Acta Astronautica 70 (2012) 23–35

Contents lists available at ScienceDirect

Acta Astronautica
journal homepage: www.elsevier.com/locate/actaastro

Evaporation and ignition of droplets in combustion chambers

modeling and simulation
V.B. Betelin a,b, N.N. Smirnov a,b,n, V.F. Nikitin a, V.R. Dushin a, A.G. Kushnirenko a,b,
V.A. Nerchenko a
a
Moscow M.V. Lomonosov State University, Leninskie Gory, 1, Moscow 119992, Russia
b
Scientific Research Institute for System Analysis of the Russian Academy of Sciences, Moscow 117218, Russia

a r t i c l e i n f o abstract

Article history: Computer simulation of liquid fuel jet injection into heated atmosphere of combustion
Received 11 December 2010 chamber, mixture formation, ignition and combustion need adequate modeling of evapora-
Received in revised form tion, which is extremely important for the curved surfaces in the presence of strong heat and
14 June 2011
mass diffusion fluxes. Combustion of most widely spread hydrocarbon fuels takes place in a
Accepted 18 June 2011
Available online 7 August 2011
gas-phase regime. Thus, evaporation of fuel from the surface of droplets turns to be one of
the limiting factors of the process as well. The problems of fuel droplets atomization,
Keywords: evaporation being the key factors for heterogeneous reacting mixtures, the non-equilibrium
Phase transition effects in droplets atomization and phase transitions will be taken into account in describing
Heat flux
thermal and mechanical interaction of droplets with streaming flows. In the present paper
Evaporation
processes of non-equilibrium evaporation of small droplets will be discussed. As it was
Non-equilibrium
Diffusion shown before, accounting for non-equilibrium effects in evaporation for many types of
Mathematical simulation widely used liquids is crucial for droplet diameters less than 100 mm, while the surface
tension effects essentially manifest only for droplets below 0.1 mm. Investigating the
behavior of individual droplets in a heated air flow allowed to distinguish two scenarios
for droplet heating and evaporation. Small droplets undergo successively heating, then
cooling due to heat losses for evaporation, and then rapid heating till the end of their lifetime.
Larger droplets could directly be heated up to a critical temperature and then evaporate
rapidly. Droplet atomization interferes the heating, evaporation and combustion scenario.
The scenario of fuel spray injection and self-ignition in a heated air inside combustion
chamber has three characteristic stages. At first stage of jet injection droplets evaporate very
rapidly thus cooling the gas at injection point, the liquid jet is very short and changes for a
vapor jet. At second stage liquid jet is becoming longer, because evaporation rate decreases
due to decrease of temperature. But combustion of fuel vapor begins which brings to
increase of heat flux to droplets and accelerates evaporation. The length of the liquid jet
decreases again and remains constant slightly oscillating.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction working cycle. Computer aided design of new engines

Most of rocket and aviation engines have pulverized
in air fuels combustion serving the base of their
n
Corresponding author at: Moscow M.V. Lomonosov State University,
Leninskie Gory, 1, Moscow 119992, Russia.
E-mail addresses: ebifsun1@mech.math.msu.su,
mech.math.msu@inbox.ru (N.N. Smirnov).
needs adequate simulation of components pulveriza-
tion, evaporation and combustion. Thus combustible
mixtures formation and deflagration or detonation initia-
tion in polydispersed fuel–air mixtures are the key
aspects providing different limitations for operation of
those engines. Computer simulation of these processes
belongs to strategic information technologies in aerospace
industry.

0094-5765/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.actaastro.2011.06.021
24 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35

Onset of detonation being very dangerous for conven- injected in liquid phase and oxidant being gaseous air
tional engines could, however, serve the basis for creating (Fig. 1). Themophysical properties of the substances were
new generation of engines—pulse detonating engines taken from the handbook [13].
(PDE). Dispersed mixtures having been formed by differ-
ent pulverizers could not be spatially uniform. However,
2. Mathematical model
in most experimental and theoretical investigations the
ignition characteristics of uniformly distributed in space
The mathematical models for simulating turbulent
mixtures were studied. To achieve uniform droplet dis-
chemically reacting flows in heterogeneous mixtures
tribution and to avoid gravitational separation of the
were described in details in [15,16]. Combustion pro-
mixture having been formed investigations under micro-
cesses in heterogeneous mixtures differ greatly from that
gravity conditions are performed [1]. Investigations of
in homogeneous mixtures, because they are governed not
acute problems of heat and mass exchange accompanied
only by chemistry but also by physical processes of
by phase transitions need adequate modeling of evapora-
combustible mixture formation, such as droplet atomiza-
tion, which is extremely important for small droplets and
tion [17–19], evaporation and diffusive mixing of fuel
sprays [2–4,12]. Combustion of hydrocarbon fuels takes
vapor with an oxidant.
place mostly in a gas-phase regime. Thus, evaporation of
The model applies both deterministic methods of
fuel from the surface of droplets turns to be one of the
continuous mechanics of multiphase flows to determine
limiting factors of the process [5–9].
the mean values of parameters of the gaseous phase and
Evaporation under terrestrial conditions is strongly
stochastic methods to describe the evolution of polydis-
influenced by gravity induced thermoconvective flows.
persed particles in it and fluctuations of parameters. Thus
Those effects mask the influence of non-equilibrium
the influence of chaotic pulsations on the rate of energy
processes in phase transitions making the proper under-
release and mean values of flow parameters can be
standing of the phenomenon very difficult in the ground-
estimated. The transport of kinetic energy of turbulent
based experiments. Besides, non-equilibrium effects have
pulsations at the same time obeys the determinestic laws
a stronger manifestation under low gravity conditions for
being the macroscopic characteristic.
interfaces of high curvature.
Averaging by Favre with the ar weight (a—volumetric
Mathematical models for individual droplets evapora-
fraction of the gas phase, r—gas density) we obtain the
tion incorporated in polydispersed mixtures modeling, are
following system for the gas phase in a multiphase flow
usually based on the assumptions of the equilibrium
[15] (the averaging bars are removed for simplicity):
character of phase transitions even for unsteady-state
evaporation regimes [5,21]. Comparison of theoretical and ! _
@t ðarÞ þ rUðar u Þ ¼ M ð1Þ
experimental data shows that this assumption being
undoubtedly valid for large droplets and flat surfaces, ! ! _ k þo
@t ðarYk Þ þ rUðarYk u Þ ¼ rU I k þ M _k ð2Þ
brings to essential errors for small droplets [4,10,11,14,20].
The aim of the study is to develop a mathematical model
! ! ! ! !_
for the non-equilibrium evaporation of droplets in sprays, @t ðar u Þ þ rUðar u  u Þ ¼ ar g arp þ rUt þ K ð3Þ
combustible mixture formation and combustion simulation
applicable to engines of different type, and, in particular, ! ! ! !
@t ðarEÞ þ rUðarE u Þ ¼ ar u U g rUp u
rocket engines conditions. The numerical simulations pre- ! !
sented in the paper will be aimed at hydrocarbon fuels rU I q þ rUðtU u Þ þ E_ ð4Þ
combustion modeling under the conditions fuel being
The Eqs. (1), (2), (3) and (4) include mass balance in
the gas phase, mass balance of kth component, momen-
tum balance and energy balance, respectively (p is the
! !
pressure, u is the fluid velocity vector, g is the gravity
!_
acceleration vector, E is the specific energy, K is the
_
specific momentum flux to gas phase, M is the specific
mass flux, E_ is the specific energy flux, t is the turbulent
!
stress tensor), I k is the turbulent diffusion flux of the kth
ro Reaction Reaction component. We have the following relationships between
the terms in Eqs. (1)–(2):
products products X X X! X
Yk ¼ 1, M _
_ k ¼ M, I k ¼ 0, o_ k ¼ 0
r
∗ k k k k

The state equations for gaseous mixture are the fol-

lowing:
X
p ¼ Rg rT Yk =Wk
k

X !2
u
Fig. 1. Schematic picture of diffusion flame surrounding fuel droplet in E¼ Yk ðcvk T þh0k Þ þ þk ð5Þ
k
2
the atmosphere of an oxidant.
 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35 25

where E is the gas energy, k is the turbulent kinetic energy, !

@t ðareÞ þ rUðar u eÞ ¼ rUðaðm þ rðnt =se ÞÞreÞ
Wk is the molar mass of kth gas component, h0k is the !
þ ðe=kÞðC1e tt : r u C2e areÞ ð11Þ
specific chemical energy, cpk, cvk is the specific heat capacity,
Yk is the mass concentration of kth gas component, T is the X
!
gas temperature, Rg is the universal gas constant. The term @t ðaryÞ þ rUðar u yÞ ¼ rUðaðm þ cpk rðnt =sk ÞÞryÞ
k
responsible for chemical transformations, o _ k is very sensi-
tive to temperature variations, as it is usually the Arrhenius þ Py þWy Dy ð12Þ
law type function for the reactions’ rates. To take into
where the production terms Py, Wy and the dissipation
account temperature variations the source term o _ k in the
term Dy are:
Eq. (2) was modeled using the Gaussian quadrature
nt 2
X
technique. Py ¼ 2arc~ p 9rT9 , Wy ¼  ok T 0 h0k ,
Let us regard the temperature being a stochastic sk k
function T with mean T and mean squared deviate X e y
y ¼ T 0 T 0 . Then, the mean value of a function having T as Dy ¼ Cg ar cpk ð13Þ
k
k ym y
independent variable could be determined as follows:
Z pffiffiffi The constants take the following values [15,16]:
f ðTÞ ¼ f ðT þ z yÞPd ðzÞdz
Cm ¼ 0:09, C1e ¼ 1:45, C2e ¼ 1:92, sd ¼ 1,
where z is a random value with zero expectation and unit st ¼ 0:9, sk ¼ 1, se ¼ 1:13,
deviate; its probability density function is Pd(z). To esti- 2
mate the integral, the minimal number of terms is used ym ¼ T =4, Cg ¼ 2:8
(namely, three) and Pd(z) is assumed to be even. In this One could see that Eqs. (10)–(12) do not contain terms
case, the formula for f(T) averaging is: responsible for particulate phase contribution to turbu-
1 pffiffiffi  1

1 pffiffiffi lence energy growth. This is due to the direct stochastic
f ðTÞ ¼ f ðTw y Þ þ 1 f ðTÞ þ 2 f ðT þ w yÞ
modeling of the particulate phase: the influence of the
2w2 w2 2w
last on the gas phase leads to stochastic behavior of the
!_
pffiffiffiof normal (Gaussian) deviate, the value of w is
In case
momentum source terms K . These terms affect the
equal to 3 (Gauss–Hermite case). Therefore, the formula
averaged gas phase velocity in the stochastic manner
above could be transformed as follows: !
and therefore the source term tt : r u is also affected.
1 pffiffiffiffiffiffi 2 1 pffiffiffiffiffiffi The motion of polydispersed droplets (particles) is mod-
f ðTÞ ¼ f ðT 3yÞ þ f ðTÞ þ f ðT þ 3yÞ
6 3 6 eled making use of a stochastic approach. A group of
In our case, the function f(T) is the Arrhenius tempera- representative model particles is distinguished. Motion of
ture dependence; the whole average for o _ k is constructed these particles is simulated directly taking into account
using combinations of these dependencies. Averaged the influence of the mean stream of gas and pulsations of
magnitudes for mass fractions and density were used in parameters in gas phase. The procedure is described in
the Arrhenius law for o _ k as the dependence of these details in papers [15,16,19].
functions is not as strong as the dependence of The number of model particles to represent real spray
temperature. is chosen by the investigator based on the actual esti-
! mated number of real particles, and the adopted real
The turbulent heat flux I q in the Eq. (4) is a sum of
two terms: particle number density distribution function in the
! ! X ! vicinity of one model particle. The minimal number of
Iq ¼ Jqþ ðcpk T þ h0k Þ I k ð6Þ model particles should be sufficient to represent all real
k particles (total injected mass conservation) so that over-
!
where J q could be interpreted as turbulent conductive whelming majority of real particles should be located
heat flux. The eddy kinematic viscosity nt is expressed within a distance from the model one comparable with
according to k-epsilon model as nt ¼ Cm ðk2 =eÞ: The turbu- the grid resolution size in solving gas dynamical problem.
lent fluxes are: Else the assumption of real particles being subjected to
T
similar forces and fluxes affecting model particle could
t ¼ aðm þ rnt Þðr! ! !
u þ r u ð2=3ÞðrU u ÞUÞ lose validity. One can see that the current mechanism
ð2=3ÞarkU ð7Þ for distinguishing model particles relies heavily on particles
spatial number density distribution, and does not take into
! account the particles number density distribution versus
I k ¼ arðD þðnt =sd ÞÞrYk ð8Þ
size. (For mono dispersed mixtures similar algorithms are
!
! X used.) However, the intention to simulate polydispersed
J q ¼ a l þ cpk rðnt =st Þ rT ð9Þ mixtures with a definite droplet size resolution could
k
introduce additional constraints from below for the minimal
The model is closed then by the equations for k, y number of model droplets. In future using multi-processor
and e: supercomputers would allow increasing the number of
! model droplets (thus decreasing the number of real droplets
@t ðarkÞ þ rUðar u kÞ ¼ rUðaðm þ rðnt =sk ÞÞrkÞ
represented by a model one). Finally one could reach the
!
þ tt : r u are ð10Þ case, that each real droplet would be represented by a
26 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35

model one, which would provide the maximal precision for function of droplet initial diameter and relative gas
representation. velocity. It could be estimated by the following formula:
v0 f ðt ¼ 0Þ
3. Droplet evaporation in a streaming flow t¼ , where a0 ¼ d ð19Þ
a0 m0

To simulate the polydispersed particulate phase a The energy equation for a droplet has the following
stochastic approach is performed, a group of representa- form:
tive droplets is distinguished. The effect of the gas mean de
stream and pulsations of parameters in the gas phase on m ¼ q þ Qs , where e ¼ cvs Ts þh0f ð20Þ
dt
the motion of droplets is considered. Gas flow properties
(the mean kinetic energy and the rate of pulsations decay) The energy of phase transitions could be determined
by the formula:
make it possible to model the droplets stochastic motion
determining the pulsations change frequency under dm
Qs ¼ hL ð21Þ
assumptions of the Poisson flow of events. dt
We model a great amount of real droplets by assem- where hL is the latent heat of evaporation. Heat flux to a
bling model droplets (their number is of the order of single droplet from the surrounding gas flow is deter-
thousands). Each model droplet is characterized by a mined as follows [5]:
vector of values, representing its location, velocity, mass (
and other properties. The following vector is determined pdl Nu ðTTs Þ, Re o 1000;
q¼ ð22Þ
for each model droplet: pd2 r9! !
v  u 9 St ðHr Hw Þ, Re Z 1000
!!!
fN,m,ms , r , v , w , o,Ts gi , i ¼ 1, . . ., Np where
When a droplet is burnt out, its mass mi ¼0, and this Cd 2=3
droplet is excluded from calculations. Nu ¼ 2 þ0:16 Re2=3 Pr1=3 , St ¼ Pr ð23Þ
! 2
The parameter w i is used to model dispersed phase
As it follows from Eqs. (20) and (21), the heat flux
turbulent pulsations. It represents the pulsation velocity
! going inside droplet and heating it qs is determined as a
vector added to the gas velocity vector u , thus a stochas-
difference between heat flux coming to a droplet from
tic force resulting from the interaction with turbulized gas
surrounding gas, and energy consumed for surface eva-
is calculated jointly with a resistance force.
poration:
The momentum equation for a single droplet motion in
the gas flow has the following form: dm
_ L;
qs ¼ qmh _ ¼
m  evaporation rate
! ! dt
dv ! ! dr !
m ¼ m g fUrp þ f d , ¼ v ð14Þ This heat flux is responsible for droplet gradual mean
dt dt
temperature increase as determined by Eq. (20). While
where the drag force could be expressed as follows:
the effect of heat conductivity coefficient in fluid could
! 1 ! ! ! ! ! ! manifest only on calculating temperature profile inside
f d ¼ Cd ðReÞrSð v þ w  u Þ9 v þ w  u 9 ð15Þ
2 droplet.
the drag coefficient being the function of Reynolds num- The non-equilibrium evaporation model is used to
ber [5] determine the evaporation rate [20]
   
24 4:4 1Ye
Cd ¼ þ pffiffiffiffiffiffi þ 0:42 bK ð16Þ m_ ¼ pdrDNulog ð24Þ
Re Re 1Yw
   pffiffiffiffiffiffiffiffiffiffiffiffiffi
r9! ! !
v þ w  u 9d
Yw ¼
WN P0 H
exp b
1

1
_
m
2pRTs
ð25Þ
Re ¼ ð17Þ
m Wp R Tb ðP0 Þ Ts p de pd2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   4=5 On approaching by the droplet surface temperature
r r T the value of boiling temperature Tb integration of
b¼ 2 , K¼
rs rs Ts (20)–(25) becomes time consuming due to enormous
where increase of time steps necessary to resolve the system.
8  Then surface temperature could be assigned the boiling
>
> g1 2 1=ðg1Þ value and a simplified formula could be used to determine
M o 1;
rs < 1 þ 2 M ,
evaporation rate on the last stages of droplet life [5]
¼   ð18Þ
r > > ðg þ 1ÞM 2 g1 ðg1ÞM2 þ 2 1=ðg1Þ
: ðg1ÞM2 þ 2 1 þ 2 2gM2 ðg1Þ , M Z 1:
dm qs q
¼ , _ s ðTb Ts Þ
qs ¼ mc
dt hL
where
! ! ! which yields
9 v þwu 9 pffiffiffiffiffiffiffiffiffiffiffi  
M¼ , c¼ gp=r  sonic velocity: dm cs ðTb Ts Þ q
c 1þ ¼ ð26Þ
dt hL hL
Gravity induced streaming flows could exist for rather
long time, while in the absence of gravity the character- Joint integration of (26) and (20) shows that internal
istic time for the streaming flow being essential is the temperature on decreasing droplet size tends to surface
 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35 27

one exponentially [5] dn, one obtains the following formulas:

Tb Ts 1
 et=t0 Adrag ¼ Nn rCd v2rel pd2n dn ;
Tb Ts0 8
d
Thus on the last stages of droplet life the mean dn ¼ ð31Þ
1 þð1=4Þðð1=2ÞCd rv2rel ð1=3Þrc v2n Þd=s
temperature could be assigned the boiling value and the
evaporation rate is determined by a simple formula The mean velocity of the cloud expansion v* could be
evaluated based on the condition of matching the two
dm q formulas for dn at We¼Wenn. The reason to perform that
¼ , T ¼ Tb ð27Þ
dt hL matching is that both formulas for breakup regimes were
The dynamic interaction of liquid droplets with the obtained from experiments, thus, indirectly the expansion
gaseous flow could bring to instability of the interface and of the cloud of droplets after the breakup should have
atomization of droplets. The criterion for liquid droplets been taken into account. On the other hand, the depen-
instability is that of the critical Weber number [17]: dence of characteristic droplets diameters on the Weber
We ¼ rv2rel d=s, where s is the surface tension at the number should be continuous. Then, finally one obtains
interface, vrel is relative velocity of a droplet versus gas. dWen
dn ¼ ð32Þ
On exceeding the critical value of the Weber number 1
8 Cd ðWeWenn ÞWen þ Wenn
droplets breakup due to vibrational instability takes place.
On essentially surpassing the critical Weber number other In modeling droplets breakup in a gas flow the inertia of
mechanisms start playing essential roles in the breakup the process should be taken into account. Fragmentation
process that brings to formation of fine mist. These main does not take place instantaneously: it needs time for small
characteristics of the atomization process could be taken droplets to separate from the initial one, i.e. it needs a
into account by approximate formula [16,18] determining definite time for the liquid bridges between the droplets to
mean diameters of droplets da originating in breakups of be established, elongated and broken [19]. The first order
initial droplets (diameter d) estimate of the breakup time gives the following formula:
  
8  1=3 d Wen 3 r 4
>
> 6a2 tn ¼ 1 þ Cd 1 ð33Þ
< d ¼ pn
> , Weo Wen ; vrel We 8 rc Cd Wen
da ¼ dWe
Wen rWe r Wenn ; ð28Þ
We ,
n
>
>
>
:d ,
n We4 Wenn ;
4. Single droplet evaporation in a streaming flow
where n is the number of droplets per volume unit, a2 is simulation
the volumetric fraction of the droplet phase, the critical
Weber numbers are determined as follows: The problem of droplet interaction with the streaming
Wen ¼ 12ð1 þLp0:8 Þ, Wenn ¼ 350 gas flow was investigated taking into account mechanical
drag, atomization of droplets, non-equilibrium heat and
where the Laplace number Lp ¼ ðdrc s=m2c Þ; rc, mc are mass transfer. A series of model problems was regarded,
liquid density and viscosity, respectively. in which gas flow and droplet initially had different
To determine the mean diameter of droplets dn after relative velocities and temperatures, and then relaxation
the breakup of a type of an explosion (We4Wenn) one took place. Those model problems are similar to that
needs to evaluate the part of the elastic energy accumu- encountered in shock wave initiation of detonation in
lated by a droplet, which was spent for the breakup. The combustible dispersed fuel–air mixtures. The two-phase
assumption, that the breakup energy was spent for the flow becomes strongly non-equilibrium behind the shock
formation of new free surface [19] makes it possible to wave, because of mechanical inertia droplets keep their
evaluate the number N and the mean diameter dn of the speed practically constant and gas accelerates on passing
formed droplets: the shock wave. Besides, due to thermal inertia, tempera-
 3 ture inside droplets practically does not change, while gas
En d temperature increases instantaneously behind the shock
N ¼ 1þ ; dn ¼ ð29Þ
spd 2 1 þ En =spd2 wave. Due to that reason gas temperature in all the
The breakup energy could be evaluated as the differ- numerical experiments was assumed to be higher than
ence between the work of the drag forces separating small the boiling temperature for liquid droplets.
droplets from the initial one, and the kinetic energy of The thermophysical properties for gas and droplets
fragments’ scattering: were assumed similar for all numerical experiments:
ambient pressure p¼1013 bar, temperature T¼ 1000 K,
Nn
X mi v2 gaseous phase – air, liquid – n-decane (C10H22). Initial
in
En ¼ Adrag  ð30Þ droplet temperature was assumed to be T¼300 K.
i¼1
2
The relaxation processes for droplets of different dia-
Assuming that the initial droplet is split into N* equal meters were regarded. Numbered lines on the successive
droplets (Nn ¼ ðd3 =d3n Þ) having equal velocities of radial figures correspond to the following initial diameters of
expansion of the cloud vn and the separation of droplets droplets (curves 1(1 mm), 2 (5 mm), 3 (10 mm), 4 (50 mm),
takes place after the droplet is moved away at a distance 5 (100 mm), 6 (500 mm), 7 (1000 mm)).
28 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35
Figs. 2–5 illustrate parameters of droplets variation in
the process of mechanical and thermal relaxation for the
set of numerical experiments assuming initial velocity
difference to be equal to 50 m/s. Fig. 2 illustrates velocity
relaxation. It is seen that on decreasing initial droplet
radius the relaxation process turns to be faster. The
decrease of the relaxation time is monotonous. Some
curves begin lower than 50 m/s because the relaxation
process is too fast for small droplets, and the time scale on
the x-axis is logarithmic beginning from a definite
small value.
Fig. 3 illustrates droplet temperature variation versus
time for droplets of different initial diameters.
The dynamics of droplet heating is the following. For
small droplets (curves 1 (1 mm), 2 (5 mm), 3 (10 mm), 4
(50 mm)) in the very beginning the temperature increases
due to external heating. On rapid decrease of droplet
relative velocity evaporation in the stream of gas brings
to a decrease of temperature. Then on decreasing droplet
Fig. 2. Velocity relaxation versus time for droplets of different dia-
meters. V0 ¼50 m/s, taking into account atomization.
Fig. 3. Mean droplet temperature variation versus time for droplets of
different initial diameters. V0 ¼ 50 m/s.
radius heat fluxes growth brings to an increase of tempera-
ture until the droplet disappears. The first increase of
temperature is higher for larger droplets, which relative
velocity decreases much slower than for small droplets
(curves 5 (100 mm), 6 (500 mm), 7 (1000 mm)). Large dro-
plets follow different scenario: The first increase of their
temperature goes as high as up to the critical value. Then
evaporation takes place at a critical temperature very
rapidly. Arrows with numbers in Figs. 3 and 5 indicate the
end of droplet lifetime. That is the reason for a larger droplet
(curves 5, 6, 7 in Figs. 4 and 6) to have a shorter lifetime.
Fig. 4 illustrates droplet diameter variation due to
evaporation. It is seen from the figure that evaporation
time increases with the increase of initial droplet dia-
meter, but the increase does not take place monoto-
nously: the increase of lifetime (curves 1, 2, 3, 4)
changes for a decrease (curves 4, 5) and then comes back
to an increase (curves 5, 6, 7). That testifies the effect of
manifestation of different mechanisms depending on the
Fig. 4. Droplet diameter variation versus time. V0 ¼50 m/s.
Fig. 5. Mean droplet temperature variation versus normalized time for
droplets of different diameters. V0 ¼50 m/s, taking into account
atomization.
 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35 29

characteristic size of droplets. The first rapid decrease of

droplet radius (curves 6, 7 in Fig. 4) is due to atomization
of large droplets. The breakup of large droplets (Fig. 4)
brings to the conditions that behavior of droplets of initial
diameter 500 mm and 1000 mm looks absolutely similar
after some time necessary for atomization to proceed.
Fig. 5 presents the results for droplet temperature
variation versus normalized time t/t (19), i.e. in the
relative time scale characterizing each droplet. As it is
seen from the figure all maxima and minima of tempera-
ture correspond to nearly the same normalized time
moments independently on initial droplet radius. The
difference between the two scenarios for droplet behavior
in the stream of heated air could be clearly distinguished
from this figure.
Figs. 6–8 present the results of similar numerical experi-
ments, but for higher initial relative velocity: V0 ¼100 m/s.
Two different scenario for droplet behavior could be dis-
Fig. 8. Mean droplet temperature variation versus time for droplets of
tinguished as well. But for the higher relative velocity different diameters. V0 ¼100 m/s, taking into account atomization.
smaller droplets already reach critical temperature during

Fig. 9. Droplet diameter variation versus time for droplets of different

Fig. 6. Velocity relaxation versus time for droplets of different diameters. initial diameters. V0 ¼50 m/s, atomization not taken into account.
V0 ¼100 m/s, taking into account atomization.

the first temperature increase (Fig. 8). Thus curve 4 belongs

already to the second group of curves, which corresponds to
the second scenario of droplet heating and evaporation.
Fragmentation is more essential for higher relative velocity:
curves 4, 5, 6, 7 converge in time due to equalizing of
droplet radii in fragmentation (Fig. 8).
Atomization of droplets changes droplet lifetime and
evaporation scenario for large droplets. Figs. 9 and 10
illustrate droplet diameter variation for droplets of dif-
ferent initial size in interaction with streaming flow not
accounting for atomization. Comparison with Figs. 4 and 7
representing results with account of atomization clearly
demonstrates the effect.

5. Spray evaporation modeling

To investigate the cloud slowing down, cooling of

Fig. 7. Droplet diameter variation versus time for droplets of different particles and heating surrounding atmosphere, atomiza-
diameters. V0 ¼ 100 m/s, taking into account atomization. tion of droplets it is necessary to apply the models for
30 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35

Figs. 11 and 12 show the model particles location for

Fig. 10. Droplet diameter variation versus time for droplets of different
diameters. V0 ¼ 100 m/s, atomization not taken into account.
multiphase media accounting for the two-way coupling
effects for droplets and gas, thermochemical and mechan-
ical destruction of fragments. Mathematical models for
the non-equilibrium polydispersed mixture flows and
droplet breakup are described in details in the papers
[19,20].
The cloud of small droplets injected into the chamber
slows down very rapidly due to the drag forces. The
deceleration for fragments is proportional to 1/r0 and
grows up with the decrease of a characteristic size r0.
The velocity variation versus distance could be estimated
by formula:
the two successive times. The initial pressure of gas inside
the containment was rather low: p0 ¼0.01 MPa (0.1 atm).
The size of circles showing model particles is much larger
than their real size, but directly proportional to it. The
color reflects particles temperature. Map of colors for the
temperature is shown in Fig. 13 relevant to Figs. 11–15.
Injection location is shown by white circle. It is seen from
Figs. 11 and 12 that temperature variation is not so big
because convective heat exchange with rarefied gas is
rather small.
On increasing pressure 100 times up to 10 atmospheres
one could see that spray interaction with surrounding gas
became more essential (Figs. 13–15). First, cooling of
droplets takes place. Second, fastest droplets experience
breakup forming clouds of much smaller droplets, which
lag behind larger ones.
Figs. 13–15 illustrate the model particles locations for
the case of a relatively high gas pressure inside the
containment (p0 ¼1 MPa (10 atm)). Gray color indicates
injected droplets and piece of jet, which exceeded the
number assigned for model particles and their tempera-
ture was not calculated. The aerodynamic drag is much
Fig. 11. Model particles location for droplet cloud co-axial jet injection
into the chamber with low gas pressure p0 ¼ 0.01 MPa (0.1 atm) along
the axis of symmetry.

 
3 r x
vrel ¼ v0rel exp  Cd ð34Þ
8 rc r0

It is seen from (34) that on decreasing droplet size the

slowing down distance also decreases. Since droplet cloud
incorporates a variety of droplets having different size
and speed its evolution could bring to a peculiar non-
uniform pattern of droplets distribution within the com-
bustion chamber.
Numerical modeling of co-axial jet injection into a gas-
filled cylindrical containment from the wall along the axis
Fig. 12. Successive stage for droplet cloud co-axial jet injection into the
was performed based on the developed mathematical chamber with low gas pressure p0 ¼0.01 MPa (0.1 atm) along the axis of
models. A cylindrical containment of 0.1 m radius and symmetry. The jet reached far end of the chamber.
0.2 m height was regarded. A cloud of small droplets,
characterized by the average diameter 0.3 mm and sto-
chastic deviations 0.05 mm was injected from left bottom
along the axis. The fragments initial temperature was
400 K with stochastic deviations 710 K; maximal velo-
city in the axial direction was assumed to be 20 m/s,
average velocity 15 m/s with stochastic deviations 5 m/s
in axial direction, average velocity in radial direction was
assumed 0 m/s with stochastic deviations 5 m/s.
The average density of liquid was assumed as rc ¼
1000 kg/m3, viscosity and surface tension in the liquid
10  3 N s and 10  2 N/m, respectively. The gas pressure
inside the containment was varied from 0.01 MPa up
Fig. 13. Model particles location for droplet cloud co-axial jet injection
to 1.5 MPa, initial temperature T0 ¼300 K, molar mass into the chamber with high gas pressure p0 ¼ 1 MPa (10 atm). Legend for
0.028 kg/mol. temperature values in K provided in right hand side.
 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35 31

Table 1
Mixture composition.

Nr 1 2 3 4 5 6 7

Component O2 C10H22 CO CO2 H2 H2O N2

Molar mass 32 142 28 44 2 18 28
(g/mol)
Potential (kJ/mol) 0  249.3  110.5  393.5 0  241.8 0

Fig. 14. Successive stage for droplet cloud co-axial jet injection into the
chamber with high gas pressure p0 ¼1 MPa (10 atm).
The molar reaction rates (in moles per volume unit)
are provided by formulas:
  
rY1 rY2
w1 ¼ K1 ðTÞ
W W
 1  2 
rY1 rY3
w2 ¼ K2 ð T Þ
W1 W3

 r  rY 
4
w3 ¼ K3 ðTÞ
W W4
  
rY1 rY5
w4 ¼ K4 ðTÞ
W1 W5
 r rY 
6
w5 ¼ K5 ðTÞ
W W6

Fig. 15. Final stage for droplet cloud co-axial jet injection into the
chamber with high gas pressure p0 ¼ 1 MPa (10 atm). The larger droplets Then the reaction rates of chemical transformations for
reached far end of the chamber, while smaller droplets are still in the air. each of the 7 components, which are present in Eq. (2),
could be determined by formulas:

more essential for the present case. On entering the
containment fragmentation of liquid droplets due to their
interaction with the atmosphere brings to a formation of
very fine droplets in the front part of the cloud (Fig. 13).
The droplets, representing smaller fragments, are illu-
strated by dots in Fig. 13. Nevertheless, the major mass of
the cloud is represented by those dots, and only a smaller
number of low velocity particles keep its initial size.
The small droplets slow down very rapidly and lose
their kinetic energy much faster than the large ones. Thus
the large droplets, that had initially much lower velocity,
come to overtake the small ones (Figs. 14 and 15).
6. Droplet cloud combustion modeling
The model chemical kinetics incorporated seven com-
ponents mixture and assumed five brutto-reactions. More
details are provided in Tables 1 and 2.
The rate of chemical reactions is provided by Arrhenius
law:
8   hydrocarbons the following parameters were taken: r ¼
< Aj exp  Taj , T ZTmj
Kj ðTÞ ¼
T
ð35Þ
: 0, T oTmj
The number of component j and respective values of
kinetic parameters are provided in Table 2. The third
column in Table 2 provides the number of component
acting as a ‘‘fuel’’ for each reaction.
Dimensionless value B in Table 2 stands for the share
of water vapor in n-decane decomposition. Its value could
vary from zero to unity.
o_ 1 ¼ W1 ðð2n þ m4 BÞw1 þ 0:5ðw2 þ w4 w3 w5 ÞÞ
o_ 2 ¼ W2 w1
o_ 3 ¼ W3 ðnw1 þ w2 w3 Þ
o_ 4 ¼ W4 ðw2 w3 Þ
o_ 5 ¼ W5 ðm2 ð1BÞw1 þ w4 w5 Þ
o_ 6 ¼ W6 ðm2 Bw1 þ w4 w5 Þ
o_ 7 ¼ 0
Numerical investigations of detonation initiation in dis-
persed hydrocarbon fuel–air mixtures after mild ignition via
DDT and by shock waves of different intensities were
performed. The vessel was 0.5 m long and 0.05 m diameter.
The aerosol occupied the cylindrical co-axial zone beginning
from 0.1 m along the axis its diameter being equal to
0.015 m. The flow was assumed to have the following initial
turbulence characteristics: energy k¼0.1 J/kg, the mixing
length l¼0.01 m, mean velocity u¼0, initial temperature
300 K. The number of model particles used in calculations
was 20,000. Validation of numerical scheme was performed
based on comparing the obtained results of numerical
simulations for the detonation wave velocities in dispersed
mixtures with available experimental data [16]. To simulate
850 kg/m3; DH¼43 MJ/kg; hL ¼200 kJ/kg; W¼140 kg/kmol.
The share of water in hydrocarbon decomposition was
assumed to be z ¼0.2. The mean droplets diameter
was assumed 100 mm, minimal diameter – 10 mm, maximal
– 200 mm. The droplets size distribution function was
assumed to be a triangular one. The initial droplets density
was varied from 5 to 20 kg/m3.
The results of simulations (Figs. 16 and 17) show that
for different fuel concentration and intensity of shock
wave initiation different scenarios of the process are
32 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35

Table 2
Chemical interactions in gaseous phase.

Nr Reaction NF Reaction formula Aj Taj Tmj

8
1 n-Decane decomposition 2 Cn Hm þ ðn2 þm
4BÞO2 -nCO 1  10 25,270 500
þm2ð1BÞH2 þ m
2 BH2 O

2 Combustion CO 3 CO þ 12O2 -CO2 5.89  106 2000 450

3 Decomposition CO2 4 CO2 þ M-CO þ 12O2 þ M 2.75  107 20,000 800
4 Hydrogen combustion 5 H2 þ 12O2 -H2 O 2  106 1500 500
5 Water vapor decomposition 6 H2 O þ M-H2 þ 12O2 þ M 2  107 20,000 800

Fig. 16. Hydrocarbon concentration for successive times of shock wave Fig. 17. Temperature maps (K) for successive stages of shock wave
propagation in dispersed mixture (a, b—initial fuel density 5 kg/m3; propagation in dispersed mixture (a, b—initial fuel density 5 kg/m3;
c, d—initial fuel density 20 kg/m3). c, d—initial fuel density 20 kg/m3).

possible. There could be formed a combustion wave
lagging behind attenuated shock. There could be onset
of detonation, or galloping detonation in the dispersed
layer. For high average density of fuel droplets within the
layer combustion does not take place inside layer, where
pressure is maximal, however high speed detonation type
process onset on the periphery is observed supported by
the piston effect due to induced vapor combustion in the
zones within the concentration limits.
Thus the results illustrate the fact, that increase of
droplet concentration above definite value inhibits the
onset of detonation in dispersed mixtures and gives birth
to a detonation mode typical for non-premixed systems.
Fig. 18 illustrates successive temperature fields in a
cylindrical chamber, into which fuel spray of liquid n-decane
is injected along the axis from the bottom wall in the form of
cylindrical jet and ignited in the heated atmosphere of the
chamber. The droplet versus diameter distribution function
has a triangular shape: minimal diameter 0.0001 m, medium
diameter 0.001 m and maximal diameter 0.002 m. Injection
velocity 100 m/s.
Fig. 19 illustrates successive fields of model droplets
location and temperature. It is seen from the figure, that
at first stage of jet injection into heated atmosphere
droplets evaporate very rapidly thus cooling the gas at
injection point. Then droplet jet is becoming longer,
because evaporation rate decreases due to decrease of
temperature. But combustion of fuel vapor begins which
brings to increase of heat flux to droplets and accelerates
evaporation. The length of the liquid jet decreases again
and remains constant slightly oscillating. The combustion
 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35 33

Fig. 18. Gas temperature fields for successive stages of fuel jet injection and combustion in a chamber filled with oxidant 40 bar.

of vapor takes place in diffusion regime as it could be seen
from maximal temperature distribution (Fig. 18), which
tracks the flame zone.
7. Conclusions
Accounting for non-equilibrium effects produces
an influence on the integral characteristics of droplet
evaporation process. The lifetime for single evaporating
droplet could be several times longer under non-equili-
brium conditions as compared with equilibrium ones.
Investigating the behavior of individual droplets in a
heated air flow allowed distinguish two scenarios for
droplet heating and evaporation. Small droplets undergo
successively heating, then cooling due to heat losses for
evaporation, and then rapid heating till the end of their
34 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35

Fig. 19. Model particles location and temperature for successive stages of fuel jet injection and combustion in a chamber filled with oxidant 40 bar.

lifetime. Larger droplets could directly be heated up to a
critical temperature and then evaporate rapidly. Atomiza-
tion affects droplet lifetime.
Investigation of polydispersed multi-velocity droplet
cloud injection into the heated atmosphere of a combus-
tion chamber showed that flow scenario essentially
depends on gas state (pressure and temperature) inside
the chamber. The fastest droplets undergo fragmentation
and slow down rapidly, while the slowest droplets remain
big and overtake small ones. This droplet size separation
effect manifests stronger for the case of higher gas density
inside the chamber.
Increase of droplet concentration above definite value
inhibits the onset of detonation in dispersed mixtures and
gives birth to a detonation mode typical for non-premixed
systems.
 V.B. Betelin et al. / Acta Astronautica 70 (2012) 23–35
The scenario of fuel spray injection and self-ignition in
a heated air inside combustion chamber has three char-
acteristic stages. At first stage of jet injection droplets
evaporate very rapidly thus cooling the gas at injection
point, the liquid jet is very short and changes for a vapor
jet. At second stage liquid jet is becoming longer, because
evaporation rate decreases due to decrease of tempera-
ture. But combustion of fuel vapor begins which brings to
increase of heat flux to droplets and accelerates evapora-
tion. The length of the liquid jet decreases again and
remains constant slightly oscillating.
The combustion of vapor takes place in diffusion
regime as it could be seen from maximal temperature
distribution, which tracks the flame zone. The diffusion
flame is located along the moving boundary of vapor jet.
Acknowledgments
The authors gratefully acknowledge financial support
from the Russian Foundation for Basic research (Grants
09-08-00284 and 10-07-00354).
References
[1] N.N. Smirnov, V.N. Pushkin, V.R. Dushin, A.V. Kulchitskiy, Micro-
gravity investigation of laminar flame propagation in monodisperse
gas-droplet mixtures, Acta Astronautica 61 (2007) 626–636.
[2] G.M. Kampher, On the limits of equilibrium regime for droplet
evaporation, Proceedings of the MADI, Moscow 71 (1974) 43–52.
[3] I.N. Ivchenko, On heat and mass transfer in evaporation and
condensation growth of small spherical droplets, Thermophysics
of High Temperatures 23 (4) (1985) 787–791.
[4] R.S. Miller, K. Harstad, J. Bellan, Evaluation of equilibrium and non-
equilibrium models for many-droplet gas-liquid flow simulations,
International Journal of Multiphase Flow 24 (6) (1998) 1025–1055.
[5] N.N. Smirnov, I.N. Zverev, Heterogeneous Combustion, Moscow
Univ. Publ., Moscow, 1992.
[6] N.N. Smirnov, Heat and mass transfer in unsteady-state diffusion
combustion of a condensed fuel surface, International Journal of
Heat and Mass Transfer 29 (5) (1986) 673–679.
35
[7] N.N. Smirnov, V.F. Nikitin, V.V. Baskakov, Unsteady-state combus-
tion of fuel droplets in weightlessness, in: Proceedings of the X-th
European, VI-th Russian Symposium on Physical Sciences in Micro-
gravity, Moscow, 1997, vol. 1, pp. 409–416.
[8] N.N. Smirnov, V.F. Nikitin, V.V. Baskakov, Droplets and polydis-
persed sprays combustion under microgravity, in: Advanced Com-
bustion Research Proceedings of the ESA Workshop, C.N.R.S.,
Orleans, 1997, 15 pp.
[9] M. Tanabe, J. Konig, O. Moriue, Ch. Eigenbrod, J. Sato, M. Kono, H.J.
Rath, Autoignition behaviour of single hydrocarbon fuel droplets,
in: Proceedings of the Drop Tower Days in Bremen, Bremen, 1996,
pp. 310–312.
[10] N.N. Smirnov, A.V. Kulchitskiy, Non-equilibrium effects in liquid
fuel droplets combustion, in: Proceedings of the 51-st IAF Congress,
Rio de Janeiro, 2000, IAF-00-J.4.09, 9 pp.
[11] A.V. Kulchitskiy, N.N. Smirnov, Accounting for non-equilibrium
effects in evaporation of solution droplet, Journal of Engineering
Thermophysics 10 (4) (2000) 293–314.
[12] G. Picker, J. Straub, Evaporation coefficient determined from vapor
bubbles growing under microgravity conditions, in: Heat Transfer
1998, Proceedings of the 11th IHTC, Kyongju, 1988, vol. 2.
[13] N.B. Vargaftic, Handbook on Thermophysical Properties for Fluids
and Gases, Science Publ., Moscow, 1963.
[14] N.N. Smirnov, V.F. Nikitin, V.A. Nerchenko, P. Andreini, G. Manzini,
Water droplet non-equilibrium interaction with heated atmo-
sphere. ‘‘La Termotecnica’’ Gennaio–Febbraio 1 (2011) 67–71.
[15] N.N. Smirnov, V.F. Nikitin, J.C. Legros, Ignition and combustion of
turbulized dust—air mixtures, Combustion and Flame 123 (1/2)
(2000) 46–67.
[16] N.N. Smirnov V.F. Nikitin, J. Khadem, Sh. Aliari Shourekhdeli, Onset
of detonation in polydispersed fuel–air mixtures, in: Proceedings of
the Combustion Institute. 2006, vol. 31, pp. 832–841.
[17] B.J. Azzopardi, G.F. Hewitt, Atomization of droplets, Multiphase
Science & Technology 9 (1997) 109–204.
[18] N.N. Smirnov, Combustion and detonation in multiphase media.
Initiation of detonation in dispersed-film systems behind a shock
wave, International Journal of Heat and Mass Transfer 31 (4) (1988)
779–793.
[19] N.N. Smirnov (Ed.), Space Debris Hazard Evaluation and Mitigation,
Taylor and Francis, London, New York, 2002.
[20] V.R. Dushin, A.V. Kulchitskiy, V.A. Nerchenko, V.F. Nikitin,
E.S. Osadchaya, Yu.G. Phylippov, N.N. Smirnov, Mathematical
simulation for non-equilibrium droplet evaporation, Acta Astro-
nautica 63 (2008) 1360–1371.
[21] R. Prud’homme, M. Habiballah, L. Matuszewski, Y. Mauriot,
A. Nicole, Theoretical analysis of dynamic response of a vaporizing
droplet to acoustic oscillations, Journal of Propulsion And Power 26
(1) (2010) 74–83.