propagation speed increases with the SHS process, by being restricted to the solid–solid
decreasing particle size in a sensitive manner [26–28]. systems. It is also suggested that the reader refers to
This inherent deficiency comes from insufficient treatment, Ref. [31] for topics about the SHS process for the
based on the homogeneous premixed-flame theory. Even if gas–solid systems, which are not covered in the present
fine powder is used, for example 0.1 mm in particle review.
diameter, each particle contains more than 10 8 atoms. This In the next section, a phenomenological description of the
situation is far from that to which the homogeneous flame propagation in the SHS process is presented. In
premixed-flame theory can be applied. That is, we cannot Section 3, the statistical counting procedure is reviewed
regard the SHS process as a homogeneous phenomenon. and its usefulness is assessed. In Section 4, theoretical
Therefore, we are urgently required to re-examine the approaches are primarily aimed at studying the SHS flame
appropriateness of describing the SHS process as the propagation in the adiabatic condition. Following this,
premixed flame propagation in the homogeneous medium. several important effects on the SHS flame propagation
In another mathematical model for the SHS process, it is (such as the heat loss, size distribution of particles, and
proposed to assume the compacted medium as laminae of external heating by electric current) are discussed in relation
reactants [29,30]. The combustion situation for the usual to flame extinction. In Section 7, the transition boundary
SHS process, however, is quite different from this model between steady and pulsating combustion, which is indis-
because the combustion wave propagates through the pensable in bulk processing, is discussed. Up to this point,
10 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 4. Schematic flame structure in the heterogeneous flame (crystallization not described) [32].

discussion is confined to the flame propagation and extinc- consideration strongly suggests that a satisfactory formula-
tion. In the next section, initiation of combustion wave is tion must take into account the nonpremixed mode of par-
discussed for two different ignition methods: igniter and ticle reaction.
radiative heat flux. Here, it may be of great importance to recognize that the
flame propagation in the SHS process is actually quite simi-
2.3. Phenomenological description lar to that in fuel spray combustion, if we identify the fuel
droplets as the nonmetal (or higher melting point metal)
In order to recognize differences between the hetero- particles and the oxygen in air as the molten metal, restrict-
geneous flame in the SHS process and the homogeneous ing ourselves to the flame zone in the combustion wave. It is
premixed flame, let us first make a detailed phenomenolo- also appropriate to recognize that across the combustion
gical consideration for the SHS flame propagation. It can be wave, dozens of nonmetal particles [33] can exist in the
considered that the SHS process typically involves reaction “reaction” zone in which the temperature is higher than
between particles of a metal and a nonmetal (or another the melting point of metal. This can easily be confirmed
metal of a higher melting point). Generally speaking, before by comparing a representative particle size, for instance
the arrival of the combustion wave, the effects of the reac- 10 mm, to the approximate “flame thickness” d of about
tion cannot be remarkable due to the low temperature and 0.3 mm, evaluated from a relation d ⬇ l=…rcu† derived by
small area of contact between particles. However, as this the phenomenological analysis for gaseous premixed flames
particle matrix is heated by the approaching combustion [34], with representative values of thermometric conductiv-
wave, the lower melting point metal melts first, resulting ity l /(r c) of 3 × 10⫺6 m2 =s and the burning velocity u of
in slurry consisting of the higher melting-point nonmetal 0.01 m/s for Ti–C system. This recognition further suggests
particles suspended in the molten metal. In addition, due that a statistical counting procedure for spray combustion,
to an increase in the area of contact, as well as that in explained in Ref. [35], is indispensable and of great use in
temperature, the reaction between nonmetal particles and describing the combustion behavior of the heterogeneous
molten metal is greatly enhanced. Then, it is anticipated flame in the SHS process. The degree of similarity is achiev-
that the subsequent reaction between them will take place able if we restrict ourselves to fundamental principles with-
mainly over the surface of the nonmetal particles; this situa- out probing too far into specific details of the physical
tion is schematically shown in Fig. 4 [32]. Provided that the processes.
nonmetal particles do not dissolve in the molten liquid, and Another important factor that can influence the SHS flame
that they are not too small, compared to the thickness of the propagation is the mass diffusion in liquid phase, which is
combustion wave, it is easily seen that the flame propagation closely related to the nonpremixed reaction of nonmetal
through the compacted medium is basically heterogeneous particles in the combustion wave. Since the mass diffusivity
in nature, involving a premixed-mode of propagation for the in the SHS process is anticipated to increase markedly over
bulk flame supported by the nonpremixed reaction of the a relatively thin, high temperature combustion wave, use of
dispersed nonmetal particles in the liquid metal. Therefore, the temperature-dependent liquid-phase mass-diffusivity in
it is reasonable to anticipate that the resulting flame struc- Arrhenius fashion [22] is appropriate and indispensable.
ture and response should be quite different from those As for the surface reaction on each nonmetal particle,
obtained by assuming a premixed-mode of reaction at the related to the nonpremixed reaction in the combustion
molecular level. That is, the present phenomenological wave, it is recognized to be quite similar to the combustion
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 11
 
of a single solid particle, having been examined in the field Ta;s
A ˆ Das exp ⫺ ; b ˆ exp…x~ † ⫺ 1; …5†
of solid combustion, if the oxygen in air is identified as the Ts
molten metal. This recognition further facilitates inclusion
of the finite-rate reaction at the particle surface when formu- by solving the quasi-steady liquid-phase conservation equa-
lation is conducted. tion in spherical coordinates
!
Finally, in closing this section, it may be informative ~ ~
to describe a simplified model [32] which has been ~ d r~2 dY M ⫹ x~ dY M ˆ 0;
⫺…r~ D† …6†
d~r dr~ dr~
adopted for investigating combustion behavior of the
heterogeneous flame in the SHS process. As shown in under the isothermal condition, with the boundary condition
Fig. 4, the combustion situation modeled is the planar, at the particle surface …~r ˆ 1†
heterogeneous flame in an infinite domain of a !
~
compacted medium, originally consisting of a mixture ~ s dY M ⫹x~ Y~ M;s ˆ ⫺x~ ;
⫺…r~ D† x~ ⬅ r~ vr r~2 ˆ AY~ M;s ;
of particles of nonmetal N, metal M, and an inert which dr~ s
can be the product P of the reaction according to …7†
nM M ⫹ nN N ! nP P; where n i is the stoichiometric coef-
ficient. For simplicity, it is assumed that there is no and that at the outer edge …~r ! ∞† of the boundary layer
reaction until the mixture has been heated to the melt- around the particle
ing point Tm of the metal, at which all the metal par- Y~ M ˆ Y~ M;e : …8†
ticles melt instantly. The reaction that follows is assumed to
take place at the particle surface at a finite rate and proceed In the above, A is the reduced surface Damköhler number,
until all of the nonmetal (or higher melting-point metal) parti- Das the surface Damköhler number ( ˆ B·R/D), b the Spald-
cles, or metal species, are consumed. ing transfer number, and other variables and parameters are
defined as
r vR Y n W
3. Formulation of the heterogeneous theory r~ ˆ ; vr ˆ r ; Y~ M ˆ M ; fst ˆ M M ;
R De fst nN WN
3.1. Surface regression rate rM D
r~ ˆ ; D~ ˆ :
rM;e De
Before describing the statistical counting procedure used
in spray combustion, let us first confirm the description of a where r is the radial coordinate, vr the radial velocity, Y~ M the
single solid particle undergoing combustion. Since it is stoichiometrically weighted mass fraction, fst the stoichio-
unnecessary to consider the velocity difference between metric mass ratio, W the molecular weight, D the mass
particles and molten metal in the SHS combustion wave, diffusivity, and the subscript s and e designate, respectively,
only combustion behavior in a quiescent fluid will be exam- the particle surface and the outer edge of the boundary layer.
ined. For a single, spherical, nonpermiable, solid particle of An introduction of the isothermal condition, which
radius R in a quiescent fluid, the rate of change of its mass enables us to put r~ ˆ 1 and D~ ˆ 1; can be justified when
caused by the surface reaction nM M ⫹ nN N ! nP P is the thickness of the bulk flame is much larger than the
expressed as particle size because we can anticipate that both a nonmetal
  solid particle and molten metal around the particle are
M_ ˆ ⫺ d rN 4 pR3 ˆ ⫺rN 4pR2 R; _ …1† heated equally as the combustion wave arrives. The use of
dt 3 the conventional constant property assumption can also be
which yields justified if attention is confined to a restricted region around
the single particle. Note that changes in thermophysical
rM D M_ properties in the bulk during the flame propagation can be
⫺RR_ ˆ x~ ; x~ ⬅ : …2†
rN 4prM DR incorporated.
The analytical expression in Eq. (4), however, is an im-
Here, x~ is the surface regression rate, which coincides with plicit expression with respect to the surface regression rate,
the definition of the nondimensional combustion rate, which is difficult to have a clear image of its dependence on
commonly used in the conventional solid combustion. other parameters. Therefore, an attempt has been made to
Then, we have obtain an explicit, approximate expression [32,36] as
x rM D  
R_ ˆ ⫺ ; xˆ x~ : …3† x~ ⬅ ln…1 ⫹ b† ⬇ ln 1 ⫹
A ~
Y M;e ; …9†
R rN 1⫹A
The specific form of x~ can be obtained as [32,36] by use of the approximate relation
Y~ M;e ⫺ b b
x~ ˆ A ; …4† ˆ 1 ⫺ exp…⫺x~ † ⬇ x~ …10†
1⫹b 1⫹b
12 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

for small x~ ; as is the case for most solid combustion. Note the overall continuity equation for the fluid is given as
that the expression in Eq. (9) for the surface regression rate 2rf ZZ
has the same form as that commonly used in droplet ⫹ 7 x ·…rf u† ˆ ⫺ _ dR dv;
rN 4pR 2 Rf …12†
2t
combustion. The error between Eqs. (4) and (9) is within
2% for Ti–C system. It may be informative to note that Y~ M;e where r f is the fluid density defined as the mass of fluid per
in Eq. (9) can change as the combustion wave propagates, unit volume of physical space. Note that the fluid and liquid
being governed by the species conservation equations, to be densities are related by the expression
mentioned in the next section.
 ZZ 4 
rf ˆ rM 1 ⫺ pR3 f dR dv : …13†
3.2. Statistical counting procedure 3

3.2.1. Change of distribution function


3.2.2.2. Species conservation When chemical reactions
In spray combustion [35] a statistical description is given
occur at the particle surface and species M is
by the distribution function
consumed at the surface reaction according to
f …R; x; v; t† dR dx dv n MM ⫹ n NN ! n PP, the species conservation equation
for component M is given as
which is the probable number of particles in the radius range
dR about R, located in the spatial range dx about x with 2
velocities in the range dv about v at time t. Here, dx and …r Y † ⫹ 7 x ·‰rf …u ⫹ UM †YM Š
2t f M
dv are the three-dimensional elements of physical space and
 ZZ
velocity space, respectively. The variables R, x, and v must n W _ dR dv;
appear in the distribution function because conditions are ˆv⫹ M M rN 4pR 2 Rf …14†
nN WN
not known well enough to permit specification of the exact
size, position, or velocity of each particle. where YM is the mass fraction of species M in the fluid,
An equation governing the time rate of change of the UM the diffusion velocity, and v the mass rate of
distribution function f, which is called the spray equation production of species M by homogeneous chemical
in spray combustion, is given as reactions in the fluid. Note that for steady flow, by
2f 2 _ use of Fick’s law, UM is expressed as
ˆ⫺ …Rf † ⫺ 7 x ·…vf † ⫺ 7 v ·…Ff †; …11†
2t 2R
UM YM ˆ ⫺D…7 x YM †: …15†
when there are no particle formation/destruction by
processes nor collisions with other particles. Here, R_
( ˆ dR/dt) is the rate of change of the particle size R at (R, 3.2.2.3. Momentum conservation The momentum
x, v, t) and F …ˆ dv=dt† the force per unit mass on this conservation equation is given as
particle. The terms in the RHS respectively represent the
changes in f resulting from the change of particle size, the 2u XN

motion of particles into and out of the spatial element dx by rf ⫹ rf u·7 x u ˆ ⫺7 x ·P ⫹ rf Yk f k


2t kˆ1
virtue of their velocity v, and the acceleration of particles in
the velocity element dv caused by the force F. Note that both ZZ 4
R_ and F are allowed to depend on R, x, v, t, and the local ⫺ rN pR 3 Ff dR dv
3
properties of the fluid.
ZZ
⫺ _ ⫺ u†f dR dv;
rN 4pR2 R…v …16†
3.2.2. Conservation equations in general form
In analyzing behavior of the combustion wave, hydrody-
namic equations for the fluid are necessary. If we restrict in which the third term in the RHS represents the average
ourselves to the situation in which the statistical fluctuations force per unit volume exerted on the particles by the
in the fluid, induced by the random motion of individual surrounding fluid, and the last term accounts for the
particles, may be neglected, these governing equations for momentum carried to the fluid by the material from the
the local average properties in the fluid are equivalent to the particles. In the above, P is the total pressure tensor
ordinary equations of fluid dynamics, with suitably added ( ˆ pE ⫹ T), related to the hydrostatic pressure p and the
source terms accounting for the average effect of the viscous stress tensor T, where E is the unit tensor. In
dispersed particles. addition, fk is the external body force per unit mass acting
on species k in the fluid.
3.2.2.1. Overall continuity By adding the mass of material
per unit volume per unit time from the particles to the fluid, 3.2.2.4. Energy conservation The energy conservation
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 13

equation is given as and the bar denotes an average over all velocities, that is
" !# " !#
1 Z_ 1 Z
2
r h ⫹
u2
⫹ 7 x · rf u hf ⫹
u2 R_ ˆ Rf dv; v ˆ vf dv: …20†
2t f f 2 2 G G
Eq. (18) can be further simplified as
2p XN
ˆ ⫺7 x ·q ⫺ 7 x ·…T·u† ⫹ ⫹ rf Yk …u ⫹ Uk †·f k 2 _ 1 2
2t …RG† ⫹  ˆ 0;
…SvG† …21†
kˆ1 2R S 2x
ZZ 4
⫺ rN pR 3 …F·v†f dR dv when the local cross-sectional area S(x) only depends on the
3 variable x. Here, v is the x component of v ; and the quantities
! _ v;
R;  and G are averages over the cross section, independent
ZZ v2
2_
⫺ rN 4pR R hin ⫹ f dR dv; …17† of the spatial coordinates normal to x.
2 By using a relation that R_ ˆ ⫺x=R in Eq. (3), Eq. (21)
becomes
where hf is the total enthalpy per unit mass of the fluid, q the
heat flux vector, and hin the total enthalpy per unit mass of 2c 2c
⫺x 
⫹ vR ˆ 0; …22†
material flowing from the vicinity of a particle to the fluid. 2R 2x
In the above, the last two terms accounts for the work done where
on the fluid by particle and the energy added to the fluid by
the material from the surface. 
SvG
c⬅ : …23†
R
3.3. Heterogeneous theory for the SHS process Upon transformation to the new independent variables
Zx  x  Zx  x 
In the SHS process, it is usual to consider a situation in s 2 ˆ R2 ⫺ 2 dx; h2 ˆ R2 ⫹ 2 dx;
which a reaction, initiated at one end of a matrix of ⫺∞ v ⫺∞ v
compacted reactive particles, moves through the adjacent …24†
unburned matrix in the form of a self-sustained combustion Eq. (22) becomes 2c=2s ˆ 0; which means that c only
wave. In this situation, even a quasi-one-dimensional treat- depends on h . Letting the subscript 0 identify conditions
ment has practical importance, and its simplicity is favor- at the unburned state, we have c ˆ c 0(h ). Since h ˆ R0
able for clarifying dominant parameters that influence the as x ! ⫺ ∞, we see from Eq. (23) that the expression
flame propagation.   
S 0 v0 R
Gˆ G …h† …25†
Sv h 0
3.3.1. Size distribution function in quasi-one-dimensional
form determines the size distribution G(R, x) at any position x in
The velocity dependence of the distribution function f, terms of the distribution G0(h ) at the unburned state …x !
which is not of primary interest, can be eliminated from ⫺∞†:
the governing equations by integrating Eq. (11) over all
velocity space. Since f ! 0 as 兩v兩 ! ∞ for all physically 3.3.2. Conservation equations for steady, one-dimensional
reasonable flows, the integral of the last term in Eq. (11) form
is zero. As for the unsteady change of the distribution func- When we restrict ourselves to steady, one-dimensional,
tion f, it is of considerable significance in spray combustion, constant-area flows in which all particles travel at the same
in relation to such phenomena as diesel-engine combustion velocity, the distribution function f is expressed as
or combustion instability in liquid-propellant rocket motors.
f …R; x; v† ˆ G…R; x†d…v ⫺ v†;
 …26†
However, in the SHS process it may be excluded from our
consideration, because a matrix of compacted reactive par- where x is the flow direction, v the velocity of all particles,
ticles through which a combustion wave propagates is and d (·) the delta function.
expected to have a certain distribution of particles when it
is produced by compaction, as far as the distribution of 3.3.2.1. Overall continuity The integral over v in Eq. (12)
particle sizes is concerned. Then, we have yields
2 _ d Z∞
…RG† ⫹ 7 x ·…vG† ˆ 0; …18† …rf u† ˆ ⫺ _ dR;
rN 4pR 2 RG …27†
2R dx 0

where the number of particles per unit volume per unit range where u is the x component of u. The RHS in Eq. (27) also
of radius is appears in
Z d Z∞
G ˆ f dv; …19† …r v†
 ˆ _ dR;
rN 4pR2 RG …28†
dx s 0
14 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

which is derived by multiplying Eq. (11) by r N(4/3)pR 3 and which yields


integrating over v and R. Note that r s is the mass of solid per ! !
u2 v2
unit spatial volume defined as rf u hf ⫹ ⫹ rs v hN ⫹ ⫹ qx ⫹ txx u ˆ constant:
2 2
Z∞ 4
rs ˆ rN pR3 G dR: …29† …37†
0 3
Note that in deriving Eq. (36), use has been made of the
Eqs. (27) and (28) show that the overall continuity equation relation
is written in the form Z∞
_ in GdR ˆ d …rs vh
rN 4pR2 Rh  N †; …38†
rf u ⫹ rs v ˆ m ˆ constant; …30† 0 dx
where m is the total mass flow rate (i.e. mass burning rate) obtained by equating
per unit area. Z∞
7 x ·…rs vh
 N† ˆ _ N G dR
rN 4pR2 Rh
0
3.3.2.2. Species conservation When there exists no ZZ  
4 2 hN
homogeneous chemical reactions in the fluid, the species ⫹ rN pR3 R_ ⫹ v·7 x hN ⫹ F·7 v hN f dR dv;
conservation Eq. (14) is simplified as 3 2R
  …39†
d nM WM d
‰rf …u ⫹ UM †YM Š ˆ …r v†;
 …31† and
dx nN WN dx s
Z∞ Z∞
which yields _ in G dR ˆ
rN 4pR2 Rh _ N G dR
rN 4pR2 Rh
0 0
dYM ZZ  
rf D ⫺ …rf uYM ⫺ rs v fst † ˆ constant; …32† 4 2h
dx ⫹ rN pR3 R_ N ⫹ v·7 x hN ⫹ F·7 v hN f dR dv:
3 2R
where UM is the x component of UM and can be expressed by …40†
Fick’s law.
Here, Eq (39) is obtained by multiplying Eq. (11) by
rN …4=3†pR3 hN and integrating over v and R, while Eq.
3.3.2.3. Momentum conservation The momentum (41) is obtained by multiplying
conservation Eq. (16), by neglecting influences of body
_ in ˆ d 4
forces fk, becomes as rN 4pR2 Rh …r pR3 hN †
dt N 3
d dp dt  
…r u2 ⫹ rs v2 † ˆ ⫺ ⫺ xx ; …33† _ N ⫹ rN 4 pR3 R_ 2hN ⫹ v·7 x hN ⫹ F·7 v hN ;
dx f dx dx ˆ rN 4pR2 Rh
3 2R
which yields …41†
rf u2 ⫹ rs v2 ⫹ p ⫹ txx ˆ constant; …34† by f and integrating over v and R. Note that Eq. (41) gives a
relation between hin and hN.
where t xx is the xx component of the viscous stress tensor T.
Note that in deriving Eq. (33), use has been made of the 3.3.3. Further simplification of the governing equations
relation In the SHS process, all particles in the combustion wave
Z∞ 4 dv are reasonably assumed to travel at the same velocity as the
rN pR3 FG dR ˆ rs v ; …35† fluid; that is, v ˆ u: Then, the overall mass conservation in
0 3 dx
Eq. (30) becomes
which is the momentum equation of the dispersed particles,
rt u ˆ m ˆ constant; …42†
derived by multiplying Eq. (11) by rN …4=3†pR3 v and
integrating over v and R, as well as the relations in Eqs. where rt ⬅ rf ⫹ rs is the total density. As for the mass
(27) and (28). consumption of solid N particles, by introducing the mass
fraction of fluid as
3.3.2.4. Energy conservation When radiative heat transfer is rf
Zˆ ; …43†
neglected, the energy conservation Eq. (17) becomes as rf ⫹ rs
" ! !#
d u2 v2 Eq. (28) reduces to
rf u hf ⫹ ⫹ rs v hN ⫹
dx 2 2 d…1 ⫺ Z† 1 Z∞ _ dR:
ˆ⫺ r 4pR2 RG …44†
dx m 0 N
dq d
ˆ⫺ x ⫺ …t u†; …36† Substituting Eqs. (3) and (25) into Eq. (44) and by
dx dx xx
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 15

integrating, we have species conservation equations, such as


    2! rf YM UM
d…1 ⫺ Z† 4p Z∞ u0 R …YM ⫹ fst †Z ⫹ ˆ …YM;0 ⫹ fst †Z0 ; …54†
ˆ⫺ rN x G0 …h† dR: …45† m
dx m 0 u h
rf YP UP
When the particles are monodispersed with number density {YP ⫺ …1 ⫹ fst †}Z ⫹ ˆ {YP;0 ⫺ …1 ⫹ fst †}Z0 ; …55†
n0, that is G0 …R† ˆ n0 d…R ⫺ R0 †; Eq. (45) reduces to m
    as well as the relation between the heat of combustion and
d…1 ⫺ Z† 4prN n0 R0 u0 1 ⫺ Z 1=3
ˆ⫺ x ; …46† enthalpies as
dx m u 1 ⫺ Z0
⫺fst hM ⫹ …1 ⫹ fst †hp ⫺ hN
where the following relation has been made use of
! nP WP hP ⫺ nM WM hM ⫺ nN WN hN
Z∞ R2  Zx  x  1=2 ˆ
nN WN
G0 …h† dR ˆ n0 R20 ⫺ 2 dx …47†
0 h ⫺∞ v
ˆ ⫺q0 ⫹ …c ⫺ cN †…T ⫺ T 0 †; …56†
and
Eq. (50) reduces to
!
rt u 1⫺Z
  Zx  x  3=2
R30 ˆ R20 ⫺ 2 dx ; …48† …l=c† dT~
r t;0 u0 1 ⫺ Z0 ⫺∞ v ˆ …T~ ⫺ T~ 0 † ⫺ …Z ⫺ Z0 †; …57†
m dx
derived by the integration of Eq. (29). where T~ is the nondimensional temperature ‰ˆ cT=q0 Š and
The species conservation equation (32) for M species is q 0 the heat of combustion per unit mass of N species. Here, it
given as is assumed that enthalpy of phase change is negligible
  because it is usually much smaller than the heat of combus-
rf D dY~ M tion, that thermal conductivity across the combustion wave
ˆ …Y~ M ⫹ 1†Z ⫺ …Y~ M;0 ⫹ 1†Z0 ; …49†
m dx is constant, and that the specific heats are the same, that is
c ˆ cN :
where Y~ M is the stoichiometrically weighted mass fraction
Note that in the nonadiabatic condition, it is necessary to
‰ˆ YM =fst Š and the constant of integration is evaluated at the
take account of heat loss L in the lateral direction, so that the
unburned state …x ! ⫺∞†: Note that the use of the Arrhe-
energy conservation equation is expressed as
nius mass-diffusivity defined as D ˆ D0 exp…⫺Td =T† is " #
indispensable for the appropriate description of the SHS d …l=c† dT~ L
process. ⫺ …T~ ⫺ T~ 0 † ⫹ …Z ⫺ Z0 † ˆ : …58†
dx m dx mq0
The momentum conservation equation has already been
expressed in Eq. (34), which states that constant pressure is a In closing this section, it should be mentioned that these
good assumption when the velocity is low and the velocity governing equations can simply be derived if it is a priori
gradient is small. assumed that N particles are the same size in counting for
The energy conservation equation (37) in the adiabatic the heat and mass balance around them [38,39]. Regardless,
condition becomes here statistical counting procedure has been used, to show
the mathematical description in a general form, which is
qx
Zhf ⫹ …1 ⫺ Z†hN ⫹ ˆ Z0 hf;0 ⫹ …1 ⫺ Z0 †hN;0 ; …50† applicable to systems in which particle size distribution is
m not monodispersed.
when u2 =2 p hf and/or hN. The enthalpy of the fluid is given
as
4. Flame propagation in the adiabatic condition
X
N
hf ˆ Yk hk ; hk ˆ h0k ⫹ c…T ⫺ T 0 †; k ˆ M; P …51† Mathematical description [32,37] of the flame propagation
kˆ1 in the SHS process has been made by use of Eq. (46) for N-
and that of the particles is consumption, Eq. (49) for M-conservation, and Eq. (57) 1 for
energy conservation, as well as the relation of constant mass
hN ˆ h0N ⫹ cN …T ⫺ T 0 †: …52†
1
It should be noted that the energy conservation equation first
As for the x component of heat flux vector, it is expressed as reported [32] was incorrect, due to inappropriate treatment in
accounting for a term representing the sensible heat of N particles.
dT XN
qx ˆ ⫺l ⫹ hk rf Yk Uk ; …53† This was pointed out and corrected in the subsequent paper. [37]
dx kˆ1 Figs. to be presented in this review have already been corrected by
use of the energy conservation Eq. (57) although quantitative trends
when there is no radiation heat transfer. Then, by the use of are the same.
16 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

burning rate expressed in Eq. (42). The combustion situation 4


rN pR30 n0 1
modeled, as shown in Fig. 4, is that of the steady, adiabatic,  3 

one-dimensional, planar, heterogeneous flame propagation in 4 3 …nN WN †=…vM WM †
r YM;0 1 ⫺ pR0 n0
an infinite domain of a compacted medium, originally consist- 3
ing of a mixture of particles of nonmetal N, metal M, and an
fst 1 ⫺ Z0
inert which can be the product P of the reaction according to ˆ ; …64†
vM M ⫹ vN N !vp P: Then the flame propagation can be exam- YM;0 Z0
ined by solving these governing equations, subject to boundary  
conditions at the unburned and burned states. 4
r YP;0 1 ⫺ pR30 n0
kˆ  3 ˆ …1 ⫺ YM;0 †Z0 ; …65†
4.1. Dominant parameters 4 3 4
r 1 ⫺ pR0 n0 ⫹ rN pR30 n0
3 3
Before describing numerical results, let us first reconfirm
dominant parameters, because it is important in analyses to use where r is the fluid density that consists of metal M and inert
parameters which can fairly represent system conditions for P. It is seen that by specifying m and k , both the mass
the initial and final states of the SHS process, as well as their fraction of N particles, 1 ⫺ Z0 ; and the mass fraction YM,0
clear definitions from the physical point of view. Furthermore, of M species in fluid at the unburned state can be specified.
in order to make comparisons between theoretical and exper- It may be noted that the present consideration of dilution,
imental results, definite relations between parameters used in through which the mass fraction of molten M is reduced due
analyses and those in experiments are strongly required. to addition of combustion product, implicitly assumes that
At the burned state, we have the product is soluble in the liquid metal, and that the dis-
solution process occurs much faster than the reaction of N
T~ ∞ ˆ Z∞ ⫺ Z0 ⫹ T~ 0 ; …59† particles. The first requirement depends on the specific
metal/nonmetal (or metal/metal) system, and can be
Z assessed from its phase diagram, while the second require-
Y~ M;∞ ˆ …Y~ M;0 ⫹ 1† 0 ⫺ 1; …60†
Z∞ ment is favored for fast rates of interfacial dissolution and
dispersion through mass diffusion, especially for small
which are obtained, respectively, from the energy conserva-
product particles. For example, in the Ti–C system, TiC
tion equation (57) and the species conservation equation
melts in the range 1918–3340 K, depending on composi-
(49) at a completely reacted state. We see that both T~ ∞
tion. Since Ti melts at about 1950 K, and the mixture
and Y~ M∞ are expressed in terms of not only their initial
temperature increases due to reaction, TiC in the combus-
values, but also the mass fractions of fluid, Z, at the burned
tion wave is expected to exist in the fluid phase.
and unburned states. As for Z∞, by counting for the mass
As another dominant parameter, we can point out the
fraction of the the remaining N particles after the completion
initial temperature T0. The effect of T0, which determines
of the reaction, we have
8 T∞ through Eq. (59), may be significant because T∞ can
>
<1 …m ⱕ 1† exponentially influence the reaction rate and mass diffusiv-
Z∞ ˆ 1 ⫺ Z0 ; …61† ity in the condensed phase. It should be noted that T0 is the
>
: Z0 ⫹ …m ⱖ 1† temperature of the unburned medium, just before the arrival
m
of the flame front of the combustion wave.
where m is the mixture ratio defined as the initial molar ratio
of nonmetal, N, to metal, M, divided by the corresponding 4.2. Burning velocity
stoichiometric molar ratio. Note that it is common in exper-
iments to specify the nature of the initial (unburned) Because the spatial coordinate x appears in none of these
compact by this mixture ratio m and the degree of dilution equations, except as d/dx, the problem can further be
k , defined as the initial mass fraction of the diluent. These reduced to having Z instead of x as the independent coordi-
experimental parameters are related to the initial mass nate, for adiabatic flame propagation. Dividing Eqs. (49)
fraction of N particles, 1 ⫺ Z0, used in analyses, as [37] and (57) by Eq. (46), and defining the nondimentional
variables as
m…1 ⫺ k†
1 ⫺ Z0 ˆ : …62†
m ⫹ fst T~ ⫺ T~ 0 Y~ M ⫺ Y~ M;0 Z ⫺ Z0
uˆ ; zˆ ; jˆ ;
This relation can be derived from the following expressions T~ ∞ ⫺ T~ 0 Y~ M;∞ ⫺ Y~ M;0 Z∞ ⫺ Z0
for their definitions [32]: …66†
 
4 we have two first-order differential equations as [32]
r 1 ⫺ pR30 n0
3 !
Z0 ˆ  ; …63† du L …r =r † T~ d …u ⫺ j†
4 3 4 ˆ 0 t;0 1=3t exp ; …67†
r 1 ⫺ pR0 n0 ⫹ rN pR30 n0 dj …1 ⫺ j† x~ T~ …1 ⫺ Z 0 †
3 3
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 17
!
dz Le 0 L0 …rt;0 =rt † 2 2T~ d diffusion-controlled limit, x~ becomes independent of R0
ˆ exp and we have ma ⬃ R⫺1 0 :
dj {Z0 ⫹ …1 ⫺ Z0 †j}…1 ⫺ j† x~ 1=3
T~
Once the specific value of L 0 is obtained by a numerical
   
Z∞ Z0z calculation, the burning velocity is then obtained as [40]
 z⫺ j⫹ ; …68†
Z∞ ⫺ Z0 1 ⫺ Z0   p s
1 ⫺ Z0 rM =rN
when the mixture ratio is smaller than or equal to unity. …u0 ·R0 † ˆ …D0 3L0 Le0 † ; …75†
Z∞ ⫺ Z0 1 ⫺ Z0
When the mixture ratio is larger than unity, we have [40]
! which is simply expressed as [32,37]
du L 0 …rt;0 =rt † T~ d …u ⫺ j†
ˆ     1=3 exp ; s
dj Z∞ ⫺ Z0 T~ …Z ∞ ⫺ Z0 † p rM =rN
1⫺ j x~ …u0 ·R0 † ˆ D0 3L0 Le0 ; …76†
1 ⫺ Z0 1 ⫺ Z0
…69†
when the mixture ratio is under-stoichiometric. We see that
dz Le 0 L0 …rt;0 =rt † 2 the burning velocity is inversely proportional to the particle
ˆ     1=3 size. The parameter u0·R0 is sometimes called the “SHS rate-
dj Z ∞ ⫺ Z0
{Z0 ⫹ …Z∞ ⫺ Z0 †j} 1 ⫺ j x~ constant”. Note that in the experiments the burning velocity
1 ⫺ Z0
u0 should be determined as the velocity at which the
!   
2T~ d Z∞ Z0z dazzling combustion wave moves normal to its surface
 exp z⫺ j⫹ : through the adjacent unburned compacted medium.
T~ Z∞ ⫺ Z0 Z∞ ⫺ Z0
As for the effect of the relative density r rel, which is also
…70† one of the important system parameters in experiments and
In the above, L 0 is the mass burning rate eigenvalue defined is defined as the ratio of the apparent density of a compacted
as medium to the theoretical maximum density r TMD, its influ-
ence on the burning velocity is anticipated to appear through
…Z∞ ⫺ Z0 † 2 m2a the thermophysical properties in the Lewis number Le0 ‰ˆ
L0 ˆ ; …71†
4p…rM D0 †…l=c†n0 R0 …l=c†=…rt D†0 Š which appears in Eqs. (75) or (76). Since the
total density r t is expressed as rt ˆ rrel ·rTMD and the
Le0 the Lewis number as
specific heat c is proportional to r rel, while the thermal
…l=c† conductivity is a weak function of r rel, the burning velocity
Le0 ˆ ; …72†
…rt D†0 is considered to be inversely proportional to the relative
density r rel, as the first approximation for the usual range
and x~ the surface regression rate as expressed in Eq. (9). In
of r rel in the SHS process.
Eq. (71), ma is the total mass burning rate in the adiabatic
condition. The boundary conditions are
4.3. Range of flammability
j ˆ 0; u ˆ um; z ˆ 0; …73†
The regime in which a solution of the present problem can
j ˆ 1; u ˆ 1; z ˆ 1: …74† exist is anticipated to coincide with the range of flammabil-
ity. If we consider the general restriction that the tempera-
Note that the cold boundary difficulty [34] can be eliminated
ture of the completely reacted state should be higher than the
by assuming that the reaction is initiated at the melting point
melting point …T∞ ⱖ Tm † [32,37], we have
Tm of the metal [32].
Therefore, it is seen that the problem is reduced to solving fst
! ⱕ m ⱕ 1; …77†
the mass burning rate ma from the two first-order equations 1⫺k
⫺1
of Eq. (67) and (68) for under-stoichiometric mixture ratios, T~ m ⫺ T~ 0
or Eqs. (69) and (70) for over-stoichiometric mixture ratios,
satisfying the four boundary conditions in Eqs. (73) and  
(74). For spray combustion, the surface regression rate x~ f
0 ⱕ k ⱕ 1 ⫺ 1 ⫹ st …T~ m ⫺ T~ 0 †; …78†
is independent of the droplet size and L 0 becomes the eigen- m
value. Since n 0 R03 is fixed for a given stoichiometry, L 0 for the under-stoichiometric mixture ratios;
implies that ma ⬇ R⫺1 0 : For the present problem, the surface
Damköhler number A, and the surface regression rate x~ ; 1⫺k
1ⱕmⱕ ⫺ fst ; …79†
depend on the particle radius R because of the existence T~ m ⫺ T~ 0
of the surface reaction; the particle radius varies as R ˆ
R0 …1 ⫺ j† 1=3 : Then the dependence of ma on R0 is more 0 ⱕ k ⱕ 1 ⫺ …m ⫹ fst †…T~ m ⫺ T~ 0 †; …80†
complex. In the limit of small R0, x~ ⬃ A ⬃ R 0 ; that is L 0/
R0 becomes the eigenvalue such that ma ⬃ R⫺1=2 0 : In the for the over-stoichiometric mixture ratios.
18 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 5. Representative flame structure for the temperature u , the carbon concentration …1 ⫺ j†; and the reaction rate dj /ds , along the
nondimensional coordinate s , for the stoichiometric mixture …m ˆ 1:0† without dilution …k ˆ 0† when the initial temperature T 0 ˆ 450 K;
the Lewis number Le0 ˆ 100; the initial Damköhler number Da0 ˆ 10 9 ; and the adiabatic mass burning rate eigenvalue L0 ˆ 6:109 × 10⫺5 :
Dashed curves are the results in the adiabatic condition [32] and solid curves are those in the nonadiabatic condition [64].

4.4. Theoretical and experimental results for Ti–C system nature of the heterogeneous flame structure supported by
particle reaction, so that the region of the particle consump-
Eqs. (67) and (68) have been numerical integrated to tion in the combustion wave can be called the “consumption
determine the mass burning rate eigenvalue for Ti–C system zone.” For the prescribed conditions with R 0 ˆ 10 mm; the
[32]. Values of the physicochemical parameters are: q ˆ flame thickness and burning velocity are predicted to be
15 MJ=kg; c ˆ 1 kJ=…kg:K†; rM ˆ 4:50 × 103 kg=m3 ; rN ˆ about 1 mm and 10 mm/s, respectively.
2:25 × 103 kg=m3 ; WN ˆ 47:9 × 10⫺3 kg=mol; WM ˆ 12:0 × Fig. 6 shows mass burning rate eigenvalue L 0 as a func-
10⫺3 kg=mol; Td ˆ 1973 K; Ta ˆ 3 × 104 K: For the mass tion of the initial surface Damköhler number Da0 ‰ˆ
diffusivity D ˆ D0 exp…⫺1:66 × 104 =T† m2 =s is used, B·R0 =D0 Š for an initial temperature T~ 0 ˆ 0:03 …T0 ˆ
following Hardt and Phung [29]. Since apparent thermo- 450 K†; with the Lewis number Le0 and the mixture ratio
metric conductivity l /(r tc) is of the order of 10 ⫺5 m 2/s at m taken as parameters. With increasing Da0, L 0 initially
3000 K [41], the Lewis number Le0 is estimated to be 50 or increases and attains a constant value for Da0 ⬎ 10 5.
more. Total densities before and after combustion are This implies that for higher values of Da0, the reaction,
assumed to be equal. Note that the thermophysical proper- and hence the mass burning rate, becomes diffusion
ties used here implicitly account for effects of compact limited, herein A=…1 ⫹ A† ! 1 in x~ which appears in
density, gases in void spaces, gas evolution, etc. Eqs. (67) and (68). Since Da0 is closely related to B·R0,
Fig. 5 shows a representative flame structure with profiles its value is also shown in the abscissa. Fig. 6 further
of temperature u , carbon mass fraction, 1 ⫺ j; and the par- shows that due to the very large value of Le0, L 0 is insen-
ticle consumption rate dj /ds , along the nondimensional sitive to Le0. It is, however, significantly increased with
distance s ˆ ‰m a …1 ⫺ Z0 †=…l=c†Šx: The dashed curves rep- increasing m because of the increased amount of
resent the results in the adiabatic condition, while solid volumetric heat generation.
curves are those of the nonadiabatic condition, to be Since L 0 is basically independent of Le0, in order to
discussed in Section 5. Position at which u ˆ u m is set at determine appropriate values of B and Le0, numerical results
s ˆ 0: With the monotonic increase in the temperature u are compared with the experimental results [33] for various
from the melting point u m to the adiabatic combustion radii (from 0.5 to 12.5 mm) of specimen and it is confirmed
temperature …u ˆ 1†; the mass fraction of carbon 1 ⫺ j [32] that close correlation can be obtained for Le0 ˆ 100
rapidly decreases from 1 to 0. The particle consumption and B ⱖ 10 3 m=s when the relative density r rel is about 0.6.
rate dj /ds exhibits the initial increase due to carbon deple- Note that the value of B ⱖ 103 m=s implies that the combus-
tion and consequent decrease due to reactants depletion, as tion is diffusion-controlled. Fig. 7 shows the effect of the
R ! 0: Its profile spans approximately the same range over initial radius R0 of carbon particle on the burning velocity u0
which the temperature increases. Compared to the typical of the stoichiometric mixture …m ˆ 1† for various B. The
premixed flame structure, characterized by a narrow, highly comparison between the predicted and experimental results
peaked reaction rate profile located close to the maximum [26–28,33,42] demonstrates the importance of heterogen-
temperature, the present result demonstrates the diffusive eity in the flame propagation process, and shows the fact
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 19

that u0 is inversely proportional to R0 as shown in Eqs. (75)


or (76). The theory breaks down as R0 ! 0; and ma ! ∞;
which corresponds to the theoretical fact that the flame
propagation mode should become homogeneous and the
combustion be reaction-controlled (cf. Fig. 6). Experimental
observation [33], however, shows that there occurs a tran-
sition from the steady, one-dimensional flame propagation
to the so-called pulsating combustion [1–3] when the par-
ticle radius is less than about 2.5 mm.
Fig. 8 shows u0 as a function of the mixture ratio m , with
R0 taken as a parameter. Comparison between the calcula-
tions and experimental results [33] is again satisfactory. The
lower limit of flammability given by Eq. (77) is 0.448, which
is close to the lowest m of about 0.47 obtained experimen-
tally [33]. Calculations have also been performed with a
constant diffusivity of D ˆ 10⫺9 m2 =s; which is representa-
tive of its value in the consumption zone. Results for the
constant diffusivity are more “curved” than the somewhat
linear variations shown in both the experimental data and
calculated results with Arrhenius mass-diffusivity. This
therefore demonstrates the need [22] to realistically include
Fig. 6. Mass burning rate eigenvalue L 0 for Ti–C system as a the temperature-dependent mass diffusivity, without which
function of the initial surface Damköhler number Da0 for the u0 becomes large because of the overestimated mass diffu-
nonmetal particle, with the Lewis number Le0 and the mixture sivity at relatively low temperatures.
ratio m taken as parameters [32]; the initial temperature T~ 0 ˆ 0:03 As for the dependence of u0 on m in its wide range, Fig. 9
(T0 ˆ 450 K). shows the SHS rate-constant u0,a·R0 as a function of 2m=…1 ⫹
m† for T0 ˆ 450 K; with the degree of dilution k taken as a

Fig. 7. Burning velocity u0 for Ti–C system as a function of the Fig. 8. Burning velocity u0 for Ti–C system as a function of the
initial radius R0 of carbon particle, with the frequency factor B taken mixture ratio m , with the initial particle radius R0 and the frequency
as a parameter [32]; the Lewis number Le0 ˆ 100; the initial factor B taken as parameters [32]; the Lewis number Le0 ˆ 100 and
temperature T~ 0 ˆ 0:03 (T0 ˆ 450 K), and the mixture ratio m ˆ the initial temperature T~ 0 ˆ 0:03 (T0 ˆ 450 K). Data points are
1:0: Data points are experimental [26–28,33,42]; solid curves are experimental [33]; solid curves are calculated with Arrhenius diffu-
calculated from theory. sivity; dashed curves are calculated with representative diffusivity.
20 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 9. Burning velocity multiplied by particle radius (SHS rate-constant) u0·R0 for Ti–C system as a function of 2m=…1 ⫹ m†; with the degree of
dilution k taken as a parameter [40]; the initial temperature T0 ˆ 450 K: Data points are experimental [33,43] for different particle radii of
R0 ˆ 5 mm and 12.5 mm.

parameter. Introduction of 2m=…1 ⫹ m†; which is a harmonic gradually decreases until k reaches the dilution limit given
mean of m and 1=m; enables us to discuss the effect of m , by Eq. (78), due to reduced volumetric heat generation.
which can change from zero to infinity, in a finite domain. Experimental results shown in Fig. 11 are those by
With increasing 2m=…1 ⫹ m†; the SHS rate-constant u0,a·R0 Bloshenko et al. [44] at T0 ˆ 973 K and those by Makino
first increases, reaches the maximum, and then decreases. et al. [33] at T0 ˆ 450 K; the experimental results
The increase in u0,a·R0 is attributed to an increase in the [26,45,46] with carbon particles of 0.1 mm in diameter are
volumetric heat generation, as well as an increase in the excluded because steady propagation cannot be expected.
heat generation rate in the combustion wave, related to the Fig. 12 shows u0 as a function of the relative density r rel
total surface area of N particles. The peak of u0,a·R0 occurs for m ˆ 1:0; with 2R0 taken as a parameter; u0 gradually
on the rich side …m ⬎ 1† of N particles, because of the decreases when 2R0 is larger than 10 mm and the steady
increase in the total surface area of N particles, in spite of
the decrease in volumetric heat generation. Further increase
in 2m=…1 ⫹ m† results in a decrease in u0,a·R0, because of the
reduced volumetric heat generation. It is also observed that
u0,a·R0 decrease with increasing k , and that dilution moves
the position of the maximum u0,a·R0 towards the stoichio-
metric mixture. Experimental data [33,43] for two different
particle sizes are also shown in Fig. 9 and fair agreement is
observed as far as the general trend and the approximate
magnitude are concerned. It is pointed out, however, that
discrepancies near the maximum u0,a·R0 may be attributed to
the assumption of constant thermophysical properties
throughout the combustion wave. Overestimation of the
range of flammability by Eqs. (77) and (79) is attributed to
the heat loss to be discussed in Section 5.
Fig. 10 shows u0 as a function of the initial temperature
T0, for m ˆ 0:6 and 1.0. The trend of increasing u0 with
increasing T0 is as expected, as the preheating effect. The
comparison with experimental data [33,44] is again satisfac-
tory.
Dilution by the final product (TiC), to control the burning
velocity and/or combustion temperature has also been inves- Fig. 10. Burning velocity u0 for Ti–C system as a function of the
tigated; in the field of material processing, this corresponds initial temperature T0, with the mixture ratio m and the frequency
to controlling grain size in materials through control of the factor B taken as parameters [32]; the Lewis number Le0 ˆ 100 and
maximum temperature. Fig. 11 shows u0 as a function of k the initial radius R0 ˆ 10 mm: Data points are experimental [33,44];
for m ˆ 1:0; with T0 taken as a parameter. It is seen that u0 solid curves are calculated from theory.
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 21

densities it is caused by reduced temperature in the combus-


tion wave due to enhanced heat transfer ahead of the
combustion wave.

4.5. Applicability of the heterogeneous theory for other


systems

Agreement with a single system is not adequate to establish


the validity of the heterogeneous theory, especially since this
theory represents a fundamentally different description of the
SHS process. To this aim, the heterogeneous theory has been
applied to compare with experimental results obtained from
the syntheses of several borides and intermetallic compounds
for which reliable experimental data are available. It is attrib-
uted that the viability of the theory will be demonstrated
through extensive, quantitative comparisons in the range of
flammability with respect to effects of mixture ratio, degree
of dilution, initial temperature, and/or particle size on the burn-
ing velocity.

4.5.1. Available experimental data


Borides and intermetallic compounds have been chosen
Fig. 11. Burning velocity u0 for Ti–C system as a function of the [47] because of the practical interest for advanced applica-
degree of dilution k , with the initial temperature T0 and the tions. For example, the boride ceramics of titanium diboride
frequency factor B taken as parameters [32]; the Lewis number (TiB2), zirconium diboride (ZrB2), and hafnium diboride
Le0 ˆ 100 and the mixture ratio m ˆ 1:0: Data points are exper-
(HfB2), due to their refractory nature and hardness, can be
imental [33,44]; solid curves are calculated from theory.
used as heating elements for high-temperature electric-
resistance furnace, as components and reinforcements for
propagation is maintained. When the particle size is small, furnaces in metal industry, and as cutting tools and abra-
pulsating combustion occurs and the trend is quite different sives. Further, TiB2 is considered an important component
from that of the steady propagation. It is reported [26] that in the fabrication of Functionally Graded Materials (FGMs),
the reduction of the burning velocity is caused by the for use as heat-shielding/structural materials in future space
reduced contact area between the molten metal and solid applications.
particles for low relative densities, as well as the reduced Alloys based on the intermetallic compound of nickel
volumetric heat generation, whereas for high relative aluminide (NiAl) possess an attractive combination of

Fig. 12. Burning velocity u0 for Ti–C system as a function of the relative density r rel, with particle diameter 2R0 taken as parameters; the Lewis
number Le0 ˆ 100 and the mixture ratio m ˆ 1:0: Data points are experimental [26,33]; open symbol designates the steady propagation and
half-open symbol the pulsating combustion.
22 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

mechanical and physical properties, such as high strength,

obtained with data from Itin et al. [53]


obtained with data from Itin et al. [53]
low density, and high thermal conductivity, that make them
ideal candidates for a number of high-temperature applica-
tions in aviation, automobile, and nuclear power engineer-

Naiborodenko and Itin [58]


Remarks with respect to Ta

ing.
Borovinskaya et al. [51]
Borovinskaya et al. [51]
Borovinskaya et al. [51]
The intermetallic compounds such as titanium cobaltide
Azatyan et al. [46]

(TiCo) and titanium nickelide (TiNi) are of interest because


of their shape-memory properties. It is even suggested [5]
that formation of these compounds by the SHS process has
distinct advantages over other conventional methods in
terms of stronger shape-restoring force and enhanced homo-
geneity.
The values of physicochemical parameters for these
reaction Ta (K)
temperature of

systems are listed in Table 2. Some of the transport par-


Activation

3:00 × 104
3:83 × 104
3:73 × 104
4:79 × 104
1:68 × 104
4:28 × 104
5:61 × 104

ameters listed in Table 3 are those in the solid phase of


the metal near the melting point, because of very limited
published data for the mass diffusivity. For the activation
energies, it has been found that they have nearly the same
Melting point
of M-species

values as those obtained from the Arrhenius plot of (u/T) vs.


(1/T) at high temperatures in which mass diffusivity can be
Tm (K)

expected to influence the burning velocity, where u is the


1973
1973
2125
2500

1767
1728
934

burning velocity and T the maximum temperature. The ther-


mometric conductivity l=…rt c† are those at the melting point
WN (kg/mol)

12:0 × 10⫺3
10:8 × 10⫺3
10:8 × 10⫺3
10:8 × 10⫺3
58:7 × 10⫺3
47:9 × 10⫺3
47:9 × 10⫺3
Mol. wt. of

of the lower melting-point metal [48]. Representative values


N-species

of the Lewis number Le0 are also listed.


In order to apply the heterogeneous theory [32], which
allows for the presence of finite-rate surface reactions,
47:9 × 10⫺3
47:9 × 10⫺3
91:2 × 10⫺3
178:5 × 10⫺3
27:0 × 10⫺3
58:9 × 10⫺3
58:7 × 10⫺3
WM (kg/mol)

kinetic parameters for the individual surface reactions are


Mol. wt. of
M-species

also required. However, because of the high temperature in


the consumption zone in the SHS process, for instance,
higher than about 2000 K for boride synthesis and 1000 K
for the intermetallic compounds, the reaction is likely to be
r N (kg/m 3) b
Density of

fully activated, such that the combustion behavior is


2:25 × 103
2:35 × 103
2:35 × 103
2:35 × 103
8:85 × 103
4:50 × 103
4:50 × 103
N-species

expected to proceed in the diffusion-controlled regime.


Satisfactory agreement in the subsequent comparisons
substantiates this assumption.
r M (kg/m 3) a

4:50 × 103
4:50 × 103
6:53 × 103
13:09 × 103
2:69 × 103
8:90 × 103
8:85 × 103
Density of
M-species

N-species represents nonmetal or higher melting-point metal.

4.5.2. Effects of mixture ratio, m


M-species represents metal or lower melting-point metal.

Fig. 13(a) shows the burning velocity u0 for Ti–B system


as a function of stoichiometry, with initial temperature T0
taken as a parameter. It has been found that flame propaga-
Specific heat
c (kJ/kg ·K)

tion becomes diffusion-controlled when the frequency factor


B exceeds 10 4 m/s. The particle radius is set to R0 ˆ 0:5 mm
1.00
0.99
0.46
0.45
1.18
0.55
0.52

in accordance with experimental data; some of the exper-


iments [45,49,50] were conducted with 1 mm boron particle
List of physicochemical parameters [47]

in diameter, others [51,52] with 0.1 mm boron particle. The


combustion
q 0 (MJ/kg)

comparison between predicted and experimental results is


Heat of

considered to be satisfactory, hence demonstrating the


15.0
12.9
14.9
16.6
2.0
2.6
2.9

importance of stoichiometry in the flame propagation


process. The trend of decreasing u0 with decreasing m is
nM M ⫹ nN N ! nP P

as expected, because of reduced volumetric heat generation,


Hf ⫹ 2B ! HfB2

Co ⫹ Ti ! TiCo
Zr ⫹ 2B ! ZrB2

Al ⫹ Ni ! NiAl
Ti ⫹ 2B ! TiB2

and the lower limits of flammability as given by Eq. (77)


Ni ⫹ Ti ! TiNi
Ti ⫹ C ! TiC

fairly agree with those of the experiments. As for the


preheating effect, it is found to be relatively small although
Reaction
Table 2

u0 increases with increasing T0 at a certain stoichiometry.


b
a

Further, since there is no discernible difference in the


A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 23

burning velocities for different sizes of the boron particles, it

Samsonov and Vinitskii [62]


Remarks with respect to D0

Naiborodenko and Itin [58]


is pointed out that there can be agglomeration for small
particles in the experiments.

Hardt and Phung [29]


Hardt and Phung [29]
Hardt and Phung [29]
Fig. 13(a)–(d) shows similar comparisons for Zr–B,
[3,45,50–52] Hf–B, [51,52] and Co–Ti [53,54] systems.
The values of R0 used are indicated in the respective figures.

Graham [63]
Graham [63]
The comparison between predicted and experimental
and Td

results, including the limits of flammability, can be consid-


ered to be satisfactory.

4.5.3. Effects of degree of dilution, k


Fig. 14(a) shows the burning velocity u0 for Ti–B system
Emissivity, e b

as a function of k for m ˆ 1:0; with T0 taken as a parameter.


The experimental results are from Refs. [45,51,52,55,56]. It
is seen that the burning velocity gradually decreases until k
0.4
0.6
0.7
0.7
0.8
0.8
0.8

reaches the dilution limit given by Eq. (78), due to reduced


volumetric heat generation, as well as the reduced mass
diffusivity caused by the reduced temperature in the
Representative Lewis

consumption zone. Fig. 14(b)–(e) shows similar compari-


sons for Zr–B [3,45,50–52], Hf–B [50–52], Al–Ni [57],
and Co–Ti [53] systems. Fair agreement is demonstrated
number Le0

as far as the trend and approximate magnitude are


concerned.
1000
100

500

10
10
4

4.5.4. Effects of initial temperature, T0


Initial temperature T0 is also one of the controllable par-
ameters in experimentation, and its effects on the burning
Touloukian et al. [41] and Touloukian et al. [48]; values at the melting point of the M-species.

velocity have been examined extensively. Fig. 15(a) shows


Thermometric conductivity

u0 for Al–Ni system as a function of T0, for stoichiometric


mixture …m ˆ 1:0†: Values of B in the diffusion-controlled
regime is 10 5 m/s and more. The particle radius is set to
R0 ˆ 10 mm in accordance with the experimental data;
l /(r t c) (m 2/s) a

Hardt and Holsinger [45] used the Ni particle of 6 mm in


1:0 × 10⫺5
1:6 × 10⫺5
1:8 × 10⫺5
1:8 × 10⫺5
7:0 × 10⫺5
8:0 × 10⫺6
1:6 × 10⫺5

diameter, Naiborodenko and Itin [58,59] used 20 mm Ni


particle, and Maslov et al. [57] used particle less than
45 mm in diameter; the relative density was about 0.4. The
trend of increasing u0 with increasing T0 is due to the
preheating effect, as expected. Fair agreement is demon-
temperature Td (K)

strated when T0 is lower than about 600 K. The deviation


at high temperatures can be attributed to the experimental
determination of T0; that is, the initial temperature T0, which
Touloukian et al. [41] and Touloukian et al. [69].
Activation

1:66 × 104
1:54 × 104
1:74 × 104
2:40 × 104
9:01 × 103
1:87 × 104
1:92 × 104

should be that before arrival of the combustion wave, could


be higher than the ambient temperature, which might be the
reported T0, because of the heat associated with the ignition
source. Fig. 15(b) shows similar comparison for Co–Ti
system. The particle radius is chosen to be R0 ˆ 30 mm:
List of transport parameters [47,66]

Experimental results are from Ref. [53]. The trend is the


factor D0 (m 2/s)
Pre-exponential

same as that of Al–Ni system.


Diffusivity

2:0 × 10⫺7
8:9 × 10⫺9
1:3 × 10⫺8
5:0 × 10⫺7
4:8 × 10⫺6
1:1 × 10⫺5
2:9 × 10⫺5

4.5.5. Effects of particle radius, R0


Particle radius R0 is the important parameter which speci-
fies the heterogeneity in the combustion wave, such that the
burning velocity can be controlled by varying the particle
size. Fig. 16(a) shows the effect of particle radius R0 on u0
Table 3

System

for Co–Ti system when m ˆ 1:0 and T0 ˆ 573 K: The


Co–Ti
Al–Ni

Ni–Ti
Hf–B
Zr–B
Ti–C
Ti–B

b
a

comparison between predicted and experimental [53] results


24 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 13. Burning velocity u0 as a function of the mixture ratio m , with the initial temperature T0 taken as parameters [47]; solid curves are
calculated from theory. (a) Ti–B system for Le0 ˆ 103 and R0 ˆ 0:5 mm; (W) data from Hardt and Holsinger [45]; (A) from Kirdyashkin et al.
[49]; ( ) from Zenin et al. [50]; (K) from Borovinskaya et al. [51]; (S) from Novikov et al. [52]. (b) Zr–B system for Le0 ˆ 500 and R0 ˆ
0:5 mm; (A) data from Merzhanov and Borovinskaya [3]; (W) from Hardt and Holsinger [45]; ( ) from Zenin et al. [50]; (K) from Borovinskaya
et al. [51]; (S) from Novikov et al. [52], (c) Hf–B system for Le0 ˆ 4 and R0 ˆ 0:5 mm; (K) data from Borovinskaya et al. [51]; (S) from
Novikov et al. [52]. (d) Co–Ti system for Le0 ˆ 10; R0 ˆ 30 mm and 50 mm; ( ) data from Itin et al. [53]; (A, W) from Itin et al. [54].

Fig. 14. Burning velocity u0 as as a function of the degree of dilution k , with the initial temperature T0 taken as a parameter [47]; stoichiometric
mixture …m ˆ 1:0†; solid curves are calculated from theory. (a) Ti–B system for Le0 ˆ 103 and R0 ˆ 0:5 mm; (W) data from Hardt and
Holsinger [45]; (K) from Borovinskaya et al. [51]; (S) from Novikov et al. [52]; ( ) from Shcherbakov and Pityulin [55]; (L) Maksimov
et al. [56]. (b) Zr–B system for Le0 ˆ 500 and R0 ˆ 0:5 mm; (A) data from Merzhanov and Borovinskaya [3]; (W) from Hardt and Holsinger
[45]; ( ) from Zenin et al. [50]; (K) from Borovinskaya et al. [51]; (S) from Novikov et al. [52]. (c) Hf–B system for Le0 ˆ 4 and R0 ˆ
0:5 mm; ( ) from Zenin et al. [50]; (K) from Borovinskaya et al. [51]; (S) from Novikov et al. [52]. (d) Al–Ni system for Le0 ˆ 10 and R0 ˆ
10 mm; (A, S) data from Maslov et al. [57] for T0 ˆ 298 K and 740 K. (e) Co–Ti system for Le0 ˆ 10 and R0 ˆ 30 mm; (W) data from Itin et al.
[53] for T0 ˆ 873 K:
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 25
26 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 15. Burning velocity u0 as a function of the initial temperature


T0 [47]; stoichiometric mixture …m ˆ 1:0†; solid curves are calcu-
lated from theory. (a) Al–Ni system for Le0 ˆ 10 and R0 ˆ 10 mm; Fig. 16. Burning velocity u0 as a function of the initial particle
(W) data from Hardt and Holsinger [45]; (A) from Maslov et al. radius R0 [47]; stoichiometric mixture …m ˆ 1:0†; solid curves are
[57]; ( ) from Naiborodenko and Itin. [58,59]. (b) Co–Ti system calculated from theory. (a) Co–Ti system for Le0 ˆ 10 and T0 ˆ
for Le0 ˆ 10 and R0 ˆ 30 mm; (W) data from Itin et al. [53] for k ˆ 573 K; data from Itin et al. [53]; open symbol designates the steady
0:0; ( ) from Itin et al. [53] for k ˆ 0:2: propagation and solid symbol the unsteady combustion. (b) Ni–Ti
system for Le0 ˆ 4 and T0 ˆ 750 K; (K) data from Bratchikov et al.
[60]; (W) from Itin et al. [61].
can be considered to be satisfactory, hence demonstrating
the essentially heterogeneous nature of the flame propaga- parametrically, by solving the two first-order equations of
tion process. Specifically, u0 is inversely proportional to R0 (67) and (68) under the four boundary conditions in Eqs.
when combustion is diffusion-controlled (say, B ⱖ (73) and (74). An approximate analytical expression for the
107 m=s†: Fig. 16(b) shows similar comparison for Ni–Ti burning velocity is also obtainable, which can explicitly
system [60,61] when m ˆ 1:0 and T0 ˆ 750 K: The inverse display the functional dependence of the burning velocity
dependence of u0 on R0 is again observed. on the various system parameters.

4.6. Approximate expression for the burning velocity 4.6.1. Derivation of the approximate expression for m ⱕ 1
To this aim, we should first note that for the usual SHS
Mass burning rate eigenvalue L 0, the square root of process the diffusion-controlled reaction prevails, as
which is the nondimensional burning velocity, has been mentioned in the previous sections, and the Lewis number
obtained numerically [32,47] and hence presented is very large, say, Le ⬎ 104 in the consumption zone. By
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 27

assuming diffusion-controlled limit and infinite Lewis and the complementary error function
number, we have
2 Z∞
x~ ⬇ ln…1 ⫹ Y~ M † …81† erfc…z† ˆ p exp…⫺t2 † dt …89†
p z

j In integrating the RHS of Eq. (83), an attempt [37] has been


z⬇ …82†
Z 0 ⫹ …1 ⫺ Z0 †j made to express u with 1 ⫺ j and …1 ⫺ j 1=3 †; and it is
reported that by comparing the numerical results, the rela-
respectively. The former is derived from Eq. (9) by setting tion
the reduced Damköhler number A to be infinity. The latter is
 
derived by setting the last part of the RHS in Eq. (68) to be 1 ⫺ um
zero, because of the excessively slow rate of diffusion as …1 ⫺ u† ˆ …1 ⫺ j† …90†
2
compared to thermometric conductivity and reaction rate.
Numerical results obtained by using Eqs. (81) and (82), fairly represents the temperature profile at high temperatures
instead of Eqs. (9) and (68), are even found to be practically (cf. Fig. 17), which is needed for the fair evaluation of the
indistinguishable. Arrhenius nature for the liquid-phase mass-diffusion, to
By introducing these relations, Eq. (67) is expressed in determine L 0. Note that the proportionality constant is the
integral form as arithmetic mean of the constants determined at j ˆ 0 and 1.
! When Eq. (90) is substituted into Eq. (83) and the first two
Z1 T~ d
exp ⫺ du terms of the following mathematical formula is used
um T~
 
rt;0 Z1 …t ⫺ j† ln…1 ⫹ z†
ˆ …1 ⫹ g†L0   dj
rt 0 f⫺j (
…1 ⫺ j† 1=3 ln 1 ⫹
j⫹g (83) z z3 z5
ˆ2 ⫹ ⫹ ⫹…
2⫹z 3…2 ⫹ z† 3
5…2 ⫹ z†5 …91†
where
)
1 Z0 …mk ⫹ f † z2n⫺1
fˆ ; gˆ ˆ …84† ⫹ ⫹ … ;
m 1 ⫺ Z0 m…1 ⫺ k† …2n ⫺ 1†…2 ⫹ z†2n⫺1

The integral Iu in the LHS of Eq. (83) is evaluated by


performing an asymptotic expansion for large values of
the Zeldovich number

T~ d …T~ ∞ ⫺ T~ 0 †
bˆ …85†
T~ ∞2

and retaining up to the second term in the expansion. That is,


by using the relation

T~ d T~ n o
ˆ d 1 ⫹ a…1 ⫺ u† ⫹ a2 …1 ⫺ u†2 ⫹ …
T~ T~ ∞
"   #
b 3 1 2
⬇ ⫹ a2 …1 ⫺ u† ⫹ ; …86†
a 4 2a

we have
p  
p 3b
Iu ⬇ p exp ⫺
2b a=b 4a
    
1 1 ⫹ 2a…1 ⫺ um †
 erfc p ⫺ erfc p …87†
2 a=b 2 a=b

where Fig. 17. Temperature profile in the reaction zone, with the mixture
ratio m taken as a parameter [37]. Solid curves are numerical results
T~ ∞ ⫺ T~ 0 and dashed curves the assumed profiles to obtain the approximate
aˆ ; …88†
T~ ∞ analytical expressions for the mass burning rate eigenvalue L 0.
28 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

p
Fig. 18. Nondimensional burning velocity L0 for Ti–C system [37]; (a) as a function of the mixture ratio m ; and (b) as a function of the degree
of dilution k . The initial temperature T0 is taken as a parameter. Solid curves are numerical results and dashed curves are those with analytical
expression.

the integral Ij in the RHS of Eq. (83) is expressed as initial temperature T0, the mixture ratio m , and the degree of
! " dilution k are operating parameters when the experimental
1 ⫹ um 1 …f ⫺ 1†
Ij ⬇ …f ⫹ g† 2 ⫺ ⫹ conditions are specified.
2 5…f ⫹ g† 2a Keeping only the leading terms in Iu and Ij in Eqs. (87)
( ! p ) and (92), respectively, the functional dependence of L 0 on
…1 ⫹ a† 3 p ⫺1 2 ⫺ a 3
 ln ⫺ 2 3 tan p ⫺ p the system parameters can be displayed as
f 3a 3
m 2 …1 ⫺ k†2 exp…⫺b=a†
…f ⫹ 2g ⫹ 1† L0 ⬃ …94†
⫹ …m ⫹ fst †‰…1 ⫹ fst † ⫺ k…1 ⫺ m†Šb…1 ⫹ um †
6b
( ! p )# As far as m and k are concerned, L0 ⬃ m2 =…m ⫹ fst † for
…b ⫺ 1†3 p ⫺1 2 ⫹ b 3
× ln ⫹ 2 3 tan p ⫺ p ; k ˆ 0; and L0 ⬃ …1 ⫺ k†p 2
for m ˆ 1: That is, the nondimen-

f ⫹ 2g 3b 3 sional burning velocity L0 decreases with decreasing m
…92† and 1 ⫺ k: Its quantitative dependence on m and 1 ⫺ k;
however, is more complicated because volumetric heat
where a3 ˆ f ⫺ 1; and b3 ˆ f ⫹ 2g ⫹ 1:
generation, which is a function of m and 1 ⫺ k; also influ-
Then, from Eqs. (87) and (92), we have
  ences a and b through T∞. The effect of T0, which deter-
r Iu I
L0 t;0 ˆ ˆ …1 ⫺ Z 0 † u ; …93† mines T∞, may also be significant because T∞ can
rt …1 ⫹ g†Ij Ij exponentially influence L 0 through a and b . Note that b
also depends on the choice of the mixture components.
which is found to be surprisingly close to that obtained by a
When we compare L 0, Ti–C for Ti–C system with L 0, Ti–B
numerical calculation. This good agreement further demon-
for Ti–B system, (L 0, Ti–B/L 0, Ti–C) 1/2 is about 4 for m ˆ
strates the importance of the describing processes well at the
1; k ˆ 0; and T0 ˆ 450 K; because of the smaller value of
hot boundary, rather than the cold boundary (cf. Fig. 17).
b for Ti–B system.
Since Eq. (93) is relatively simple, its use could facilitate
analysis and interpretation.
4.6.3. Comparisons with numerical results
4.6.2. General behavior Applicability of the analytical expression has been exam-
An explicit form of L 0 provides useful insight into the ined by comparing it with the numerical results [32,47] for
effects of the reduced parameters (a , b , u m, f , and g ), TiC, TiB2, ZrB2, HfB2, and TiCo systems in Sections 4.4 and
which depend on the system parameters (fst, Td, Tm, T0, m , 5.5, and a fairpagreement
 has been reported in Ref. [37]. Fig.
and k ) used in the experiments. Among these parameters, 18(a) shows L0 for Ti–C system as a function of m , with
the stoichiometric mass ratio fst, the activation temperature T0 taken as a parameter. Solid and dashed curves are respec-
Td in the mass diffusivity, and the melting point Tm of the tively numerical and analytical results.pExcept
 near stoichio-
lower melting-point metal are physicochemical parameters metry …m ˆ 1†; both results show that L0 decreases until m
when the two components of the mixture are specified. The reaches the limit given by Eq. (77), and that they mostly
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 29

p
Fig. 19. L0 for Ti–B system. Notation is the same as that in Fig. 18.

agree well with each other. Although the trend in the analy- then, we have
tical expression near m ˆ 1 is opposite to that of the numer-  
r I fI u
ical result, the quantitative comparison can be considered to L 0 t;0 ˆ …Z∞ ⫺ Z0 † u ˆ ; …96†
be fair (⬍7% error). This discrepancy is attributed to the rt Ij …1 ⫹ g†Ij
overestimated temperature profile given by Eq. (90), which
causes an overestimation of Ij , and hence an underestima- where the relation has been made use of in Eq. (61). The
tion of L 0. Fig. 18(b) shows the similar comparison with integral Iu is the same as that in Eq. (87), while the integral
respect to k . The analytical expression again agrees well Ij is expressed as [40]
with the numerical results, to be within 7%. Figs. 19 and ! !
1 ⫹ um f⫹g
20 show similar comparisons for Ti–B and Co–Ti Ij ⬇
systems. The extent of agreement is similar to that for 2 f2
Ti–C system. "
…3 ⫺ 5f† ⫺ 3…1 ⫺ f†a2 …f ⫹ g†
× 2…1 ⫺ a2 † ⫹ ⫹
10…f ⫹ g† 3b
4.6.4. Approximate expression for m ⱖ 1 ( !3 !
2…f ⫹ g† b⫺1 p 2⫹b
When the mixture ratio is over-stoichiometric, the  ln ⫹2 3 tan⫺1 p
normalized concentration z of M-species is f ⫹ 2g a⫹b 3b
!)#
Z∞j p ⫺2a ⫹ b
z⬇ ; …95† ⫺ 2 3 tan⫺1 p : (97)
Z 0 ⫹ …Z∞ ⫺ Z0 †j 3b

p
Fig. 20. L0 for Co–Ti system. Notation is the same as that in Fig. 18.
30 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

p
Fig. 21. L0 for Ti–C system as a function of m when T0 ˆ 450 K; with k taken as a parameter. Solid curves are numerical results and dashed
curves are those with analytical expressions.

p
Dependence of L0 for Ti–C system on m in its wide range this aim, it is required to extend the adiabatic theory to allow
is shown in Fig. 21, with k taken as a parameter. The extent for the volumetric heat loss, to properly describe flame
of agreement in m ⱖ 1 is similar to that in m ⱕ 1: extinction and the various parameters that characterize it.
It will be emphasized here that not the reaction, but the
4.7. Some remarks on the flame propagation in the adiabatic temperature sensitive Arrhenius liquid-phase mass-diffusion
condition is responsible for the flame extinction, because in the SHS
process, combustion is basically diffusion controlled.
Calculated results for several ceramics and intermetallic
compounds, such as Ti–C, Ti–B, Zr–B, Hf–B, Al–Ni, Co–
5.1. Governing equations and boundary conditions
Ti, Ni–Ti systems, using consistent sets of physicochemical
parameters, respectively, show satisfactory agreement with A one-dimensional model with heat loss has been used to
experimental results in the literature, if we use the hetero- discover the dominant parameters which determine the limit
geneous theory in the adiabatic condition. An important of steady flame propagation. Although heat loss perpendi-
relation, that the burning velocity is inversely proportional cular to the propagation direction might lead to the curva-
to the nonmetal particle size, has been discovered. In ad- ture of the combustion wave, a choice of this model
dition, these comparisons suggest that the nonmetal (or simplifies the formulation and hence meets the present
higher melting-point metal) particles in these systems purpose. Therefore, regardless of the existence of heat
react in the diffusion-controlled limit, indicating that finite loss, the governing equations (except for the energy conser-
rate chemical reaction is not the controlling factor in the vation equation) have the same form as those in the adia-
propagation of the bulk flame. Furthermore, it has been batic condition, that is Eq. (46) for N-consumption and Eq.
suggested that concentration of the molten metal is mainly (49) for M-conservation [64]. The energy conservation
affected through consumption, because of the excessively equation is now not Eq. (57), but Eq. (58).
slow rate of diffusion compared to the thermometric Defining the nondimensional variable and parameter
conductivity and reaction rate, due to excessively large  
Lewis number in the combustion wave. These identifica- …Z ∞ ⫺ Z0 †ma …l=c†L
sˆ x; F…u† ˆ 2 0 …98†
tions further enable us to obtain an approximate expression …l=c† m a q …T~ ∞ ⫺ T~ 0 †;
for burning velocity, which can show clear dependence of
in addition to those in Eqs. (66), (71) and (72), we have the
the burning velocity on the dominant parameters.
governing equations as
" #
dz ~  m 
Le 0 …rt;0 =rt † exp…T~ d =T†
5. Flame propagation in the nonadiabatic condition ˆ
ds Z0 ⫹ …Z∞ ⫺ Z0 †j ma
The heterogeneous theory [32] described in Section 4 is    
Z∞ Z0z
an adiabatic one, and hence is incapable of describing the  z⫺ j⫹ ; …99†
experimentally observed phenomena of flame extinction due Z∞ ⫺ Z0 Z∞ ⫺ Z0
to heat loss from the high-temperature combustion process.    
From a practical point of view, extinction is to be properly d2 u m d u⫺j F…u†
⫺ ˆ ; …100†
described and hence controlled in the SHS process. Toward ds 2 m a ds Z ∞ ⫺ Z0 …Z ∞ ⫺ Z0 † 2
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 31
" #
dj x~
ˆ
ds …m=m a †L0 …rt;0 =rt † exp…T~ d =T†
~
    1=3
Z∞ ⫺ Z0
 1⫺ j : …101†
1 ⫺ Z0
Eqs. (99)–(101), together with the mass burning rate m, are
to be solved subject to the boundary conditions
sˆ0: u ˆ u m ; z ˆ 0; j ˆ 0; …102†

du F…u† dz dj
s!∞: ˆ⫺ ; ˆ 0; ˆ 0:
ds …m=m a †…Z∞ ⫺ Z0 † ds ds
…103†
In the same manner as that in the adiabatic flame propa- Fig. 22. The representative profiles of the diffusion, consumption,
gation, the cold boundary difficulty is eliminated by speci- convection, and loss terms in the energy conservation equation,
fying that the reaction is initiated at the melting point of M. along the nondimensional distance [64]. Conditions are the same
The downstream boundary condition for u , based on as those in the nonadiabatic situation in Fig. 5; m=ma ˆ 0:6 and C ˆ
convection-loss balance, requires a short domain of integra- 3:347 × 10 ⫺2 :
tion, and is preferred over the statement of T ˆ T 0 …u ˆ 0†;
which would require an excessively large domain of inte-
gration. If we further assume the heat loss to be Newtonian,
it is expressed as F…u† ˆ C·u; where C is called the heat Here, ma,st is the mass burning rate for the stoichiometric
loss parameter. mixture in the adiabatic condition. Results exhibit the char-
acteristic extinction turning point behavior, with the upper
5.2. Characteristics in the nonadiabatic condition branch being the stable solution and the turning point desig-
nating the state of extinction. With decreasing m or increas-
Values of the physicochemical parameters employed are ing k , the mass burning rate is gradually reduced, until
those for Ti–C system and the Lewis number Le0 is set to be extinction occurs. Since not only m/ma,st but also (ma/
100 [32] as described in Section 4.4. Representative flame ma,st) 2·C strongly depends on the mass burning rate m, the
structure for the stoichiometric mixture …m ˆ 1† without parameters m/ma and C are usually used to obtain the
dilution …k ˆ 0† has already been shown in Fig. 5 by solid extinction condition; the values of (m/ma)cr corresponding
curves, when the nonadiabatic situation is close to the to flame extinction ranges from 0.57 to 0.70, depending on
extinction state. In contrast to the adiabatic case in which m and k .
temperature increases monotonically to the adiabatic flame Here it may be informative to note the significant impli-
temperature, the temperature gradually decreases to the cation of Eq. (81). Since extinction is frequently associated
ambient value after the maximum flame temperature. with the nonlinear, highly temperature-sensitive nature of
Decrease in the temperature u reduces the particle consump- the Arrhenius kinetics, the disappearance of its influence in
tion rate dj /ds and hence the normalized burning-rate m/ma. the diffusion-controlled limit implies that extinction as
The reactant consumption zone is then broadened due to shown in Fig. 23 is actually caused by an alternate
heat loss. factor, which also has a nonlinear influence on flame
By obtaining profiles of the four terms constituting Eq. propagation. This factor is the liquid mass-diffusivity
(100), which, respectively, represent diffusion d 2u /ds 2, D ‰ˆ D 0 exp…⫺Td =T†Š which varies in a sensitive, nonlinear,
convection ⫺‰…m=ma †=…Z∞ ⫺ Z0 †Š…du=ds†; consumption Arrhenius manner with temperature. Thus in the SHS
‰…m=ma †=…Z∞ ⫺ Z0 †Š…dj=ds†; and heat loss ⫺F…u†=…Z ∞ ⫺ process, even in the diffusion-controlled limit, combustion
Z0 †2 ; a further examination has been made to identify the behavior, as well as the particle regression rate, is still an
flame structure for the nonadiabatic case in Fig. 5. As shown Arrhenius function of temperature, with the large activation
in Fig. 22, the combustion wave consists of a relatively thin temperature for diffusion, Td, exerting the dominant
zone of diffusion and consumption/convection, which is temperature sensitivity.
followed by a much thicker zone of convection and heat Fig. 24(a) and (b), respectively, show the effect of m and
loss. The existence of an extended zone of downstream k on the critical heat loss parameter Ccr ; with the initial
heat loss is similar to the nonadiabatic homogeneous temperature T0 taken as a parameter. The change in Ccr is
gaseous flame. gradual when either m or k is away from the flammability
Fig. 23 shows the variation of the normalized mass burn- limit, while Ccr decreases rapidly as this limit is approached.
ing rate m/ma,st as a function of the normalized heat loss It is also seen that Ccr increases with increasing T0 for a given
C ⴱ ‰ˆ …m a =ma;st †2 ·CŠ with m and k taken as parameters. m or k . This suggests that even if a heat loss exists, flame
32 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 23. Normalized mass burning rate m/ma,st as a function of the normalized heat loss C ⴱ ˆ …m a =ma;st †2 ·C; with the mixture ratio m and the
degree of dilution k taken as parameters when the initial temperature T 0 ˆ 450 K [64]. The values of numerically obtained (m/ma,st)cr
corresponding to the turning point are indicated by the horizontal solid lines; those of analytical ones are indicated by the dashed lines.

propagation can be sustained when T0 is high. Note that Eliminating s from Eq. (104) with Eq. (101) and conducting
there exists a maximum in C cr at high T0. integration, we have
    !
exp‰⫺b…1 ⫺ u†Š m 2 rt;0 T~ d
ˆ L0 exp
5.3. Turning-point determined by the thermal theory b ma rt T~ ∞
The heterogeneous SHS flame propagation has some Zj …105†
u⫺j
special features, as shown in Eqs. (81) and (82), which ×     1=3 dj;
0 Z∞ ⫺ Z0
can be contrasted with the homogeneous gaseous flame …Z ∞ ⫺ Z0 † 1 ⫺ j x~
1 ⫺ Z0
propagation. Nevertheless, an explicit relation for the state
of extinction can be derived [64] in the same manner as that where the relation has been used in Eq. (86), retaining up to
for the homogeneous premixed flame [65]. the first term. When je ⬇ 1 at the downstream edge of the
In the thin diffusion consumption/convection zone, consumption zone, by taking account of the adiabatic con-
contribution of the heat loss term is very small, such that dition, that is u ˆ j ˆ 1 and m ˆ m a ; we obtain the follow-
Eq. (100) becomes ing relation to the lowest order:
    
du m u⫺j m 2
⬇ : …104† ˆ exp‰⫺b…j e ⫺ ue †Š: …106†
ds ma Z ∞ ⫺ Z0 ma

Fig. 24. Critical heat loss parameter C cr [64]: (a) as a function of the mixture ratio m when the degree of dilution k ˆ 0; and (b) as a function of
k when m ˆ 1:0: The initial temperature T0 is taken as a parameter.
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 33

Fig. 25. Comparisons between numerical and analytical solutions for the critical value of the mass burning rate ratio (m/ma)cr [64]; (a) as a
function of m ; and (b) as a function of k . The initial temperature T0 is taken as a parameter.

On the other hand, if we integrate and evaluate Eq. (100) at where


the downstream edge of the consumption zone, we have
hrad ˆ es SB …T 2 ⫹ T02 †…T ⫹ T0 † …111†
…V=b†
…j e ⫺ ue † ˆ ; …107† is a radiative heat-transfer coefficient, e the emissivity, and
…m=m a † 2
s SB the Stefan–Boltzmann constant. As for the heat loss
where (V /b ) is expressed as parameter C , it is expressed as
  Zu e  " # 2!
V f…u† l 4es SB …T 2 ⫹ T02 †…T ⫹ T0 † R0
ˆ du ⫹ f…ue †; …108† Cˆ : …112†
b 0 u rtc …rt c†…u0;a R0 †2 2r

by the use of the upstream relation …du=ds† ˆ …m=m a †  Note here that C depends not only on the physicochem-
‰u=…Z∞ ⫺ Z0 †Š and the downstream boundary condition ical parameters, but also on the particle radius R0 and the
(103). compact diameter 2r. It is also inferred that R 02 =…2r†; which
Combining Eqs. (106) and (108), we obtain the relation is sometimes called the “Heat Loss Index”, can be a useful
    parameter in correlating size effects related to the heat loss,
m 2 m 2 from a practical point of view. This parameter R20 =…2r† can be
ln ˆ ⫺V; …109†
ma ma understood as follows, from a physical point of view; the
increase in R0 decreases the total surface area of N particles
from which we have …m=m a †cr ˆ e ⫺1=2 ˆ 0:6065 at the turn-
in the combustion wave, thereby reducing the volumetric
ing point, corresponding to the state of extinction. Fig. 23
heat generation rate; the decrease in r increases the
shows that this analytical extinction state fairly represents
surface-volume ratio of the compact, thereby enhancing
the numerical value, as the first approximation.
heat loss rate.
Fig. 25(a) and (b) further show (m/ma)cr as functions of m
and k , with T0 taken as a parameter. Although the numerical
and analytical solutions for (m/ma)cr differ by 15%, this 5.4.1. Range of flammabilty and extinction limit
difference is reduced to less than 4% in determining C cr. Predicted results are compared with experimental data for
A more precise treatment will be given in Section 5.6. Ti–C system [64]. In addition to the values of physicochem-
ical parameters, described in Section 4.1, use has been made
of l=…rt c† ˆ 10⫺5 m2 =s [41], e ˆ 0:4 [41], and s SB ˆ
5.4. Experimental comparisons for Ti–C system 5:67 × 10 ⫺11 kJ=…m2 :s:K4 †: The compact radius is set to be
r ˆ 9 mm and its relative density 0.6, in accordance with the
There exist several sources of experimental data which
experiments [33,43]. As for the heat loss parameter C , it is
can be compared with the calculated results, as in
evaluated at the maximum temperature of the combustion
Refs.[64,66], by keeping as many parameters as fixed
wave.
possible. For the heat loss, only thermal radiation is consid-
Fig. 26(a) shows the range of flammability when T 0 ˆ
ered, because of the high temperature in the consumption
450 K; the abscissa is the mixture ratio m and the ordinate
zone, say, higher than 2000 K. Then the lateral heat loss for
the particle radius R0. Extinction is anticipated to occur
a cylindrical compact with radius r can be expressed as
when C in Eq. (112) reaches the critical heat loss parameter
 
4h rad C cr in Fig. 24(a). We see that while the range of flamm-
Lˆ …T ⫺ T0 †; …110†
2r ability spans from 0.448 to 1 when R 0 ! 0; it contracts with
34 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 26. Nonadiabatic combustion behavior for Ti–C system [64]; Fig. 27. Nonadiabatic combustion behavior for Ti–C system [64].
solid curves are calculated from theory; data points are experimental (a) Range of flammability with respect to the degree of dilution k
[33,43]. (a) Range of flammability with respect to the mixture ratio and the particle radius R0; notation is the same as that in Fig. 26(a).
m and the particle radius R0; (W) designates the steady propagation, (b) u0·R0 as a function of k , with R0 taken as a parameter.
(Z) the pulsating combustion, (K) the flame extinction during the
propagation, and ( × ) the non-ignition. (b) SHS rate-constant u0·R0
as a function of m , with R0 taken as a parameter. dilution k is shown in Fig. 27(a), and Fig. 27(b) shows u0·R0
as a function of k . The same comments as those for Fig.
26(a) and (b) can be made.
increasing R0. It also shows that the combustion wave Fig. 28(a) shows the range of flammability for various m
cannot propagate through a compact when R0 is larger and k , with R0 taken as a parameter [66]. The range of
than 20.9 mm. Experimental results [33,43] are also shown flammability becomes narrow with increasing R0. Exper-
in Fig. 26(a). Outside the range of flammability, even if the imental results [33,43] in Fig. 28(a) are those of R0 ˆ
flame propagation might be initiated by a powerful ignition 5 mm: Fig. 28(b) shows a similar comparison [40] for the
source, it loses its self-sustained nature and is extinguished dependence of the range of flammability on m in its wide
in the course of the propagation. Fair agreement is demon- range when R0 ˆ 12:5 mm: In the range from m ˆ 1 to 2, the
strated for particles smaller than 7.5 mm, while discrepancy range of flammability with respect to k is wider than that for
becomes large for R0 ˆ 12:5 mm: This discrepancy is attrib- the stoichiometric mixture. This is attributed to the increase
uted to the effect of smaller particles, because of the exis- in u0,a·R0 (cf. Fig. 9), which reduces C , and hence retards the
tence of particle-size distribution. It is also pointed out that flame extinction. Discrepancy between the predicted and
the extinction limit obtained here corresponds to that for the experimental results for the under-stoichiometric mixtures
steady flame propagation. When the process occurs in an …m ⬍ 1† is attributed to the effect of smaller particles
unsteady mode, due to disappearance of stability of the because of the existence of particle size distribution. Dis-
steady propagation, the limit of self-sustained combustion crepancy for the over-stoichiometric mixtures …m ⬎ 1†
could be modified accordingly. Fig. 26(b) shows the SHS might be attributed to an initiation of a radial crack, which
rate-constant u0·R0 in the adiabatic condition as a function of inevitably appears in pulsating combustion [1–3] resulting
m . The agreement is again fair, as far as the trend and in a laminated structure with separable layers.
approximate magnitude are concerned, suggesting that Munir and Lai [67] measured the boundaries of the SHS
u0·R0 is one of the important parameters in correlating for Ti–C system at several ambient (or initial) temperatures.
experimental results. Their test specimen (19 mm in diameter and 0.65 in relative
The range of flammability with respect to the degree of density) is composed of tens of layers (5 mm in thickness)
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 35

with increasing initial temperature T0, as expected, due to


preheating. This suggests that even if there is heat loss,
flame propagation can be sustained when T0 is high. It is
also seen that for a given T0, the range of flammability
narrows with increasing R0.
Shkiro and Borovinskaya [26] reported experimental data
for Ti–C system with varying degree of dilution k and
compact diameter 2r. The relative density is 0.56. Fig.
30(a) shows the range of flammability as functions of k
and 2r, with T0 taken as a parameter. The carbon particle
size and the initial temperature are estimated to be R 0 ˆ
2:5 mm and T0 ˆ 300–450 K; respectively, by examining
the dependence of the reported burning velocity on k and
T0. It is seen that with increasing k , the compact diameter in
which the combustion wave can propagate increases first
gradually and then rapidly. Further, an increase in T0
expands the range of flammability. Since it has been found
[66] that their experimental data correspond well to the
calculated result for T0 ˆ 300 K when k is less than about
0.2, while the comparison is better for T0 ˆ 450 K near the
limit of flammability, by examining the dependence of u0 on
k and T0, further confirmation has been made by investigating
the effect of the compact diameter 2r on the burning velocity
u0. Fig. 30(b) shows that u0 decreases with decreasing 2r

Fig. 28. Range of flammability for Ti–C system as functions of m


and k , with R0 taken as a parameter: (a) for R0 ˆ 5 mm [66]; (b) for
R0 ˆ 12:5 mm [40]. Data points are experimental [33,43] and nota-
tion is the same as that in Fig. 26(a).

with a gradual increase in the degree of dilution. The average


diameter of the carbon particles is reported to be 5 mm. Fig. 29
shows the limit of dilution with the radius of carbon particles,
R0, taken as a parameter. The range of flammability expands

Fig. 30. Nonadiabatic combustion behavior for Ti–C system [66].


(a) Range of flammability as functions of k and the compact
Fig. 29. Range of flammability for Ti–C system as functions of k diameter 2r, with T0 taken as a parameter, when R0 ˆ 2:5 mm;
and T0, with R0 taken as a parameter [66]. Data points are exper- data points are experimental [26] and notation is the same as that
imental [67]; (W) designates the steady propagation, (L) the in Fig. 26(a). (b) Burning velocity u0 as a function of 2r, with k
unsteady propagation, and ( × ) the flame extinction. taken as a parameter; T0 ˆ 450 K:
36 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

because the heat loss rate is enhanced due to an increase in constant u0·R0 and the heat loss index R20 =…2r† is demon-
the surface-to-volume ratio. Fair agreement is demonstrated strated, as the correlating parameters for experimental
between the experimental and predicted results. results.

5.5. Experimental comparisons for other systems


5.4.2. Burning velocity under heat loss condition
Here, let us investigate the effect of the heat loss on the The nonadiabatic theory can also be applied to other
burning velocity. As pointed out in Section 5.4, not only the experimental results obtained from the synthesis of several
physicochemical properties, but also sizes of N particles and borides and intermetallic compounds for which reliable
test specimen influence C through Eq. (112), hence the experimental data are available. Through extensive compar-
burning velocity. Fig. 31(a) shows the SHS rate-constant isons [64,66] for the range of flammability with respect to
u0·R0 as a function of 2r=R 02 ; which is the inverse of the effects of dominant parameters (m , k , T0, R0, and 2r) on the
heat loss index, for T0 ˆ 450 K and k ˆ 0; with m taken burning velocity u0, it has been demonstrated that the theory
as a parameter [40]. With decreasing 2r=R20 ; u0 ·R0 decreases is indeed viable, thereby allowing the synthesis of the
because of the increase in the heat loss rate, and reaches the diverse set of experimental results within a coherent frame-
extinction limit shown by a dashed curve. We see that work of description. In the same manner as that for the Ti–C
effects of the compact diameter 2r and the particle radius system, only thermal radiative heat loss is considered,
R0 on the burning velocity u0 can generally be discussed by because of the high temperature in the combustion wave,
the use of parameters 2r=R20 and u0·R0. Experimental results say, higher than about 1000 K.
[33,43] for various R0 are also shown in Fig. 31(a). Fair
agreement is shown, as far as the general trend and approx- 5.5.1. Boride synthesis
imate magnitude are concerned. Fig. 31(b) shows a similar Munir and Sata [68] measured the boundary of SHS in the
comparison for the stoichiometric mixture …m ˆ 1†; with k boride synthesis of Ti, Zr, and Hf at several ambient (or
taken as a parameter. Again, usefulness of the SHS rate- initial) temperatures. The test specimen was prepared in
the same manner as that explained for Fig. 29, with amor-
phous boron of about 1 mm diameter and crystalline boron
particles whose diameter is less than 42 mm. Fig. 32(a)
shows the limit of dilution for Ti–B system [64] obtained

Fig. 32. Range of flammability for Ti–B system; (a) as functions of


Fig. 31. SHS rate-constant u0·R0 for Ti–C system as a function of 2r/ the degree of dilution k and the initial temperature T0, with the
R02 for T0 ˆ 450 K [40]: (a) with m taken as a parameter; and (b) initial particle radius R0 taken as a parameter [64]; and (b) as func-
with k taken as a parameter. Data points are experimental [33,43] tions of k and R0, with T0 taken as a parameter [66]. Data points are
for various sizes of N particles. experimental [68] and notation is the same as that in Fig. 29.
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 37

parameter. The same trend as that in Fig. 27(a) can be


observed and the range of flammability expands with
increasing T0. Experimental data with crystalline boron
particles have been plotted for R0 ˆ 5 mm because sieved
powders of ⫺325 mesh (⬍44 mm) size are estimated to have
about 10 mm in diameter. The deviation at large particles
could be caused by the difference of crystallization and/or
the existence of small particles, as well as the experimental
determination of T0, as inferred in Section 4.5.
Figs. 33 and 34 show similar comparisons for Zr–B
system [45,50,68] and Hf–B system [68]. Extent of agree-
ment is similar to that for the Ti–B system.

5.5.2. Synthesis of intermetallic compounds


Naiborodenko and Itin [58] measured the burning velo-
cities for stoichiometric Al–Ni system by varying the
diameter of compact whose relative density is 0.4. The
diameter of the Ni powder was reported to be of the order
of 20 mm. Fig. 35 shows that u0 decreases with decreasing
compact diameter because of the enhanced heat loss rate;
the boundary of the steady propagation is shown by a dashed
curve [66]. The agreement is fair for R0 ˆ 15 mm:
Fig. 33. Range of flammability for Zr–B system [66]; (a) as func- Itin et al. [53] measured the burning velocities for Co–Ti
tions of k and T0, with R0 taken as a parameter; data points are system by varying the compact diameter and the particle
experimental [68] and notation is the same as that in Fig. 29; (b) size of Ti. The compact for the stoichiometric mixture had
as functions of k and R0, with T0 taken as a parameter; datum for
a relative density of 0.6; diameters of Ti particles were from
k ˆ 0 is from Hardt and Holsinger [45]; data for k ˆ 0:3 and 0.5 are
100 to 160 mm; initial temperature was 573 K. Fig. 36(a)
from Zenin et al. [50].
and (b), respectively, show the range of flammability and the
burning velocity [66]. We see that the range of flammability
by the use of physicochemical parameters in Table 2 and the expands with increasing compact diameter and that the
transport parameters in Table 3. The same trend as that in burning velocities agree well for the 130 mm particles,
Fig. 29 is observed. Data points are experimental with boron which is representative of the average particle diameters.
particles of about 1 mm diameter. It is pointed out that the
limit of flammability observed by Munir and Sata [68] could
have been affected by other effects not considered in the
theory such as agglomeration of the particles, as inferred
in Section 4.5.2.
Fig. 32(b) shows the range of flammability for Ti–B
system [66] as functions of k and R0, with T0 taken as a

Fig. 35. Burning velocity u0 in the nonadiabatic condition for Al–Ni


system as a function of the compact diameter 2r, with the initial
Fig. 34. Range of flammability for Hf–B system as functions of k particle radius R0 taken as a parameter [66] the mixture is stoichio-
and T0, with R0 taken as a parameter [64]. Data points are exper- metric and the initial temperature T0 ˆ 300 K; data are from
imental [68] and notation is the same as that in Fig. 29. Naiborodenko and Itin [58].
38 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 37. Nonadiabatic combustion behavior for Ni–Ti system with


stoichiometric mixture at T0 ˆ 573 K [66]. (a) Range of flammabil-
Fig. 36. Nonadiabatic combustion behavior for Co–Ti system with ity as functions of 2r and R0; data are from Bratchikov et al. [60];
stoichiometric mixture at T0 ˆ 573 K [66]. (a) Range of flammabil- datum for R0 ˆ 25 mm is from Itin et al. [71]; notation is the same
ity as functions of 2r and R0; data are from Itin et al. [53] and as that in Fig. 29. (b) Burning velocity u0 as a function of 2r, with R0
notation is the same as that in Fig. 29. (b) Burning velocity u0 as taken as a parameter; data points are experimental [60].
a function of 2r, with R0 taken as a parameter; data points are
experimental [53].
for the temperature, as the outer solutions, in the upstream
Bratchikov et al. [60] measured the burning velocities for and downstream regions where the consumption term dj /ds
stoichiometric Ni–Ti system with a relative density of 0.6 is exponentially small:
by varying the compact diameter and the particle size of Ti.    
C C a0
The initial temperature was 573 K. Fig. 37(a) shows the u ˆ 1 ⫹ 1 ⫹ 22 ⫹ … exp s …s ⬍ 0†;
range of flammability as functions of 2r and R0 [66]. Particle b b …Z∞ ⫺ Z0 †
size of Ti is estimated from their related paper [70] which …113†
states that Ti particles were from 100 to 160 mm in diameter.
     
Fig. 37(b) shows that the burning velocities are correlated C C 1 bC
fairly well with the predictions for R0 ˆ 100 mm: u ˆ 1 ⫹ 1 ⫹ 22 ⫹ … exp ⫺ s
b b …Z∞ ⫺ Z0 † a 0

5.6. Approximate expression of the heat loss parameter …s ⬎ 0†: (114)

Extinction of the combustion wave in the SHS process Here, b is the Zeldovich number defined in Eq. (85),
has been obtained numerically and hence presented par-
s m a a
ametrically. An approximate analytical expression for the s ˆ ; ˆ a0 ⫹ 1 ⫹ 22 ⫹ …; …115†
extinction criterion [72] which explicitly displays its func- b ma b b
tional dependence on the various system parameters, is also
and C1 ; C2 ; … are constants.
obtainable by referring to an analytical criterion for the
As for the inner solution, with the inner variables being
“premixed” SHS flame propagation obtained by Kaper et
al. [73] by the use of matched asymptotic expansion method. t1 t y y
uˆ1⫹ ⫹ 22 ⫹ …; j ˆ y 0 ⫹ 1 ⫹ 22 ⫹ …;
The analysis and results presented herein is an extension of b b b b
those described in Section 5.3. …116†
h
sˆ ;
b
5.6.1. Asymptotic expansion analysis
Following Ref. [73], we have the following expressions we obtain the following equations for the first-order inner
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 39

structure have
!
d 2t1 a0 dy0 2bC
⫹ ˆ 0; …117† exp ⫺ ˆ a20 k…1 ⫺ M† ⫹ M; …125†
dh2 …Z∞ ⫺ Z0 † dh a 02

  "   # which has nearly the same form as that presented by Kaper
r dy 3b b 1 at 2 et al. [73] except for the factor k. The parameter M identified
bL0 t;0 a0 0 ˆ exp ⫺ ⫺ ⫺ 1
rt dh 4a a 2 b in Ref. [74] is called the melting parameter because it corre-
2   3 sponds to the energy to heat the unburned mixture from T0 to
Z∞ ⫺ Z0 Tm. Consequently, 1 ⫺ M is the normalized driving force for
   1=3 f ⫺ y
Z∞ ⫺ Z0 6 1 ⫺ Z0  0 7
× 1⫺ y0 ln 6
41 ⫹  7;
5
the adiabatic flame propagation. The turning point
1 ⫺ Z0 Z ⫺ Z0
g⫹ ∞ y0 corresponding to the flame extinction, then, appears at
1 ⫺ Z0 dC=da 02 ˆ 0; which yields
…118†
Xcr …1 ⫹ ln Xcr † ˆ M; …126†
where a is defined in Eq. (88) and f ˆ 1=m: where ln Xcr ˆ ⫺2bC cr =…a0 †cr2 ;
as well as the following rela-
Matching the outer and inner solutions in the limit of h ! tions for the critical parameters at extinction:
^∞; we have
⫺Xcr ln Xcr X …ln Xcr † 2
a0 …a0 †2cr ˆ ; C cr ˆ cr : …127†
h ! ⫺∞ : t1 ! C1 ⫹ h; y0 ! 0; …119† k…1 ⫺ M† 2bk…1 ⫺ M†
…Z∞ ⫺ Z0 †
Note that hte maximun flame temperature is given by
h ! ∞ : t1 ! C1 ; y0 ! 1; …120† umax ˆ 1 ⫹ …ln Xcr =b†:
Knowing the critical heat loss parameter C cr and the
where the constant C1 is given by the second-order inner maximum flame temperature u max, the critical size of N
problem as particles and the critical compact diameter for the extinction
are determined through
2bC !
C1 ˆ ⫺ : …121†
a 02 R 02 C cr …rt c†…u0;a R0 † 2
ˆ ; …128†
2r cr ‰l=…rt c†Š4es SB …Tmax 2 ⫹ T 2 †…T
0 max ⫹ T0 †
By using the matching conditions, Eq. (117) can be inte-
grated once to yield derived from Eq. (112) when only the thermal radiative heat
loss from the cylindrical surface of a compact with radius r
dt1 a0
ˆ …1 ⫺ y0 †: …122† is considered [64,66]. It is informative to recall that R20 is
dh …Z∞ ⫺ Z0 † closely related to the heat generation rate in the consump-
Eliminating h from Eq. (122) with Eq. (118) and conducting tion zone and 2r the heat loss rate. In addition, the SHS rate-
integration, we have constant in the adiabatic condition, which is the product of
" the burning velocity u0,a and the initial radius R0 of N
  #
…Z∞ ⫺ Z0 † ZC1 3b b 1 a t1 2 particles, is expressed as
exp ⫺ ⫺ ⫺ dt1 !
bL0 …rt;0 =rt †a20 ⫺ ∞ 4a a 2 b   
2 rM 1 ⫺ Z0 Iu
…u0;a ·R0 † ˆ 3Le0 D0
2
; …129†
Z1 …1 ⫺ y0 † rN Z∞ ⫺ Z0 Ij
ˆ 2   3 dy0 :
Z∞ ⫺ Z0
0
   1=3 f ⫺ y0 by combining Eqs. (75) and (96).
Z∞ ⫺ Z0 6 7
1⫺ y0 ln 6
41⫹  1 ⫺ Z0  7 5
1 ⫺ Z0 Z∞ ⫺ Z0
g⫹ y0 5.6.2. Correction term for the heat loss parameter
1 ⫺ Z0
…123† Predictions based on Eq. (126) have been found to deviate
substantially from the numerical results shown in Section
If we note here that L 0 is expressed in Eq. (93), that Iu in Eq. 5.4. The discrepancies are at least 50% for the heat loss
(87), and that Ij in Eq. (92) for m ⬉ 1 and Ij in Eq. (97) for parameter C cr, and are even more for the critical particle
m ⱖ 1; then Eq. (123) becomes size and/or compact diameter. Since this discrepancy is
" !#   expected to be caused by ignoring the O(b ⫺1) term in
2a 02 b
1⫺ bIu exp obtaining the RHS of Eq. (118), which would have modified
1 ⫹ um a the RHS of Eqs. (125) and (126) by an amount ⫺O‰…1 ⫺
Z0    M†=bŠ; we are therefore required to add the term (1 ⫺ M)/b
at
ˆ exp t1 1 ⫺ 1 dt1 ⬇ 1 ⫺ exp…C1 †: …124† to Eq. (126) as
C1 b  
1⫺M
X cr …1 ⫹ ln Xcr † ˆ M ⫺ : …130†
Putting k ˆ 2=…1 ⫹ um † and 1 ⫺ M ˆ bIu exp…b=a†; we b
40 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

have been used. If we further use …1 ⫹ um †=2 ⬃ 1; a ⬃ 1; and


T~ d …T~ ∞ ⫺ T~ 0 † T~ d T~ d
bˆ ⬃ ˆ
~
T∞2 ~
T∞ ⫺ T0 ~ Z∞ ⫺ Z0
  
1 ⫺ Z0 m ⫹ fst ~
ˆ T ; …135†
Z∞ ⫺ Z0 m…1 ⫺ k† d
we have the following relations with respect to m and k :
 ( ! )
1 m e⫺2 m
C cr ⬃ 1⫺ …m ⱕ 1†;
2eT~ d m ⫹ fst T~ d m ⫹ fst
…136†

 ( ! )
1 1 e⫺2 1
C cr ⬃ 1⫺ …m ⱖ 1†
Fig. 38. Critical value of the normalized heat loss, Xcr ˆ 2eT~ d m ⫹ fst T~ d m ⫹ fst
exp‰⫺2bC cr =…a0 †cr2 Š; as a function of the melting parameter M, …137†
with the measure of activation energy relative to thermal energy,
b , taken as a parameter [72]. Solid curves are numerical results and when there is no dilution …k ˆ 0†; when the mixture is
dashed curves are results by the explicit approximate analytical stoichiometric …m ˆ 1†; we have
expression.  ( ! )
1 1⫺k e⫺2 1⫺k
C cr ⬃ 1⫺ : …138†
2eT~ d 1 ⫹ fst T~ d 1 ⫹ fst
With this modification, the discrepancies can be reduced
to less than 20%. Furthermore, Xcr is now explicitly That is, C cr increases with increasing m when m ⱕ 1; while it
expressed in terms of M as decreases with increasing k or m when m ⱖ 1; although with
progressively diminishing rate. The effect of T0, which deter-
    1=2 p !
1 1⫺M 1 ⫹ 2e mines T∞, also appears because T∞ influences C cr through a
Xcr ˆ 1 ⫹ 2e M ⫺ ⫹ 1⫺ M2; and b .
e b e
…131†
5.6.4. Applicability of the analytical expression
which yields a discrepancy of less than 1% (cf. Fig. 38), in Applicability of the analytical expression has been exam-
the range of practical values of M. ined by comparing it with the numerical results for the
syntheses of TiC, TiB2, ZrB2, HfB2, and AlNi, because of
fair agreement between numerical and experimental results,
5.6.3. General behavior of the heat loss parameter as shown in Sections 5.4 and 5.5. Fig. 39(a) shows the
An explicit form of C cr provides useful insight into the critical value of the heat loss parameter C cr for Ti–C system
effects of the reduced parameters (a , b , and u m), which as a function of m , with k ˆ 0 and T0 taken as a parameter.
depend on the system parameters (fst, Td, Tm, T0, m , and Solid and dashed curves are respectively numerical and
k ), in the same manner as that for the mass burning rate analytical results. Fig. 39(b) shows C cr as a function of k ,
eigenvalue L 0 in Section 4.6.2. Keeping only the leading with m ˆ 1 and T0 taken as a parameter. The analytical
terms, the functional dependence of C cr on the system par- result mostly agrees with the numerical result, within 17%
ameters can clearly be displayed as error.
Fig. 40(a) shows the critical value of the heat loss index
   ‰R 02 =…2r†Šcr as a function of m , with k ˆ 0 and T0 taken as a
1 ⫹ um 2a…e ⫺ 1† ⫺ e
C cr ⬃ 1⫺ : …132† parameter. The agreements between the numerical and
4eb b
analytical results are generally good, except near the stoi-
Here the following relations chiometric state at which the analytical result shows a
reverse trend, resulting in about 18% under-prediction at
  m ˆ 1: The discrepancy here is attributed to the overestima-
b 2a
…1 ⫺ M† ˆ bIu exp ⬃1⫺ …133† tion in Ij in obtaining the approximate expression of L 0, as
a b
mentioned in Section 4.6. Fig. 40(b) shows ‰R 02 =…2r†Šcr as a
and function of k , with m ˆ 0:95 and T0 taken as a parameter.
When m is 0.95 or less, agreement is good, with less than
 
1 e…2a ⫺ 1† 14% error.
Xcr ⬃ 1⫹ …134† Figs. 41 and 42 show similar comparisons for Hf–B and
e b
Al–Ni systems, as examples, when the degree of dilution is
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 41

Fig. 39. Critical heat-loss parameter, C cr, for Ti–C system [72]; (a) as a function of the mixture ratio m , with the initial temperature T0 taken as a
parameter when the degree of dilution k ˆ 0; and (b) as a function of k , with T0 taken as a parameter when m ˆ 1: Solid curves are numerical
results and dashed curves are those with the analytical expression.

varied. The extent of agreement, including Ti–B and Zr–B features of the heterogeneity in the SHS process. A useful
systems, is similar to that for Ti–C system. parameter called the “heat loss index”, which is defined as
the square of the nonmetal particle size divided by the
5.7. Some remarks on the flame propagation under heat loss compact diameter, has also been identified. Furthermore,
condition under the appropriate and realistic assumptions of diffu-
sion-limited reaction and infinite Lewis number, an approx-
A nonadiabatic, heterogeneous theory for the SHS process imate criterion for the extinction has been derived, which can
with heat loss has been formulated and the range of flammabil- yield useful insight into manipulating the flame extinction.
ity can be determined from the characteristic extinction turn-
ing-point. A particularly significant identification is that the
process responsible for its temperature sensitivity and ultimate 6. Other aspects of the nonadiabatic flame propagation
extinguishment is the Arrhenius liquid-phase mass-diffusion.
Fair agreement between predicted and experimental results In this section, several other important effects, such as the
for the extinction limits, in trend and in approximate magni- bimodal particle distribution, representative length of the
tude, suggests that this theory has captured the essential cross-section of a compacted medium, on the SHS flame

Fig. 40. Critical heat-loss index [R02/(2r)]cr for Ti–C system [72]; (a) as a function of m , with T0 taken as a parameter when k ˆ 0; and (b) as a
function of k , with T0 taken as a parameter when m ˆ 0:95: Notation is the same as that in Fig. 39.
42 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 41. Critical values of parameters for stoichiometric Hf–B system as a function of k , with T0 taken as a parameter: (a) C cr; (b) [R02/(2r)]cr.
Notation is the same as that in Fig. 39.

propagation and/or extinction are described. Also described possibly more desirable, packing. Since not only the
here is the correspondence between the heterogeneous increase in the burning velocity but also the extension
theory and the homogeneous premixed-flame theory, from of the flammability limit is anticipated by reducing the
the analytical point of view. Effects of the electric field on particle size, for a fixed amount of the dispersed phase,
the flame propagation and/or extinction are also described in it is considered to be desirable to have small particle
this section. size for the propagation of the combustion wave.
However, preparation of the small particles can be
6.1. Bimodal particle dispersion quite energy intensive. Therefore, preparation of the
bimodal mixture, with only a fraction of the N particles
Analyses of the heterogeneous flame, hitherto having small sizes, can be less energy intensive without
described, have assumed a monodisperse distribution substantially affecting the burning intensity. Further,
of the nonmetal particles. However, there exist practical since SHS processes have been observed to possess a
interests in having a bimodal particle distribution. For variety of flame instability phenomena, which are invari-
example, such a distribution would allow smaller par- ably dependent on the burning intensity and hence the
ticles to fill in part of the inter-particle spacing between particle size, such instabilities could be modulated for
large particles, hence allowing for a denser, and bimodal mixtures. In view of the above considerations,

Fig. 42. Critical values of parameters for stoichiometric Al–Ni system as a function of k , with T0 taken as a parameter: (a) C cr; (b) [R02/(2r)]cr.
Notation is the same as that in Fig. 39.
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 43

formulation has been extended to bimodal systems the mass burning rate eigenvalue becomes
under nonadiabatic situations [75,76]. " #
…Z∞ ⫺ Z0 †2 m2a
L0 ˆ
4p…rM D0 †…l=c†…n0 R0 †E
6.1.1. Equivalent particle radius
  " #
The problem of interest is basically the same as that Z∞ ⫺ Z0 2 …1 ⫺ Z0 †u20;a …RL;0 =F†2
described in Section 5.1. The bimodal system for N particles ˆ …145†
1 ⫺ Z0 3Le0 D20 …rM =rN;L †
is considered to consist of small particles of density r N,S,
initial radius RS,0, and number density nS,0, and large parti- where RL,0/F is the equivalent particle radius and
cles with the corresponding parameters r N,L, RL,0, and nL,0. ! !2
In spite of the existence of the bimodal particle distribution, rN;L RL;0
1⫹ PG
the governing equations and boundary conditions are iden- rN;S RS;0
F ˆ
2
: …146†
tical to those for the monomodal case, if we introduce the …1 ⫹ G† 2=3 …1 ⫹ P3 G† 1=3
equivalent particle radius and number density as (n0R0)E and
express the conservation equation for the N-Consumption We see that u0,a·(RL,0/F) is independent of the bimodality,
formally as [75,76] and that Eq. (143) suggests P ˆ O…R S;0 =RL;0 †: Note that
G=…1 ⫹ G† is the mass fraction of small particles and can
    
d…1 ⫺ Z† 4prM D u 1 ⫺ Z 1=3 be called the degree of small particle content. As for the
ˆ⫺ …n0 R0 †E 0 x : heat loss parameter C , it is given as
dx m u 1 ⫺ Z0
…139†  " #" 2 #
l 4es SB …T 2 ⫹ T02 †…T ⫹ T0 † …RL;0 =F†2
Cˆ :
Thus, our primary concern is to find out the specific form of rtc …rt c†…u20;a …RL;0 =F†2 2r
(n0R0)E in Eq. (139), as a function of the bimodal par- …147†
ameters. The description for the bimodal system has first
been made for the system with the same kind of N particles It depends not only on the physicochemical parameters but
[75] and then extended for the system with different kinds of also on the equivalent particle radius RL,0/F and the compact
N particles [76]. diameter 2r.
The parameter (n0R0)E can be obtained by using the size
distribution function 6.1.2. Experimental comparisons for spherical carbon
particles
rN G0 …R† ˆ rN;S nS;0 d…R ⫺ RS;0 † ⫹ rN;L nL;0 d…R ⫺ RL;0 †; There exist several sources that can be compared with the
…140† numerical results. Makino et al. [43,77,78] using various
radii of spherical carbon particles, measured the range of
where d (·) is the delta function. By substituting it into the flammability and the burning velocity of the SHS process
general N-consumption equation in Eq. (45), we have for Ti–C system. The representative Lewis number of Le0 ˆ
" ! !2 # 25; which corresponds to l=…rt c† ˆ 5 × 10⫺6 m2 =s and has
rN;L RL;0 been determined through correlation similar to that in
…n0 R0 †E ˆ 1 ⫹ PG
rN;S RS;0 Section 4.4 with experimental results [43,77,78] (0.6 rela-
  1=3 tive density) of monomodal system, has been used in numer-
1⫹G ical calculations [75] because of the fair correlation for
 …n L;0 RL;0 † …141†
1 ⫹ P 3G various radii of spherical carbon particles and various
compact diameters, as shown in Figs. 43 and 44.
where
Fig. 43(a) shows the range of flammability for the mono-
r N;S nS;0 RS;0
3 modal system with respect to the mixture ratio m and the
Gˆ …142† heat loss index R20 =…2r†; with T0 taken as a parameter [79].
rN;L nL;0 R3L;0
Experimental results [43,77,78] for various particle radii
is the mass ratio, P ˆ …hS =RS;0 †=…hL =RL;0 † the size ratio, and (1.5, 2.5, 5, 10, and 15 mm) and various compact diameter
hi =Ri;0 …i ˆ S; L† the normalized particle size defined as (5, 7, 10, and 18 mm) at T0 ˆ 300 K show that not only the
range of flammability but also the range of pulsating
!2   combustion extends with decreasing R20 =…2r†: Fig. 43(b)
hi 2 Zx x
ˆ1⫺ 2 dx …i ˆ S; L†: …143† shows the similar comparison with respect to the degree
Ri;0 Ri;0 ⫺ ∞ u
of dilution k and R20 =…2r†: Fair agreement is demonstrated
If we further take account of the relation for the mass frac- between the predicted and experimental results.
tion of N particles as The choice of Le0 ˆ 25 for the spherical carbon particles
requires further comments because Le0 ˆ 100 has been used
 
4 in comparisons in Section 4.4 and 5.4. As a remarkable
rt;0 …1 ⫺ Z0 † ˆ prN;L …1 ⫹ G†…n L;0 RL;0
3
† …144†
3 difference, compared to the irregular particles in Refs.
44 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 43. Range of flammability for Ti–C system with monomodal particles; (a) with respect to the mixture ratio m and the heat-loss index R02/
(2r), with the initial temperature T0 taken as a parameter [79]; and (b) with respect to the degree of dilution k and the heat-loss index R02/(2r).
Data points are experimental at T0 ˆ 300 K [43,77,78]: (W) designates the steady-propagation; (Z) the pulsating combustion; (S) the quasi-
steady, multipoint combustion; (K) the extinction during the flame propagation, and ( × ) the non-ignition.

[33,43], the spherical particles are reported to have smooth function of G=…1 ⫹ G † for T 0 ˆ 300 K and m ˆ 1; with k
surfaces and very narrow size distribution [77]. This taken as a parameter. Owing to the decrease in the equiva-
suggests that the value of Le0 ˆ 100 for the irregular par- lent particle radius, u0,a increases first gradually and then
ticles implicitly accounts for the effects of surface roughness rapidly with increasing G=…1 ⫹ G †: At a certain G=…1 ⫹ G †;
and the relatively wide size distribution because these u0,a decreases with increasing k because of the reduced
effects, which appear in (n0R0)E in Eq. (139), are grouped amount of chemical heat generation. The dashed curve is
into l /c when only the particle size is known. the extinction limit under heat loss conditions. A fair degree
Fig. 44(a) shows the SHS rate-constant u0·R0 for the of agreement is shown.
monomodal system as a function of the inverse of the heat The appropriateness of the equivalent particle radius can
loss index R20 =…2r† at T0 ˆ 300 K; with m taken as a para- further be shown in Fig. 46. It is, however, important to note
meter [79]. The trend is the same as that in Fig. 31(a). Data that since the present formulation is based on the statistical
points are experimental [43,77,78] and the agreement is analysis, a few tens of large particles should exist at least
again considered to be satisfactory. Fig. 44(b) shows a simi- across the consumption zone, even for the bimodal system.
lar comparison, with k taken as a parameter. Violation of this restriction results in the breakdown of the
Having confirmed the appropriateness of the choice of present formulation, such that discussion with the equivalent
several physicochemical parameters, comparisons for the particle radius becomes less useful.
bimodal system have been conducted, with the size ratio
P taken to be P ˆ RS;0 =RL;0 : Fig. 45(a) shows the range 6.1.3. Experimental comparisons for diamond particles
of flammability for the bimodal system with respect to the There exists another source for the bimodal system.
degree of small particle content G=…1 ⫹ G † and the degree of Makino et al. [77] measured the range of flammability and
dilution k , for T 0 ˆ 300 K; m ˆ 1; RS;0 ˆ 1:5 mm; RL;0 ˆ the burning velocity of the SHS process for Ti–diamond
10 mm; and 2r ˆ 18 mm: It is seen that the range of flamm- system, as a basic research to fabricate the diamond
ability, which spans from k ˆ 0 to 0.178 when G=…1 ⫹ G † ˆ embedded TiC/Ti composites by the SHS process, to
0; gradually extends with increasing G /(1 ⫹ G ) and finally achieve advanced cutting tools and/or abrasives. In this
spans from k ˆ 0 to 0.400 when G=…1 ⫹ G † ˆ 1: Extension fabrication, the SHS flame must propagate through a
of the range of flammability is attributed to the decrease in medium of compacted particles of titanium, carbon (or
the equivalent particle radius RL,0/F. The limit of dilution graphite), and diamond, without participation of the
…k ˆ 0:446† without heat loss suggests that even at G=…1 ⫹ diamond particles in the SHS process.
G † ˆ 1: the heat loss affects the range of flammability. It is By comparing the predictions [76] with the experimental
seen that as far as the trend and approximate magnitude are results [43,77] it has been determined to use the physico-
concerned, we can fairly predict the experimentally chemical parameters, described in Sections 4.4 and 5.4.1,
obtained extinction limit [43,77]. except for r N ˆ 3:51 × 10 3 kg=m3 : A Lewis number of
Fig. 45(b) shows u0,a in the adiabatic condition as a Le0 ˆ 100 is used because of the fair correlation with the
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 45

Fig. 45. Combustion behavior for Ti–C system with bimodal par-
ticles [75]; (a) Range of flammability with respect to the degree of
small particle content G=…1 ⫹ G † and the degree of dilution k , for
T 0 ˆ 300 K; m ˆ 1; RS;0 ˆ 1:5 mm; and RL;0 ˆ 10 mm; data points
are experimental [43,77] and notation is the same as that in Fig. 43.
(b) Burning velocity u0,a in the adiabatic condition as a function of
G=…1 ⫹ G †; with k taken as a parameter; notation is the same as that
in Fig. 44.
Fig. 44. SHS rate-constant u0·R0 for Ti–C system with monomodal
particles, as a function of the inverse of the heat loss index R02/(2r) at
to k . Data points are experimental [43] and the same
T0 ˆ 300 K [79]; (a) with m taken as a parameter and (b) with k
taken as a parameter. Data points are experimental [43,77,78]; open
comments as those for Fig. 47 can be made.
symbol designates the steady propagation and half-open symbol the Having confirmed the fair agreement for the monomodal
pulsating combustion. system, it was intended to make experimental comparisons
for the bimodal system. Fig. 49(a) shows the range of
flammability with respect to G=…1 ⫹ G † and k , and Fig.
49(b) shows u0,a as a function of G=…1 ⫹ G †; for the stroichio-
experimental results [43,77] (0.6 relative density) for the metric mixture …m ˆ 1† at T 0 ˆ 300 K with RS;0 ˆ 3 mm;
various radii of diamond particles, as shown in Figs. 47
and 48. The compact diameter 2r is 18 mm in accordance
with the experiments.
Fig. 47(a) shows the range of flammability for mono-
modal compacts with respect to m and R0, for T0 ˆ
300 K and k ˆ 0: Experimental results [43,77] for the
steady propagation, quasi-steady multipoint combustion,
extinction, and non-ignition show that the range of
flammability extends with the decreasing R0. Fig.
47(b) shows u0,a·R0 as a function of m , with R0 taken
as a parameter. Compared to the Ti–C system, the burn-
ing velocity for Ti–diamond system is reduced because
of the increase in the nonmetal density r N, as shown in
Eq. (76), which suggests that u0,a is inversely propor-
tional to (r N) 1/2. The increase in r N, thus, reduces the Fig. 46. SHS rate-constant u0,a·(RL,0/F) in the adiabatic condition as
total surface area of particles in the combustion wave, a function of the degree of dilution, with T0 taken as a parameter.
and hence, suppresses the volumetric heat-generation Data points are experimental [43,77] and notation is the same as that
rate. Fig. 48 shows a similar comparison with respect in Fig. 44.
46 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 47. Combustion behavior for Ti–diamond system with mono- Fig. 48. Combustion behavior for Ti–diamond system with mono-
modal particles [76]. (a) Range of flammability with respect to m modal particles [76]. (a) Range of flammability with respect to k
and R0, for T0 ˆ 300 K without dilution …k ˆ 0†; data points are and R0, for stoichiometric mixture …m ˆ 1† at T0 ˆ 300 K; data
experimental [43,77] and notation is the same as that in Fig. 43. points are experimental [43,77] and notation is the same as that in
(b) u0,a·R0 as a function of m , with R0 taken as a parameter. Fig. 43. (b) u0,a·R0 as a function of k , with R0 taken as a parameter.

RL;0 ˆ 15 mm; and rN;S ˆ rN;L : Extent of agreement


between the predicted [76] and the experimental [77] results present bimodal system with different densities of N parti-
is similar to that of Fig. 45. cles. Data points are experimental [77] and agreement is
The appropriateness of the equivalent particle radius fair, as far as the trend is concerned.
RL,0/F can be further shown by comparing the exper-
imental results of R0 ˆ 8 mm for the monomodal system
in Fig. 48(a) with the predicted results for the bimodal 6.2. Representative length of the cross-sectional area
system of G=…1 ⫹ G † ˆ 0:463; which corresponds to
R L;0 =F ˆ 8 mm; in Fig. 49(a). As mentioned in Section Thus the far, shape of a compacted medium has been
6.1.2, this discussion with the equivalent particle radius considered to be cylindrical in evaluating the radiative
becomes less useful when a number of large particles heat loss from the side surface of the compacted medium.
across the combustion wave becomes too small to apply Then, the representative length of the cross-sectional area
the statistical counting method. has been the compact diameter, as has appeared in the heat
In order to further study a fundamental aspect of the loss parameter C in Eq. (112). Here the choice of a rep-
practical fabrication of the diamond embedded TiC/Ti resentative length for the noncircular cross-sectional area is
composite, the bimodal system with large diamond and examined, with a rectangular compact taken as an example.
small carbon particles has been used. Here, the diamond When the radiative heat loss was introduced into the
particles are large enough to prevent the SHS flame propa- formulation in Section 5.1, the heat loss L per unit time
gation, even for the stoichiometric mixture. Fig. 50(a) shows and per unit volume was considered. When a sample speci-
the range of flammability with respect to G=…1 ⫹ G † and k , men has a cross-sectional area S and perimeter p, the heat
and Fig. 50(b) shows u0,a as a function of G=…1 ⫹ G †; for the loss per unit time is expressed as LSd , where d is the thick-
stoichiometric mixture …m ˆ 1† at T 0 ˆ 300 K with RS;0 ˆ ness of a volume considered along the x-coordinate, say, the
2:5 mm; RL;0 ˆ 15 mm; rN;S ˆ 2:25 × 10 3 kg=m3 ; and thickness of the combustion wave or the thickness of the
rN;L ˆ 3:51 × 103 kg=m3 : The same trends as those shown consumption zone. As for the heat loss per unit time trans-
in Fig. 45(a) and (b) are obtained, respectively, for the ferred from the surface to the ambience by radiation, it is
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 47

Fig. 49. Combustion behavior for Ti–diamond system with bi- Fig. 50. Combustion behavior for Ti–C–diamond system [76]. (a)
modal particles [76]. (a) Range of flammability with respect to Range of flammability with respect to G=…1 ⫹ G † and k , for m ˆ 1 at
G=…1 ⫹ G † and k , for m ˆ 1 at T 0 ˆ 300 K with RS;0 ˆ 3 mm; T 0 ˆ 300 K with RS;0 ˆ 2:5 mm; RL;0 ˆ 15 mm; rN;S ˆ 2:25 × 10 3
RL;0 ˆ 15 mm; and rN;S ˆ rN;L ; data points are experimental [77] kg/m 3, and rN;L ˆ 3:51 × 103 kg/m 3; data points are experimental
and notation is the same as that in Fig. 43. (b) Burning velocity u0,a [77] and notation is the same as that in Fig. 43. (b) Burning velocity
in the adiabatic condition as a function of G=…1 ⫹ G †; with k taken u0,a in the adiabatic condition as a function of G=…1 ⫹ G †; with k
as a parameter; notation is the same as that in Fig. 44. taken as a parameter; notation is the same as that in Fig. 44.

expressed as es ST …T 4 ⫺ T04 †pd: Then, we have Fair agreement is again obtained, as far as the trend and
approximate magnitude are concerned.
es ST …T 4 ⫺ T04 †
Lˆ ; …148† 6.3. Correspondence between the heterogeneous theory and
S=p
the homogeneous theory
by which the heat loss parameter is expressed as
Much of the theoretical analyses on the flame propagation
 " # !
l 4es SB …T 2 ⫹ T02 †…T ⫹ T0 † R20 in SHS, as reviewed in Ref. [24], are based on the classical
Cˆ : …149†
rtc …rt c†…u0;a R0 †2 4S=p treatment of the premixed flame propagation, characterized
by a large activation energy Arrhenius reaction rate and
When the cross-sectional area is circular with radius large Lewis number heat and mass diffusion. Although
r, we have S ˆ pr2 and p ˆ 2pr; which yields 4S=p ˆ these analyses have successfully captured the temperature-
2r: When the cross-sectional area is rectangular with a sensitive nature of the flame propagation, and consequently
in length and b in width, we have S ˆ ab and p ˆ 2…a ⫹ have been able to describe such empirically observed
b†; which yields 4S=p ˆ 2ab=…a ⫹ b†: It may be informa- phenomena as abrupt flame extinction due to heat loss,
tive to note that this is a harmonic mean of a and b, and instability of the combustion wave in the form of pulsat-
and is called the equivalent diameter in the field of the ing and spinning propagation, as explained in Refs. [4–9],
heat-transfer engineering. there are several fundamentally unsatisfactory aspects as
Fig. 51(a) and (b) shows the range of flammability for Ti– mentioned in the previous sections. While it is pleasing
C system as a function of m and k , respectively. Values of that the heterogeneous theory has been successful in
the physicochemical parameters are as those in Sections 4.4 explaining the experimental results, especially the tempera-
and 5.4.1 Data points are experimental [80] at T0 ˆ 450 K ture sensitivity, one must nevertheless confront the some-
with various a and b, as well as the various radii of (irregu- what curious result that the characterization of an Arrhenius
lar) carbon particles. Appropriateness of the choice of the diffusion process can be similar to that of an Arrhenius
representative length is demonstrated. Fig. 52(a) shows u0 as reaction process. To resolve this interesting correspondence,
a function of m , with k taken as a parameter, and Fig. 52(b) we shall demonstrate [81] that the governing equations for
shows u0 as a function of k , with m taken as a parameter. the heterogeneous theory can indeed be cast in a form which
48 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 51. Range of flammability for Ti–C system, with T0 taken as a parameter; (a) with respect to m and R20 =‰2ab=…a ⫹ b†Š; and (b) with respect to
k and R20 =‰2ab=…a ⫹ b†Š: Data points are experimental [80] and notation is the same as that in Fig. 43.

can be considered to be functionally identical to that The boundary conditions are as follows:
describing the large activation-energy premixed-flame
x ˆ 0 : T ˆ Tm ; 1 ⫺ Z ˆ 1 ⫺ Z0 …156†
propagation in a condensed medium.
To this aim, the N-consumption equation (46) must first
d L d…1 ⫺ Z†
be rewritten as x!∞: …cT† ˆ ⫺ ; ˆ 0: …157†
  dx …rt u† dx
d…1 ⫺ Z† T
rt u ˆ ⫺rt B~ w exp ⫺ d ; …150† We see that a set of governing equations and boundary
dx T
conditions is exactly the same as that for the conventional
where homogeneous theory (cf. Ref. [4]). Although differences are
!   ! minor from a mathematical point of view, this identification
3 rM rt
B~ ˆ …1 ⫺ Z0 †D0 ; …151† is valuable for the physical interpretation of the combustion
R20 rN rt;0 behavior. First, it should be noted that w in the “reaction”
 1=3   term is a function of the mass fraction of solid N particles,
1⫺Z A ~
wˆ ln 1 ⫹ YM ; …152† 1 ⫺ Z: In addition, if we note that the SHS proceeds under
1 ⫺ Z0 1⫹A diffusion-controlled situation, as is the case for usual SHS,
  we have A=…1 ⫹ A† ! 1; which suggests that the mass diffu-
1⫺m
…1 ⫺ Z0 † ⫹ …1 ⫺ Z† sivity in the Arrhenius fashion exerts nonlinear influence on
m
Y~ M ˆ : …153† flame propagation and/or extinction. That is, the activation
1 ⫺ …1 ⫺ Z† temperature which can influence the nonlinear, highly
In deriving Eq. (153), the following relation has been used temperature-sensitive nature of the SHS flame propagation
  is not the activation temperature Ta for the reaction, but Td
Z
…1 ⫹ Y~ M † ˆ …1 ⫹ Y~ M;0 † 0 ; …154† for the liquid-phase mass-diffusivity. As for the effect of
Z particle size, it has been incorporated into the frequency
~ tacitly.
factor B;
derived from Eq. (49) under the assumption of the negligible
Since it has been confirmed that the set of the governing
diffusion effect. The particle consumption term in the RHS
equations and boundary conditions in the heterogeneous
of Eq. (150) therefore corresponds to the “reaction” term in
theory is identical to that in the conventional homogeneous
the homogeneous premixed-flame theory, with B~ being the
theory, results of theoretical analyses hitherto obtained by
“frequency” factor, w the “reactant concentration”, and Td
the homogeneous premixed-flame theory, as reviewed in
the “activation temperature”. As for the energy conservation
Refs. [23,24], can be anticipated to be qualitatively valid
equation, from Eq. (58), we have
if we take note of the differences involved.
    
d l d d T
…cT† ⫺ rt u …cT† ⫹ q0 rt B~ w exp ⫺ d
dx c dx dx T 6.4. Field activated SHS

ˆ L: (155) For low exothermic systems, in order to sustain flame


A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 49

described in Section 5.1, except for the energy conservation


equation, which is now not Eq. (58) but
 
d dT
l ⫺ mc…T ⫺ T0 † ⫹ mq0 …Z ⫺ Z0 † ˆ L ⫺ P …158†
dx dx

when it is expressed in dimensional form [89] where P is the


energy supplied by the electric field. When we consider the
situation that only N particles participate in the external
heating due to electric field, the energy supply rate P is
expressed as

scE 2
P ˆ …1 ⫺ Z† ; …159†
2

where E is the electric field (V/m) and s c the electric


conductivity (W ⫺1/m). It may be informative to note that
the electric field induced by the alternative current must
p be
evaluated as the effective value with a factor of 1= 2: In
addition, it is anticipated that when the temperature is lower
than the melting point of M species, there is insufficient
energy supply by electric current, because of the contact
resistance among particles. When the combustion product,
such as SiC has been produced, its electric resistance is too
high to allow for electric current.
By introducing a nondimensional parameter

…l=c†s c E 2
Hˆ ; …160†
m2a q0
Fig. 52. SHS rate-constant u0·R0 for Ti–C system as a function of
the inverse of the heat loss index R20 =‰2ab=…a ⫹ b†Š at T0 ˆ 450 K;
in addition to other variables and parameters in Sections 4.2
(a) with m taken as a parameter; and (b) with k taken as a parameter.
and 5.1, Eq. (158) becomes
Data points are experimental [80] and notation is the same as that in
Fig. 44. !
d du m=m a m=ma
⫺ u⫹ j
ds ds Z∞ ⫺ Z0 T~ ∞ ⫺ T~ 0

C·u H…1 ⫺ j†
ˆ ⫺ ; …161†
propagation, it is usual to heat a compacted medium exter- …Z ∞ ⫺ Z0 † 2 …Z ∞ ⫺ Z0 †…T~ ∞ ⫺ T~ 0 †
nally by preheating and/or by applying an electric field.
Since preheating in a long period of time could induce where H can be called the heat-input parameter. The other
production of undesired compounds which might inhibit governing equations are the same as those in Eqs. (99) and
initiation of the combustion wave, it is reported to be prefer- (101). The boundary conditions are also the same as those in
able [82,83] for low exothermic systems to apply the electric Eqs. (102) and (103).
field, which can also exert a great influence on the flame
propagation [84,87] especially when the electric conductiv-
ity in the combustion wave is much higher than those in the 6.4.2. Temperature profiles outside the combustion wave
unburned and burned states. Although there exist several As mentioned in Section 5.2, in the limit of large Zeldo-
analytical [88] and/or numerical [84,86] works on this vich number, melting, diffusion, and consumption/convec-
subject based on the homogeneous premixed-flame theory, tion are confined to a thin layer in the neighborhood of the
heterogeneity involved has not been explored. In this combustion wave. Outside this layer, the diffusion and
section, focusing on the effect of the electric field on the consumption/convection terms are exponentially small. If
burning velocity, let us examine it by use of the heteroge- we consider the situation that the combustion wave locates
neous theory. at s ˆ s f and that the electric field exerts influence only in
the region 0 ⬍ s ⬍ s f ; temperature profile outside the
combustion wave is expressed as
6.4.1. Heat-input parameter
The problem of interest is basically the same as that s ⱕ 0 : u ˆ A u exp…lu s†; j ˆ 0; …162†
50 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

0 ⬍ s ⬍ s f : u ˆ Ah exp…lu s† ⫹ B h exp…ld s† (170) as


 
Hs f
H…Z ∞ ⫺ Z0 † …T~ ∞ ⫺ T~ 0 † ˆ …Z∞ ⫺ Z0 † 1 ⫹ ; …172†
⫹ ; j ˆ 0; (163) Z∞ ⫺ Z0
C…T~ ∞ ⫺ T~ 0 †
" #
sf …1 ⫺ H†…Z∞ ⫺ Z0 †
s ⱖ s f : u ˆ C exp…ld s†; j ˆ 1; …164† ˆ ln ; …173†
Z∞ ⫺ Z0 …T~ m ⫺ T~ 0 † ⫺ H…Z∞ ⫺ Z0 †
where
in the limit of C ! 0: We see that the adiabatic flame
s)
(
m=ma 4C temperature increases with increasing strength of the elec-
lu ˆ 1⫹ 1⫹ ; tric field. It may be informative to note that the relation in
2…Z∞ ⫺ Z0 † …m=m a † 2
…165† Eq. (59), which gives a relation between the temperature and
( s) the mass fraction of fluid cannot be used any more when heat
m=ma 4C
ld ˆ 1⫺ 1⫹ : is supplied externally.
2…Z∞ ⫺ Z0 † …m=m a † 2

Relations among constants Au, Ah, Bh, and C are deter- 6.4.3. Effective range of electric field
mined by use of the continuities of temperature and heat flux In order for the electric field to be effective for increasing
at s ˆ 0; respectively, the continuity of temperature at the combustion temperature, the flame location must be posi-
combustion wave …s ˆ s f †; and the jump condition of du / tive, which requires that the logarithmic term in Eq. (173)
ds in the combustion wave as must be positive. From this requirement, the range in which
    the heat-input parameter H can exist is determined as
du du m=ma
⫺ ˆ ; …166† T~ m ⫺ T~ 0
ds s f⫺ ds s f⫹ T~ ∞ ⫺ T~ 0 0ⱕH⬍ ⬍ 1: …174†
Z∞ ⫺ Z0
which is obtained by integrating Eq. (161) and evaluating at
s ˆ s f : Then, we have In the same manner, when there exists heat loss, we have
from Eq. (171) the following relation between C and H.
1 m=ma ( )
Au ˆ exp…⫺lu s f †
lu ⫺ ld T~ ∞ ⫺ T~ 0 1 …m=m a † 2 H…Z∞ ⫺ Z0 †
⫺ ⫺
2 4 T~ m ⫺ T~ 0
ld H…Z∞ ⫺ Z0 †
⫺ ‰1 ⫺ exp…⫺lu s f †Š; …167† v
( )
lu ⫺ ld C…T~ ∞ ⫺ T~ 0 † u
u …m=ma †2 …m=ma †2 H…Z ⫺ Z †
  ⫺t ⫹ ∞ 0

1 m=ma ld H…Z∞ ⫺ Z0 † 8 8 T~ m ⫺ T~ 0
Ah ˆ ⫹
lu ⫺ ld T~ ∞ ⫺ T~ 0 lu ⫺ ld C…T~ ∞ ⫺ T~ 0 † ( )
1…m=m a † 2 H…Z∞ ⫺ Z0 †
⬍C⬍⫺ ⫺
 exp…⫺lu s f †; …168† 2 4 T~ m ⫺ T~ 0
v
( )
lu H…Z∞ ⫺ Z0 † u
Bh ˆ ⫺ …169† u …m=ma †2 …m=ma †2 H…Z ⫺ Z †
lu ⫺ ld C…T~ ∞ ⫺ T~ 0 † ⫹t ⫹ ∞ 0
: …175†
8 8 T~ m ⫺ T~ 0
1 m=ma Note that when C has the value of the limit, length between
Cˆ exp…⫺ld s f †
lu ⫺ ld T~ ∞ ⫺ T~ 0 the location of melting and that of the combustion wave
becomes infinite.
lu H…Z∞ ⫺ Z0 †
⫺ ‰1 ⫺ exp…⫺ld s f †Š: …170†
lu ⫺ ld C…T~ ∞ ⫺ T~ 0 † 6.4.4. Experimental comparisons for Si–C system
Values of physicochemical parameters employed are
If we further note that the temperature at s ˆ 0 is the
those for Si–C system: q 0 ˆ 6:23 MJ=kg; c ˆ 1 kJ=…kg=K†;
melting point, we can determine the flame location s f as
r M ˆ 2:34 × 103 kg=m3 ; rN ˆ 2:25 × 103 kg=m3 ; WM ˆ
2 m ld 3 28:1 × 10⫺3 kg=mol; WN ˆ 12:0 × 10⫺3 kg=mol; Tm ˆ
6 ⫹ H…Z ∞ ⫺ Z † 7 D ˆ 3:3 ×
1 ma C 0 1681 K: For the mass diffusivity,
sf ˆ ln 6
4
7;
5 10⫺5 exp…⫺3:39 × 104 =T† m2 =s is used [90]. As for the
lu ~ ~ l
…T m ⫺ T 0 †…lu ⫺ ld † ⫹ d
H…Z ∞ ⫺ Z0 † representative Lewis number, Le0 ˆ 1; because experimen-
C
…171† tal results [85] have been correlated for this value.
Fig. 53 shows the SHS rate-constant u0·R0 for stoichio-
which is obtained from Eq. (167) by setting Au ˆ um : The metric Si–C system, as a function of the square root of the
flame temperature in the adiabatic condition, which is indis- heat-input parameter H. A solid curve is u0,a·R0 in the adia-
pensable for nondimensionalization, is obtained from Eq. batic condition and a dashed curve is that at the extinction
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 51

In this context, the heterogeneous theory is indeed viable in


examining the SHS process. Correspondence between the
heterogeneous theory and the homogeneous premixed-flame
theory has further been examined. In spite of the minor
differences, from the viewpoint of the mathematics, the
important identification related to the physical interpreta-
tion, pointed out in the previous sections, has been recon-
firmed.

7. Boundary between steady and pulsating combustion

So far, flame propagation and/or extinction in steady


combustion has been examined. However, SHS flame
propagation can be either steady or nonsteady, depending
on the system parameters. Since nonhomogeneity in
combustion products induced by nonsteady propagation is
often responsible for the weaknesses in the solid materials
synthesized, understanding of the mechanisms that trigger
Fig. 53. SHS rate-constant u0·R0 for stoichiometric Si–C system, as
instabilities is expected to give appropriate directions for
a function of the square root of the heat-input parameter H [89]; a
controlling the SHS process.
solid curve is that in the adiabatic condition and a dashed curve is
that at the limit of flame propagation. Data points are experimental The nonsteady flame propagation can assume two char-
[85]; (S) designates data for rrel ˆ 0:52 and (W) for rrel ˆ 0:64: acteristic modes [3,7,9] namely pulsating and spinning
combustion. In pulsating combustion, the combustion
limit. A range surrounded by these curves corresponds to wave travels in a planar but pulsating manner as a result
that for flame propagation under electric field. With increas- of the one-dimensional instability of the planar combustion
ing H, up to the limit determined by Eq. (174), u0·R0 wave to longitudinal perturbations. In spinning combustion,
increases first gradually and then rapidly. In the abscissa, the combustion wave is nonplanar, and one or more hot
for convenience, the strength of electric field is shown, spots are observed to move along a spiral path over the
obtained by use of the electric conductivity of carbon, s c ˆ surface of the sample specimen as a consequence of the
1:21 × 10 4 ⫹ 6:25…Tr ⫺ 450† (W ⫺1/m), based on the data in instability of the one-dimensional oscillatory to transverse
the literature, [91] with the representative temperature Tr perturbations. It has been observed that spinning combus-
taken as the arithmetic mean of the melting point of Si tion occurs in low exothermic systems with excess amounts
and the maximum combustion temperature. The initial of either one of the reactants, or the diluent, or the large
carbon radius R0 of 2 mm is evaluated from the reported particles, at low preheat temperatures. As for pulsating
burning velocity [85] that was sustained in the lowest elec- combustion, it has been observed in energy intensive
tric field of about 0.8 kV/m. Data points are experimental systems with stoichiometric mixtures of small particles at
[85] and fair degree of agreement is demonstrated, as far as elevated preheat temperatures.
the trend and approximate magnitude are concerned. Since rapid synthesis is preferred from a practical point of
It is also reported that the characteristic extinction turn- view, and it can be accomplished by using small particles
ing-point behavior described in Section 5.2 can exist when and/or high preheat temperatures, pulsating combustion
the heat-input parameter H is less than about 0.2, while it frequently occurs and results in materials with a laminated
ceases to exist and the flame extinction occurs at the limit in structure [3] (cf. Fig. 3), which cannot be used as a bulk. As
Eq. (175) when H ⬎ 0:2: It may be informative to note that such, the study and control of transition from steady
since the heat-input parameter H depends on the particle combustion to pulsating combustion are of greater practical
size, only the heterogeneous theory could captured the relevance than those of the transition from steady combus-
heterogeneous feature of the field activated SHS. tion to spinning combustion.
In order to identify and gain an understanding on the
6.5. Some remarks on the several, other, important factors boundary governing the transition from steady to pulsating
combustion, we are required to extend the steady heteroge-
Related to the SHS flame propagation and/or extinction, neous theory of SHS flame propagation developed in Refs.
other important factors, such as the bimodal particle distri- [32,64] to allow for a nonsteady effect, as has been done in
bution, the representative length of the cross-section, and Ref. [92]. In the following, first by considering the material
the field activation are examined. It has been shown that strength and flame instability, description for the boundary
these effects on the flame propagation and/or extinction between steady and pulsating combustion is made in terms
are also closely related to the particle size of the nonmetal. of the system parameters, which consist not only of the
52 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

physicochemical parameters, but also of the operating par- 7.2. Linear stability analysis
ameters, such as the mixture ratio, degree of dilution, parti-
cle size, and initial temperature. From this identification, 7.2.1. Basic solution for planar propagation
operating parameters at which steady flame propagation is Prior to the linear stability analysis, let us first confirm the
maintained, even at the adiabatic condition, are obtained, to governing equations and boundary conditions. By introdu-
elucidate the necessary condition for pulsating combustion, cing the nondimensional time defined as
in order to identify the useful parameter in suppressing
…Z∞ ⫺ Z0 † 2 m2a t
pulsating combustion. Finally, transverse effects related to tˆ ; …178†
…l=c†rt
heat loss from the side surface of the compact are examined
to assess the range of steady combustion. in addition to the other variables and parameters introduced
in Sections 4.2 and 5.1, the energy conservation equation
7.1. Criterion for the appearance of cracks and the N-concentration equation are expressed as

Pulsating combustion is characterized by the periodic 2 …m=m a † 2 22u F…u†


…j ⫺ u† ⫹ …j ⫺ u† ⫹ ˆ ;
movement of the combustion wave, with the laminated 2t …Z∞ ⫺ Z0 † 2s 2s 2 …Z ∞ ⫺ Z0 † 2
structure of the combustion product being caused by the …179†
presence of cracks in the radial direction. Although it has  
been pointed out [51] that both thermal and mechanical 2j …m=m a † 2 Z0
⫹ j⫹
effects induce pulsating combustion, the main concern has 2t …Z∞ ⫺ Z0 † 2s Z∞ ⫺ Z0
been directed to the thermal effect, and various numerical     1=3
x~ Z∞ ⫺ Z0
and analytical investigations have been conducted, as ˆ 1⫺ j :
reviewed by Margolis [24]. However, the appearance of …Z∞ ⫺ Z0 †L0 …rt 0=rt † exp…T~ d =T†
~ 1 ⫺ Z0
radial cracks in the combustion products implies that …180†
mechanical effects caused by pressure increase in pores
The boundary conditions are as follows:
[93,94] and/or thermal expansion, can exert strong influence
on the combustion behavior. When there exists temperature s ! ⫺∞ : u ˆ 0; j ˆ 0; …181†
distribution along the axial direction, there appears not only
a normal stress corresponding to macroscopic pressure sˆ0: u ˆ u max ; …182†
distribution along the axial direction, due to desorbed gas
in pores, but also a shear stress corresponding to a difference 2u F…u† 2j
s!∞: ˆ⫺ ; ˆ 0: …183†
of thermal expansions in the radial direction. When strains 2s …m=m a †…Z∞ ⫺ Z0 † 2s
caused by these stresses are kept below a certain value Ksb,
The mass burning rate m…ˆ r t u†; subject to heat loss L,
determined by the material fracture toughness, as well as the
would emerge from the solution as an eigenvalue, with ma
coefficient of thermal expansion, radius of the curvature of
and L 0 being, respectively, the mass burning rate and the
the pores, compact diameter, and so on, we can anticipate
mass burning rate eigenvalue in the steady, adiabatic limit.
that radial cracks can be suppressed and steady combustion
The SHS rate-constant u0·R0 is obtained from this eigen-
can be maintained. Since both of these stresses are related to
value.
a difference of the temperature gradients along the axial
In conducting the linear stability analysis, we are first
direction, and the difference is the maximum at the propa-
required to obtain the basic solution u 0(s ) and a0. Since
gating combustion wave, the criterion that steady combus-
the perturbation quantities are O(b ⫺1), the basic solution
tion is maintained can be expressed as [95]
     is exactly the same as that for the asymptotic “outer” solu-
2T 2T tion in the steady state, as presented in Eqs. (113) and (114),
⫺ ⱕ Ksb : …176†
2x 0⫺ 2x 0⫹ in the limit of large Zeldovich number b . In Ref. [73], the
basic solution is expressed in a slightly different form as
In evaluating the temperature gradients, an attempt [92]
has been made to apply and extend the linear stability analy- u ˆ C exp…l u s†; j ˆ 0; …s ⬍ 0†; …184†
sis within the framework of the heterogeneous theory, by
linearizing the problem about the basic solution and seeking u ˆ C exp…l d s†; j ˆ 1; …s ⬎ 0†; …185†
harmonic solutions through the perturbation quantities u p(s ,
where
t ) and ap(t ) as   
1 a0 1
u ˆ u 0 …s† ⫹ u p …s; t†; …m=m a † ˆ a0 ⫹ ap …t†; …177† lu ˆ 1⫹ ;
2 Z∞ ⫺ Z0 C
as has been made by Margolis [24,74] and Kaper et al. [73]    …186†
1 a0 1
based on the homogeneous premixed-flame theory. Note ld ˆ 1⫺ ;
2 Z∞ ⫺ Z0 C
that the perturbation quantities are O(b ⫺1), where b is the
Zeldovich number. and C is the maximum temperature in the normalized form,
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 53

expressed as In examining the linear stability of the steady planar SHS


s flame, it is the standard to linearize the nonlinear problem
4C (Eqs. (179)–(183)) regarding the basic solution (Eqs. (184),
C 1 ⫹ 2 ˆ 1; …187†
a0 (185), (187) and (188)) and to seek harmonic solutions.
Substituting the perturbation quantities in Eq. (177) into
which is obtained from the jump in du /ds at s ˆ 0: Note Eq. (190), we obtain
that by conducting a similar treatment in Eq. (125), we have
the relation
  2u p a0 2up ap du0
m 2 exp ‰b…C ⫺ 1†Š ⫺ M ⫹ ⫹
ˆ …188† 2t …Z∞ ⫺ Z0 † 2s …Z∞ ⫺ Z0 † ds
ma k…1 ⫺ M†:
2 2up C· u p
Having obtained the basic solution, we can now evaluate ˆ ⫺ ; …192†
2s 2 …Z∞ ⫺ Z0 † 2
the difference of the temperature gradients in the criterion in
Eq. (176). Then, we have
l=c
a0 …Z∞ ⫺ Z0 † ⫹ O…b⫺1 † ⱕ ‰a0 …Z∞ ⫺ Z0 †Šsb ⬅ Ksb up ˆ 0 as s ! ^∞: …193†
ma …q0 =c†
…189†
for steady combustion. We see that the appearance of cracks Here, a relation between the perturbed quantities is given as
in pulsating combustion can be primarily controlled by [73]
a0 …Z∞ ⫺ Z0 † because the perturbed O(b ⫺1) term is much ( )
smaller than the first term. Note here that Z∞ ⫺ Z0 is related B b M
to the heat of combustion, which depends on the mixture a p ˆ Bup …0; t†; ˆ 1⫹ 2 …194†
a0 2 a0 k…1 ⫺ M†
ratio m , degree of dilution k , and initial temperature T0, and
that a0 is related to the heat loss which depends on the
particle size R0 and the compact diameter 2r. When Z∞ ⫺ where use has been made of Eq. (188). Note that, since 1 ⫺
Z0 is kept constant, we can anticipate that steady combustion M is the normalized driving force for the flame propagation,
occurs in the range a0 ⬍ …a0 †sb ; which is equivalent to C ⬎ as mentioned in Section 5.6.1, B/a0 is a measure of the
Csb ; if we recall the relationship between C and a0. When energy generated in the combustion wave compared to the
the heat loss condition is kept constant, the steady combus- driving force for the nonadiabatic flame propagation, that is
tion range is Z ∞ ⫺ Z0 ⬍ …Z∞ ⫺ Z0 †sb ; which yields m ⬍ msb a measure of the driving force for instability.
for under-stoichiometric mixture …m ⬍ 1† and m ⬎ msb for A normal mode solution for u p(s , t ) is then found to be a
an over-stoichiometric mixture …m ⬎ 1† at a certain k ; for scalar multiple of
constant m , k ⬎ ksb :
Eq. (189) further suggests that when Z∞ ⫺ Z0 ⬍ …Z∞ ⫺ " ! !
1 a0 B
Z0 †sb ; steady combustion can be maintained, even if the u p …s; t† ˆ exp …ivt† ⫺ l C
iv Z∞ ⫺ Z0 a0 u
perturbed O(b ⫺1) term is positive; likewise, when Z∞ ⫺
Z0 ⬎ …Z∞ ⫺ Z0 †sb ; steady combustion cannot be maintained ( ! ! )
1 a0 B
even if the perturbed term is negative. In this context, the × exp…lu s† ⫹ 1 ⫹ lu C
iv Z∞ ⫺ Z0 a0
condition for no amplification, nor attenuation of the
perturbed term is anticipated to provide information about #
the transition boundary, and then we are required to obtain × exp …pu s† ; …s ⬍ 0†; (195)
the specific form of the perturbed term in Eq. (189).

7.2.2. Asymptotic solution and linear stability boundary " ! !


In obtaining asymptotic solution, we should note that the 1 a0 B
u p …s; t† ˆ exp…ivt† ⫺ l C
first order inner solution is identical to that for the adiabatic iv Z∞ ⫺ Z0 a0 d
combustion when the Zeldovich number is large, because of ( ! ! )
a higher order effect of heat loss, while the outer problem is 1 a0 B
× exp …ld s† ⫹ 1 ⫹ lC
influenced by both heat loss and unsteadiness. The outer iv Z∞ ⫺ Z0 a0 d
problem is then given by [92] #
2u …a0 † 2u 2 2u C·u × exp…pd s† ; …s ⬍ 0†; (196)
⫹ ˆ ⫺ ; …190†
2t …Z∞ ⫺ Z0 † 2s 2s 2
…Z ∞ ⫺ Z0 † 2

u ˆ 0 as s ! ^∞: …191† where


54 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74
8 9
s
 <   =
1 a0 1 Z∞ ⫺ Z0 2
pu ˆ 1⫹ ⫹4 …iv† ;
2 Z∞ ⫺ Z0 : C 2 a0 ;

…197†
8 s9
 <   =
1 a0 1 Z∞ ⫺ Z0 2
pd ˆ 1⫺ ⫹4 …iv† :
2 Z∞ ⫺ Z0 : C 2 a0 ;

…198†
Having obtained the perturbation quantities u p(s ,t ) and
ap(t ), we can now evaluate the perturbed term in Eq. (176)
as
" !2 !
1 a0 B
a 0 …Z∞ ⫺ Z0 † exp…ivt† ⫺
2…iv† Z∞ ⫺ Z0 a0
( !2 ! )
1 a0 B
⫹ 1⫹ C …199†
2…iv† Z∞ ⫺ Z0 a0 Fig. 54. Location of the neutral stability boundary [92].
v
!2 #
u
u 1 Z ⫺ Z0
t 2 ⫹ 4 ∞
where Xsb is determined by solving the neutral stability
…iv† :
C a0 condition
 
Since the jump condition with respect to 2u=2s at s ˆ 0 b X
gives iv as [73] 2 X⫺M
8 s 9
! 2" ( !2 ! )
<  2=
1 a0 1 B B 2 1 2
iv ˆ ⫺ ⫺ ⫹4 C ˆ s 1 ⫹ 1 ⫹ 1 ⫺ ln X ; …206†
8 Z∞ ⫺ Z0 C2 a0 a0 2 : 4 b ;
1 ⫺ ln X
!v
u

!2 ! # b
1 B u t 1 B B
^ ⫺ ⫹ ⫺8 C ; with X ⬅ exp…⫺2bC=a 02 † ˆ a20 k…1 ⫺ M† ⫹ M: In obtaining
C a0 C a0 a0
Eq. (206), use has also been made of Eqs. (187) and (194).
…200† Fig. 55 shows the neutral stability boundary in the b –M
we see that the perturbed term attenuates in the regime plane, with X being taken as a parameter.
0 s1 p ! Once the critical values of the heat-loss parameter C sb,
 
B 1 A 1 6 and the flame temperature u sb, are obtained, the critical size
0ⱕ ⬍ 2C@1 ⫹ 1 ⫹ ; 1 ⱕ ⬍ ;
a0 4C4 C 2 of N particles and compact diameter for steady combustion
…201†
  p !
B 1 6 1
0ⱕ ⬍ ; ⬍ ; …202†
a0 C…1 ⫺ C2 † 2 C

because of the negative value of the real part of iv , as shown


in Fig. 54. Restricting ourselves to the neutral stability
boundary on which there is neither attenuation nor amplifi-
cation of the burning velocity, we have a set of critical
values as
Xsb ⫺ M
…a0 †2sb ˆ ; …203†
k…1 ⫺ M†

…a0 †sb2
C sb ˆ ⫺ ln Xsb ; …204†
2b

ln Xsb
usb ˆ 1 ⫹ …205†
Fig. 55. Neutral stability boundary in the b –M plane [92].
b
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 55

2r is 18 mm. The solid and dotted curves are respectively


the numerically [64] and analytically [72] obtained
extinction limits. The dashed curves are the neutral
stability boundaries, with l /(r tc) taken as a parameter.
Results show that, except near stoichiometry …m ˆ 1†;
the range of instability with respect to R0 is reduced
with decreasing mixture ratio. When m ⬍ 0:673;
instability is suppressed, even for R0 ! 0; because of
the decrease in the volumetric heat-generation with
decreasing m . We also see that the instability can be
suppressed with increasing l=…rt c†: This is attributed to
the fact that heat transferred from the consumption zone
increases with increasing l=…rt c†: Likewise, a reduction
of T0 is also expected to suppress instability, although
the range of flammability is reduced. The slight increase
in the range of instability with decreasing m near stoi-
chiometry is attributed to the overestimation in Ij in Eq.
(92), as noted in Section 4.6. Experimental results [77]
also show that the range of pulsating combustion
extends with decreasing R0. The decrease in R0
increases the total surface area for surface reactions in
the combustion wave, and hence, makes the mixture
more energy intensive. Since instability is invariably
dependent on the burning intensity, mixtures with
small particles have wider ranges for pulsating combus-
tion. Experimental results suggest that the limit of
Fig. 56. Ranges of flammability and instability for Ti–C system instability predicted by the present analytical result is
with l=…rt c† taken as a parameter [92]; the initial temperature T0 ˆ
fair.
300 K and the compact diameter 2r ˆ 18 mm; (a) as functions of
Fig. 56(b) shows the similar comparison with respect
the mixture ratio m and the radius R0 of N particles, with l=…rt c†
taken as a parameter; and (b) as functions of the degree of dilution k to the degree of dilution k . We see that the range of
and R0. In the region below the dashed curve, pulsating combustion instability with respect to R0 is narrowed with increas-
proceeds. In the region between the solid (or dotted) and dashed ing k and that instability is suppressed even at R 0 ! 0
curves, stable and steady propagation occurs. Data points are when k ⬎ 0:280: Again, fair agreement is demonstrated
experimental [43,77] and notation is the same as that in Fig. 43. with the experimental results [43,77] in the trend and
the approximate magnitude.
Fig. 56(a) and (b) show that the predicted limit of instabil-
are determined through ity with l=…rt c† ˆ 5 × 10 ⫺6 m2 =s fairly represents the
! experimental results, while the limit of extinction has been
R 02 C sb …rt c†…u0;a R0 † 2 obtained with l=…rt c† ˆ 2:5 × 10⫺6 m2 =s: This difference
ˆ …207†
2r sb ‰l=…rt c†Š4es SB …Tsb2 ⫹ T02 †…Tsb ⫹ T† may be caused by the different temperatures at which
these two phenomena occur; that is, the temperature at
Note that the SHS rate-constant u0,a·R0 in the adiabatic which the pulsating combustion ceases to occurs is about
condition is given by Eq. (129). 200 K higher than that for flame extinction.
Fig. 57 shows the ranges of flammability and instability
7.2.3. Experimental comparisons for the boundary with respect to m and k , with R0 ˆ 1:5 mm at T0 ˆ 300 K:
In order to demonstrate the appropriateness of the transi- The solid curve is the limit of flammability [64] the dashed
tion boundary determined above, a comparison has been curve the boundary for R0 ˆ 1:5 mm obtained by numeri-
made with Ti–C system taken as an example. Values of cally solving Eq. (206) with l=…rt c† ˆ 5 × 10⫺6 m2 =s; and
the physicochemical parameters are the same as those the dotted curve the boundary for the adiabatic condition to
described in Sections 4.4 and 5.5, while the Lewis number be mentioned in the next section. It is seen that the pulsating
Le0 ˆ 25 in accordance with experiments with spherical combustion can be suppressed with decreasing m and/or
carbon particles. increasing k because of the decrease in the volumetric
Fig. 56(a) shows the ranges of flammability and heat-generation. Fair agreement is again demonstrated
instability with respect to the mixture ratio m and the radius between the predicted and experimental [77] results for
R0 of N particles for an initial temperature T0 of 300 K. In spherical carbon particles (3 mm in diameter). Since the
accordance with the experiments [77] the compact diameter condition for these experimental results is considered to
56 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 57. Ranges of flammability and instability for Ti–C system as functions of m and k , for T0 ˆ 300 K and 2r ˆ 18 mm [95]. The solid curve
is the limit of flammability [64], the dashed curve the numerically obtained boundary [92] for R0 ˆ 1:5 mm; and the dotted curve the predicted
result [95] in the adiabatic condition. Data points are experimental [77] for R0 ˆ 1:5 mm and notation is the same as that in Fig. 43.

be close to the adiabatic condition, as shown in Fig. 56(a) Since T∞ ⫺ T0 is related to the system parameters as [40]
and (b), heat-loss effect on this boundary is relatively small. 8
>
> m…1 ⫺ k†
>
< m ⫹ fst …m ⱕ 1†
7.2.4. Boundary at the adiabatic condition T~ ∞ ⫺ T~ 0 ˆ Z∞ ⫺ Z0 ˆ ; …210†
>
> …1 ⫺ k†
In the SHS process, reduction of particle size, which >
: …m ⱖ 1†
reduces heat-loss effect, facilitates the occurrence of pulsat- m ⫹ fst
ing combustion when the system parameters, such as the when there is no external heating, we can obtain the critical
physicochemical and operating parameters in Sections values for these operating parameters.
4.6.2 or 5.6.3, are given. However, even if we use small Fig. 58(a) shows the boundary between steady combus-
particles, there is a possibility that steady combustion is tion and pulsating combustion in the adiabatic condition, as
maintained because appearance of pulsating combustion functions of the mixture ratio m and the initial temperature
also depends on the operating parameters, such as the T0, when there is no dilution …k ˆ 0†: The solid curve is the
mixture ratio m , degree of dilution k , and initial temperature limit of flammability numerically obtained [64] and the
T0, through the heat of combustion when the physicochem- dashed curve for kˆ~ 0 is the boundary obtained with Eq.
ical parameters for a specific system are given. Such that, (208) [95]; the chain curve for k~ ˆ k~lower is the boundary
operating parameters at which steady combustion proceeds when the transverse effect is the largest, which will be exam-
even at the adiabatic condition (i.e. X ˆ 1) has been ined in the next section. We see that the pulsating combus-
obtained in order to elucidate the necessary condition for tion regime expands with increasing T0, and that the steady
pulsating combustion, under which the only way effective in combustion regime contracts. Experimental results
suppressing the appearance of pulsating combustion is [33,43,77] for small carbon particles, with diameters less
increasing size of N particles. In order for Xsb to be unity, than 3 mm, are also shown because these data are considered
the critical values of b and M must satisfy to be influenced very little by the heat loss (cf. Fig. 57). Fair
  p agreement is demonstrated as far as the general trend is
b=2
ˆ 2 ⫹ 5; …208† concerned. Fig. 58(b) shows the transition boundary as func-
1 ⫺ M sb
tion of k and T0, when the system is stoichiometric. The
which has already been identified in the field of stability extent of agreement is similar to that in Fig. 58(a).
analysis. If we recall the expression 1 ⫺ M ˆ
bIu exp…b=a† in Section 5.6.1 and Iu in Eq. (87), we can 7.3. Transverse effects on the range for steady combustion
obtain …T∞ ⫺ T0 †sb because a and b are defined by use of
T∞ ⫺ T0 : Note that when the normalized melting-point u m is The transition from steady combustion to pulsating
smaller than about 0.7, the following equation fairly rep- combustion is in general accompanied by the appearance
resents the relation between a and b . of nonplanar spatial perturbations. Through various analyses
p p for examining the transition boundary by use of the multi-
5 ⫹ 2 p exp…y2 † erfc…y† b=a dimensional media with cylindrical [74,96,97] or rectangu-
a⬇ p ; y⬅ : …209†
2 y 2 lar [98–100] shapes, it is reported that this kind of
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 57

Fig. 58. Ranges of flammability and instability for Ti–C system [95]; (a) as functions of m and T0; (b) as functions of k and T0. The solid curve is
the limit of flammability [64] numerically obtained, while the dashed curve for k~ ˆ 0 and the chain curve for k~ ˆ k~lower are the numerically
obtained boundaries [95] between steady and pulsating combustion; k~ ˆ 0 means that there is no transverse effect, while k~ ˆ k~lower means the
transverse effect is the largest. Data points are experimental [33,43,77] when R0 is less than 1.5 mm; notation is the same as that in Fig. 43.

perturbation can be properly described by introducing over- given by solving


all transverse wave-number k~ into the linear stability analy- b

X

4
sis. When this effect is taken into account, the boundary ˆ s ; …213†
2 X⫺M 2
under heat-loss conditions is given by the following relation: 1 ⫺ ln X
[73,74] b
    3 ~ which is specified as k~lower : For the
1 b X appears at a certain k;
C⫺ stoichiometric Ti–C system, we have b=2 ˆ 2:29 and M ˆ
C 2 X⫺M
     2 0:240 at T0 ˆ 300 K: In this system, Xsb is 0.5792 when
1 1 b X
⫹ ⫺4C C ⫺ ⫹ 2 ⫹ 4k~2
C C 2 X⫺M
      
1 1 1 b X
⫺ ⫹ 4C C ⫺ ⫹ 4k~2
C C C 2 2 X⫺M
 2
1
⫺ ⫹ 4k~2 ˆ 0: (211)
C2

When a set of parameters (b , a , M, and k) ~ is given, the


critical value of Xsb can be determined. As reported in
Refs. [73,74] and as shown in Fig. 59, with an increasing
k~ for fixed Xsb, b /2 first decreases, reaches a minimum, and
then increases. The minimum value of b /2 exists at k~ deter-
mined by solving
1
4k~2 ˆ 2 ⫺ : …212†
C2
We see in Fig. 59 that b /2 decreases with decreasing Xsb
when k~ is fixed.
It may be common in practical situations that the system
parameters such as b , a , and M are given by specifying
combinations of metal and nonmetal species, initial
temperature, and so on. Thus, it is of interest to examine
~ with b /2 taken as a
changes in Xsb due to an increase in k; Fig. 59. Neutral stability boundary for stoichiometric Ti–C system,
~
constant. With increasing k for fixed b /2, Xsb determined as functions of the wave-number k;~ the half of the Zeldovich number
from Eqs. (211) and (212) first increases, reaches a maxi- b , and the normalized heat loss Xsb at the critical condition. The
mum, and then decreases. The maximum of Xsb, which is initial temperature T0 is 300 K and there is no dilution.
58 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

there are no effects of the transverse wave-number …k~ ˆ 0†;


while Xsb ˆ 0:6412 for kˆ ~ k~lower ˆ 0:4489; as shown in Fig.
59. This result suggests that the steady combustion regime is
the broadest when k~ ˆ k~lower :
The boundary between the steady and pulsating combus-
tion determined with k~ ˆ k~lower without heat loss …X ˆ 1† is
also shown in Fig. 58(a) and (b). Although k~ depends on the
compact diameter and the transverse mode of the combus-
tion wave, because it is influenced by the boundary con-
ditions in the radial direction such as the heat-loss
condition on the side surface of the compact, we see that
the difference between the boundary for k~ ˆ 0 and that for
k~ ˆ k~lower is relatively small.
Fig. 60(a) shows the ranges of flammability and instabil-
ity for Ti–C system at T0 ˆ 300 K as functions of m and
R20 =…2r† [95]. The solid curve is the limit of flammability
numerically obtained [64] the dotted curve is that analyti-
cally obtained [72]; the dashed curve is the boundary
between the steady and pulsating combustion for k~ ˆ 0;
and the chain curve is that for k~ ˆ k~lower ; by use of Eqs.
(206) and (211), respectively. Data points are experimental
[43,77,78] for various particle radii (1.5, 5, 10, 15 mm) and
compact diameters (3, 5, 7, 10, 18 mm). We see that the
range of instability is extended with decreasing R20 =…2r†
because the effect of heat loss is reduced due to the increase
in the volume-to-surface ratio. Fair agreement between
the predicted and the experimental results suggests that
the heat loss index R02/(2r) is useful in correlating
experimental results. Fig. 60(b) shows the similar plot
for k and R20 =…2r†:
Finally, it may be informative to note that bifurcation
[74,96,97,99] of pulsating and spinning combustion, and
chaotic combustion [98,100] can occur at a certain con-
dition, because of the trend between b and k; ~ as shown in
Fig. 59. Topics about bifurcation and the chaotic combus-
tion, which are related to the nonlinear stability, are not
Fig. 60. Ranges of flammability and instability for Ti–C system for
covered here; readers are recommended to refer Ref. [24].
T0 ˆ 300 K [95]; (a) as functions of m and R20 =…2r†; and (b) as
functions of k and R20 =…2r†: Data points are experimental
[43,77,78] for various particle radii and various compact diameter;
7.4. Some remarks on the boundary of steady combustion notation is the same as that in Fig. 43. The solid curve is the limit of
flammability numerically obtained [64], the dotted curve is that
Use of small particles enhances the SHS flame propaga- analytically obtained [72], and the dashed curve for k~ ˆ 0 and the
tion speed and necessarily induces pulsating combustion, chain curve for k~ ˆ k~lower are the numerically obtained boundaries
such that the identification obtained here offers the limit [95].
of the use of small particles when other system parameters
are kept unchanged. As for the effects of transverse wave 8. Initiation of the combustion wave
number, which are related to heat loss on the side surface of
the compacted medium on the location of the transition Besides flame propagation, initiation of the combustion
boundary, it has been found to be relatively small. Further, wave (which is indispensable for the SHS process) has also
appropriateness of the transition boundary, which has been been investigated extensively because the energy that must
demonstrated by comparing the predictions with experimen- be supplied externally in the SHS process is basically that
tal results in the literature, suggests that the heterogeneous for initiating the combustion wave. Although much of the
theory has captured the essential, heterogeneous feature of theoretical accomplishment about the initiation of the
the SHS flame propagation and that the results are of prac- combustion wave has been reviewed by Barzykin et al.
tical importance in controlling the combustion behavior to [101] and Barzykin [102] all are based on the homogeneous
be steady. premixed-flame theory, and hence the heterogeneity has not
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 59

over a short period when the activation energy (or tempera-


ture) is large. Let tc denote an ignition delay time, at which
temperature rises rapidly. During most of time up to the
ignition, t ⬍ tc ; the specimen experiences only inert heating
because heat-generation term is exponentially small due to
large activation energy, and reactant consumption can be
neglected. Then, Eq. (214) becomes as
Fig. 61. Model used for analyzing initiation of the SHS combustion
 
wave in the specimen, induced by another SHS flame that has 2T~ 2T~ 1 4hrad ~
passed through the igniter [103]. aj ˆ ⫹ …T ⫺ T~ 0 †; …j ˆ ig; sp†;
2x2 2t …rt †j c·2r
…216†
been clarified. The heterogeneous theory, which has
succeeded in describing dependence of the flame-propaga-
where
tion process on the particle size, is also expected to be
applied for examining ignition phenomena. First, we shall
consider the initiation of the combustion wave induced by T~ ˆ …cT†=q0 ; a ˆ l=…rt c†: …217†
an igniter [103] in order to examine heterogeneity that has
not been examined in a series of analyses [104–110]. After The boundary conditions at x ˆ 0 are as follows:
that, initiation of the combustion wave by use of the radia-
! !
tive heat flux will be examined.
~ ig ˆ …T†
~ sp ; 2T~ 2T~
…T† l ˆ l : …218†
2x 2x
8.1. Flame initiation induced by igniter ig sp

One-dimensional, planar, heterogeneous flame initiation As the initial conditions, we have


has been considered in a semi-infinite medium, which is
" #
heated by an SHS flame passed through the other semi-infi- …m†ig
nite medium, called an igniter; cf. Fig. 61. It is supposed that x ⬍ 0 : …T~ ⫺ T~ 0 † ˆ …T~ max ⫺ T~ 0 † exp x ; Z ˆ Z0 ;
…l=c†sp
the region x ⬎ 0 is the igniter with thermophysical proper-
ties of l ig, (r t)ig, and cig, and that the region x ⬍ 0 is the …219†
medium to be ignited, called a specimen, with l sp, (r t)sp, and
csp. The contact surface is set to be x ˆ 0 and the moment x ⬎ 0 : …T~ ⫺ T~ 0 †
that the SHS flame reaches the surface is t ˆ 0: The igniter "   #
and the specimen are compacted media respectively, 1 4hrad …220†
ˆ …T~ max ⫺ T~ 0 † exp ⫺ x ; Z ˆ Z∞ ;
consisting of particles of nonmetal (or higher melting- …m†ig c·2r
point metal) N, lower melting-point metal M, and an inert
I that can be the product P of the reaction between N and M which are obtained for the steady SHS flame propagation
according to nM M ⫹ nN N ! nP P: Monodisperse size distri- (e.g. Ref. [73]), under an assumption of large activation
bution is considered for N particles, with an initial radius R0 energy for the rate-limiting process, by noting that melting,
and number density n0. diffusion, and consumption/convection are confined to a thin
Governing equations for these media, from Eqs. (179) and layer in the combustion wave. Putting
(180), are expressed in a dimensional form as follows.
Energy: …Z ∞ ⫺ Z0 †sp …ma †sp x …Z∞ ⫺ Z0 †sp2 …ma †2sp t
sˆ ; tˆ ;
…l=c†sp …l=c†sp …rt †sp
2 22 T 2Z
rt …cT† ˆ l 2 ⫹ q0 rt ⫺ L: …214†
2t 2x 2t
T~ ⫺ T~ 0 T~ ⫺ T~ 0
uˆ ; us ˆ max ;
N-consumption: …T~ ∞ ⫺ T~ 0 †sp …T~ ∞ ⫺ T~ 0 †sp
! 1=3
2Z rt 1⫺Z
rt ˆ …4prN n0 R0 † x : …215†
2t rt;0 1 ⫺ Z0 …m†ig =…ma †sp C
k~ ˆ ; w~ ˆ ;
…Z∞ ⫺ Z0 †sp ~ ∞ ⫺ Z0 †sp2
k…Z

8.1.1. Inert stage !


Ignition processes can usually be characterized by a a sp …4hrad † R20 l
Cˆ ; L ˆ p
gradual increase in temperature followed by a rapid increase …rt c†sp …u0 ·R0 †sp2 2r a
60 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

and using the Laplace transformation, we have diffusive region near the contact surface, stretched variables
!
us s2 c ˆ b…u ⫺ u I †; X ˆ ⫺bKz …s ⫺ s c †; …224†
uI ˆ exp ⫺ ~
⫺ w~ k t
‰1 ⫹ …Lsp =Lig †Š 4t are introduced, where b is the Zeldovich number defined as
( ! " ! 2#
1 L sp s p T~ d …Z∞ ⫺ Z0 †
× 1⫹ p ⫹ k~ t bˆ : …225†
2 Lig
exp
2 t T~ c2
! ! Retaining only nonvanishing terms on the exponential, we
s ~
p 1 L sp
× erfc p ⫹ k t ⫺ 1⫺ have
2 t 2 Lig
d2 c D
" ! 2# ˆ ⫺ t ec⫺X ; …226†
s p dX 2 2
× exp ⫺ p ⫹ k~ t
2 t !
! 2 x~ c
Dt ˆ ; …227†
s ~
p bKz2 …Z∞ ⫺ Z0 †L0 …rt;0 =rt † exp…T~ d =T~ c †
× erfc ⫺ p ⫹ k t
2 t
which is subject to
" s! 2 #  
s a ig dc
⫹ exp ⫺ p ⫹ w~ t …c† Xˆ0 ˆ 0; ˆ 0: …228†
2 t asp dX X!∞
s!) Eq. (226) can be solved analytically and we have
s a ig
" ( !2 )#
× erfc ⫺ p ⫹ w~ t : (221)
2 t asp 1 ⫺1 ⫹ C e ⫺X
c ˆ X ⫹ ln 1⫺ ;
Dt 1 ⫹ C e⫺X
Here, L ig and L sp are the heat penetration coefficients. In the p …229†
following, temperature rise in the inert, heating stage is 1 ⫺ 1 ⫺ Dt
Cˆ p
designated with the subscript I. Note that temperature 1 ⫹ 1 ⫺ Dt
response for a finite size of the igniter can also be obtained
(e.g. Ref. [111]). Since there is no solution for Dt ⬎ 1; the
in the same manner, as shown by Strunia et al. [106]
critical condition is as follows:
!
8.1.2. Transition stage 2 x~ c
Dt ⬅ ˆ 1; …230†
In analyzing the transition from inert heating to vigorous bKz2 …Z∞ ⫺ Z0 †L0 …rt;0 =rt † exp…T~ d =T~ c †
reaction, temperature deviation must be taken into account
because the heat generation term becomes important as where
!
temperature rises. Since departure from the inert heating is 2uI us
anticipated to occur at time and position such that uI ⫺ …uI †c Kz ˆ ˆ exp…⫺w~ k~tc †
2s c
1 ⫹ …Lsp =Lig †
is close to zero, we have the lowest order solution as
(
uI ˆ …uI †c ⫹ Kt …t ⫺ tc † ⫹ Kz …s ⫺ s c †; …222† p
× k~ exp…k~ 2 tc † erfc…k~ tc †
by expanding Eq. (221), where Kt ˆ …uI =ut†c ; Kz ˆ
s ! s!)
…2uI =2s† c : Note that (·)c designates the condition corre- aig aig aig
sponding to the rapid increase in temperature, which occurs ⫺ w~ exp w~ 2 tc erfc w~ t : (231)
asp asp asp c
at t ˆ tc and s ˆ s c ; …uI †c is the maximum temperature in
the inert heating. Since temperature at a certain position in
Recalling definitions of parameters, we see that the ignition
the specimen first increases and then decreases, the most
delay time can be expressed as functions of not only physi-
plausible condition for the ignition is that for 2uI =2t ˆ 0;
cochemical parameters, but also operating parameters char-
at which the position has its maximum temperature.
acteristic to the SHS process, such as the mixture ratio,
The inner problem for the asymptotics is then as
degree of dilution, initial temperature, and particle radius,
2u 2 2u x~ through the eigenvalue L 0 and Z∞ ⫺ Z0 :
ˆ ⫹ : …223†
2t 2s z2 …Z∞ ⫺ Z0 †L0 …rt0 =rt † exp…T~ d =T†
~ It may be informative to note that the criterion used by
Averson et al. [104] and/or Strunia et al. [106] was Dt ˆ 2
Here, reactant consumption and heat loss are ignored, as is which was obtained by Enig [112] by the use of the so-called
the case for usual ignition phenomena. Identification of Eq. integral method by equating a transient–reactive relation at
(223) infers that the present problem parallels the earlier the flame initiation and a transient–diffusive relation in the
development [104–110]. In analyzing the reactive– inert heating. Such that, compared to the criterion in
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 61

We see that the ignition delay time is proportional to the


square of the particle size, which has not been captured in
previous theories.

8.1.4. Experimental comparisons for the ignition delay time


In numerical calculations, the situation considered is that
both the igniter and the specimen consist of mixtures of
titanium and carbon particles, although their mixture ratios
and degrees of dilution are different. In accordance with
experiments [113] the initial temperature before arrival of
the SHS flame is set to be 450 K, the compact diameter is
18 mm, and particle radius of carbon particles in the igniter
is 5 mm. The Lewis number (Le0)ig for the igniter is set to be
100 [32] while that for the specimen, (Le0)sp, is 25 [74]
because different kinds of carbon powder are used in an
experiment; that in the igniter has an irregular shape,
while that in the specimen has a spherical shape. Ratio of
Fig. 62. Ignition delay time as a function of the mixture ratio m ig in the heat penetration coefficients L sp/L ig is set to be 0.5,
the igniter, with the emissivity e taken as a parameter [103]. Data accordingly.
points are experimental [113]. Fig. 62 shows the ignition delay time tc as a function of
the mixture ratio m ig in the igniter, with the emissivity e
Eq. (230) by the asymptotics, the ignition delay time is taken as a parameter, which is equivalent to varying heat
overestimated by a factor of 2 when k~ ! ∞ and w~ ˆ 0: loss rate. The particle radius (R0)sp in the specimen is 10 mm.
With decreasing m ig, tc increases first gradually, then
8.1.3. Limiting solution rapidly, and reaches the ignition limit, due to the decrease
It may be informative to consider a limiting situation for in heat supplied from the igniter. We see that the decrease in
the heat loss makes tc shorter, because the extent of heat
k~ ! ∞ and w~ ˆ 0; …232†
used for initiating the combustion wave becomes large. A
which corresponds to the flame initiation caused by a spon- further decrease in the heat loss brings about a situation at
taneous contact with a hot substance in adiabatic condition. which the ignition limit locates beyond the extinction limit
The critical condition yields explicit dependence of the igni- for the combustion wave in the igniter, because of the heat
tion delay time on other parameters as loss during the flame propagation [64]. Data points are
experimental for …R0 †sp ˆ 10 mm [113]. We see that the
T~ d …T~ max ⫺ T~ 0 †2 R20 exp…T~ d =T~ c † ignition delay times are fairly correlated with e ˆ 1:0
tc ˆ :
2 ~
2p‰1 ⫹ …Lsp =Lig †Š T c 3D0 …1 ⫺ Z0 †sp …rM =rN †x~ c
2
when m ig is higher than about 0.8, while they are correlated
…233† with e ˆ 0 as m ig decreases. This is attributed to the

Fig. 63. Ignition delay time tc for e ˆ 1, with (R0)sp in the specimen taken as a parameter [103]; (a) as a function of the mixture ratio m sp; and (b)
as a function of the degree of dilution k sp. Data points are experimental [113].
62 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

establishment of the combustion wave is impossible, even


though initiated. Data point are experimental [113] and a
fair degree of agreement is demonstrated. Fig. 63(b) shows
the similar comparison with respect to k sp. It is also reported
that the same trends can be observed as those in Fig. 63(a)
and (b) by reducing the heat loss, instead of the particle size.
As for the effects of particle size on the ignition phenom-
ena, little attention has been paid in the previous theoretical
works, as was reviewed [101,102]. However, particle size is
expected to exert great influences in the flame initiation no
less than in the flame propagation and/or extinction, because
it is closely related to the total surface area which partici-
pates in reaction [32,64]. Fig. 64 shows tc as a function of
…R0 †2sp ; with e taken as a parameter when mig ˆ 0:8 and
msp ˆ 1: As shown in Eq. (233), tc is proportional to
…R0 †2sp when e ˆ 0; because of the large value of k; ~ say,
2
Fig. 64. Ignition delay time tc as a function of (R0)sp, with e taken as
larger than about 10. With increasing heat loss, tc becomes
a parameter [103].
large, because the extent of heat used for initiating combus-
tion wave becomes small. At a certain heat loss, with
maximum temperature of the SHS flame that has passed increasing …R0 †2sp ; tc increases first gradually, then
through the igniter because influence of heat loss can be rapidly, and reaches the ignition limit. The increase in
strongly suppressed as the flame temperature decreases by tc is attributed to the decrease in the total surface area
reducing m ig in the igniter. Note that tc becomes short as for the reaction, caused by the increase in …R0 †sp : Note
…R0 †sp in the specimen is reduced, because a decrease in the that the flame extinction limit with respect to the parti-
particle size increases the total surface area which partici- cle radius is 20.9 mm when the compact diameter 2r is
pates in the reaction. 18 mm [64].
Fig. 63(a) shows tc for e ˆ 1:0 as a function of m sp in the Fig. 65(a) shows the ignitable range as functions of m sp
specimen with …R0 †sp taken as a parameter. The mixture ratio and R20 =…2r†: When there is no heat loss …e ˆ 0†; the ignitable
m ig in the igniter is 0.8. With decreasing m sp, tc for …R0 †sp ˆ range coincides with the flammable range because ignition
10 mm increases first gradually, then rapidly, and reaches is limited only by the flame extinction after the establish-
the ignition limit because of a decrease in the heat generated ment of the combustion wave. The ignitable range is
during the flame initiation. As …R0 †sp is reduced, tc becomes contracted as m sp is reduced. In addition, it narrows drasti-
short. In addition, with decreasing m sp, tc gradually increases cally as the heat loss increases, because of the appearance of
and before reaching the ignition limit, encounters the flame the ignition limit. Experimental results [113] in Fig. 65(a)
extinction limit [64] corresponding to …R0 †sp ; beyond which also represent a fair agreement. Fig. 65(b) shows the similar

Fig. 65. Ignitable range with heat loss taken as a parameter [103]; (a) as functions of the mixture ratio m sp and the heat loss index ‰R20 =…2r†Šsp ; and
(b) as a functions of the degree of dilution k sp and the heat loss index ‰R20 =…2r†Šsp : Data points are experimental [113]. (W) designates the flame
initiation and ( × ) the non-ignition.
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 63

comparison with respect to k sp and again demonstrates a fair where


degree of agreement. r! s!
4w~ cig 4w~
1⫹ 1⫹ ⫺ 1⫹ 1⫹
k~ csp k~ig
CA ˆ ⫺ s ;
8.1.5. Expression for the ignition energy cig 4w~
2 1⫹
Concerning the ignition delay time tc, fair agreement has csp k~ig
been demonstrated between the predicted and experimental …237†
results. Now, let us evaluate the ignition energy [114] for r! s!
4w~ cig 4w~
establishing combustion wave in the specimen, by integrat- 1⫹ 1⫹ ⫺ 1⫺ 1⫹
k~ csp k~ig
ing heat flux through the contact surface [104,106,108]. CB ˆ s ;
Note here that the ignition energy consists of two parts: cig 4w~
2 1⫹
one is the energy supplied after arrival of the SHS flame, csp k~ig
which has passed through the igniter, at the contact surface;
and the other is supplied during the flame propagation in the …l=c†sp ~
k~ig ˆ k: …238†
igniter before arrival of the SHS flame at the contact surface. …l=c†ig
By the use of the temperature profile in the inert stage and
the ignition delay time, the energy Ec, supplied after the In deriving Eq. (236), a temporal variation of temperature in
arrival of the SHS flame, is obtained as the specimen is first obtained as
0 q 1
" k~ ⫹ k~ 2 ⫹ 4w~ k~
us B C
E~ c ˆ
1 ~ tc } erfc…k~ptc †
exp{…k~2 ⫺ w~ k† s ⱕ 0 : u ˆ B u exp @
2
sA; …239†
…1 ⫹ Lsp =Lig † k~ ⫺ w~
s
q # p where
k~ aig =asp
⫺1 ⫹ {1 ⫺ erfc… w~ k~tc †} ⫺ s!
w~ w~ …aig =asp † ⫺ k~ cig 4w~ 1
B u ˆ 2us 1⫹ ; …240†
" ( ! ) s! csp k~ig D
2 aig ~ aig
× exp w~ ⫺ w~ k tc erfc w~ t ⫺1 ( s! s!)
asp asp c
4w~ cig 4w~
Dˆ 1⫹ 1⫹ ⫺ 1⫺ 1⫹
s q #! ~k csp k~ig
w~ aig
⫹ {1 ⫺ erfc… w~ k~tc †} ; q
k~ asp ! ( s!
k~ig ⫹ k~ig2 ⫹ 4w~ k~ig 4w~
…234† × exp sf ⫺ 1⫹ 1⫹
2 k~
where s!) q !
cig 4w~ k~ig ⫺ k~ig2 ⫹ 4w~ k~ig
…ma †sp Ztc  2t  ⫺ 1⫹ 1⫹ exp sf ;
E~ c ⬅ Ec ˆ ⫺ du: …235† csp k~ig 2
…l=c†sp q …rt †sp
0 2s sˆ0
0
…241†

It may be informative to note that the expression of Ec when the propagating SHS flame in the igniter locates at
obtained by Strunia et al. [106] is slightly different from s ˆ s f …⬎0†: After that by assuming constant speed of the
that of Eq. (234) because their expression was derived flame propagation, the following relation has been used
under an assumption that there is no temperature distribu- …rt †sp
s f ˆ ⫺k~ t …242†
tion in the compacted medium, although they have taken the …rt †ig
temperature distribution into account in obtaining the igni-
tion delay time. which is equivalent to xf ˆ ⫺…u†ig t:
As for the energy E0, supplied before the arrival of the
SHS flame, although it has not been taken into account in the 8.1.6. Results for the ignition energy
literature [106] is expressed as In numerical calculations, the situation considered is the
same as that in Section 8.1.4. Fig. 66 shows the ignition
s! energy E as a function of m ig in the igniter, with e taken
1 …rt †ig 4w~ us
E~ 0 ˆ 1⫹ 1⫹ as a parameter. It is seen that with decreasing m ig, the igni-
2 …rt †ig k~ CB tion energy E increases first gradually and then rapidly until
" ! ( s!⫺1 )# the ignition limit, and that the contribution of E0 supplied
X∞
CA
n
2 4w~
× ⫺ 1 ⫹ …2n ⫹ 1† 1 ⫹ ; before arrival of the SHS flame is relatively large. In ad-
nˆ0
CB k~ig k~ig dition, in spite of the decrease in the heat of combustion in
…236† the igniter, the energy E0 increases with decreasing m ig. This
64 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

trend is similar to that for the ignition delay time tc in Fig.


63(a). Fig. 67(b) shows the similar results with respect to
k sp. The same trends as those in Fig. 67(a) and (b) can be
observed by reducing heat loss, instead of reducing particle
size.
Fig. 68(a) shows E as a function of …R0 †sp2 ; with e taken as
a parameter when mig ˆ 0:8 and msp ˆ 1: The general trend
is similar to that for tc in Fig. 64. Fig. 68(b) shows the similar
results for E at mig ˆ 0:65: The same trend as that in Fig.
68(a) is observed although the ignition energy is increased
and the ignitable range is contracted. This is attributed to the
decrease in the maximum temperature in the SHS flame that
has passed through the igniter, due to the decrease in m ig in
the igniter.

8.2. Flame initiation by use of radiative heat flux


Fig. 66. Ignition energy E as a function of m ig in the igniter, with e
taken as a parameter [114]. The specimen is stoichiometric with Combustion wave in the SHS process can also be initiated
…R0 †sp ˆ 10 mm: Energy E0 is that supplied before the arrival of by use of radiative heat flux, instead of using igniters.
the SHS flame in the igniter. Although numerical results based on the heterogeneous
theory are reported [115] it is considered that an extension
is attributed to the gradual heating of the specimen due to of the heterogeneous theory for the ignition phenomena
slow burning velocity of the SHS flame in the igniter. As for should be conducted by including analytical results related
the energy supplied after the arrival of the SHS flame, it to the ignition of reactive condensed materials, such as
increases first gradually and then rapidly, with decreasing explosives and solid propellants [116–120]. For this aim,
m ig, due to a decrease in the maximum temperature of the it is considered the situation that the front surface of a
SHS flame in the igniter. It is also seen that at a certain m ig, a compacted medium experiences partial, stepwise heating
decrease in the heat loss makes the ignition energy smaller, by the radiative heat flux [121–123]. In analyzing the
because extent of heat used for initiating combustion wave inert stage, use has been made of the formulation presented
becomes large. A further decrease in the heat loss brings by Niioka and Hasegawa [124]. In the analysis for the tran-
about a situation at which the ignition limit locates beyond sition stage, related to the initiation of the combustion wave,
the extinction limit. results of the asymptotics [116] have been introduced into
Fig. 67(a) shows E for e ˆ 1:0 as a function of m sp in the the heterogeneous theory. Comparisons between predicted
specimen, with …R0 †sp taken as a parameter. The general and experimental results are also conducted.

Fig. 67. Ignition energy E under heat loss condition …e ˆ 1†; with (R0)sp in the specimen taken as a parameter [114]; (a) as a function of m sp in the
specimen; (b) as a function of k sp in the specimen. The mixture ratio m ig in the igniter is 0.8.
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 65

Fig. 69. Model used for analyzing the SHS flame initiation by use of
radiative heat flux [124].

with large activation energies, the compacted medium


experiences only inert heat conduction during most of the
time that t ⬍ tc because the heat generation term, related to
reactant consumption, is exponentially small due to the large
activation energy. Then, the equation of heat conduction in
the axisymmetric coordinate becomes as

2u 22 u 22 u 1 2u
ˆ ⫹ ⫹ ⫺ fu; …243†
2t 2s z2 2s r2 s r 2s r

which is subject to the initial and boundary conditions as

…I:C:† t ˆ 0; u ˆ 0; …244†

2u ~
…B:C:† s z ˆ 0; 兩s r 兩 ⱕ a;
~ ˆ ⫺Q; …245†
2s z

2u
s z ˆ 0; 兩s r 兩 ⬎ a;
~ ˆ 0; …246†
2s z

s z ! ∞; s r ! ∞; u ˆ 0; …247†
Fig. 68. Ignition energy E as a function of …R0 †2sp ; with e taken as a
parameter [114]; (a) the mixture ratio m ig in the igniter is 0.8; (b) where
mig ˆ 0:65:
…Z∞ ⫺ Z0 †ma xi Q
si ˆ ; …i ˆ z; r†; Q~ ˆ ;
…l=c† …Z∞ ⫺ Z0 † 2 q0 ma
8.2.1. Model definition and inert stage
The problem of interest is the radiative heating of the …Z∞ ⫺ Z0 †ma a
a~ ˆ ;
front surface of an axisymmetric semi-infinite compacted …l=c†
medium (Fig. 69), originally consisting of a mixture of N
particles, metal M, and an inert I that can be the product P of
the reaction nM M ⫹ nN N ! nP P. The front surface of the C
fˆ ;
compacted medium is subjected to a partial, stepwise heat- …Z ∞ ⫺ Z0 † 2
ing with a constant heat flux Q for time t ⬎ 0; the diameter  " # 2!
of heating area is 2a. Monodisperse size distribution is l 4es SB …T 2 ⫹ T02 †…T ⫹ T0 † R0
Cˆ :
considered for N particles, with an initial radius R0 and rtc …rt c†…u0;a R0 †2 2r
number density of n0.
As described in Section 8.1.1, in the ignition processes Noted that the inert problem described has a known solution
66 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

and can be written as [123,125] variables are to be introduced.


( q !
Q~ a~ Z∞ c ˆ b…u ⫺ u I †; …253†
uI ˆ J …ls r †J1 …la†
~ exp ⫺ s z …l 2 ⫹ f†
2 0 0
s ˆ bKt …t ⫺ tc †
q! q! ( ! !)
sz
× erfc p ⫺ e t…l ⫹ f† ⫺ exp s z …l 2 ⫹ f†
2
1 x~ T~ d
2 t ⫹ ln p 2 exp ⫺ ;
bKz …Z∞ ⫺ Z0 †L0 …rt;0 =rt † T~ c
q!)
sz dl …254†
× erfc p ⫹ …l 2 ⫹ f† p ; (248)
2 t l2 ⫹ f
X ˆ bKz s z;c ; …255†
where Jn is the Bessel function of the first kind of order n. where b is the Zeldovich number defined in Eq. (225). By
Note that temperature rise in the inert, heating stage is desig- retaining only the first nonvanishing terms on the exponen-
nated with the subscript I. tial, the energy conservation Eq. (223) yields
d 2c 1
8.2.2. Transient stage ˆ ⫺ p exp…c ⫺ X ⫹ s†; …256†
In the same manner as that described in Section 8.1.2, the dX 2 b
transition stage can be analyzed. The only difference is the which is subject to
position at which departure from the inert heating occurs;  
that is, s z ˆ s r ˆ 0: Therefore, Eq. (248) is expanded as dc
ˆ c…∞; s† ˆ c…X; ⫺∞† ˆ 0: …257†
dX Xˆ0
uI ˆ …uI †c ⫹Kt …t ⫺ tc † ⫺ Kz s z;c ; …249†
The condition as s ! ⫺∞ is obtained from matching to the
where inert stage.
( By using c ⫺ X as the dependent variable, Eq. (256) is
Q~ p
integrated to produce
…uI †c ˆ p 2 ⫺ 2erfc … ftc †
2 f   s
!) dc 2
p p ˆ 1 ⫺ 1 ⫹ p {exp…c 0 ⫹ s† ⫺ exp…c ⫺ X ⫹ s†};
a~ dX b
⫺ exp…⫺a~ f† erfc p ⫺ ftc
2 tc …258†
!
p a~ p where c 0 ˆ c…0; s†: This solution must match to that of the
⫹ exp…a~ f† erfc p ⫹ ftc ; (250) transient-diffusive region which has the spatial variable
2 tc
p
p bKt
( !) v ˆ bK t s z ˆ X: …259†
  bKz
2uI Q~ a~ 2
Kt ˆ ˆ p 1 ⫺ exp ⫺ exp…⫺ftc †;
2t c ptc 4tc Application of matching and introduction of the time vari-
able
…251†  
Kz
! s~ ˆ s ⫹ ln p …260†
2uI ~
Kt
⫺Kz ˆ ˆ ⫺Q: …252†
2s z c results in the transient–diffusive problem as
 
Terms relevant to s r can be ignored because temperature 2c 2 2c 2c  
ˆ ; ˆ ⫺exp c…0; s~† ⫹ s~ ;
distribution in the radial direction is flat around s r ˆ 0: 2~s 2v 2 2v vˆ0 …261†
Here, (·)c designates the condition corresponding to t ˆ tc
and s z ˆ s r ˆ 0: c…∞; s~† ˆ c…v; ⫺∞† ˆ 0;
Since the independent variables in the transition stage are which has already been solved by Liñán and Williams
just t and s z, the use of a set of one-dimensional, transient ÿ  [116].
It is reported that the thermal runaway, that is c 0; s~ ! ∞;
governing equations in the heterogeneous theory [92,95] can occurs at a finite time s~ ˆ ⫺0:431: Since it is defined that
be justified. If we further ignore the heat loss effect, as is the t ˆ t c at ignition, Eqs. (254) and (260) yield an implicit
case for usual inner problems related to ignition, the energy formula for the ignition delay time as
conservation equation coincides with Eq. (223). ! !
This identification infers that the present problem closely 1 x~ T~
p exp ⫺ d ˆ e ⫺0:431
parallels the earlier development [116] for the ignition of Kz bKt …Z∞ ⫺ Z0 †L0 …rt0 =rt † T~ c
solid propellants. In analyzing the reactive-diffusive region
near the surface of the compacted medium, new stretched ˆ 0:650: (262)
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 67

We see that the ignition delay time can be expressed as


functions of not only Td, Q, and a, as pointed out by Niioka
and Hasegawa [124] but also the operating parameters char-
acteristic to the SHS process, such as the mixture ratio m ,
the degree of dilution k , the initial temperature T0, and the
initial particle radius R0. This can be seen from the definition
of t , which is expressed as
    !
rt;0 rM t
t ˆ 3D0 L0 …1 ⫺ Z0 † …263†
rt rN R20

by using Eq. (71), such that tc ⬃ tc R20 for a given stoichi-


ometry and/or a degree of dilution, once t c is obtained by
solving Eq. (262). Note also that …1 ⫺ Z0 † ˆ m…1 ⫺ k†=…m ⫹
fst †; shown in Eq. (62).
The dependence of the nondimensional parameters on the
operating parameters, in the same manner as that in Eq.
(263), is represented as follows. For Q~ and a~ we have
Fig. 70. Ignition delay time tc for Ti–C system with R0 ˆ 5 mm;
…QR0 †
Q~ ˆ with the radiative heat flux Q taken as a parameter [121]; (a) as a
q0 r t;0 D0 …Z∞ ⫺ Z0 †{3Le0 L0 …1 ⫺ Z0 †…rM =rN †}1=2 function of the mixture ratio m ; (b) as a function of the degree of
dilution k . Data points are experimental [126]; open symbol desig-
⬃ …QR0 †; …264† nates the ignition and flame establishment, while solid symbol
designates the ignition but extinguishment.
  1=2    
3L0 r a a
a~ ˆ …1 ⫺ Z0 † M ⬃ ; …265†
Le0 rN R0 R0 with the radiative heat flux Q taken as a parameter. The
initial radius R0 of N particles in the compacted medium is
which suggest that for a given stoichiometry and/or a degree set to be 5 mm. We see that the ignition delay time increases
of dilution, both Q~ and a~ depend on the particle radius R0 with decreasing m , due to a decrease in heat generated
even when the heat flux Q and the radius a of the heating during the flame initiation. A further decrease in m makes
area are fixed, as are the cases for usual experimental con- the combustion wave impossible to propagate, for the same
ditions. The energy supplied per unit area in the heating reason as that in Fig. 63(a). As for the effect of Q, we see that
surface, required for initiating the combustion wave, is tc increases with decreasing Q, because its decrease prolongs
given by the heating period required for the front surface to be heated
  1=2  
1 3L0 r Qtc to the ignition temperature. Data points are experimental
Q~ tc ˆ 0 …1 ⫺ Z0 † M [126] for different heat fluxes (3.3 and 2.57 MW/m 2) and
q rt …Z∞ ⫺ Z0 † Le0 rN R0
  a fair degree of agreement is demonstrated. Here, it may be
Qtc informative to note that in general, tc for the radiative
⬃ ;
R0 heating is much longer than that for the igniter. This may
…266† be attributed to the fact that heat supply by the SHS flame
with high temperature can be made much more rapidly than
while the total energy supplied is simply by the radiative heat flux. Fig. 70(b) shows a similar
!
Qtc a comparison with respect to the degree of dilution k . The
~
…Qtc †…pa~ † ⬃
2
: …267†
R30 same comment can be made as that in Fig. 70(a).
Fig. 71 shows tc for m ˆ 1 as a function of R0, with Q
It is clear that the particle radius R0 is an essential parameter taken as a parameter. We see that tc increases with increas-
affecting initiation of the combustion wave. ing R0, in the same reason as that in Fig. 64. It is important to
emphasize that such a significant parametric dependence
8.2.3. Results and experimental comparisons cannot be described by the homogeneous premixed-flame
In calculations, use has been made of physicochemical theory. The decrease in Q causes tc longer, because of the
parameters for Ti–C system [32,64] as described in Sections prolonged heating period of time.
4.4 and 5.4.1; a Lewis number of Le0 ˆ 25 is used [75] in Fig. 72 shows tc for m ˆ 1 with R0 ˆ 5 mm; as a function
accordance with an experiment. Let us first examine the of the radius a of the heating area, with Q taken as a par-
situation without heat loss, and then with heat loss. ameter. With decreasing a, tc becomes longer, first gradually
and then rapidly, due to a decrease in the heating area at the
8.2.3.1. Situation without heat loss Fig. 70(a) shows the front surface.
ignition delay time tc as a function of the mixture ratio m , Fig. 73(a) and (b) shows tc as functions of m and k ,
68 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 71. Ignition delay time tc for Ti–C system of the stoichiometric
mixture …m ˆ 1†; as a function of R0, with Q taken as a parameter
[121]. Data points are experimental [126].

respectively, with Q and a taken as parameters. Because of


the strong effect in decreasing the heating area, as shown in
Fig. 72, tc becomes longer. In addition, this effect becomes
remarkable as m decreases or k increases, because the
increase in m or decrease in k reduces the heat generated
during the flame initiation.
Fig. 74 shows tc for m ˆ 1 as a function of R0, with Q and
a taken as parameters. It is further confirmed that the effect
of the heating area on tc is strong, as well as the effect of R0.

8.2.3.2. Situation with heat loss Now, let us examine effects

Fig. 73. Ignition delay time tc for Ti–C system with R0 ˆ 5 mm; as a
function of k , with Q and a taken as parameters [122]; (a) as a
function of m ; and (b) as a function of k . Data points are exper-
imental [126].

of heat loss on the ignition delay time. Since the ignition


delay time t c is determined by solving the criterion in Eq.
(262), dependence of the LHS of Eq. (262) on the heat loss
Fig. 72. Ignition delay time tc for Ti–C system of m ˆ 1 with R0 ˆ f is to be examined, with t c taken as a parameter. As shown
5 mm; as a function of the radius a of the heating area, with Q taken in Fig. 75, with increasing f , the LHS of the criterion,
as a parameter [121]. Data points are experimental [126]. designated by F, first decreases slightly and then increases
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 69

which initiation of the combustion wave cannot be realized.


The state of the turning point can be determined by dF=df ˆ
0; which yields
duc t=2
ˆ⫺ ⬃ 0: …268†
df Z∞ ⫺ Z0
b⫹
T~ c
It is informative to note that as a first approximation,
p a~
ft ˆ p …269†
2 t
can be a solution for determining the state of the turning
point.
Fig. 76 shows the ignition delay time tc as functions of the
mixture ratio m and degree of dilution k , with the radiative
heat flux Q and the heat loss taken as parameters. The con-
ditions are the same as those in Fig. 70. Note that the emis-
sivity e represents the extent of the heat loss. It is seen that,
as the emissivity e increases, which results in an increase in
the heat loss, the ignitable range becomes narrow although
the ignition delay time itself is not influenced very much by
the heat loss.
Fig. 74. Ignition delay time tc for Ti–C system of m ˆ 1 with R0 ˆ Fig. 77 shows tc for the stoichiometric mixture as a func-
5 mm; as a function of R0, with Q and a taken as parameters [122]. tion of the radius R0, with the radiative heat flux Q and the
Data points are experimental [126]. emissivity e taken as parameters. The results clearly demon-
strate that the system ignitability can be profoundly affected
by the particle size in the presence of heat loss. Since heat
rapidly. When tc ˆ 2:4317; F for f ˆ 0 is 0.650, which loss is invariably present, the existence of the loss induced
corresponds to the value for the thermal runaway in the ignition limit is particularly significant in assessing the
adiabatic limit. Another solution for tc ˆ 2:4317 exists global performance of a synthesis system.
when f is about 0.093, at which heat loss is quite large
and it is considered to be unrealistic. It is further seen that 8.3. Some remarks on the flame initiation
for F ˆ 0:650; there exist two solutions for 2:4317 ⱕ tc ⬍
2:7252; one solution at tc ˆ 2:7252; and there is no solution In the SHS process, the energy required is basically for
for tc ⬎ 2:7252: Thus, there exists an ignition limit, beyond initiating SHS flame, such that effects of various parameters

Fig. 75. Behavior of the thermal runaway criterion on the heat loss [123], for the stoichiometric mixture of Ti–C system with R0 ˆ 5 mm; heated
by the radiative heat flux of 3.3 MW/m 2, with 10 mm in heating diameter.
70 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Fig. 76. Ignition delay time tc for Ti–C system, with the radiative
heat flux Q and the emissivity e taken as parameters [123]; (a) as a Fig. 77. Ignition delay time tc for Ti–C system of the stoichiometric
function of the mixture ratio m ; (b) as a function of the degree of mixture …m ˆ 1†; as a function of the particles radius R0, with Q and
dilution k . Conditions are the same as those in Fig. 70.
e taken as parameters [123].

on the flame initiation are indispensable for practical utility.


Investigation based on the heterogeneous theory suggests premixed flame in the homogeneous medium. So that
that the ignitable range is limited by the heat loss during heterogeneous features of the combustion wave have been
flame initiation when the particle size is large and the unresolved in previous theoretical works.
compact diameter is small; while it is the heat loss during Recognizing the limitation of a homogeneous premixed-
flame propagation that confines the ignitable range when the flame theory in describing the heterogeneous nature related
particle size is small and the compact diameter is large. This to the SHS process (especially on the influence of particle
identification provides useful insight in controlling ignition size), the existing theory on flame propagation in fuel sprays
phenomena in the SHS process by varying particle size, has been extended to describe such processes. Specific
which is expected to make a contribution in applying the modifications included are finite rate of reaction at the par-
SHS process for near-net-shape fabrications by use of the ticle surface and Arrhenius mass diffusion in the liquid
SHS process. phase of molten metal, so that the particle consumption
rate controlled by both diffusion and reaction can be incor-
porated into the formulation, and prediction and character-
9. Concluding remarks
ization of such important bulk flame propagation as the
flame propagation speed, the range of flammability, and
In the following the current status of SHS research on
the flame extinction limit, have successfully been made in
flame propagation, extinction, and initiation by use of the
relation to the heterogeneity in the combustion wave.
heterogeneous theory is summarized, and areas which
Calculated results for several representative ceramics and
require further research are suggested.
intermetallic compounds, such as Ti–C, Ti–B, Zr–B, Hf–B,
9.1. Summary of the present survey Al–Ni, Co–Ti, Ni–Ti systems, using consistent sets of
physicochemical parameters, respectively, show satisfac-
From a phenomenological consideration of flame propa- tory agreement with experimental results in the literature.
gation in the SHS process, it may appear in the combustion An important relation that the burning velocity is inversely
wave as a slurry-like configuration, which consists of the proportional to the nonmetal particle size has been
higher melting-point nonmetal particles and the molten confirmed. These comparisons further suggest that the
metal. This consideration leads to the important nonmetal (or higher melting-point metal) particles in these
identification about the SHS process that, although the systems might react in the diffusion-controlled limit, indi-
flame propagation in the bulk appears like that of the cating that finite rate chemical reaction is not the controlling
premixed flame, this bulk flame propagation is eventually factor in the propagation of bulk flame. Under such situa-
supported by the nonpremixed reaction of the dispersed tions, the Arrhenius liquid-phase mass-diffusivity exerts the
nonmetal particles in the molten metal. This nonpremixed dominant temperature sensitivity on the flame propagation
mode has not been incorporated in formulations in the in the SHS process. In addition, because of the excessively
previous theoretical works based on the theory for the slow rate of diffusion compared to the thermometric
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 71

conductivity and reaction rate (suggested by the excessively differences are minor from the mathematical point of
large Lewis number in the combustion wave), it has been view, important identification which has already been
clarified that concentration of the molten metal is mainly pointed out for the physical interpretation has been recon-
affected through consumption. firmed. So that qualitative validity of theoretical results
A nonadiabatic, heterogeneous theory for the SHS based on the homogeneous premixed-flame theory has
process with volumetric heat loss, has further been formu- been shown, if the differences involved are noted.
lated and shown that the characteristic extinction turning The field activated SHS, which is effective especially for
point exists in the same manner as that in the premixed low exothermic systems to sustain flame propagation, can
flame. It has also been found that the range of flammability also be examined by use of the heterogeneous theory, by
under heat loss conditions can well be described, in terms of introducing an additional term due to external heating by
its dependence on stoichiometry, dilution, initial tempera- electric current. It has been shown that the external heating
ture, particle size, and compact diameter. A particularly enhances the burning velocity and extends the range of
significant identification is that the process responsible for flammability, as expected. Significant identifications
its temperature sensitivity and ultimate extinguishment is obtained are that extent of these effect is strongly influenced
the Arrhenius liquid-phase mass-diffusion. The fair agree- by the particle size, as well as the physicochemical par-
ment between predicted and experimental results for the ameters, and that the range of flammability is still restricted
extinction limits in trend and in approximate magnitude even under external heating conditions, neither of which
suggests that this theory has captured the essential features have been captured by the homogeneous premixed-flame
of the nonadiabatic heterogeneous flame propagation in the theory. In this context, the heterogeneous theory is indeed
SHS process. A useful parameter called the heat loss index, viable in examining the field activated SHS.
which is defined as the square of the nonmetal particle size The study for the transition boundary from steady to
divided by the compact diameter, has also been found. pulsating combustion by use of the neutral stability analysis,
Based on these identifications, not only an approximate has also been conducted and it has been shown that there
expression for burning velocity in the adiabatic condition, exists a limit in the use of small particles, which is an
but also for extinction criterion in the nonadiabatic con- important suggestion for near-net-shape fabrication by use
dition, has been derived under the appropriate and realistic of the SHS process. As for the effects of transverse wave
assumptions of diffusion-limited reaction and infinite Lewis number, which are related to heat loss on the side surface, on
number. Effects of the dominant parameters on the burning the location of the boundary, it has been found to be rela-
velocity and/or extinction criterion have been clarified, tively small from a practical point of view.
yielding useful insight into manipulating the SHS flame In addition to the flame propagation and/or extinction,
propagation and extinction. Since a fair degree of agreement initiation of the combustion wave has been examined
with numerical results has been demonstrated by examining because energy required is basically that for initiating
applicability of these expressions, these expressions are SHS flame and effects of various parameters on the flame
expected to have general, qualitative and quantitative utility, initiation are indispensable for practical utility. It has been
in each case. found that the ignitable range is limited by heat loss during
Bimodal particle dispersion for the nonmetal particles has flame initiation when the particle size is large and/or
also been taken into account in the nonadiabatic heteroge- compact diameter is small; while it is heat loss during
neous theory. It has resulted that the mathematical problem flame propagation that confines the ignitable range when
can be rendered identical to the monomodal case by intro- the particle size is small and/or compact diameter is large.
ducing an equivalent particle radius expressed as a function This identification provides useful insight to controlling
of the bimodal parameters. This identification has further ignition phenomena in the SHS process by varying particle
provided useful insight into manipulating the flame propa- size, which is expected to make various contributions in
gation and/or extinction by making the system bimodal, practical applications.
especially in recognizing that because of the increase in
the total surface area of particles in the combustion wave, 9.2. Area for further research
the reduced equivalent particle size in the bimodal system
increases the burning velocity and broadens the range of In view of the recent interest to produce Functionally
flammability. Furthermore, flame instability, usually caused Graded Materials (FGMs) by use of the SHS process, a
by the high energy-release rate in the combustion wave candidate for research is the examination of flame
when the preparation is made with small particles, can propagation in a medium with distributed reactant concen-
even be suppressed by increasing the equivalent particle trations, not only in the longitudinal direction but also in the
size in the bimodal system. lateral directions. In this situation, contour of the combus-
The heterogeneous theory which has succeeded in captur- tion wave is no longer planar and the combustion wave
ing the essential feature of the heterogeneity in the SHS spreads two-dimensionally, because of the distributed heat
process has further been examined to elucidate differences generation rate and/or two-dimensional temperature distri-
from the homogeneous premixed-flame theory. Although bution, caused by the distributed reactant concentrations.
72 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

Since flame strength is not constant along the curved Acknowledgements


combustion wave, the concept of flame stretch might be of
great use in fundamental understanding of this flame propa- It is a pleasure to acknowledge the continuing interest and
gation and its related extinction. support that the author has received for this research on
Another candidate for research related to producing flame propagation, extinction, and/or initiation of the
FGMs is that of examining flame propagation for multicom- combustion wave in the SHS process, through a series of
ponent system. Since both diffusion-controlled combustion the Grant-in-aid for Scientific Research of the Ministry of
for some components and reaction-controlled combustion Education, Science, Sports, and Culture, the Iwatani Naoji
for other components can proceed simultaneously in the Foundation’s Research Grant, the Research Grant of Tokai
combustion wave, recognition of dominant parameters for Foundation for Promoting Industrial Technologies, and the
each component and/or each competing reaction might be Mechanical Industry Development and Assistance Founda-
indispensable for conducting theoretical works and funda- tion’s Research Grant, Japan.
mental understanding. This identification would further
facilitate exploration of such combustion mechanism as
bulk flame propagation, which proceeds prior to the other
bulk flame propagation, observed for multicomponent References
system.
In order to gain fundamental understanding for the [1] Merzhanov AG, Borovinskaya IP. Dokl Akad Nauk SSSR
SHS flame propagation, it might be of great importance 1972;204:366 (in Russian).
to incorporate results of microscopic observations into [2] Merzhanov AG, Filonenko AK, Borovinskaya IP. Dokl Akad
Nauk SSSR 1973;208:892 (in Russian).
formulations of theoretical works, because information
[3] Merzhanov AG, Borovinskaya IP. Combust Sci Technol
indispensable for constructing the heterogeneous theory 1975;10:195.
is that of microscopic observations. In this context, [4] Merzhanov AG. Archivum Combustionis 1981;1:23.
further accumulation of experimental results through [5] Frankhouser WL, Brendley KW, Kieszek MC, Sullivan ST.
microscopic observations is strongly advised, in order Gasless combustion synthesis of refractory compounds. Park
to refine the heterogeneous theory. Ridge: Noyes, 1985.
In addition to microscopic observations, experiments in [6] Munir ZA. Ceram Bull 1988;67:342.
the gravity free environment might be of great value in [7] Munir ZA, Anselmi-Tamburini U. Mater Sci Rep
obtaining further fundamental understanding, although this 1989;3:277.
experiments cannot be easily conducted on earth. For the [8] Merzhanov AG. Self-propagating high-temperature synth-
esis: twenty years of search and findings. In: Munir ZA,
SHS process, natural convection would exert little influence
Holt JB, editors. Combustion and plasma synthesis of high-
in removing heat from the combustion wave, compared to temperature materials, New York: VCH Publishers, 1990. p.
that of thermal radiation, but density difference between the 1–53.
nonmetal particles and the molten metal in the combustion [9] Varma A, Lebrat J-P. Chem Engng Sci 1992;47:2179.
wave (by yielding local change of stoichiometry, thickness [10] ISMAN, Self-propagating high-temperature synthesis of
of the combustion wave, etc.), could exert influences, inorganic compounds. In: ISMAN leaflet, Institute of Struc-
because of the existence of gravity. tural Macrokinetics, Russian Academy of Science, Cherno-
Although not specifically covered in the present golovka, 1993.
review, the field activated SHS might also prove a [11] Merzhanov AG, Dvoryankin AV, Strunina AG. Dokl Akad
fertile area of research because this is indispensable Nauk SSSR 1982;267:869 (in Russian).
[12] Strunina AG, Dvoryankin AV, Merzhanov AG. Combust
for low exothermic systems to initiate and/or sustain
Explos Shock Waves 1983;19:158.
combustion wave. In addition to the effects of external [13] Novozhilov BV. Dokl Akad Nauk SSSR 1961;141:151 (in
heating, if influences of electric and/or magnetic fields Russian).
on profiles of reactant concentrations could be utilized [14] Merzhanov AG. Combust Flame 1969;13:143.
in producing advanced materials, useful and multi- [15] Maksimov EI, Shkadinskii KG. Combust Explos Shock
functional characteristics would be generated by inducing Waves 1971;7:392.
heterogeneity in composition and/or texture. [16] Aldushin AP, Martem’yanova TM, Merzhanov AG, Khaikin
Finally, it may be informative to point out the similarity BI, Shkadinskii KG. Combust Explos Shock Waves
between spray combustion and the SHS process, as has been 1973;9:531.
mentioned in the present review. This similarity also [17] Vol’pert VA, Vol’pert AI, Merzhanov AG. Combust Explos
Shock Waves 1983;19:435.
suggests that the SHS process might be used for the basic
[18] Khaikin BI, Khudyaev SI. Dokl Phys Chem, Acad Sci USSR
research of spray combustion, especially for identifying 1979;245:225.
effects of dominant parameters on the flame propagation, [19] Aldushin AP, Khaikin BI, Shkadinskii KG. Combust Explos
extinction, and/or initiation, because influence of thermal Shock Waves 1976;12:725.
expansion which usually conceal those of the other [20] Firsov AN, Shkadinskii KG. Combust Explos Shock Waves
parameters is relatively weak in the SHS process. 1987;23:288.
A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74 73

[21] Boddington T, Cottrell A, Laye PG. Combust Flame [52] Novikov NP, Borovinskaya IP, Merzhanov AG. Combust
1989;76:63. Explos Shock Waves 1974;10:157.
[22] Margolis SB, Williams FA. Flame propagation in solids and [53] Itin VI, Bratchikov AD, Merzhanov AG, Maslov VM.
high-density fluids with Arrhenius reactant diffusion. In: Combust Explos Shock Waves 1981;17:293.
Sandia Report, SAND90-8508, Sandia National Labora- [54] Itin VI, Bratchikov AD, Merzhanov AG, Doronin VN.
tories, 1990. Combust Explos Shock Waves 1982;18:536.
[23] Merzhanov AG, Khaikin BI. Prog Energy Combust Sci [55] Shcherbakov VA, Pityulin AN. Combust Explos Shock
1988;14:1. Waves 1983;19:631.
[24] Margolis SB. Prog Energy Combust Sci 1991;17:135. [56] Maksimov YuM, Merzhanov AG, Raskolenko LG, Pak AT,
[25] Dunmead SD, Readey DW, Semler CE, Holt JB. J Am Lepakova OK. Dokl Akad Nauk SSSR 1986;286:911 (in
Ceram Soc 1989;72:2318. Russian).
[26] Shkiro VM, Borovinskaya IP. Study of combustion mechan- [57] Maslov VM, Borovinskaya IP, Merzhanov AG. Combust
ism for mixtures of titanium with carbon. In: Merzhanov AG, Explos Shock Waves 1976;12:631.
editor. Combustion processes in chemical technology and [58] Naiborodenko YuS, Itin VI. Combust Explos Shock Waves
metallurgy. Akad. Nauk SSSR, OIKhF, 1975. p. 253–8 (in 1975;11:293.
Russian). [59] Naiborodenko YuS, Itin VI. Combust Explos Shock Waves
[27] Nekrasov EA, Maksimov YuM, Ziatdinov MKh, Shteinberg 1975;11:626.
AS. Combust Explos Shock Waves 1978;14:575. [60] Bratchikov AD, Merzhanov AG, Itin VI, Khachin VN,
[28] Deevi SC. J Mater Sci 1991;26:2662. Dudarev EF, Gyunter VE, Maslov VM, Chernov DB. Sov
[29] Hardt AP, Phung PV. Combust Flame 1973;21:77. Powder Metall Met Ceram 1980;19:5.
[30] Armstrong R. Combust Sci Technol 1990;71:155. [61] Itin VI, Khachin VN, Bratchikov AD, Gyunter VE, Chernov
[31] Merzhanov AG. Combust Sci Technol 1995;105:295. DB. Inorg Mater 1980;16:1337.
[32] Makino A, Law CK. Heterogeneous flame propagation in the [62] Samsonov GW, Vinitskii IM. Handbook of refractory
self-propagating high-temperature synthesis (SHS) process: compound. New York: Plenum Press, 1980.
theory and experimental comparisons. In: The 24th Sympo- [63] Graham D. Summary of impurity diffusion in the beta phase
sium (International) on Combustion. The Combustion Insti- of titanium. In: Wheeler Jr. JA, Winslow FR, editors. Diffu-
tute, Pittsburgh. 1992. p. 1883–91. sion on body-centerd cubic metals. Metals Park: American
[33] Makino A, Araki N, Kuwabara T. Trans JSME Ser B Society for Metal, 1965. p. 27–34.
1992;58:1925 (in Japanese); JSME Int’l J Ser. B (English [64] Makino A, Law C K. Propagation and diffusion-limited
Translation) 1994;37:576. extinction of nonadiabatic heterogeneous flame in the SHS
[34] Williams FA. Theory of laminar flames. Combustion theory. process. In: The Twenty-Fifth Symposium (International) on
2nd ed. Menlo Park: Benjamin/Cummings, 1985. p. 130–81. Combustion. The Combustion Institute, Pittsburgh, 1994, p.
[35] Williams FA. Spray combustion. Combustion theory. 2nd ed. 1659–67.
Menlo Park: Benjamin/Cummings, 1985. p. 446–84. [65] Williams FA. Ignition. Extinction, and flammability limit. in
[36] Makino A. Combust Flame 1992;90:143. combustion theory. 2nd ed. Menlo Park: Benjamin/
[37] Makino A, Law CK. Combust Flame 1995;101:551. Cummings, 1985. p. 265–93.
[38] Kanury AM. Metall Trans A 1992;23:2349. [66] Makino A, Law CK. J Am Ceram Soc 1996;79:3097.
[39] Huque A, Kanury AM. Combust Sci Technol 1993;89:27. [67] Munir ZA, Lai W. Combust Sci Tech 1992;88:201.
[40] Makino A. Ann Chim Fr 1995;20:143. [68] Munir ZA, Sata N. Int J SHS 1992;1:355.
[41] Touloukian YS. Thermophysical properties of high tempera- [69] Touloukian YS, DeWitt DP. Thermophysical properties of
ture solid materials. Thermophysical properties research matter, vol. 8. New York: IFI/Plenum Press, 1972.
center, Purdue University, vol. 5, New York: MacMillan, 1967. [70] Itin VI, Bratchikov AD, Doronin VN, Pribytkov GA. Sov
[42] Kirdyashkin AI, Maksimov YuM, Merzhanov AG. Combust Phys J (English translation) 1981;24:1134.
Explos Shock Waves 1981;17:591. [71] Itin VI, Naiborodenko YuS, Bratchikov AD, Butkevich NP,
[43] Makino A, Kita K, Araki N. Nensho no Kagaku to Gijutsu Korostelev SV, Sholokhova LV. Sov Phys J (English trans-
(Combust Sci Technol, Jpn) 1993;1:165 (in Japanese). lation) 1975;18:408.
[44] Bloshenko VN, Bokii VA, Borovinskaya IP, Merzhanov AG. [72] Makino A, Law CK. Int J SHS 1995;4:25.
Combust Explos Shock Waves 1984;20:676. [73] Kaper HS, Leaf GK, Margolis SB, Matkowsky BJ. Combust
[45] Hardt AP, Holsinger RW. Combust Flame 1973;21:91. Sci Technol 1987;53:289.
[46] Azatyan TS, Mal’tsev VM, Merzhanov AG, Seleznev VA. [74] Margolis SB, SIAM. J Appl Math 1983;43:351.
Combust Explos Shock Waves 1977;13:156. [75] Makino A, Law CK. Combust Sci Technol 1995;106:193.
[47] Makino A, Law CK. J Am Ceram Soc 1994;77:778. [76] Makino A. SHS flame propagation and extinction for titanium
[48] Touloukian YS, Powell RW, Ho CY, Nicolaou MC. Thermal and diamond system: theory and experimental comparisons. In:
diffusivity, Thermophysical properties of matter, The TPRC Chan SH, editor. Transport phenomena in combustion, vol. 1.
Data Series, vol. 10. New York: IFI/Plenum Press, 1973. Washington: Taylor and Francis, 1996. p. 678–88.
[49] Kirdyashkin AI, Maksimov YuM, Merzhanov AG. Combust [77] Makino A, Yamada T, Araki N. Nensho no Kagaku to
Explos Shock Waves 1981;17:591. Gijutsu (Combust Sci Technol, Jpn) 1995;3:41 (in Japa-
[50] Zenin AA, Merzhanov AG, Nersisyan GA. Combust Explos nese).
Shock Waves 1981;17:63. [78] Makino A, Sugiyama K, Araki N. Nensho no Kagaku to
[51] Borovinskaya IP, Merzhanov AG, Novikov NP, Filonenko Gijutsu (Combust Sci Technol, Jpn) 1996;4:93 (in Japanese).
AK. Combust Explos Shock Waves 1974;10:2. [79] Makino A. Energy Convers Mgmt 1997;38:1043.
74 A. Makino / Progress in Energy and Combustion Science 27 (2001) 1–74

[80] Makino A, Izumi H, Araki N. Nensho no Kagaku to Gijutsu [104] Averson ÉA, Barzykin VV, Merzhanov AG. Dokl Phys
(Combust Sci Technol, Jpn) 1998;6:65 (in Japanese). Chem, Acad Sci USSR 1968;178:1.
[81] Makino A, Law CK. Submitted for publication. [105] Strunina AG, Martem’yanova TM, Barzykin VV, Ermakov
[82] Knyazik VA, Merzhanov AG, Solomon VB, Shteinberg AS. VI. Combust Explos Shock Waves 1974;10:449.
Combust Explos Shock Waves 1985;21:333. [106] Strunina AG, Vaganova NI, Barzykin VV. Combust Explos
[83] Yamada O, Miyamoto Y, Koizumi M. J Mater Res Shock Waves 1977;13:707.
1986;1:275–9. [107] Strunina AG, Ermakov VI, Barzykin VV. Combust Explos
[84] Feng A, Munir ZA. Metall Mater Trans B 1995;26:581. Shock Waves 1978;14:441.
[85] Feng A, Munir ZA. Metall Mater Trans B 1995;26:587. [108] Strunina AG, Ermakov VI, Averson ÉA. Combust Explos
[86] Feng A, Munir ZA. J Am Ceram Soc 1996;79:2049. Shock Waves 1979;15:484.
[87] Munir ZA. Int J SHS 1997;6:165. [109] Strunina AG, Ermakov VI, Kostin SV, Barzykin VV.
[88] Filimonov IA, Kidin NI. SHS-process in external electric Combust Explos Shock Waves 1981;17:351.
fields. In: The 24th Symposium (International) on [110] Strunina AG, Firsov AN, Kostin SV. Combust Explos Shock
Combustion. The Combustion Institute, Pittsburgh, 1992. Waves 1981;17:500.
p. 1893–8. [111] Law CK. Int J Heat Mass Transfer 1978;21:1363.
[89] Makino A. Investigation of effects of electric field on the [112] Enig JW. Proc R Soc A 1968;305:205.
combustion behavior in the SHS process by use of the hetero- [113] Yamada T, Makino A, Araki N. Measurement of ignition
geneous theory. In SHS-99, Moscow, Russia, 1999. energy in the SHS process for Ti–C system. In: Proceedings
[90] Thermophysical Properties Handbook. Japan Society of of the 32nd Japanese Symposium on Combustion. The Japanese
Thermophysical Properties, editor. Tokyo: Yokendo, 1990. Section of the Combustion Institute, Sendai, 1994. p. 579–81.
p. 232 (in Japanese). [114] Makino A. Initiation of SHS flame induced by another SHS
[91] Touloukian YS. Thermophysical properties of high tempera- flame—determination of the ignition energy. In: Proceedings
ture solid materials, vol. 1. Thermophysical properties of the Second International Symposium on Advanced Energy
research center, Purdue University. New York: Macmillan, Conversion Systems and Related Technologies, RAN98,
1965. p. 85 1998. p. 242–3.
[92] Makino A, Law CK. Pulsating instability in the nonadiabatic [115] Kanury AM, Hernandez-Guerrero A. Combust Sci Technol
heterogeneous SHS flame: theory and experimental compar- 1995;104:181.
isons. In: The 26th Symposium (International) on Combus- [116] Liñán A, Williams FA. Combust Sci Technol 1971;3:91.
tion, The Combustion Institute, Pittsburgh, 1996. p. 1867– [117] Linan A, Williams FA. Combust Flame 1972;18:85.
74. [118] Liñán A, Crespo A. Combust Sci Technol 1972;6:223.
[93] Shkadinsky KG, Shkadinskaya GV, Matkowsky BJ, Volpert [119] Niioka T, Williams FA. Combust Flame 1977;29:43.
VA. Combust Sci Technol 1992;88:271. [120] Niioka T. Combust Sci Technol 1978;18:207.
[94] Butakov AA, Vaganov DA, Leont’ev SN. Combust Sci [121] Makino A. Initiation of the heterogeneous flame in the self-
Technol 1995;106:137. propagating high-temperature synthesis (SHS) process
[95] Makino A, Law CK. On the transition boundary from steady induced by radiative heat flux. In: ASPACC99, Taiwan,
to pulsating combustion in SHS flames. In: The 27th Sympo- 1999.
sium (International) on Combustion, The Combustion Insti- [122] Makino A. Initiation of the heterogeneous SHS by radiative
tute, Pittsburgh, 1998. p. 2469–76. heat flux. In: Proceedings of the 37th Japanese Symposium
[96] Margolis SB, Kaper HG, Leaf GK, Matkowsky BJ. Combust on Combustion. The Japanese Section of the Combustion
Sci Technol 1985;43:127. Institute, Chiba, 1999. p. 395–6.
[97] Margolis SB. Matkowsky. SIAM J Appl Math 1988;48:828. [123] Makino A, Law CK. Transient radiative initiation of the
[98] Margolis SB. Proc R Soc London A 1991;433:131. heterogeneous flame in SHS: theory and experimental
[99] Margolis SB, SIAM. J Appl Math 1991;51:693. comparisons. In: The 28th International Symposium on
[100] Margolis SB. Combust Sci Technol 1993;88:223. Combustion, The Combustion Institute, Edinburgh, 2000
[101] Barzykin VV, Merzhanov AG, Strunina AG. Ignition of (submitted for presentation).
heterogeneous systems containing condensed reaction [124] Niioka T, Hasegawa S. Ignition time of locally heated reac-
products. In: The 23rd Symposium (International) on tive materials. In: Rep. Inst. Fluid Sci., vol. 1, Tohoku
Combustion, The Combustion Institute, Pittsburgh, 1990. p. University, 1990.
1725–31. [125] Carslaw HS, Jaeger JC. Conduction of Heat in Solids. 2nd ed.
[102] Barzykin VV. Pure Appl Chem 1992;64:909. Oxford: Clarendon Press, 1959. p. 264.
[103] Makino A. Dependence of the ignition delay time on various [126] Kaise T, Makino A, Araki N. Effects of Various parameters
parameters in the SHS process initiated by another SHS flame. on the ignition delay time in the SHS process for Ti–C
In: Proceedings of the 5th ASME/JSME Joint Thermal Engi- system. In Proceedings of the 34th JSME Conference in
neering Conference, AJTE99-6321, San Diego, CA, 1999. Tohoku Section, Sendai, 1999. p. 127–8.

Vous aimerez peut-être aussi