Chinese Journal of Aeronautics, (2022), 35(5): 319–332

Chinese Society of Aeronautics and Astronautics

& Beihang University
Chinese Journal of Aeronautics
cja@buaa.edu.cn
www.sciencedirect.com

Detailed modeling of aluminum particle combustion

– From single particles to cloud combustion in
Bunsen flames
Jiarui ZHANG a, Oliver T. STEIN b, Tien D. LUU b, Ali SHAMOONI b,
Zhixun XIA a, Zhenbing LUO a, Likun MA a,*, Yunchao FENG a,
Andreas KRONENBURG b

a
College of Aerospace Science and Engineering, National University of Defense Technology, Changsha 410073, China
b
Institut für Technische Verbrennung, Universität Stuttgart, Stuttgart 70174, Germany

Received 23 August 2021; revised 14 September 2021; accepted 18 October 2021

Available online 17 December 2021

KEYWORDS Abstract A numerical model for aluminum cloud combustion which includes the effects of inter-
Aluminum particle combus- phase heat transfer, phase change, heterogeneous surface reactions, homogeneous combustion,
tion; oxide cap growth and radiation within the Euler–Lagrange framework is proposed. The model is
Bunsen flame; validated in single particle configurations with varying particle diameters. The combustion process
Burning time; of a single aluminum particle is analyzed in detail and the particle consumption rates as well as the
Burning rate; heat release rates due to the various physical/chemical sub-models are presented. The combustion
Particle cloud time of single aluminum particles predicted by the model are in very good agreement with empirical
correlations for particles with diameters larger than 10 lm. The prediction error for smaller parti-
cles is noticeably reduced when using a heat transfer model that is capable of capturing the transi-
tion regime between continuum mechanics and molecular dynamics. The predictive capabilities of
the proposed model framework are further evaluated by simulating the aluminum/air Bunsen
flames of McGill University for the first time. Results show that the predicted temperature distri-
bution of the flame is consistent with the experimental data and the double-front structure of the
Bunsen flame is reproduced well. The burning rates of aluminum in both single particle and particle
cloud configurations are calculated and compared with empirical correlations. Results show that
the burning rates obtained from the present model are more reasonable, while the correlations,
when embedded in the Euler–Lagrange context, tend to underestimate the burning rate in the com-
bustion stage, particularly for the considered fuel-rich flames.
Ó 2021 Chinese Society of Aeronautics and Astronautics. Production and hosting by Elsevier Ltd. This is
an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

* Corresponding author.
E-mail address: malikun@nudt.edu.cn (L. MA).
https://doi.org/10.1016/j.cja.2021.12.005
1000-9361 Ó 2021 Chinese Society of Aeronautics and Astronautics. Production and hosting by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
320 J. ZHANG et al.

1. Introduction determine the burning rate of the particles.9,11,16 In this case,

macro parameters required for the correlations, such as the
Aluminum is widely used as an additive in solid rocket propel- gas temperature and oxygen concentration, that refer to the
lants1,2 and pyrotechnics3 owing to its outstanding characteris- far-field conditions of the combustion chamber in the original
tics such as high energy density and combustion temperature, experiments are approximated by the state of the gas that
environmentally benign products, and good stability. In the immediately surrounds the target particle.22,16 The such-
past few decades, significant progress has been made in studies obtained governing scalars are then used to calculate the con-
of ignition and combustion of single isolated aluminum parti- sumption rate of the particles at every time step, which may
cles,4 and Dn models with very low computational cost have lead to significant errors.
been established for nano-5 and micron-sized6 particles, respec- To overcome these limitations, the objective of the present
tively. Since the typical mass fraction of aluminum particles in study is to establish a detailed numerical model for aluminum
high-energy materials is less than 20%,7 the interactions particle cloud combustion in the Euler–Lagrange framework,
among different metal particles can typically be neglected.8 which is based on fundamental physics and avoids the use of
In this case, correlation models for particle ignition9 and com- very general empirical correlations for key quantities, such as
bustion10 based on experimental data of single aluminum par- ignition temperature and burning time. The model considers
ticles are applicable to the simulation of solid propellant the effects of interphase heat transfer, phase change, heteroge-
combustion.11 neous surface reactions and homogeneous chemistry, as well as
In recent years, a process referred to as metal fuel cycle12 oxide cap growth. By simultaneously integrating these physico-
has been proposed, in which metal fuel represented by alu- chemical sub-models, their interaction is directly computed,
minum13 acts as a clean and renewable energy carrier to miti- for example the competing effect of evaporation and heteroge-
gate climate change. One of the related energy conversion neous surface reactions for particle consumption, among
methods is to burn the metal fuel directly with air in a combus- others. Another contribution of this work is that the radiative
tor to provide clean high-grade heat.14 In the scenario of heat transfer is modeled by solving the full 3D radiative trans-
metal-fuelled burner, the aluminum particles provide the pri- fer equation, and the effects of dispersed particles on radiation
mary fuel, the concentration of which can be as high as 600 emission, absorption and scattering are all included. The new
model is then used to study single particle combustion and
g=m3 , resulting in an average inter-particle spacing of about
Bunsen flames of aluminum particles in detail, and the exper-
13 particle diameters15; therefore, thermal feedback between
imental aluminum Bunsen flame is reproduced numerically for
particles cannot be neglected in particle cloud combustion.
the very first time, which acts as a solid validation case for the
Moreover, experiments suggest that rich metal-fuel suspen-
present model.
sions are usually required to achieve stable combustion,15
and under this circumstance, the conversion process of most
particles is supposed to occur in the near-zero-oxygen region.16 2. Modeling
Therefore, the diffusion of the evaporated gaseous aluminum
and subsequent gas-phase homogeneous reaction should be 2.1. Modeling of dispersed phase
considered along with the heterogeneous reaction.
In the field of metal particle suspension combustion, recent The dispersed phase particles under consideration feature a
experimental investigations have shed some light on the com- spherical core of solid aluminum (Al) surrounded by a thin
bustion characteristics and global flame properties,17,18 but surface layer of aluminum oxide (alumina, Al2 O3 ). The Al core
detailed information on flame structure, mass and heat transfer has an effective diameter dp;Al , while the total particle diameter
processes is difficult to infer from experiments. However, such comprised of Al and Al2 O3 is denoted as dp . It is noted that
information can be obtained from numerical simulations of there are two formation routes of the final product of the oxi-
metal particle cloud combustion. Huang et al.9 performed a dation process, Al2 O3 , in the present study. The first is Hetero-
detailed analysis which evaluated the effect of particle size at geneous Surface Reaction (HSR), which leads to the newly
the nano and micron scale on aluminum dust flames. Later, formed Al2 O3 directly condensing on the particle surface and
a one-dimensional steady-state model was proposed by Khalili remaining a part of the Lagrangian dispersed phase. The sec-
et al.19, in which the effects of thermophoresis and Brownian ond type of Al2 O3 is the final product of homogeneous Al oxi-
motion were included. Recently, Han and Sung16 proposed a dation, which exists as a finely dispersed smoke and attributes
detailed numerical model including radiation and particle sur- to the continuous Eulerian phase. Moreover, it is common
face reactions, and the propagation process of aluminum dust practice to assume that Al particles with diameters larger than
flames under various conditions in a half-open channel was 20 lm burn in the diffusion-controlled regime, while for nano-
studied. Then, Zou et al.20 extended the model in Ref.16 to cap- sized Al particles heterogeneous surface reactions become
ture the combustion properties of aluminum nanoparticle more important, which indicates kinetically-controlled com-
clouds. They approximated the flame speed by the propagation bustion.23 The diameter of the Al particles used in the target
speed of the ignition front, with the latter being defined by the Bunsen flame is about 5.6 lm, and their combustion mode
minimum gas temperature that can ignite a single aluminum should therefore be under the joint control of diffusion and
particle.21 chemical kinetics. Hence, both heterogeneous surface reactions
An inherent limitation of many previous numerical studies and homogeneous gas reactions, as well as diffusion processes
is the direct application of experimental correlations derived in the gas phase are considered in the combustion process of
from single particle combustion measurements in particle the Al particles in the present study.
cloud models. For instance, particle burning time obtained The dispersed particles are described in the Lagrangian
from experimental data has been used in previous studies to framework and their interaction with the gas phase is charac-
Detailed modeling of aluminum particle combustion 321

terized by the transfer of mass, momentum and energy. More- to the transition regime for heat transfer. Therefore, the Kava-
over, particle collisions are neglected in the present study. This nau correlation28 for the Nusselt number in the transition
assumption is reasonable because the dust concentration of the regime is employed in the present study
Al dust flame under consideration is about 500 g=m3 , corre-
sponding to an average particle spacing of about 14 particle Nup;Ranz
Nup;Kava ¼ ð8Þ
diameters. 1 þ 3:42Nup;Ranz ReMa
p Prp
The evolution of the particle temperature, Tp , is written as
dTp where Ma is the Mach number evaluated within the Eulerian
mp cp;p ¼ Q_ p;inter þ Q_ p;melt þ Q_ p;HSR þ Q_ p;evap þ Q_ p;rad ð1Þ cell surrounding the particle. Eq. (8) is taken as an alternative
dt
to Eq. (7). Unless specified otherwise, the Kavanau model is
with the particle mass mp and the specific heat of the particle
used for all simulations throughout this paper.
cp;p . Q_ p;inter ; Q_ p;melt ; Q_ p;HSR ; Q_ p;evap and Q_ p;rad are source terms
of particle temperature due to interphase heat transfer, melt- 2.2.2. Melting model
ing, HSR, evaporation and radiation respectively, and their
details will be presented in Section 2.2. When the particle temperature reaches the melting point of
Assuming that only drag force acts on the Al particle, the aluminum or alumina and the corresponding solid mass frac-
particle momentum equation reads tion of the particle is greater than zero, the left hand side of
Eq. (1) is zero and the melting rate can be calculated as
dup 3qg CD
 
¼ ug  up ug  up  ð2Þ Q_ p;melt Q_ p;inter þ Q_ p;HSR þ Q_ p;evap þ Q_ p;rad
dt 4qp dp m_ p;melt ¼  ¼ ð9Þ
hp;melt hp;melt
where qg ; qp ; ug and up are gas density, particle density, gas
velocity and particle velocity, respectively. CD is the drag coef- where hp;melt is the latent heat of fusion.29
ficient calculated as24
 
24 1 2.2.3. Surface reaction model
CD ¼ 1 þ Re2=3 Rep 6 1000 ð3Þ
Rep 6 p The HSR model used in the present study is the one proposed
by Gurevich et al.30 . The consumption rate of Al in the dis-
with the particle Reynolds number Rep computed as
  persed phase by HSR reads
Rep ¼ qug  up dp =l ð4Þ 
Ea

m_ Al
p;HSR ¼ Ap;eff qs Yox;s Ar exp  ð10Þ
where l is dynamic gas viscosity. RTp
The change of the particle mass is due to evaporation and
where Ap;eff is the surface area of the particle free of oxide film,
HSR, and can therefore be expressed as
which is thought to be the effective area for HSR, R is the uni-
m_ p ¼ ðm_ p;evap þ m_ p;HSR Þ ð5Þ versal gas constant, qs the gas density on particle surface, and
with the models for m_ p;evap and m_ p;HSR presented in Section 2.2. Yox;s the mass fraction of oxidizer on the particle surface, i.e.
oxygen in the present study. The Arrhenius coefficients for
the surface reaction,
2.2. Aluminum combustion sub-models
Al þ 0:75ðO2 þ 3:76N2 Þ ! 0:5Al2 O3 þ 2:82N2 ð11Þ
The models to describe the principal physical and chemical
are set to Ar ¼ 1:5  104 m/s and Ea ¼ 83:72 kJ/mol, respec-
processes occurring in aluminum particle combustion are pro-
tively. The effective particle surface Ap;eff in Eq. (10) is assumed
vided in the present section, considering interphase heat trans-
to be the surface of the Al core, the diameter of which is calcu-
fer, melting, evaporation, surface reaction, and radiation.
lated as16
sffiffiffiffiffiffiffiffiffiffiffiffiffi
2.2.1. Interphase heat transfer model 3 6mp;Al
dp;Al ¼ ð12Þ
Q_ p;inter in Eq. (1) is the interphase heat transfer rate calculated pqAl
by

 where mp;Al is the mass of aluminum in the particle, qAl is the
Q_ p;inter ¼ pdp kg Tg  Tp Nup ð6Þ density of aluminum.
where kg is the thermal conductivity, and Tg and Tp are the The only product of HSR except of N2 is condensed phase
temperatures of the gas and particle, respectively. Nup is the Al2 O3 , which is assumed to be directly deposited on the parti-
Nusselt number that can be calculated from the Ranz- cle surface. Therefore, the mass rate of change in Eq. (5) can be
Marshall model25 calculated as
WAl2 O3 Al
Nup;Ranz ¼ 2 þ 0:6Re1=2 1=3
p Prp ð7Þ m_ p;HSR ¼ m_ Al
p;HSR  0:5 m_ p;HSR ð13Þ
WAl
where Prp is the particle Prandtl number.
and the source term of particle temperature in Eq. (1) due to
When particle sizes are of the order of microns, the treat-
HSR is obtained as
ment of the surrounding gas as a continuum may become ques-
tionable.26 Following Ref. 26, particles with a diameter of Q_ p;HSR ¼ m_ Al
p;HSR hp;HSR ð14Þ
5.6 lm, which is the size of the aluminum particles employed
in the Bunsen flame experiments27 targeted here, are subject where hp;HSR is the heat release from the surface reaction,
which is calculated as
322 J. ZHANG et al.

WAl2 O3 where fAl2 O3 is the volumetric fraction of alumina smoke, which

hp;HSR ¼ hAl ðTp Þ  0:5 hAl2 O3 ðTp Þ ð15Þ
WAl is approximated by its mole fraction in the gas. The absorption
with Wi and hi denoting the molar mass and total enthalpy of coefficient of alumina, jAl2 O3 , is obtained from
species iði ¼ Al; Al2 O3 Þ. ln ð1  eAl2 O3 Þ
It is noted that a detailed surface reaction model can be jAl2 O3 ¼  ð22Þ
L
employed by considering different species on the particle sur-
where L is the path length, and eAl2 O3 the emissivity of alumina,
face.31 However, coupling this process into the present
which is calculated by polynomial fitting using the experimen-
Euler–Lagrangian method requires solving dozens of Arrhe-
tal data from Ref. 34. The path length L is calculated as33
nius equations for each particle, which is unrealistic for future
simulations of large-scale burner. Vd
L ¼ 3:6 ð23Þ
Ad
2.2.4. Evaporation model
where Vd is the volume of the radiation domain and Ad is the
Considering the effect of the oxidation cap on the effective sur- surface area of the domain where radiation is absorbed.
face area of particle evaporation, the evaporation rate of the The radiative heat exchange term in Eq. (1) is calculated as
molten particle is described by 
Q_ p;rad ¼ ep Ap;s r H4r  T4p ð24Þ
m_ p;evap ¼ pð1  bÞdp;Al ShqDF ln ð1 þ BM Þ ð16Þ
where b is the fraction of the aluminum sphere covered by the where ep ¼ 0:12 is the particle emissivity,35 Ap;s is the particle
oxide cap (Section 2.2.6), and DF is the diffusion coefficient of surface area, and Hr is the Eulerian phase radiation tempera-
the fuel vapor in the mixture. Note that here the effective ture, which is calculated as
diameter dp;Al given in Eq. (12) is employed to consider the hin-  1=4
G
dering effect of the oxide film on evaporation, instead of using Hr ¼ ð25Þ
4r
the entire particle diameter dp . Sh is the Sherwood number. BM
is the Spalding mass transfer number defined as where r is the Stefan–Boltzmann constant. G is the incident
YF;s  YF;1 radiation, which is calculated by the discrete ordinates method
BM ¼ ð17Þ 36,37
in the present study.
1  YF;s
where YF;s and YF;1 are the mass fractions of the evaporated 2.2.6. Oxide cap growth model
fuel at the droplet surface and in the far-field, respectively. When the particle temperature exceeds the melting point of
The mass fraction of fuel vapor at the particle surface is calcu- alumina, the oxide layer slides on the surface of the liquid alu-
lated as minum core to form an oxide cap due to surface tension. In the
XF;s WF present study, oxide cap growth model proposed in Refs. 38,39
YF;s ¼ ð18Þ
XF;s WF þ ð1  XF;s ÞWnonF is employed to consider the effect of the oxide cap on the reac-
tive surface of the Al particle. Based on the geometric relation-
where WF and WnonF are the molar mass of the fuel vapor and ship shown in Fig. 1,39 the following equations can be
non-vapor gas, respectively. X denotes the mole fraction. The obtained:
fuel vapor mole fraction at the surface is obtained from
Raoult’s law VAl2 O3 ¼ 43 pR32  13 pðR1  d1 Þ2 ð2R1 þ d1 Þ
ð26Þ
PF;sat  13 pðR2  d2 Þ2 ð2R2 þ d2 Þ
XF;s ¼ ð19Þ
Pa
d21 ¼ R21  h2 ð27Þ
where PF;sat is the fuel vapor saturation pressure at droplet
temperature obtained from Ref. 32, and Pa is the ambient pres-
d22 ¼ R22  h2 ð28Þ
sure. Therefore, the energy source term due to evaporation in
Eq. (1), Q_ p;evap , can be obtained as
Q_ p;evap ¼ m_ p;evap hp;evap ð20Þ
where hp;evap is the latent heat of evaporation.

2.2.5. Radiation model

Compared with the simplified 1D radiation models used in pre-
vious studies,16,20 radiation is modeled by solving the full 3D
Radiative Transfer Equation (RTE)33 in the present study
under the assumption of grey particles and grey gas with neg-
ligible scattering. The condensed phase alumina is assumed to
exist as a very fine smoke,16 which we model as a continuous
Eulerian species with an emissivity much higher than the
remaining gas phase. Therefore, the gas absorption/emission
coefficient, jg , is calculated as
jg ¼ fAl2 O3 jAl2 O3 ð21Þ
Fig. 1 Geometry of aluminum droplet with oxide cap.39
Detailed modeling of aluminum particle combustion 323

  
h¼
R1 R2
ð29Þ @ðqg ht Þ @ qg ht uj @ @ht
d1 þ d2 þ ¼ qg ag þ S_ h;s þ S_ g;rad ð37Þ
@t @xj @xj @xj
where R1 is the particle radius of the aluminum core, which where ht is the total enthalpy (formation + sensible) and ag
can be calculated from Eq. (12), and VAl2 O3 is the volume of
thermal diffusivity. S_ g;rad is the source term due to radiation.
the Al2 O3 cap, which is computed dynamically via the Al2 O3 
mass accumulation term through heterogeneous surface reac- S_ g;rad ¼ 4jg r H4r  T4g ð38Þ
tions. The remaining unknowns, R2 ; d1 ; d2 and h, can be
obtained by solving Eqs. (26)–(29), then the angle a is calcu- The source term S_ h;s in Eq. (37) represents the energy
lated as exchange with the solid phase, which reads
 
h N 
a ¼ arcsin ð30Þ 1 X
R1 S_ h;s ¼  Q_ p;inter  m_ np;evap hg;evap  m_ Al;n
p;HSR hg;HSR ð39Þ
Vcell n¼1
and the solid angle of the oxide cap # can be calculated as
# ¼ 2pð1  cos aÞ ð31Þ where Q_ p;inter is the heat exchange with the solid phase through
The fraction of the aluminum sphere covered with oxide, interphase heat transfer obtained by Eq. (6). hg;evap is the total
which is used in Eq. (16), is calculated as enthalpy of the evaporated gas from the particle. hg;HSR is the
enthalpy change due to oxygen consumption by HSR,
# 1
b¼ ¼ ð1  cos aÞ ð32Þ hg;HSR ¼ #st hO2 ðTg Þ ð40Þ
4p 2
and updated at every time step. where #st is the stoichiometric oxygen mass fraction required
for HSR, and hO2 is the total enthalpy of oxygen taken at tem-
2.3. Modeling of carrier gas phase perature Tg .
The transport equation for species mass reads


  
2.3.1. Governing equations of gas phase @ qg Yk @ qg Yk uj @ @Yk
þ ¼ qg Dk þ S_ Yk ;p
The total gas mass follows @t @xj @xj @xj


@qg @ qg uj þ S_ Y ;reac ð41Þ
þ ¼ S_ m;s ð33Þ k

@t @xj
where Yk is the mass fraction of species k. Dk is the mass dif-
where uj is the jth component of the gas velocity vector. The fusivity of species k, which is obtained by qDk ¼ l=Sc, and Sc
source term S_ m;s represents the mass exchange with the solid is the Schmidt number assumed to be 0.7. The source terms
phase S_ Yk ;p and S_ Yk ;reac account for mass exchange with the dispersed
N   phase and the production or consumption of species k by
1 X MO2 Al;n
S_ m;s ¼ m_ np;evap  0:75 m_ p;HSR ð34Þ homogeneous reactions, respectively.
Vcell n¼1 MAl
2.3.2. Homogeneous chemistry
where Vcell is the volume of the Eulerian cell where the particles
are located. N is the total number of particles in the cell. MO2 Table 1 shows the homogeneous chemistry adopted for the
and MAl are the molar mass of oxygen and aluminum, respec- present study. It is composed of 8 species and 11 chemical reac-
tively. m_ np;evap is the evaporation rate of aluminum obtained by tions.9 Since Al2 O3 does not prevail as a stable gas phase
species,9 an irreversible reaction is used to describe the transi-
Eq. (16). m_ Al;n
p;HSR is the consumption rate of aluminum calcu- tion from gaseous to condensed phase Al2 O3 , denoted as
lated by Eq. (10). The superscript nindicates the nth particle Al2 O3 ðlÞ in Table 1. As stated above, the condensed alumina
in the cell. Since a one-step HSR is used in present study(Eq. from the homogeneous reaction is assumed to exist as a very
(11)) and the product of the HSR is assumed to directly con- fine smoke and attributed to the Eulerian gas phase, which is
dense on the particle, the second term on the right hand side a common practice in aluminum particle combustion.9,31,16
of Eq. (34) accounts for the oxygen consumption by HSR.
Momentum conservation follows
3. Numerical details
@ðqg ui Þ @ ðqg ui uj Þ @p
@t
þ @xj ¼  @x
h  i ð35Þ 3.1. Computational configurations
i

þ S_ ui ;s
@ui @uj @uk
þ @x@ j l @x j
þ @x i
 2l
3
dij @x k
3.1.1. Single aluminum particle
where p is the pressure, dij is Kronecker delta. The momentum
To evaluate the behavior of the various aluminum sub-models
exchange with the solid phase, S_ ui ;s , is computed as as well as their interaction in the overall combustion model, we
1 X N
dðmnp unp;i Þ consider the combustion of a single particle in a hot coflow
S_ ui ;s ¼  ð36Þ environment. The computational domain features a cuboid
Vcell n¼1 dt
and it spans Lx ¼ 105dp ; Ly ¼ Lz ¼ 65dp . As shown in Fig. 2,
where unp;i is the ith component of the velocity vector of the nth a single aluminum particle is fixed in the flow field at
particle in the cell. ð27:5dp ; 32:5dp ; 32:5dp Þ. According to Ref. 40, the thickness of
Considering the assumption of a unity Lewis number, the the initial oxide layer (Al2 O3 ) of the particle with diameters
energy equation can be written as varying from 0.01 lm to 25 lm is generally between 2 nm
324 J. ZHANG et al.

Table 1 Al/O mechanisms, k ¼ A  Tn exp ðE=RTÞ.

No. Reaction A(cm3 /(mols)) n E(cal/mol)
1 Al+O2 = AlO + O 13 0 159.95
9.7210
2 Al + O + M = AlO + M 3.001017 1 0
3 AlO+O2 = AlO2 +O 4.621014 0 19885.9
4 Al2 O3 = Al2 O2 +O 3.001015 0 97649.99
5 Al2 O3 = AlO2 +AlO 3.001015 0 126999.89
6 Al2 O2 = AlO + AlO 1.001015 0 117900
7 Al2 O2 = Al+AlO2 1.001015 0 148900
8 Al2 O2 = Al2 O+O 1.001015 0 104249.94
9 AlO2 = AlO + O 1.001015 0 88549.86
10 Al2 O = AlO + Al 1.001015 0 133199.94
11 Al2 O3 = Al2 O3 ðlÞ 1.001014 0 0
Notes: k is the reaction rate, A is pre-exponential factor, T is the temperature, R is the universal gas constant, and E is activation energy. 1 cal
= 4.1858 J.

Fig. 2 Computational domain and boundary conditions for

single aluminum particle combustion.

and 4 nm; therefore, it is assumed that the thickness of the

oxide layer of all particles is 4 nm in the present study. The
domain is discretized into 211313 cells, which results in a
mesh resolution of 5dp . The initial gas temperature in the
domain is set to 2500 K. Preheated air flows in from the left Fig. 3 Computational domain and boundary conditions for
boundary with a fixed velocity(U) of 0.4 m/s and a tempera- aluminum particle Bunsen flame.
ture(T) of 2500 K. Zero-gradient outflow conditions are used
for all other boundaries.
the same velocity as the central jet.41 The width of the nozzle
3.1.2. Aluminum Bunsen flame wall is 1 mm in the present configuration. The dust concentra-
The second configuration considered in this study is Al cloud tion at the nozzle outlet is 500 g=m3 , which is the same as the
combustion established on the Bunsen burner of Goroshin experimental operating condition.27 Moreover, all reported
et al. 27, in which an aluminum/air flame was stabilized above quantities of the Al Bunsen flame are averaged within 0.1 s
a nozzle with an exit diameter of 20 mm. To limit the compu- after flame stabilization.
tational cost of the simulations, an axi-symmetric geometry, as
demonstrated in Fig. 3, is used in the present study. The com- 3.2. Numerical setup
putational domain is a wedge with an angle of 3° to represent a
cylinder with a total radius of 40 mm and a length of 80 mm. All simulations reported in this study are carried out using the
The domain is discretized into 200 grid points in the radial open-source CFD package OpenFOAM42 v1912. The underly-
direction and 400 grid points in the axial direction, resulting ing reacting two-phase solver is based on our previous work
in a mesh resolution of 200 lm. As shown in Fig. 3, the alu- documented in Ref. 43 and has been extended to consider the
minum particles are carried by air and have the same inlet detailed Al modeling framework described in Section 2 for
velocity(Ux ) of 0.5 m/s and temperature of 300 K as the carrier the present contribution. There are nearly 130000 particles in
gas (red area). The coflow of air (blue area) is assumed to have the computational domain of the Al Bunsen flames, resulting
Detailed modeling of aluminum particle combustion 325

in an average number of about 5 particles per cell. The simu-

lation cost of the Al Bunsen flame is approximately 5000
CPU hours, in which the time step is set to 5106 s.
The transport equations are spatially discretized with a
finite volume method. The convection terms in the momentum
equation are discretized with a second-order linear upwind dif-
ferencing scheme, while those in the scalar equations are dis-
cretized with a second-order limited linear differencing
scheme. The diffusive terms are discretized with an unbounded
second-order central differencing scheme. For time integra-
tion, the second-order implicit backward scheme is used.

4. Results and discussion

4.1. Single particle combustion

4.1.1. Combustion process analysis

Fig. 4 shows the temporal evolution of several particle-related
properties during the combustion process of a single aluminum
particle with a diameter of 10 lm. Fig. 4 (a) presents the vari-
ation of temperature, diameter, mass and liquid aluminum
content of the particle. m  p and dp are the mass and diameter
of the particle non-dimensionalized by the corresponding ini-
tial values. Yp;AlðlÞ is the mass fraction of aluminum in liquid
phase. Fig. 4 (b) shows the variation of
_ p;HSR and m
Yp;AlðsÞ ; Yp;Al2 O3 ðlÞ ; Yp;Al2 O3 ðsÞ ; m _ p;evap . Yp;AlðsÞ is the
mass fraction of solid aluminum. Yp;Al2 O3 ðlÞ and Yp;Al2 O3 ðsÞ are
the mass fractions of alumina in the liquid and solid phase.
_ p;HSR ; m
m _ p;evap are the Al consumption rates due to HSR and
evaporation, respectively, and both are normalized by the ini-
tial particle mass. Fig. 4 (c) shows the source terms in the par-
ticle energy equation, Eq. (1), and Q_ p;evap ; Q_ p;HSR ; Q_ p;inter ; Q_ p;rad
are source terms due to evaporation, HSR, interphase heat
transfer and radiation, respectively.
As demonstrated in Fig. 4 (a), the particle is initially heated
up to the melting point of Al, 933 K, and its temperature
remains constant until melting is complete, that is,
Yp;AlðsÞ ¼ 0, indicated in Fig. 4 (b). Similarly, when the particle
temperature reaches the melting point of alumina around
0.4 ms, as shown in the inset of Fig. 4 (a), melting of alumina
proceeds at a constant temperature of 2327 K, followed by fur-
ther heat-up of the particle. Fig. 4 (c) reveals that the main
heat source for heating and melting the particle mainly comes
from the interphase heat transfer between the two phases, but
this contribution becomes continuously weaker as the particle
temperature rises. The density of Al decreases with increasing
temperature, and mass loss due to evaporation is small in the
early stage of preheating; therefore, the particle size increases
continuously for 0.14 ms < t < 0.42 ms. As soon as liquid
Al is present in the particle (t > 0:15 ms), i.e., Yp;AlðlÞ > 0,
HSR occurs and soon becomes the main contributor for heat- Fig. 4 Combustion process of a single particle with dp ¼ 10 lm.
ing the particle.
As shown in Figs. 4 (b) and (c), the Al consumption rate
due to HSR along with its heat release rate increase dramati- on the particle temperature. When Tp is lower than the melting
cally at t = 0.3 ms, which in turn increases the particle temper- point of alumina, solid Al2 O3 is deposited on the particle, as
ature to more than 2500 K. Since the product of the surface indicated from 0.3–0.4 ms by the red line in Fig. 4 (b). At
reaction, Al2 O3 , is assumed to condense on the particle, addi- about t = 0.4 ms, right after m _ p;HSR reaches its maximum, liq-
tional mass is gained on the particle, which also accelerates the uid Al begins to evaporate from the particle and the evapora-
increase of the particle diameter, as shown in Fig. 4 (a). tion rate increases sharply as the particle temperature increases
Whether solid or liquid Al2 O3 is produced by HSR depends towards the boiling point of Al. Under the assumption of
326 J. ZHANG et al.

vapor–liquid equilibrium on the particle surface, a higher par- evaporation mainly comes from HSR. Since a unity Lewis
ticle temperature leads to a higher Al vapor mass fraction on number is employed in the present study, the Ranz-Marshall
the particle surface; therefore, the mass fraction of the oxygen model also predicts a higher Sherwood number, resulting in
on the surface decreases. As a result, the rate of HSR is a higher evaporation rate during combustion (not shown).
decreased noticeably, as can be seen at t ¼ 0:4 ms in Fig. 4 This is the reason for the somewhat lower particle mass in
(b). Around the same time, as demonstrated in Fig. 4 (a), when the post-combustion stage when the Ranz-Marshall model is
the evaporation rate is large enough so that the mass consump- used.
tion rate by evaporation exceeds the mass increase rate by To evaluate the effect of the particle diameter on combus-
HSR, the particle mass and diameter begin to decrease. Corre- tion time, eight cases of single aluminum particle combustion
spondingly, the sink term due to evaporation in the particle with different diameters, varying from 1 lm to 80 lm, are sim-
energy Eq. (1) reaches its peak at t ¼ 0:43 ms, followed by a ulated. Fig. 6 presents the combustion time obtained in the
decrease of the particle temperature. The beginning of the present study, compared with the burning time correlations
combustion stage is defined as the particle temperature reach- in Refs. 6,44. It should be noted that these two correlations
ing this local maximum (t ¼ 0:43 ms). This moment also corre- have been derived for aluminum particles of different sizes,
sponds to the maximum evaporation rate, as indicated in i.e., the first one for particles with diameters greater than
Fig. 4 (b). 10 lm,6 and the second one for nano- as well as sub-micron-
During the combustion stage, for 0.43 ms < t < 1.1 ms, the sized particles.44 The results in Fig. 6 indicate that the present
particle is found to be near thermal equilibrium, in which the model can accurately predict the burning time for Al particles
particle temperature is relatively stable. In this period, all heat with diameters greater than 10 lm. Moreover, the difference
generated by the HSR, interphase heat transfer and radiative between the combustion time calculated using the Kavanau
heat transfer is used to evaporate the liquid aluminum in the and Ranz-Marshall models for interphase heat transfer is neg-
particle. As shown in Fig. 4 (c), HSR contributes to the major ligible for particles with diameters larger than 10 lm. How-
part of the heat required for Al evaporation, while radiative ever, the difference between the model predictions can be as
heat transfer is two orders of magnitude smaller than the other high as 39.1% for a particle with diameter of 1 lm, which indi-
heat source terms, indicating that radiation can be neglected cates that the transition regime heat transfer should be consid-
for the studied single particle case. ered for Al particles with diameters smaller than 10 lm.
The end of the combustion stage is marked at t ¼ 1:1 ms in It can be seen from Fig. 6 that although the prediction error
Fig. 4, in which Yp;AlðlÞ ¼ 0. In the final post-combustion stage, of the combustion time calculated using the Kavanau model is
all Al in the particle has been consumed and turned into liquid greatly reduced compared with the results from the uncorrected
Al2 O3 (Fig. 4 (b)), and the particle temperature starts to rise interphase heat transfer model, there is still a noticeable gap
under the heating of the hot gas environment. between the present study and the correlation in Ref. 44. One
Fig. 5 compares the simulation results for the combustion possible reason for these deviations may be that the influence
of a single aluminum particle with a diameter of 10 lm using of the unresolved gas near the particle surface, which is the
different interphase heat transfer models. Subscripts ‘‘Ranz” inherent limitation of the Euler–Lagrange framework, becomes
and ‘‘Kava” indicate the Ranz-Marshall(Eq. (7)), and Kava- more significant as the particle diameter decreases. Moreover,
nau(Eq. (8)), correlations, respectively. The results demon- the present one-step HSR model may not be refined enough,
strate that, for the studied conditions, the Ranz-Marshall compared with the complex polymorphic phase transformation
model predicts a stronger heat exchange rate between the par- processes45 and chemistry reactions31 reported in the literature.
ticle and surrounding gas, i.e., the Nusselt number of the par- These two issues will be addressed in future work.
ticle, which leads to an earlier ignition. After the particle is
ignited, the combustion process is weekly dependent on the 4.1.2. Combustion rate analysis
gas temperature,23 and the heat required for aluminum particle The in-depth analysis of Al particle burning shown in the pre-
vious Section 4.1.1 has demonstrated that the combustion of

Fig. 6 Combustion time calculated in present simulations with

Fig. 5 Effect of interphase heat transfer model on particle different interphase heat transfer models, compared with the
combustion process with dp ¼ 10 lm. correlations by Beckstead6 and Huang et al.44 .
Detailed modeling of aluminum particle combustion 327

single Al particles is a complex process controlled by both heat

and mass transfer, and it is doubtful that simply using empir-
ical correlations can accurately describe the entire combustion
process. Therefore, a comparative investigation of the burning
rates calculated by the present model and burning time corre-
lations from previous studies is carried out. The burning rate
of the aluminum particle in the present model is calculated as
m_ Al ¼ m_ Al _ p;evap
p;HSR þ m ð42Þ
which is the combined mass conversion rate from HSR and
evaporation. Alternatively, burning rates based on the empiri-
cal correlations by Beckstead6 or Huang et al.44 can be
obtained as16
mp;Al
m_ Al;k ¼ k ¼ Beckstead; Huang ð43Þ
tb;k
where tb;k is the burning time of the aluminum particles esti- Fig. 7 Temporal variation of particle burning rate during the
mated as combustion process of a single aluminum particle with
d1:8 dp ¼ 10 lm.
p
tb;Beckstead ¼ 0:00735 ð44Þ
XO2 T0:2
g
particle conversion process. Therefore, in the combustion
where dp is in lm, XO2 is the oxygen mole fraction, and stage, a large amount of aluminum vapor is evaporated from
tb;Beckstead is in ms, or as the particle, which acts to reduce the oxygen concentration
in the local cell. Moreover, a flame located near the particle
d0:3
p
tb;Huang ¼ ð45Þ will consume oxygen and reduce its availability for HSR even
5:5  4 Eb =ðRTg Þ
10 e XO2 further. As a result, the local cell values considered for the
where dp is in cm, tb;Huang is in s, Eb ¼ 73:7 kJ/mol. A priori burning time correlations in the Euler–Lagrange approach
analysis is carried out to directly compare the results from tend to have a higher temperature and lower oxygen concen-
the empirical correlations to the burning rate obtained from tration compared with far-field conditions. Considering Eqs.
the present model. Hence, instead of simulating another two (43)–(45), the following estimation can be made: Comparing
cases with the correlation-based combustion models, the conventional far-field conditions (Tg = 2500 K, XO2 =
parameters in Eqs. (44)-(45) are dynamically evaluated during 0.21) with typical near-field values during particle combustion
the simulation using the present model. Such comparison is (Tg = 3000 K, XO2 = 0.1), the particle burning rate for a
able to show the difference in the burning rates obtained from 10 lm particle is 50.61% or 14.01% lower when using the local
different models under the exact same condition at each time cell values in Eq. (44) or Eq. (45), respectively. Hence the
step. Hence, Tg and XO2 are approximated by the gas temper- underestimation of the burning rates shown in Fig. 7 seems
ature and oxygen concentration of the Eulerian cell in which reasonable.
the particle resides and, along with the particle diameter, these
quantities are updated at every time step. 4.2. Aluminum Bunsen flame
Fig. 7 compares the burning rates calculated by the differ-
ent methods, and the results indicate that different models will As a more stringent test of our newly developed Euler–
lead to completely different combustion rates. The dashed line Lagrange framework for Al combustion, we simulate particle
indicates the correlation-based combustion rates in the pre- cloud combustion stabilized on the laminar Bunsen burner
heating stage, and solid lines show the correlation-based com- from McGill University.27 Fig. 8 first compares the predicted
bustion rates in the combustion stage. All burning rates are AlO distribution with a photograph of the experimental flame.
normalized by the initial particle mass. As can be seen in YAlO is the average mass fraction of AlO. The horizontal red
Fig. 7, the correlation-based methods tend to underestimate line shown on the right of Fig. 8 is the sampling line for mea-
the burning rate during the entire combustion stage, compared surements in the experiment at a height of 12 mm above the
with the present simulation. The main reason for this bias of burner. It is also used for data extraction in the present simu-
the burning rate is the different method of evaluating the gov- lation. AlO is selected for this comparison because it is a crit-
erning variables that affect the rate. Specifically, the parame- ical intermediate of aluminum/air combustion and high levels
ters in the empirical formulae for the combustion time, Eqs. of AlO are characteristic for regions with strong homogeneous
(44) - (45), are supposed to reflect the macroscopic parameters chemical activity.16 Moreover, its spectral emissivity is rela-
of the entire oxidation environment where the aluminum par- tively strong, and it is usually used as a characteristic signal
ticles are burned (‘‘far-field” conditions). However, as already of aluminum gas phase combustion.46,47 As can be seen in
stated above, in the present Euler–Lagrange simulations, the Fig. 8, the predicted flame structure is generally consistent with
governing parameters describing the oxidation environment the experimental observation, and the experimental double-
surrounding the particle are the spatially-averaged scalars over front structure with an inner (conical) and outer (jet-type)
the Eulerian cell in which the particle is located. Although the flame is reproduced successfully.
cell volume is much larger than the particle size, the oxygen Fig. 9 illustrates the structure of the Al Bunsen flame.
concentration and all other scalars are locally affected by the Fig. 9 (left) presents a snapshot of the average gas temperature
328 J. ZHANG et al.

domain, indicating that these particles are not fully burnt. A

more detailed analysis of the evolution of such inner and outer
particles is presented below.
Fig. 10 compares the gas phase temperature predicted by
the present simulation and the measured condensed phase tem-
perature distribution at a height of 12 mm above the burner for
the target Bunsen flame. TExp refers to the temperature mea-
surement in Ref. 27, and Tg is the mean gas temperature pre-
dicted by simulation. YAl ; YO2 and YAl2 O3 ðlÞ are the predicted
mean mass fractions of gaseous aluminum, oxygen, and con-
densed alumina, respectively. The temperature of the con-
densed phase was obtained by fitting the continuum spectra
of the dust flame in the 400–900 nm spectral range to Planck’s
law, and the measurement error is about 5%, as reported in
Fig. 8 Structure of aluminum Bunsen flame (left: predicted AlO Ref. 27, which has been used to determine the width of the
distribution; right: photograph of aluminum Bunsen flame in Ref. error bar in Fig. 10. The condensed phase temperature mea-
27). sured in the experiment has been used for our comparison.
This is because it is the only quantitative data in Ref. 27, the
spatial distribution of which has been recovered from the
raw measurement data by means of Abel deconvolution.
Moreover, the condensed alumina is modeled as a continuous
Eulerian phase in the present study, so that the condensed and
gas phase are at thermal equilibrium. The results indicate that
the predicted temperature agrees well with the measurements
and the radial profile of temperature is also well reproduced.
The ‘‘M” shape of the temperature distribution is attributed
to the existence of the secondary outer diffusion flame, which
has been explained in the previous discussion on Fig. 9. As
shown in Fig. 10, the oxygen mass fraction remains constant
in the inner part of the flame, for 0 mm < x < 3 mm, indicat-
ing that no homogeneous reaction happens and that the
injected two-phase mixture is still fresh at this height and
radial location. Moving outwards, the maximum value of
AlO is found near x ¼ 6 mm, which basically coincides with
the inner reaction zone and the location of the peak tempera-
ture. For 5 mm < x <12 mm, gaseous Al first increases, reaches
its peak and then decreases again. This Al was formed by evap-
Fig. 9 Contour of average gas temperature Tg (left) and average oration from particles that have crossed the inner flame front
mass fraction of aluminum YAl (right). and cannot undergo further chemical reaction in this region
due to a lack of oxygen. A second peak of AlO is formed
and the particles are colored by their instantaneous tempera-
ture. Fig. 9 (right) shows the average distribution of gaseous
Al and the particles are colored by their liquid Al mass frac-
tion. Similar to the AlO distribution shown in Fig. 8, the gas
temperature profile also presents a double-front structure, with
an inner Flame 1 and outer Flame 2. This is because the target
Bunsen flame is a fuel-rich flame with a dust concentration of
500 g=m3 , and major amounts of gaseous Al evaporated from
the dispersed phase can be observed in the middle of the
domain downstream of Flame 1. Moreover, an outer diffusion
flame fed by the mixing of vaporized Al from the center and
oxygen from the co-flow is formed and referred to as Flame
2. Focusing on the dispersed phase, particles that are injected
near the edge of the nozzle at around x = 10 mm (outer parti-
cles, yellow trajectory) enter the flame front in a relatively
short time and are ignited earlier. In contrast, particles injected
via the central part of the nozzle (inner particles, green trajec-
tory) undergo a longer preheating stage. In the downstream
region of the flame, outer particles have a higher burning rate
than the inner particles, and there is still residual liquid Al Fig. 10 Temperature and species mass fraction distribution at a
remaining in many inner particles near the outlet of the height of 12 mm above burner of Bunsen flame from Ref. 27.
Detailed modeling of aluminum particle combustion 329

around x ¼ 11:5 mm and coincides with a local temperature trajectories of which are illustrated in Fig. 9, are extracted
maximum, which corresponds to the position of the outer from the particle cloud and their properties versus time are
flame front. The final Al conversion product, Al2 O3 ðlÞ, is shown in Fig. 11 and Fig. 12 repectively. The time on the
formed between the two flame fronts, and transported out- abscissas corresponds to residence time of the particles inside
wards due to thermal expansion and mass diffusion. the computational domain.
In order to get a better understanding of the particle char-
acteristics inside the Al dust flame, two typical particles, the

Fig. 11 Combustion process of ‘‘Particle 1” in Bunsen flame, the Fig. 12 Combustion process of ‘‘Particle 2” in Bunsen flame, the
trajectory of which is shown in Fig. 9 by green dashed line. trajectory of which is shown in Fig. 9 by yellow dashed line.
330 J. ZHANG et al.

As can be seen in Fig. 11, Particle 1, injected in the central pared with the correlation-based models from Refs 6,9. Similar
region of the nozzle, does not fully convert inside the domain. to the case of the single particle that is shown in Fig. 7, the
This can be inferred from the fact that there is still a significant correlation-based methods underestimate the burning rate of
amount of liquid aluminum remaining in the particle at 73 ms, the aluminum particles in the Bunsen cloud as well, and this
which corresponds to the time shortly before the particle leaves bias of the prediction of the burning rate is increased com-
via the domain exit. Due to a lack of oxygen in the central part pared with the single particle combustion case. The target Bun-
of the fuel-rich Bunsen flame, as already discussed in the con- sen flame is fuel-rich; therefore, the oxygen concentration is
text of Fig. 9, the reaction rate of HSR decreases to zero almost zero in the region between the two reaction zones, as
quickly(Fig. 11 (b)). Fig. 12 illustrates the evolution of the demonstrated in Fig. 10. Hence, the instantaneous burning
outer ‘‘Particle 2”, injected at the edge of the Bunsen flame, time of the particles passing through this zone predicted by
the preheating stage of which is much shorter compared with Eq. (44) or Eq. (45) can be very large, resulting in extremely
Particle 1 injected at the center, and three sub-stages can be small burning rates in the combustion stage. On the other
clearly identified, as shown in Fig. 12 (a). Because of a lack hand, the burning rate calculated by the present model also
of oxygen in this Bunsen flame, HSR only occurs during the shows a decreasing trend during combustion, but with a long
beginning of the combustion stage of these two particles, as tail, and the reason for the greater decrease in the burning rate
demonstrated in Fig. 11 (b) and Fig. 12 (b). Therefore, the in the combustion stage, compared to the single particle case, is
interphase heat transfer becomes the main contributor to the that the evaporation rate is reduced due to the decrease of the
heat required for evaporation in the combustion stage, as aluminum vapor concentration gradient in the fuel-rich flame.
shown in the inset of Fig. 12 (c), in which the curves represent-
ing these two source terms almost coincide. 5. Conclusions
The burning rates of the two aluminum particles that were
tracked along the dust flame are shown in Fig. 13, and com- In the present work, a numerical model for describing the
interphase heat transfer, phase change, homogeneous and
heterogeneous reaction, oxide cap growth and radiation pro-
cesses involved in the conversion of micron-sized aluminum
particles is proposed. The model is employed to simulate the
combustion of single aluminum particles with varying diame-
ters and an aluminum Bunsen flame in the framework of the
Euler–Lagrange method. A detailed analysis of the combus-
tion process of single aluminum particles is conducted, in
which the consumption rates of aluminum due to HSR and
evaporation, as well as the heat release rates due to the various
sub-models are presented. The predictions of combustion time
of single aluminum particles agree well with available empirical
trends for particles with diameters larger than 5 lm, while the
prediction gap for smaller particles is noticeably reduced by
using a model for the transition regime of interphase heat
transfer. The double-front structure of the experimental alu-
minum Bunsen flame is reproduced well in the present simula-
tion, and the predicted temperature distribution of the flame is
consistent with the experimental measurements. By comparing
the evolution of the particle burning rates during combustion,
it is found that the burning rates obtained from the empirical
correlations deviate from the rate predicted by the present
Euler–Lagrange simulation. This is because the oxygen con-
centration and gas temperature required for the correlations
are taken from the Eulerian cell surrounding the particles,
which results in an underestimation of the particle burning rate
in the combustion stage, especially for particles in a fuel-rich
environment. The presented generalized modeling framework
is unaffected by the limits of empirical correlations and tai-
lored to simulations of aluminum particle suspension flames.
The model shall be used to predict solid rocket engine combus-
tion in future work, which will require further extensions to
consider the complex operating conditions therein.

Declaration of Competing Interest

The authors declare that they have no known competing

Fig. 13 Temporal variation of particle burning rates during
financial interests or personal relationships that could have
combustion process of two typical aluminum particles in Bunsen
appeared to influence the work reported in this paper.
flame.
Detailed modeling of aluminum particle combustion 331

Acknowledgements 19. Khalili H, Madani SA, Mohammadi M, et al. Mathematical

modeling of premixed counterflow combustion of a submicron-
sized aluminum dust cloud. Combust Explos Shock Waves 2019;55
This work was supported by the National Natural Science
(1):65–73.
Foundation of China (No. 51706241), Hunan Provincial Nat- 20. Zou X, Wang N, Wang J, et al. A numerical investigation on
ural Science Foundation of China (Nos. 2020JJ4665 and heterogeneous combustion of aluminum nanoparticle clouds.
2021JJ30775), and Hunan Provincial Innovation Foundation Aerosp Sci Technol 2021;112 :106604.
for Postgraduate, China(No. CX2019-0050). The support pro- 21. Zou X, Wang N, Liao L, et al. Prediction of nano/micro
vided by China Scholarship Council (No. 201903170201) dur- aluminum particles ignition in oxygen atmosphere. Fuel 2020;266
ing the visit of Jiarui Zhang to University of Stuttgart is :116952.
gratefully acknowledged. 22. Han DH, Shin JS, Sung HG. A detailed flame structure and
burning velocity analysis of aluminum dust cloud combustion
using the Eulerian-Lagrangian method. Proc Combust Inst 2017;36
References (2):2299–307.
23. Sundaram DS, Puri P. Yang V.A general theory of ignition and
1. Hao X, Zhang H, Hou X, et al. Radiative properties of alumina/ combustion of nano- and micron-sized aluminum particles.
aluminum particles and influence on radiative heat transfer in solid Combust Flame 2016;169:94–109.
rocket motor. Chin J Aeronaut 2022; Available from: https://doi. 24. Yuen MC, Chen LW. On drag of evaporating liquid droplets.
org/10.1016/j.cja.2021.05.024 Combust Sci Technol 1976;14:147–54.
2. Zhao X, Xia Z, Ma L, et al. Research progress on solid-fueled 25. Ranz WE, Marshall WR. Evaporation from drops. Chem Eng
Scramjet. Chin J Aeronaut 2022; Available from: https://doi.org/ Prog 1952;48(3):141–6.
10.1016/j.cja.2021.06.002. 26. Mohan S, Trunov MA, Dreizin EL. Heating and ignition of metal
3. Ben Moussa R, Proust C, Guessasma M, et al. Physical mecha- particles in the transition heat transfer regime. J Heat Transfer
nisms involved into the flame propagation process through 2008;130(10) :104505.
aluminum dust-air clouds: A review. J Loss Prev Process Ind 27. Goroshin S, Mamen J, Higgins A, et al. Emission spectroscopy of
2017;45:9–28. flame fronts in aluminum suspensions. Proc Combust Inst 2007;31
4. Zhou Y, Liu J, Liang D, et al. Effect of particle size and oxygen (2):2011–9.
content on ignition and combustion of aluminum particles. Chin J 28. Crowe CT, Schwarzkopf JD, Sommerfeld M, et al. Multiphase
Aeronaut 2017;30(6):1835–43. flows with droplets and particles. 2nd ed. Boca Raton: CRC Press;
5. Sundaram DS, Yang V, Zarko VE. Combustion of nano 2011.
aluminum particles (Review). Combust Explos Shock Waves 29. Chase M. NIST-JANAF thermochemical tables. Melville: American
2015;51(2):173–96. Institute of Physics; 1998.
6. Beckstead MW. Correlating aluminum burning times. Combust 30. Gurevich MA, Ozerova GE, Stepanov AM. Heterogeneous
Explos Shock Waves 2005;41(5):533–46. ignition of an aluminum particle in oxygen and water vapor.
7. Pang WQ, DeLuca LT, Fan XZ, et al. Combustion behavior of Combust Explos Shock Waves 1970;6(3):291–7.
AP/HTPB/Al composite propellant containing hydroborate iron 31. Glorian J, Gallier S, Catoire L. On the role of heterogeneous
compound. Combust Flame 2020;220:157–67. reactions in aluminum combustion. Combust Flame
8. Chen Y, Guildenbecher DR, Hoffmeister KNG, et al. Study of 2016;168:378–92.
aluminum particle combustion in solid propellant plumes using 32. Alcock CB, Itkin VP, Horrigan MK. Vapour pressure equations
digital in-line holography and imaging pyrometry. Combust Flame for the metallic elements: 298–2500 K. Can Metall Q 1984;23
2017;182:225–37. (3):309–13.
9. Huang Y, Risha GA, Yang V, et al. Effect of particle size on 33. Modest MF. Radiative heat transfer. 3rd ed. Pittsburgh: Academic
combustion of aluminum particle dust in air. Combust Flame Press; 2013.
2009;156(1):5–13. 34. Lynch P, Krier H, Glumac N. Emissivity of aluminum-oxide
10. Feng Y, Xia Z, Huang L, et al. Ignition and combustion of a single particle clouds: Application to pyrometry of explosive fireballs. J
aluminum particle in hot gas flow. Combust Flame 2018;196:35–44. Thermophys Heat Transf 2010;24(2):301–8.
11. Maggi F, Dossi S, DeLuca LT. Combustion of metal agglomerates 35. Harrison J, Brewster MQ. Simple model of thermal emission from
in a solid rocket core flow. Acta Astronaut 2013;92(2):163–71. burning aluminum in solid propellants. J Thermophys Heat Transf
12. Bergthorson JM. Recyclable metal fuels for clean and compact 2009;23(3):630–4.
zero-carbon power. Prog Energy Combust 2018;68:169–96. 36. Fiveland WA, Jamaluddin AS. Three-dimensional spectral radia-
13. Trowell KA, Goroshin S, Frost DL, et al. Aluminum and its role tive heat transfer solutions by the discrete-ordinates method. J
as a recyclable, sustainable carrier of renewable energy. Appl Thermophys Heat Transf 1991;5(3):335–9.
Energy 2020;275 :115112. 37. Rieth M, Proch F, Rabacal M, et al. Flamelet LES of a semi-
14. Bergthorson JM, Goroshin S, Soo MJ, et al. Direct combustion of industrial pulverized coal furnace. Combust Flame 2016;173:39–56.
recyclable metal fuels for zero-carbon heat and power. Appl 38. King MK. Aluminum combustion in a solid rocket motor
Energy 2015;160:368–82. environment. Proc Combust Inst 2009;32(2):2107–14.
15. Julien P, Whiteley S, Soo M, et al. Flame speed measurements in 39. Wang J, Wang N, Zou X, et al. Modeling of micro aluminum
aluminum suspensions using a counterflow burner. Proc Combust particle combustion in multiple oxidizers. Acta Astronaut
Inst 2017;36(2):2291–8. 2021;189:119–28.
16. Han DH, Sung HG. A numerical study on heterogeneous 40. Eisenreich N, Fietzek H, del Mar Juez-Lorenzo M, et al. On the
aluminum dust combustion including particle surface and gas- mechanism of low temperature oxidation for aluminum particles
phase reaction. Combust Flame 2019;206:112–22. down to the nano-scale. Propellants Explos Pyrotech 2004;29
17. Lomba R, Laboureur P, Dumand C, et al. Determination of (3):137–45.
aluminum-air burning velocities using PIV and laser sheet 41. Julien P, Soo M, Goroshin S, et al. Combustion of aluminum
tomography. Proc Combust Inst 2019;37:3143–50. suspensions in hydrocarbon flame products. J Propul Power
18. Zhang J, Xia Z, Ma L, et al. Experimental study on aluminum 2014;30(4):1047–54.
particles combustion in a turbulent jet. Energy 2021;214 :118889.
332 J. ZHANG et al.

42. Weller HG, Tabor G, Jasak H, et al. A tensorial approach to 45. Trunov MA, Schoenitz M, Dreizin EL. Effect of polymorphic
computational continuum mechanics using object-oriented tech- phase transformations in alumina layer on ignition of aluminium
niques. Comput Phys 1998;12(6):620–31. particles. Combust Theory Model 2006;10(4):603–23.
43. Shamooni A, Debiagi P, Wang B, et al. Carrier-phase DNS of 46. Bucher P, Yetter RA, Dryer FL, et al. PLIF species and
detailed NOx formation in early-stage pulverized coal combustion ratiometric temperature measurements of aluminum particle
with fuel-bound nitrogen. Fuel 2021;291 119998. combustion in O2 , CO2 and N2 O oxidizers, and comparison with
44. Huang Y, Risha GA, Yang V, et al. Combustion of bimodal nano/ model calculations. Symp (Int) Combust 1998;27(2):2421–9.
micron-sized aluminum particle dust in air. Proc Combust Inst 47. Peuker JM, Lynch P, Krier H, et al. On AlO emission
2007;31(2):2001–9. spectroscopy as a diagnostic in energetic materials testing.
Propellants Explos Pyrotech 2013;38(4):577–85.