Vous êtes sur la page 1sur 25

Dual-mode nonlinear instability analysis of a confined planar liquid sheet

sandwiched between two gas streams of unequal velocities and prediction of droplet
size and velocity distribution using maximum entropy formulation
Debayan Dasgupta, Sujit Nath, and Dipankar Bhanja

Citation: Physics of Fluids 30, 044104 (2018); doi: 10.1063/1.5022346


View online: https://doi.org/10.1063/1.5022346
View Table of Contents: http://aip.scitation.org/toc/phf/30/4
Published by the American Institute of Physics
PHYSICS OF FLUIDS 30, 044104 (2018)

Dual-mode nonlinear instability analysis of a confined planar liquid sheet


sandwiched between two gas streams of unequal velocities and prediction
of droplet size and velocity distribution using maximum entropy
formulation
Debayan Dasgupta, Sujit Nath,a) and Dipankar Bhanja
Department of Mechanical Engineering, National Institute of Technology Silchar, Assam 788 010, India
(Received 14 January 2018; accepted 27 March 2018; published online 12 April 2018)

Twin fluid atomizers utilize the kinetic energy of high speed gases to disintegrate a liquid sheet into
fine uniform droplets. Quite often, the gas streams are injected at unequal velocities to enhance the
aerodynamic interaction between the liquid sheet and surrounding atmosphere. In order to improve
the mixing characteristics, practical atomizers confine the gas flows within ducts. Though the liquid
sheet coming out of an injector is usually annular in shape, it can be considered to be planar as the
mean radius of curvature is much larger than the sheet thickness. There are numerous studies on
breakup of the planar liquid sheet, but none of them considered the simultaneous effects of confine-
ment and unequal gas velocities on the spray characteristics. The present study performs a nonlinear
temporal analysis of instabilities in the planar liquid sheet, produced by two co-flowing gas streams
moving with unequal velocities within two solid walls. The results show that the para-sinuous mode
dominates the breakup process at all flow conditions over the para-varicose mode of breakup. The
sheet pattern is strongly influenced by gas velocities, particularly for the para-varicose mode. Spray
characteristics are influenced by both gas velocity and proximity to the confining wall, but the former
has a much more pronounced effect on droplet size. An increase in the difference between gas veloc-
ities at two interfaces drastically shifts the droplet size distribution toward finer droplets. Moreover,
asymmetry in gas phase velocities affects the droplet velocity distribution more, only at low liquid
Weber numbers for the input conditions chosen in the present study. Published by AIP Publishing.
https://doi.org/10.1063/1.5022346

I. INTRODUCTION regime. Mitra et al.22 also reported the significant deviation


between the results obtained from linear stability analysis and
Atomization of liquid jets aided by co-flowing gas streams
experimental results mainly near the sheet breakup region. A
has been studied by several researchers and is widely used
detailed nonlinear study of planar liquid sheets in the presence
in many industrial applications such as chemical combustion,
of moving gas streams was presented by Nath et al.27 Yang
gas turbine, fire suppression system, spray cooling, etc.18 The
et al.35 discussed the effects of surrounding gas velocity on the
strong shear action between the moving air streams and the
sinuous mode of instability for planar liquid sheets moving in
liquid jet allows the formation of fine sprays even at low liq-
a co-flowing inviscid gaseous medium. Deviation of the gas
uid flow rates and pressures. Lefebvre16 described such a type
to liquid velocity ratio from unity resulted in a larger growth
of atomization as air blast atomization. Herrero et al.8 and
rate, higher initial amplitude, and shorter breakup length. Yan
Rostami et al.30 performed a temporal linear stability analy-
et al.34 performed a three-dimensional nonlinear spatial stabil-
sis of viscous liquid sheets exposed to inviscid moving gas
ity analysis of annular viscous liquids flowing in gas medium
streams. Both the studies elucidated the importance of air
with axial velocity. The study identified the shear effect result-
velocity in influencing the breakup time and droplet diameter.
ing from the gas-to-liquid velocity difference as the primary
Matas et al.20 carried out the experimental and linear analyt-
source of instability. In certain applications such as fuel prepa-
ical investigation of planar mixing layers with a fast moving
ration in jet engines and the spray drying process, the outer
gas stream and a slower parallel liquid stream. Spatial analysis
and inner gas velocities are kept unequal to provide higher
showed a steady increase in the growth rate with an increase
instability.
in the gas velocity. However, the linear investigation of the
A study on the effects of unequal gas velocities on spray
above studies cannot accurately predict the nonlinear behav-
characteristics was pioneered by Li.17 He suggested that the
ior of the waveforms when the wave amplitude becomes too
presence of asymmetry in gas velocity led to two modes of dis-
high. Asare et al.1 emphasized the importance of non-linear
turbances, viz., the para-sinuous mode (phase difference θ  0)
theory to predict the growth rate at a later stage of the unstable
and para-varicose mode (phase difference θ  π). The para-
sinuous mode was more unstable at a higher Weber number
a) Author
(We), and the para-varicose mode dominated the disintegration
to whom correspondence should be addressed: sujitnath2008@
gmail.com and sujit@mech.nits.ac.in. Tel.: +91 3842 224879/242273. process at a lower Weber number. Ibrahim and Jog9 performed
Fax: +91 3842 224797. Mobile: +91 7896287211. a three-dimensional nonlinear temporal instability analysis of

1070-6631/2018/30(4)/044104/24/$30.00 30, 044104-1 Published by AIP Publishing.


044104-2 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

annular liquid sheets subjected to unequal gas velocities with The literature reveals that there are several studies on
one gas stream quiescent. Nath et al.26 studied the influence confinement and unequal gas velocities, but none of them
of non-zero unequal gas velocities on the breakup of planar considered their combined simultaneous effect on spray char-
inviscid liquid sheets. There is also a plethora of experimental acteristics. The present study aims at investigating the effects
investigations on the breakup of liquid sheets in the pres- of unequal gas velocity on disintegration of the planar invis-
ence of moving air streams. Marmottant and Villermaux19 in cid liquid sheet in the presence of two finite thickness air
their experimental investigation on different types of liquids streams. It extensively analyzes the effects of several parame-
sheared by fast moving gases observed that the atomization ters such as the gas liquid density ratio, liquid Weber number,
process is associated with two types of instabilities: Kelvin- gas to liquid velocity ratio, asymmetry in gas phase veloci-
Helmholtz type and Rayleigh-Taylor type. The diameter of ties, and confinement thickness on the para-sinuous and para-
the droplets formed after ligament breakup was calculated varicose mode of instabilities. It also identifies, for a com-
from their surface area using exposure photographs. The liq- bination of parameters, the minimum wall separation that
uid volume contained in one ligament was thereby obtained is required for liquid sheet disintegration and the maximum
by summing the volume of all droplets. Jiang et al.11 experi- wall separation beyond which an increase in the confine-
mentally obtained a correlation for the Sauter mean diameter ment height has a negligible effect on the breakup length.
(SMD) with gas velocity, liquid velocity, liquid jet diameter, Sections II and III of the present paper present the math-
and liquid/gas mass flux ratio for a high speed annular air ematical formulation and corresponding solution procedure,
jet. Avulapati and Venkata2 and Déjean et al.5 experimentally respectively. Section IV gives the mathematical modeling for
investigated the effects of the gas to liquid mass ratio, liquid the droplet size and velocity distribution using maximum
flow rate, and shearing intensity of air on liquid sheets for an air entropy formulation. Section V is devoted to the analysis of
assist atomizer and air blast atomizer, respectively. Leboucher results. Mathematical terms and calculations which could not
et al.15 and Duke et al.6 illustrated the effects of inner and outer be shown in Secs. II and III for brevity are presented in the
gas jets on disintegration of annular liquid sheets. Mezhericher Appendixes.
et al.21 produced very fine droplets by disintegrating bubbles
with compressed air jets and studied the spray characteristics
II. MATHEMATICAL FORMULATION
by using a high speed visualization technique. Warncke et al.33
compared the numerical and experimental findings of atom- In most of the common practical injectors, even though
ization of prefilming air blast atomizers, using the high speed the liquid sheet coming out of the injector is annular in shape,
shadowgraphy technique and direct numerical simulation, the radius of curvature is usually much larger than the sheet
respectively. thickness. Hence the annular configuration can be reason-
In certain practical applications of liquid jets such as liquid ably simplified to a planar sheet. Cousin and Dumouchel4
rocket engines and gas turbine engines, addition of confine- showed that viscosity had a negligible effect on the growth
ment in injectors significantly improves mixing efficiencies of sinusoidal perturbations if We2g /Rel < 0.001, where Weg
and spray characteristics. Subramaniam and Parthasarathy32 and Rel represent the gas Weber number and liquid Reynold’s
performed temporal stability analysis of viscous liquid sheets number, respectively. Moreover, high velocity gas streams in
subjected to inviscid incompressible gas jets within a tube. atomizers also lead to high values of the Froude number,
Juniper 12 and Rees and Juniper 29 studied the effects of con- resulting in less effect of gravity. As a result, both the liq-
finement on the convective and absolute transition of two- uid and the gas are considered to be inviscid and the effect
dimensional jets and wakes. Chatterjee et al.3 performed a of gravity is neglected in the present study. The present study
temporal analysis of instability in confined annular, swirling considers a 2-dimensional steady planar liquid sheet having
inviscid liquid sheets exposed to swirling gas flows on both thickness 2h, density ρl , surface tension σ l , and velocity ul .
sides. Again, the presence of outer confinement enhanced the Two gas streams of density ρg and velocities ug1 and ug2 sur-
growth rate of instability, as compared to the limiting case of round the liquid sheet. The subscripts 1 and 2 represent the
outer air stream thickness tending to infinity. Mohanta et al.23 upper and lower liquid-gas interface, respectively, as depicted
and Fu et al.7 performed temporal stability analysis of coaxial in Fig. 1.
confined jets with heat and mass transfer at the interface. Both Following the model adopted by Nath et al.,25 the study
the results showed that confinement had a destabilizing effect considers a simple form of confinement with two parallel solid
on the liquid sheet in the presence of heat and mass transfer. walls having a separation distance 2H between them. The
However, all these studies on confined flows are limited to lin- abscissa is chosen parallel to the liquid flow, along the center
ear analysis. Kan and Yoshinaga13 investigated the linear and line of the liquid sheet, and the ordinate is chosen normal to the
nonlinear instability of inviscid planar sheets in moving gas liquid flow. The liquid flow sandwiched between gas streams
streams surrounded by solid walls. Although their study con- with unequal velocities is characterized by phase difference θ
sidered the effects of several flow conditions, the influence of between the two surface waves generated at the two gas liq-
gas velocity on sheet behavior was ignored. A detailed study uid interfaces. This phase difference is caused by the presence
on the effects of confinement on the breakup of planar inviscid of asymmetry in gas velocities in the upper and lower gas-
liquid sheets in co-flowing gas streams was performed by Nath liquid interfaces.17 As phase difference θ approaches zero, the
et al.25 They found that the presence of confinement signifi- mode of instability is called para-sinuous mode, and when θ
cantly decreased the breakup time and caused an increase in approaches π, it is called para-varicose mode. At undisturbed
the droplet diameter. state, a constant velocity is chosen along the x-direction and
044104-3 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

FIG. 1. Schematic of the planar liquid sheet sandwiched between two gas streams with unequal velocities confined within two parallel walls for (a) para-sinuous
and (b) para-varicose modes of instability.

zero velocity is chosen along the y-direction. The physical Kinematic boundary conditions
parameters in the governing equations and interface conditions
are non-dimensionalised using the following scales: ϕl1,y − η j1,t − ϕl1,x η j1,x = 0, (4)

[length, time, density, velocity] = [h, h/ul , ρl , ul ]. ϕgj1,y − η j1,t − ϕgj,x η j,x = 0. (5)

The formulation closely resembles the approach of Nath Dynamic boundary condition
et al.27 However, the solution procedure and initial conditions 1 1 1  
are significantly different and are motivated by the approach of − ρ Uj2 + ρϕgj,t − ϕl,t + ρ ϕ2gj,x + ϕ2gj,y
2 2 2
Ibrahim and Jog.9 The far boundary condition of the gas lay-
1  2  (−1) j
η j,xx
ers is taken as per Nath et al.25 Both the phases are assumed − ρ ϕl,x + ϕ2l,y =  3/2 . (6)
to be incompressible and irrotational. Considering the inter- 2 
We 1 + η 2 j,x
face to be a material surface, axial interface velocity shall be
equal to local fluid velocity which in turn yields the kine- The initial amplitude η 0 = 0.1 is taken as the perturbation
matic boundary condition. The dynamic boundary condition parameter keeping consistency with other literature stud-
is derived by balancing the surface tension force of the liquid ies.10,22,27 The surface deformation η j (x, t) is expressed in
and the jump in normal stress across the interface due to pertur- power series of η 0 as
bation. The new position of the interfaces after perturbation is ∞
given as
X
η j (x, t) = η on η jn (x, t). (7)
y(x, t) = (−1)j+1 + η j (x, t). (1) n=1
Here, j = 1 and 2 represent the upper and lower liquid gas
η oo is not included in the expansion as it represents the known
interface, respectively, and η represents the surface defor-
unperturbed interfaces.
mation. Initial velocity potentials at unperturbed state are
Assuming η jn and all its derivatives are of the same order
represented as ϕl0 = x and ϕgj0 = U j x for liquid and gas
of magnitude, perturbed velocity potential for the liquid and
phase, respectively, where non-dimensionalised gas velocity is
gas phase is expressed in terms of power series of η 0 as
U j = ug j /ul .
The velocity potential for liquid (ϕl ) and gas (ϕg ) and ∞
X
surface deformation (η j ) shall satisfy the governing equation ϕl (x, t) = η on ϕln (x, y, t), (8)
and boundary conditions. The subscripts x and y represent the n=1
partial derivative with respect to space, and t represents the X∞

partial derivative withrespect to time.We and ρ are the liquid ϕgj (x, t) = η on ϕgjn (x, y, t). (9)
phase Weber number We = ρl ul2 h/σ and gas-liquid density n=1

ratio (ρ = ρg /ρl ), respectively. The first- and second-order governing equations and bound-
Governing equations ary conditions are obtained by comparing the power of η 0
ϕl1,xx + ϕl1,yy = 0, −1 + η 2 ≤ y ≤ 1 + η 1 , (2) in the expanded expressions. The boundary conditions are
applicable only at the perturbed interface. Following the work
of Jazayeri and Li,10 perturbed velocity potential can be
ϕg,xx + ϕg,yy = 0, 1 + η 1 ≤ y ≤ H, − H ≤ y ≤ −1 + η 2 . (3) represented as

η j2 η j3
ϕ = ϕ + η ϕ
j y y=(−1)j+1 + ϕ
yy y=(−1)j+1 + ϕyyy + · · ·. (10)
y=(−1)j+1 +ηj y=(−1)j+1 2! 3! y=(−1)j+1
044104-4 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

Jazayeri and Li10 showed that the third-order surface deforma- Dynamic boundary condition
tion of sinuous waves was sinuous which made the interface
(−1)j
more distorted but does not contribute to thinning or breakup. ρ ϕgj2,t − ϕl2,t + ρ Uj ϕgj2,x − ϕl2,x − η j2,xx
Yang et al.36 in their analysis on sinuous disturbances on two-    We 
dimensional planar viscous sheets also mentioned that in terms = − ρ η j1,t ϕgj1,y + η j1 ϕgj1,yt + η j1,t ϕl1,y + η j1 ϕl1,yt
of disintegration and breakup time, the third-order perturba- 1    
tion gave similar conclusion as the second-order perturbation. − ρ ϕ2gj1,x + ϕ2gj1,y − ρ Uj η j1,x ϕgj1,y + η j1 ϕgj1,yx
2
Ibrahim and Jog9 also retained terms up to second order and
obtained good agreement with the experimental results. Hence, 1 2   
+ ϕl1,x + ϕ2l1,y + η j1,x ϕl1,y + η j1 ϕl1,yx . (21)
the current study also performs nonlinear analysis up to second 2
order. Finally, the Gaster transformation is applied to trans- Boundary conditions at the solid walls
form temporally growing disturbance into the spatial growth
rate. After complex algebraic calculation, the first- and second- ϕgj2,y = 0 at y = ±H. (22)
order governing equations and boundary conditions for the Following the work of Ibrahim and Jog ,9 the initial conditions
liquid and gas phase are obtained as follows: for the first-order analysis are considered as
First-order analysis (η 0 ):
Governing equations Para-sinuous mode: η j1 (x, 0) = cos(kx), η j1,t (x, 0) = 0, (23)
ϕl1,xx + ϕl1,yy = 0, −1 ≤ y ≤ 1, (11)
Para-varicose mode: η j1 (x, 0) = (−1)j+1 cos(kx), η j1,t (x, 0) = 0.
ϕgj1,xx + ϕgj1,yy = 0, 1 ≤ y ≤ H, − H ≤ y ≤ −1. (12)
(24)
Kinematic boundary conditions
For the second-order analysis, the initial conditions for both
ϕl1,y − η j1,t − η j1,x = 0, (13) para-sinuous and para-varicose modes are considered as
ϕgj1,y − η j1,t − Uj η j1,x = 0. (14)
η j2 (x, 0) = cos(kx), η j2,t (x, 0) = 0, (25)
Dynamic boundary condition
where k represents the dimensionless wave number.
(−1)j+1
ρϕgj1,t − ϕl1,t + ρUj ϕgj1,x − ϕl1,x = η j1,xx . (15)
We
III. SOLUTION PROCEDURE
Boundary conditions at the solid walls
First-order analysis
ϕgj1,y = 0 at y = ±H. (16) Following Ibrahim and Jog,9 the first-order deformations
Second-order analysis (η 20 ): for upper and lower interfaces are considered, respectively, as
Governing equations
η 11 (x, t) = A1 (t) exp(ikx) + c.c., (26)
ϕl2,xx + ϕl2,yy = 0, −1 ≤ y ≤ 1, (17) η 21 (x, t) = B1 (t) exp(ikx) + c.c., (27)
ϕgj2,xx + ϕgj2,yy = 0, 1 ≤ y ≤ H, − H ≤ y ≤ −1. (18) where c.c. indicates the complex conjugate. Equations (26)
Kinematic boundary conditions and (27) are substituted in governing equations and kinematic
boundary conditions Eqs. (11)–(14). The wall condition used
ϕl2,y − η j2,t − η j2,x = η j1,x ϕl1,x − η j1 ϕl1,yy , (19)
in Eq. (16) is applied to the resulting equation to finally obtain
ϕgj2,y − η j2,t − Uj η j2,x = η j1,x ϕgj1,x − η j1 ϕgj1,yy . (20) the velocity potential for liquid and gas phases

cosh(ky)  
ϕl1 =
    
A1,t + ikA1 exp (ikx) + c.c. − B1,t + ikB1 exp (ikx) + c.c.
2k sinh(ky)
sinh(ky)       
+ A1,t + ikA1 exp (ikx) + c.c. + B1,t + ikB1 exp (ikx) + c.c. + Cl1 (t), (28)
2k cosh(ky)
( " ky
e (1 + e2k(h−y) ) 
# )
ϕg11 =

(A1,t + ikU1 A1 ) exp(ikx) + Cg11 (t), (29)
kek (1 − e2k(h−1) )
e (1 + e2k(h+y) ) 
( " −ky # )
ϕg21 =

(B1,t + ikU2 B1 ) exp(ikx) + Cg21 (t). (30)
kek (e2k(h−1) − 1)

The functions C l1 (t), C g11 (t), and C g21 (t) do not appear in the Equations (26)–(30) are substituted in the first-order
perturbed velocity field or the dispersion equation. Hence their dynamic boundary condition used in Eq. (15) to obtain the
exact forms are not important for the present analysis. following expressions:
044104-5 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

a11 A1,tt + b11 A1,t + c11 A1 + d11 B1,tt + e11 B1,t + f11 B1 = 0, (31) Now applying the initial conditions for para-sinuous and
para-varicose modes for upper and lower interfaces and per-
a21 A1,tt + b21 A1,t + c21 A1 + d21 B1,tt + e21 B1,t + f21 B1 = 0. (32) forming Laplace transform, the following expressions are
obtained for F A (s) and F B (s), for para-sinuous and para-
Details of the expression are given in Appendix A. varicose modes:

Para-sinuous
" #
(a11 + d11 ) (b11 + e11 )  2
 
s + d s + e s + f

21 21 21

 




 " 2 2 



#

 (a21 + d21 ) (b21 + e21 ) 
2
 

− s

 + d11 s + e11 s + f11  

2 2
FA (s) =
 
, (33)
a11 s2 + b11 s + c11 d21 s2 + e21 s + f21 − a21 s2 + b21 s + c21 d11 s2 + e11 s + f11
  
" #
(a11 + d11 ) (b11 + e11 )   
s + a21 s2 + b21 s + c21 


 



 " 2 2 



#

 (a21 + d 21 ) (b 21 + e 21 ) 
2
 

− s

 + a11 s + b11 s + c11  

2 2
FB (s) =
 
. (34)
d11 s + e11 s + f11 a21 s + b21 s + c21 − d21 s + e21 s + f21 a11 s2 + b11 s + c11
2 2 2
  

Para-varicose
" #
(a11 − d11 ) (b11 − e11 )  2
 
s + d s + e s + f

21 21 21

 




 " 2 2 



#

 (a21 − d21 ) (b21 − e21 ) 
2
 

− s

 + d11 s + e11 s + f11  

2 2
FA (s) =
 
, (35)
a11 s2 + b11 s + c11 d21 s2 + e21 s + f21 − a21 s2 + b21 s + c21 d11 s2 + e11 s + f11
  
" #
(a11 − d11 ) (b11 − e11 )   
s + a21 s2 + b21 s + c21 


 



 " 2 2 



#

 (a 21 − d 21 ) (b 21 − e 21 ) 
2
 

− s

 + a11 s + b11 s + c11  

2 2
FB (s) =
 
. (36)
2 2 2
d11 s + e11 s + f11 a21 s + b21 s + c21 − d21 s + e21 s + f21 a11 s2 + b11 s + c11
  

The first-order dispersion equation is obtained from the Equations (33) and (35) can be expressed in terms of first-
denominator of Eq. (33) or Eq. (34) and is expressed as order complex roots as
P1a P2a P3a P4a
s4 M1 + s3 M2 + s2 M3 + sM4 + M5 = 0. (37) FA (s) = + + + .
(s − iω11 ) (s − iω11 ) (s − iω21 ) (s − iω21 )
(40)
ω11 , ω̄11 , ω21 , and ω̄21 are the complex root of the first-order
Using inverse Laplace transform of Eq. (40), expression for
dispersion equation (37). The complex roots are represented
A1 (t) is obtained as
with angular frequency α and growth rate of perturbation β
such that A1 (t) = P1a eiω11 t + P2a eiω11 t + P3a eiω21 t + P4a eiω21 t . (41)

ω11 = iα1 + β1 , (38) Detailed forms of P1a , P2a , P3a , and P4a for both the modes
are given in Appendix A.
ω21 = iα2 + β2 . (39) Similarly, the expression for the surface distribution for
lower interface is expressed as
Growth rates β 1 and β 2 are plotted against wave number k to B1 (t) = R1 A1 (t), (42)
obtain the critical wave number k critical corresponding to the  
maximum growth rate β max for both the modes of instabilities. where R1 = −(a11 s2 + b11 s + c11 )/(d11 s2 + e11 s + f11 ) is a
k critical is considered as the characteristic wave number for the constant.
higher-order analysis. The details of the calculation and the Finally, Eqs. (41) and (42) are substituted in Eqs. (26) and
expressions for the constants M 1 , M 2 , M 3 , and M 4 are given (27) to obtain the final expression for the surface deformation
in Appendix A. at upper and lower gas liquid interfaces respectively.
044104-6 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

Second-order analysis η 22 (x, t) = B2 (t) exp(2ikx) + c.c. (44)


The second-order surface deformations at upper and lower Second-order velocity potentials for liquid and gas phases are
interfaces are given, respectively, as obtained following a procedure similar to the first-order anal-
ysis. Expressions for the second-order liquid and gas velocity
η 12 (x, t) = A2 (t) exp(2ikx) + c.c., (43) potentials are given as

( )
cosh(2ky) sinh(2ky)
ϕl2 = (Q1 − Q2 ) + (Q1 + Q2 ) exp (2ikx) + c.c. + Cl2 (t), (45)
4k sinh(ky) 4k cosh(ky)
 " 2ky 4k(h−y) )
#  (A2,t + 2ikU1 A2 )  
e (1 + e

   
ϕg12 =
  
k(1 + e 2k(h−1) )  exp(ikx) + Cg12 (t), (46)
2ke 2k (1 − e4k(h−1) )   
A ikU A

2A +


 − 1 1,t 1 1  

 (1 − e2k(h−1) )
   
 
 " −2ky #  (B2,t + 2ikU2 B2 )
4k(h−y) ) 
 
e (1 + e

  
ϕg22 =
 
 exp(ikx) + Cg22 (t).
2k (e4k(h−1) − 1) 
 k(1 + e2k(h−1) ) (47)
2ke
 
2B B + ikU B



 −
 (e2k(h−1) − 1) 1 1,t 2 1  

 
 
The exact form of C l2 (t), C g12 (t), and C g22 (t) is skipped for the same reason as explained in the first-order analysis. Q1 and Q2
are presented in Appendix B. As both para-sinuous and para-varicose modes are considered simultaneously for the second-order
analysis, A1 (t) can be expressed as
A1 = Ā eiω11 t + B̄ eiω11 t + C̄ eiω21 t + D̄ eiω21 t , (48)
where Ā , B̄, C̄, and D̄ are constants. Putting Eqs. (43), (45), (46), and (48) in the dynamic boundary condition Eq. (21), the
following expressions is obtained for the upper interface:
a12 A2,tt + b12 A2,t + c12 A2 + d12 B2,tt + e12 B2,t + f12 B2
= P1a exp(i2ω11 t) + P2a exp(i2ω11 t) + P3a exp(ω11 + ω11 )t + P4a exp(i2ω21 t)
+ P5a exp(i2ω21 t) + P6a exp(ω21 + ω21 )t + P7a exp(ω11 + ω21 )t + P8a exp(ω11 + ω21 )t
+ P9a exp(ω21 + ω11 )t + P10
a
exp(ω21 + ω11 )t. (49)
Expressions for P1a to P10
a are obtained with MATHEMATICA 8 and are not shown here for brevity.

Similar to the first-order analysis, using the second-order initial conditions for both para-sinuous and para-varicose modes
and performing Laplace transform, the following expressions are obtained for upper and lower interfaces:
   
Pa d22 s2 + e22 s + f22 − Pb d12 s2 + e12 s + f12
FA (s) =      , (50)
a12 s2 + b12 s + c12 d22 s2 + e22 s + f22 − a22 s2 + b22 s + c22 d12 s2 + e12 s + f12
   
Pa a22 s2 + b22 s + c22 − Pb a12 s2 + b12 s + c12
FB (s) =      , (51)
d12 s2 + e12 s + f12 a22 s2 + b22 s + c22 − d22 s2 + e22 s + f22 a12 s2 + b12 s + c12
where Pa = (P1a + P2a + · · · + P10
a ) and P b = (P b + P b + · · · + P b ). Details of the expressions a , b , c , d , e , and f
1 2 10 12 12 12 12 12 12 for
the upper interface and expressions a22 , b22 , c22 , d 22 , e22 , and f 22 for the lower interface are given in Appendix B.
A second-order dispersion equation is obtained from the denominator of Eq. (50) or Eq. (51) and is expressed as
s4 M21 + s3 M22 + s2 M23 + sM24 + M25 = 0. (52)
Details of Eq. (52) are given in Appendix B and the equation is solved to obtain the second-order complex roots
ω12 , ω̄12 , ω22 , and ω̄22 . Now, Eq. (50) can be expressed in terms of first- and second-order complex roots as
C1a C2a C3a C4a C5a
FA (s) = + + + +
(s − iω12 ) (s − iω12 ) (s − iω22 ) (s − iω22 ) (s − 2iω11 )
C6a C7a C8a C9a C10a
+ + + + +
(s − 2iω11 ) (s − (ω11 + ω11 )) (s − 2iω21 ) (s − 2iω21 ) (s − (ω21 + ω21 ))
C11a C12a C13a C14a
+ + + + . (53)
(s − (ω11 + ω21 )) (s − (ω11 + ω21 )) (s − (ω11 + ω21 )) (s − (ω11 + ω21 ))
Details of C 1a to C 14a are given in Appendix B. Using inverse Laplace, the expression for A2 (t) is obtained as
A2 (t) = C1a eiω12 t + C2a eiω12 t + C3a eiω22 t + C4a eiω22 t + C5a e2iω11 t + C6a e2iω11 t
+ C7a e(ω11 +ω11 )t + C8a e2iω21 t + C9a e2iω21 t + C10a e(ω21 +ω21 )t
+ C11a e(ω11 +ω21 )t + C12a e(ω11 +ω21 )t + C13a e(ω11 +ω21 )t + C14a e(ω11 +ω21 )t . (54)
044104-7 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

Using a similar procedure, the expression for B2 (t) is obtained as

B2 (t) = C1b eiω12 t + C2b eiω12 t + C3b eiω22 t + C4b eiω22 t + C5b e2iω11 t + C6b e2iω11 t + C7b e(ω11 +ω11 )t + C8b e2iω21 t + C9b e2iω21 t
+ C10b e(ω21 +ω21 )t + C11b e(ω11 +ω21 )t + C12b e(ω11 +ω21 )t + C13b e(ω11 +ω21 )t + C14b e(ω11 +ω21 )t . (55)

Finally, the expression for second-order surface deforma- Thus, the total number of droplets formed per unit time in
tion for the upper and lower interface can be obtained by the spray with diameter “i” ranging between 1 to m and velocity
substituting Eqs. (54) and (55) in Eqs. (43) and (44). m Pk
“j” ranging between 1 to k is expressed as N =
P
ni,j .
i=1 j=1
The probability of droplets having diameter d i and veloc-
IV. MODELING OF DROPLET SIZE AND DROPLET ity vj is expressed as Pij = nij /N. Hence,
VELOCITY DISTRIBUTION
m X
X k
Calculation of the droplet size and velocity distribution Pi,j = 1. (58)
requires information about breakup time and ligament area, i=1 j=1
which is obtained by the nonlinear analysis of liquid sheet
Equation (58) is known as the normalization condition.
disintegration. The second-order instability analysis presents
The respective Shannon entropy31 for the joint probability
the breakup length (l) and interface profile at the time of
function is expressed as
breakup, which are used to compute the mass mean diameter
and the source terms for the constraint equations for the max- m X
X k
imum entropy formulation model. The ligaments produced S = −K Pi,j ln(Pi,j ), (59)
after sheet breakup further breaks down into droplets at full i=1 j=1
wavelengths, following the Rayleigh instability.28 The crit- where K is the Boltzmann constant. Other constraints cho-
ical length can be expressed in terms of ligament diameter sen for the present model are the mass balance, momentum
(d cl ) as balance, and energy balance equations.
So, the constraint conditions considering the normal-
λ r = 4.5dcl . (56)
ized condition, mass balance, momentum balance, and energy
The mean droplet diameter is obtained from Eq. (56) using balance are given in Eqs. (60)–(63), respectively, as
conservation of mass as m X
X k
f √ g Z= Pi,j − S1 , (60)
dm = C 2.1325 A , (57) i=1 j=1

m X
k
where C = a factor that depends upon the type of nozzle and X
Y= Pi,j Di3 − S2 , (61)
spatial location in the spray;14 A = ligament area.
i=1 j=1
The modeling is based on maximum entropy formulation
m X
k
which closely resembles to that of Nath et al.24 However, the X
calculation of source terms in momentum and energy balance X= Pi,j Di3 Vj − S3 , (62)
i=1 j=1
follows a different approach as the current study considers both
unequal gas velocities and confinement. m X
X k m X
X k
Maximum entropy formulation W= Pi,j Di3 Vj2 + B Pi,j Di2 − S4 , (63)
The maximum entropy formulation states that the prob- i=1 j=1 i=1 j=1
ability of droplets having a particular size and velocity dis- where S 1 = 1 and S 2 = 1 (source terms for the normal-
tribution depends on the Shanon entropy of the system, sub- ized condition and mass balance condition, respectively);
jected to constraint conditions. The most likely occurring size Di = d i /d m (non-dimensional droplet diameter); V j = vj /ul
and velocity distribution belong to the set of droplets which (non-dimensional droplet velocity); B = 12σ/ρl ul2 dm ; σ = sur-
exhibits the highest Shanon entropy. face tension coefficient of liquid. Considering that gas veloc-
The individual droplet diameter of class i and veloc- ities at both the interfaces are higher than the liquid velocity,
ity of class j is represented as d i and vj , respectively. ni,j momentum and energy transfer are considered to take place
= total number of droplets per unit time with diameter d i and from gas to liquid. The source terms for momentum balance
velocity vj . (S 3 ) and energy balance (S 4 ) are given as

1 1
S3 = 1 + ρ(U1 − 1)2 LCf 1 + ρ(U2 − 1)2 LCf 2 if U1 > 1 and U2 > 1,
4 4
1 1
S4 = 1 + ρ(U1 − 1)3 LCf 1 + ρ(U2 − 1)3 LCf 2 if U1 > 1 and U2 > 1,
2 2
044104-8 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

where C f 1 and C f 2 = drag coefficient on the liquid at the Calculation of volumetric probability density of droplet size
upper interface and lower interface, respectively; L = non- and droplet velocity
dimensional breakup length (l/h). The calculation of source Following Lefebvre,16 the volumetric probability density
terms is not shown here for brevity and has been included in of droplets with diameter d i is calculated as
Appendix C.
The value of Pi,j that maximizes entropy S is given as ∂Qi /∂di = (Qi − Qi−1 )/(di − di−1 ), (66)

∂S ∂Z ∂Y ∂X ∂W where Qi =
i P
k
Pc,j Dc3 = (total volume of liquid with non-
+ λ1 + λ2 + λ3 + λ4 = 0,
P
(64)
∂Pi,j ∂Pi,j ∂Pi,j ∂Pi,j ∂Pi,j c=1 j=1
dimensional diameter less than d i )/(total volume of liquid).
where λ 1 , λ 2 , λ 3 , and λ 4 are Lagrange multipliers. The Similarly, the volumetric probability density of droplets
Newton-Rhapson scheme is used to obtain these Lagrange with velocity V j is calculated as
multipliers. The converged solution of λ 1 , λ 2 , λ 3 , and λ 4 is
   
substituted to obtain Pij , at different droplet diameters and ∂Q 0j /∂Vj = Q 0j − Q 0j−1 /Vj − Vj−1 , (67)
droplet velocities as given in Eq. (65),
   j P
m
Pi,j = exp − 1 + λ 1 + λ 2 Di3 + λ 3 Di3 Vj + BDi2 . (65) where Qj0 = Pi,c Di3 = (total volume of liquid with
P
c=1 i=1
Once Pi,j is evaluated, the probability distribution functions non-dimensional velocity less than V j )/(total volume of
for droplet size and velocity are evaluated as follows. liquid).

FIG. 2. Variation of growth rate with wave number at various confinements at U 1 = 4, U 2 = 4, ρ = 0.001, We = 40: (a) for sinuous mode and (b) for varicose
mode.

FIG. 3. Variation of growth rate with wave number without confinement and various U 2 at U 1 = 0, ρ = 0.001, We = 1000: (a) for para-sinuous mode and (b) for
para-varicose mode.
044104-9 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

V. RESULT AND DISCUSSION modes of instabilities, respectively, which show excellent


A. Model validation
agreement with the results of Nath et al.25
Figures 3(a) and 3(b) show the corresponding variations
The present model has been validated with the studies of for the unconfined configuration with still gas on one side
Nath et al.25 and Kan and Yoshinaga.13 Figures 2(a) and 2(b) and moving gas on the other side of the liquid sheet for para-
show the variation of growth rate ( β) with wave number (k) sinuous and para-varicose modes of instability, respectively.
for the planar liquid sheet surrounded by gas streams of equal In all the cases, the dispersion curve, the critical wave number
velocities in a confined environment for sinuous and varicose (k critical ) corresponding to maximum growth rate ( β max ), and

FIG. 4. Variation of maximum growth rate and the corresponding critical wave number with confinement height (H) at different We for U 1 = 2, ρ = 0.001:
[(a) and (b)] U 2 = 0.5, [(c) and (d)] U 2 = 2, [(e) and (f)] U 2 = 4.
044104-10 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

the range of unstable wave numbers show excellent agreement upon flow conditions. But the para-varicose mode shows a
with the results of Kan and Yoshinaga13 for the same input monotonic decrease in k critical with an increase in H before
parameters. achieving an asymptotic value at high values of H. Similar to
β max , k critical also increases with an increase in We for both
B. First-order analysis
the modes. It can also be seen from Figs. 4(b), 4(d), and 4(f)
The subscript “s” and “v” in all the subsequent dia- that typical k critical ranges between 0.01 and 0.1; i.e., the range
grams represent the para-sinuous and para-varicose mode, of the most unstable wave number is very small for both the
respectively. From the first-order analysis, we get the value modes. Since the thickness of the film for the planar sheet is
of maximum growth rate β max and corresponding wave num- also very small, it can be deduced that wavelengths (λ = 2π/k)
ber k critical . Figures 4(a), 4(c), and 4(e) and Figs. 4(b), 4(d), of the unstable waves are long. It can also be noticed that at
and 4(f) show the variation of β max and corresponding k critical unequal gas velocities [Figs. 4(b) and 4(f)], k critical for the para-
with half confinement height (H), respectively, at different sinuous mode is relatively lower than the para-varicose mode,
Weber numbers (We) for different gas to liquid velocity ratios but at equal gas velocities at two interfaces [Fig. 4(d)], k critical
(U 2 ), namely, U 2 = 0.5, 2, 4 keeping U 1 = 2. for the varicose mode is lower than the sinuous mode. The
It can be seen that for both the modes, β max decreases with implication is that the para-sinuous mode becomes unstable at
an increase in H and finally becomes asymptotically insensi- longer wavelengths in the case of unequal gas velocities. But
tive to the change in H at its high values. An increase in We as gas velocities at two interfaces become equal, the varicose
increases β max for both the modes of instabilities. In the case mode becomes unstable at longer wavelengths.
of the para-sinuous mode, β max almost becomes invariant to
the change in H at high values of U 2 and low We. The value C. Second-order analysis
of β max is almost one order higher for the para-sinuous mode For certain nozzle exit conditions, instabilities may grow
than the para-varicose mode at all values of U 2 . The critical in such a way that the sheet touches the solid walls before
wave number (k critical ) represents the wave number at which breakup can takes place. Such conditions are represented by
the growth rate ( β) becomes maximum. k critical for the para- dotted lines, while the solid lines indicate breakup as shown in
sinuous mode may increase or decrease with H depending Figs. 5 and 6. Figure 5 shows the variation of breakup time and

FIG. 5. Variation of breakup time with half confinement height at different ρ for U 1 = 2, We = 50: (a) U 2 = 0.5, (b) U 2 = 2, (c) U 2 = 4, and (d) U 2 = 6.
044104-11 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

FIG. 6. Variation of breakup time with half confinement height (H) at different We for U 1 = 2, ρ = 0.001: (a) U 2 = 0.5, (b) U 2 = 2, (c) U 2 = 4, and (d) U 2 = 6.

touching time with H at different gas to liquid density ratios (ρ) on the breakup time for the para-varicose mode when the gas
taken for four different values of U 2 = 0.5, 2, 4, and 6 velocities at two interfaces are equal. It can also be seen that an
keeping U 1 = 2. Breakup time reduces with an increase in increase in U 2 causes reduction in the breakup time for both
ρ for both the modes. Several other literature studies10,36 have the modes. High values of U 2 induce high gas inertia interac-
also concluded that an increase in the gas to liquid density tion with the liquid surface resulting in a shorter breakup time
ratio produces a more distorted interface resulting in shorter which is more prominent in the sinuous mode.
breakup time. It is also evident from Fig. 5 that there is no mode The present study identifies two critical confinement
shift point and the para-sinuous mode dominates the breakup heights, viz., CH-1 and CH-2. CH-1 denotes the minimum
process at all flow conditions. confinement height that allows disintegration of the liquid
Figure 6 shows the variation of breakup time and touching sheet without the interfaces touching the solid walls, whereas
time with H at different values of We for different values of U 2 CH-2 denotes the maximum confinement height beyond which
= 0.5, 2, 4, and 6 keeping U 1 = 2. An increase in We increases breakup time for confined and unconfined configurations
the aerodynamic interaction at two interfaces and consequently becomes the same. Figure 7(a) shows the relation between
reduces breakup time for both the modes. Similar to Fig. 5, CH-1 and gas velocity ratio (U 2 ) at We = 50 for different values
the breakup process is dominated by the para-sinuous mode of ρ. For the para-sinuous mode, an increase in U 2 decreases
at all input conditions. Figures 5 and 6 indicate the existence CH-1 until it reaches a minimum value. Any further increase in
of a critical height beyond which confinement has a negligi- U 2 increases the value of CH-1. The value of U 2 correspond-
ble effect on the breakup time. Once the breakup has taken ing to the minimum value of CH-1 decreases with an increase
place, further increase in the confinement height has a minor in ρ. However, the minimum value of CH-1 almost remains
effect on the breakup time for the para-sinuous mode. How- invariant to the change in ρ as evident from Fig. 7(a).
ever in the case of the para-varicose mode, at U 1 = U 2 = 2, On the other hand, in the case of the para-varicose mode,
the breakup time keeps on increasing with an increase in the asymmetry in gas velocity plays a significant role on the value
confinement height even after breakup has taken place. It indi- of CH-1 but does not contribute much change at higher values
cates that the effect of confinement has more significant effect of U 2 . It can be seen that CH-1 increases drastically as the gas
044104-12 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

FIG. 7. Variation of CH-1 with gas velocity ratio U 2 for U 1 = 2 (a) at various ρ and We = 50 and (b) at various We and ρ = 0.001.

liquid velocity ratios (U 1 and U 2 ) at two interfaces become breakup. But at higher values of U 2 , any further increase in
unequal. It indicates that for the para-varicose mode with the difference between U 1 and U 2 has a negligible effect on
unequal gas velocities at two interfaces, the solid walls need CH-1. However, the initial rate of increment of CH-1 with
to be separated sufficiently enough from each other to achieve U 2 decreases with an increase in ρ. Also from Fig. 7(a), it is

FIG. 8. Sheet profile for the para-sinuous mode at the instant of breakup/touch for U 1 = 2, ρ = 0.001, We = 50, H = 8 at (a) U 2 = 2, (b) U 2 = 6, (c) U 2 = 8,
(d) U 2 = 10.
044104-13 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

observed that at U 1 , U 2 , CH-1 decreases with an increase in breakup can take place only when non-linear effects come into
ρ; however, at equal velocities (U 1 = U 2 = 2), CH-1 is less play. From Fig. 8(a), it can be observed that at low values of
sensitive to the change in ρ. Figure 7(b) shows the relation gas velocity (U 2 = 2), sheet profiles predicted by linear and
between CH-1 and gas velocity (U 2 ) at ρ = 0.001 and different non-linear analyses are almost identical. This indicates less
values of We. In the case of the para-sinuous mode, the value pronounced non-linear effects and reduced sheet distortion.
of U 2 at which CH-1 becomes minimum, decreases with an Consequently, the disturbance amplitude grows with time and
increase in We. Similar to ρ, We also has a negligible effect on the interfaces touch the solid walls before breakup can take
the minimum value of CH-1. In case of para-varicose mode, the place. At U 2 = 6 and U 2 = 8, the non-linear effects become
value of CH-1 at which CH-1 becomes asymptotically insensi- more significant causing initial distortion and early breakup.
tive to the change in U 2 decreases drastically with an increase However, with further increase in U 2 (U 2 = 10), the sheet
in We. At U 1 , U 2 , CH-1 decreases with an increase in We. distortion is replaced by more deformation resulting in high
However, at equal velocities, i.e., U 1 = U 2 = 2, CH-1 almost wave amplitude. As a result, the interfaces touch the solid
remains invariant to the change in We. surface before breakup can take place, thereby indicating a
To explain the variation of CH-1 with gas velocity (U 2 ) higher value of CH-1.
for the para-sinuous mode further, sheet profiles at different Similar to the para-sinuous mode, the effect of unequal
values of U 2 have been presented in Fig. 8. The value of half gas velocity on CH-1 for the para-varicose mode can also be
confinement thickness (H) has been chosen slightly higher explained by sheet profiles at different values of U 2 and H.
than the minimum value of CH-1 for the analysis. Here, τ 1 Figure 9(a) shows that for equal gas velocities at two inter-
and τ 2 represent the breakup/touching time obtained from the faces (U 1 = U 2 = 2), the sheet deformation is visibly less and
first- and second-order analysis, respectively. Linear analysis the sheet profiles are almost identical for both linear and non-
predicts that the separation between the two interfaces almost linear analyses. It can also be seen that breakup is primarily
remains constant at all stages of sheet development for the caused by the phase shift and can take place even at lower
para-sinuous mode. Consequently, for the para-sinuous mode, values of confinement thickness (H = 10). However, Fig. 9(b)

FIG. 9. Sheet profile for the para-varicose mode at the instant of breakup/touch for U 1 = 2, ρ = 0.001, We = 50 at (a) U 2 = 2, H = 10, (b) U 2 = 3, H = 10,
(c) U 2 = 8, H = 180, and (d) U 2 = 10, H = 180.
044104-14 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

shows that when U 1 , U 2 , there is a significant difference in the liquid sheet touches the solid walls before breakup can
the sheet profile predicted by linear and non-linear analyses. take place. So, CH-1 shows a sudden increase in the value, as
As U 2 changes from 2 to 3, the nonlinear effects become more values of U 1 and U 2 change from equal to unequal. Also, the
pronounced resulting in high sheet deformation. As a result, first-order analysis completely ignores high wave amplitudes

FIG. 10. Variation of CH-2 with gas velocity ratio U 2 at U 1 = 2 (a) for various densities at We = 50 and (b) for various We at ρ = 0.001.

FIG. 11. Sheet profile for the para-sinuous mode at the instant of breakup for the unconfined configuration for We = 50, ρ = 0.001, U 1 = 2 at (a) U 2 = 2,
τ = 2049.91, (b) U 2 = 5, τ = 119.41, (c) U 2 = 8, τ = 22.75, and (d) U 2 = 10, τ = 12.47.
044104-15 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

generated due to asymmetry in gas velocity and predicts value of U 2 beyond which a further increase in U 2 does not
breakup at much lower value of H. This highlights the affect the sheet profile as well as CH-1 value. In Fig. 9(a), it can
importance of non-linear analysis for the para-varicose mode. also be seen that when there are equal gas velocities at upper
Figures 9(c) and 9(d) also show that the sheet profile almost and lower interfaces, both the fundamental and first harmonic
remains identical for U 2 = 6 and U 2 = 8, indicating a critical of the varicose mode is varicose. But, as the gas velocities at

FIG. 12. Evolution of the sheet profile for the sinuous mode with and without confinement for We = 10, ρ = 0.001, U 1 = 2, U 2 = 6, H = 30 at non-dimensional
time interval: (a) τ = 100, (b) τ = 150, (c) τ = 200, (d) τ = 250, (e) τ = 270, and (f) τ = 283.63 (confined breakup time).
044104-16 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

the two interfaces become unequal [Figs. 9(b)–9(d)], the para- into the para-sinuous mode. Figures 10(a) and 10(b) show the
varicose mode evolves into a para-sinuous mode. The reason variation of CH-2 with U 2 at different values of ρ and We,
can be contributed to asymmetry in gas velocities at the two respectively. In the case of the para-sinuous mode, variation
interfaces which gives rise to non-linear terms that shift the of CH-2 with ρ and We is similar to that of CH-1. A comparison
phase difference (θ) from “close to π” to “close to zero” and between Figs. 7 and 10 shows that CH-2 approaches CH-1 as
thereby changes the first harmonic of the para-varicose mode U 2 increases and almost becomes equal to CH-1 at high values

FIG. 13. Evolution of the sheet profile for the varicose mode with and without confinement for We = 50, ρ = 0.001, U 1 = 2, U 2 = 6, H = 170 at non-dimensional
time interval: (a) τ = 100, (b) τ = 300, (c) τ = 600, (d) τ = 650, (e) τ = 700, and (f) τ = 714.68 (confined breakup time).
044104-17 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

of U 2 . Therefore, it can be concluded that for the para-sinuous values of We. The effect of confinement height on the droplet
mode, the effect of confinement becomes negligible at high val- size distribution becomes significant only at lower values of
ues of U 2 . In the case of the para-varicose mode, CH-2 almost We. It is observed that at We = 5 and 10, droplets size reduce
remains invariant with U 2 for higher values of We. However, with an increase in wall separation. The maximum value of
Fig. 10(b) shows that at low Weber numbers (We = 10), droplet distribution function also shows an increase as the con-
CH-2 increases marginally with U 2 before becoming constant finement walls further move apart from each other. However,
at around U 2 = 4. the effect of confinement becomes increasingly less significant
Figure 11 shows the sheet profile for the unconfined con- at higher Weber numbers and droplet size almost becomes
figuration at the instant of breakup for the para-sinuous mode invariant to confinement thickness at We = 50. On the other
at different values of U 2 . At low values of U 2 (U 2 = 2), the sheet hand, Figs. 14(b), 14(d), and 14(f) show variation in the droplet
undergoes sufficient deformation before breakup [Fig. 11(a)]. velocity distribution with H at different values of We. It can be
Similar sheet behavior can be achieved by the confined con- seen that the velocity distribution almost remains invariant to
figuration only when the solid walls are sufficiently away from change in wall separation.
each other. Therefore, in order to accommodate high amplitude From Table I, it is also observed that for unequal gas
of deformation, CH-2 values are relatively higher at low values phase velocity condition, the breakup time increases with
of U 2 . As U 2 increases, non-linear effects become increasingly the confinement height, but the ligament area and, hence,
significant in causing breakup. Figures 11(b) and 11(c) show mean-droplet-diameter decreases. Therefore, unequal velocity
that for U 2 = 5 and U 2 = 8, the sheet profile exhibits lower in confined environment promotes finer droplets. A higher
amplitude of deformation and high sheet distortion, causing Weber number encourages further finer droplets. Again if
early breakup. This explains lower values of CH-2 at rela- confinement height is decreased, breakup will decrease but
tively high values of U 2 . However, at U 2 = 10 [Fig. 11(d)], the larger droplets will be produced.
absence of initial distortion allows deformation amplitude to Figures 15(a), 15(c), and 15(e) show variation of droplet
grow sufficiently before breakup can take place, thereby indi- size with gas velocity U 2 at different Weber numbers. Keeping
cating high values of CH-2. Therefore, from Fig. 11, it can be
reconfirmed that for the para-sinuous mode there exists a crit- TABLE I. Result of nonlinear stability analysis for ρ = 0.001, U 1 = 2 at
ical value of U 2 for which the CH-2 value will be minimum different We and various values of half confinement thickness H.
and beyond which or below which CH-2 value will be always
Cross-sectional
more. Non-dimensional area of ligament
Figure 12 shows instantaneous sheet profiles for the para- We H U1 U2 breakup time after breakup (A)
sinuous mode of breakup with unequal gas velocities at two
interfaces. Here, H is chosen between CH-1 and CH-2 so that a 5 51.41 2 6 446.162 249.234
comparison with the unconfined configuration can be shown at 5 80 2 6 464.238 199.403
5 200 2 6 465.556 192.087
different instants of non-dimensional time (τ). For the chosen
parameters, confined and unconfined sheets exhibit almost a 10 26.799 2 6 280.021 128.202
similar sheet pattern at early stages of wave propagation. How- 10 40 2 6 285.983 98.159
ever, the difference in wave amplitude between the confined 10 60 2 6 286.710 93.764
and unconfined configuration grows with time and the confine-
50 4.587 2 6 58.429 7.681
ment effects become more predominant at the later stages of 50 5 2 6 58.66 7.634
sheet development. It can be seen from Fig. 12(f) that the final 50 7 2 6 58.919 7.579
sheet breakup takes place due to the sheet deformation. Faster
growth of wave amplitude causes relatively earlier breakup for
the confined configuration than H → ∞ (τ = 286.17). The con- TABLE II. Result of nonlinear stability analysis for ρ = 0.001, U 1 = 2 at
fined configuration also shows higher values of wave length different We and gas velocity ratio U 2 .
(λ) as compared to H → ∞.
Figure 13 shows the sheet profile at different instants Cross-sectional area
Non-dimensional of ligament after
of non-dimensional time (τ) for the para-varicose mode
We H U1 U2 breakup time breakup (A)
with unequal gas velocities at two interfaces. Similar to the
para-sinuous mode, the value of H for the particular flow con- 5 500 2 2 12806.643 2607.13
dition has been considered in between CH-1 and CH-2. For the 5 120 2 4 1337.46 555.542
chosen input parameters, confinement seems to have a negli- 5 60 2 6 456.3891 225.123
gible effect on the sheet profile. Sheet profiles both at initial 5 30 2 8 209.409 125.639
and later stages of the breakup process are almost identical. 10 260 2 2 6313.431 1306.275
The breakup times for confined and unconfined configurations 10 70 2 4 871.422 304.861
are also almost the same, i.e., at τ = 714.68 and τ = 714.92, 10 30 2 6 283.62 116.33
respectively. 10 15 2 8 126.357 58.715
The para-sinuous mode dominates the breakup process in
50 65 2 2 1935.567 314.002
the presence of asymmetry in gas velocities at two interfaces.
50 15 2 4 287.037 62.203
Hence, all the data mentioned in Tables I and II correspond to 50 4.587 2 6 58.66 7.6343
the para-sinuous mode. Figures 14(a), 14(c), and 14(e) show 50 6.138 2 8 22.753 3.854
variation of the droplet size distribution with H atdifferent
044104-18 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

FIG. 14. Effect of half confinement thickness on droplet size and velocity distribution for U 1 = 2, U 2 = 6, ρ = 0.001: [(a) and (b)] We = 5, [(c) and (d)]
We = 10, and [(e) and (f)] We = 50.

all other conditions constant, higher We shifts the distribu- droplet velocity distribution with U 2 for different values of
tion curve toward smaller droplets’ size. A shift in the droplet We. The effect of gas velocity on the droplet velocity distribu-
size distribution toward smaller droplets can also be observed tion is more prominent at lower values of We. Droplet velocity
with an increase in U 2 . An increase in the gas velocity leads increases as U 2 increases from U 2 = 2 to 4. However, further
to higher aerodynamic interaction between the bulk liquid increase in U 2 causes the velocity distribution curve to shift
and surrounding gas which subsequently results in smaller toward left. Keeping all other conditions constant, the rate of
droplet diameter and higher surface energy after atomiza- decay on either side of the peak reduces with an increase in
tion. Figures 15(b), 15(d), and 15(f) show the variation of U 2 . This shows that the variation in the velocity distribution
044104-19 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

FIG. 15. Effect of gas velocity ratio U 2 on droplet size and velocity distribution for U 1 = 2, ρ = 0.001: [(a) and (b)] We = 5, [(c) and (d)] We = 10, [(e) and (f)]
We = 50.

function is higher at higher values of U 2 . As U 2 reduces, most in higher aerodynamic interaction at the interfaces, produc-
of the droplets travel with the same velocity as exhibited at the ing finer droplets. Figures 14(b), 14(d), 14(f), 15(b), 15(d),
atomizer exit. and 15(f) show a shift in the velocity distribution toward right
Figures 14(a), 14(c), 14(e), 15(a), 15(c), and 15(e) show with an increase in We. At lower We, the increased surface
that an increase in We shifts the droplet size distribution curve force is balanced by surface energy at the expense of liquid
to left toward the finer droplets. Higher values of We result kinetic energy. Reduction in liquid kinetic energy subsequently
044104-20 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

produces low velocity droplets. This effect becomes more droplet velocity. The rate of decay on either side of the peak
prominent at higher values of U 2 (U 2 = 6, 8). From Table II reduces with an increase in asymmetry in the gas to liquid
also, it is evident that as the asymmetry of two gas phase velocity ratio at two interfaces, thereby indicating a less varia-
velocity increases breakup time as well as ligament area and tion in the velocity distribution at a higher gas to liquid velocity
hence, mean-droplet-diameter also decreases, resulting finer ratio U 2 .
droplets.
APPENDIX A: CONSTANTS OF FIRST-ORDER
VI. CONCLUSION ANALYSIS

In this study, a nonlinear temporal analysis has been con- Constants of Eqs. (31) and (32) are given as
ducted to study the effects of unequal gas velocity on an coth(k) tanh(k) coth [(−1 + h) k]
inviscid planar liquid sheet in a confined environment. The a11 = + +ρ , (A1)
2k 2k k
first- and second-order governing equations and boundary
b11 = i coth(k) + 2i ρU1 coth [(−1 + h) k] + i tanh (k), (A2)
conditions have been derived for the para-sinuous and para- k2 1 1
varicose mode of instability using the perturbation expansion c11 = − k coth (k)−k ρU12 coth [(−1 + h) k]− k tanh (k),
We 2 2
technique. The results show that the maximum growth rate (A3)
decreases with an increase in confinement height at all input coth(k) tanh (k)
d11 = − + , (A4)
parameters. The study identifies two critical heights, CH-1 and 2k 2k
CH-2. While CH-1 stands for the minimum height that leads e11 = −i coth (k) + i tanh (k), (A5)
to breakup without interfaces touching the solid surface, CH-2 1 1
denotes the maximum height beyond which confinement has f11 = k coth (k) − k tanh (k), (A6)
2 2
a negligible effect on the breakup time. In the case of the para- coth(k) tanh (k)
sinuous mode, CH-1 decreases with an increase in the gas a21 = − + , (A7)
2k 2k
velocity until a critical value beyond which it increases with
b21 = −i coth (k) + i tanh (k) , (A8)
further increase in the gas to liquid velocity ratio. For the para-
1 1
varicose mode, the same parameter increases drastically as the c21 = k coth (k) − k tanh (k) , (A9)
gas to liquid velocity ratio at two interfaces became unequal. 2 2
Further increase in the gas to liquid velocity difference between coth(k) tanh(k) coth [(−1 + h) k]
d21 = + +ρ , (A10)
two interfaces has a negligible effect on the CH-1. Moreover, 2k 2k k
the presence of confinement results in shorter breakup time e21 = i coth(k) + 2i ρU2 coth [(−1 + h) k] + i tanh (k), (A11)
as compared to the unconfined condition. However, further k2 1 1
increase in the wall separation beyond CH-1 causes an increase f21 = − k coth (k)−k ρU22 coth [(−1 + h) k]− k tanh (k) .
We 2 2
in the breakup time. The breakup process is dominated by the (A12)
para-sinuous mode at all flow conditions. There is a minor Constants of the first-order dispersion Eq. (37) are given as
reduction in droplet size due to the increase in confinement
M1 = a11 × d21 − d11
2
, (A13)
thickness at low liquid Weber numbers. However, the droplet
size distribution becomes increasingly invariant to confine- M2 = a11 × e21 + b11 × d21 − 2 × d11 × e11 , (A14)
ment thickness with an increase in the liquid Weber number.
Again, the droplet velocity distribution is almost insensitive M3 = a11 × f21 + c11 × d21 + b11 × e21 − e211 − 2 × d11 × f11 ,
to the variation in confinement thickness at all flow condi- (A15)
tions. An increase in difference in gas velocities drastically M4 = b11 × f21 + c11 × e21 − 2 × e11 × f11 , (A16)
shifts the droplet size distribution toward finer droplets. Fur-
thermore, the droplet velocity increases when the gas to liquid M5 = c11 × f21 − 2
f11 . (A17)
velocity ratio at two interfaces became unequal. However, fur- Unknowns of Eq. (41) in the case of the para-sinuous mode is
ther increasing difference in gas velocity ratios reduces the given as

 
ω11 (a11 + d11 ) (b11 + e11 ) d21 ω11 + e21 ω11 + f21
 2 
 
+


 


2 2

 

   
ω 2 ω
 − ω11 (a21 + d21 ) +




 (b 21 + e 21 ) d11 11 + e11 11 + f 11 





P1a =
 2 2 
, (A18)
(ω11 − ω̄11 ) (ω11 − ω21 ) (ω11 − ω̄11 )
 

 ω̄ (a
11 11 + d 11 ) (b11 + e 11 ) d21 ω̄ 2 + e ω̄ + f
11 21 11 21 

+


 


2 2

 

   
ω̄ 2 ω̄
 − ω̄11 (a21 + d21 ) +




 (b 21 + e 21 ) d 11 11 + e 11 11 + f 11






P2a =
 2 2 
, (A19)
(ω̄11 − ω11 ) (ω̄11 − ω21 ) (ω̄11 − ω̄21 )
044104-21 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

 

 ω 21 (a 11 + d 11 ) (b11 + e11 ) d21 ω21 2 +e ω +f
21 21 21 

+


 


2 2

 

   
ω 2 ω
 − ω21 21 + d21 +




 (a ) (b 21 + e 21 ) d 11 21 + e 11 21 + f 11






P3a =
 2 2 
, (A20)
(ω21 − ω11 ) (ω21 − ω̄11 ) (ω21 − ω̄21 )
 

 ω̄ 21 (a 11 + d 11 ) (b11 + e11 ) d21 ω̄21 2 + e ω̄ + f
21 21 21 

+


 


2 2

 

   



 ω̄ 21 (a 21 + d 21 ) (b 21 + e 21 ) d 11 ω̄ 2
21 + e 11 ω̄ 21 + f 11




 −
 + 

P4a =
 2 2 
. (A21)
(ω̄21 − ω11 ) (ω̄21 − ω̄11 ) (ω̄21 − ω21 )
Unknowns of Eq. (41) in the case of the para-varicose mode are given as
 

 ω 11 (a 11 − d 11 ) (b11 − e11 ) d21 ω11 2 +e ω +f
21 11 21 

+


 


2 2

 

   



 ω 11 (a21 − d 21 ) (b21 − e21 ) d 11 ω 2 +e ω +f
11 11 11 11




 −
 + 

P1a =
 2 2 
, (A22)
(ω11 − ω̄11 ) (ω11 − ω21 ) (ω11 − ω̄11 )
 
ω̄11 (a11 − d11 ) (b11 − e11 ) d21 ω̄11 + e21 ω̄11 + f21
 2 
 
+


 


2 2

 

   

 ω̄ (a − d ) (b21 − e21 ) d 11 ω̄ 2 + e ω̄ + f
11 11 11 11


 − 11 21 21

 

 + 

P2a =
 2 2 
, (A23)
(ω̄11 − ω11 ) (ω̄11 − ω21 ) (ω̄11 − ω̄21 )
 
ω21 (a11 − d11 ) (b11 − e11 ) d21 ω21 + e21 ω21 + f21
 2 
 
+


 


2 2

 

   

 ω (a − d ) (b21 − e21 ) d 11 ω 2
21 + e11 ω 21 + f11


 − 21 21 21

 

 + 

P3a =
 2 2 
, (A24)
(ω21 − ω11 ) (ω21 − ω̄11 ) (ω21 − ω̄21 )
 
ω̄21 (a11 − d11 ) (b11 − e11 ) d21 ω̄21 + e21 ω̄21 + f21
 2 
 
+


 


2 2

 

   
ω̄ 2 ω̄
 − ω̄21 (a21 − d21 ) +




 (b21 − e21 ) d 11 21 + e11 21 + f11






P4a =
 2 2 
. (A25)
(ω̄21 − ω11 ) (ω̄21 − ω̄11 ) (ω̄21 − ω21 )

APPENDIX B: CONSTANTS OF SECOND-ORDER ANALYSIS


Unknowns of Eq. (45) are given as

 coth(k) [(A1 (t) + ikA1 ) − (B1 (t) + ikB1 )] 



Q1 = (A2,t + 2ikA2 ) − kA1 

 + tanh(k) [(A1 (t) + ikA1 ) + (B1 (t) + ikB1 )]  , (B1)

 
 coth(k) [(A1 (t) + ikA1 ) − (B1 (t) + ikB1 )] 
Q2 = (B2,t + 2ikB2 ) − kB1 
 
 − tanh(k) [(A1 (t) + ikA1 ) + (B1 (t) + ikB1 )]  . (B2)

 
The second-order constants of Eqs. (50) and (51) are given as
coth (2k) tanh (2k) ρ coth [(−1 + h) 2k]
a12 = + + , (B3)
4k 4k 2k
b12 = i coth (2k) + 2i ρU1 coth [(−1 + h) 2k] + i tanh (2k), (B4)
4k 2
c12 = − k coth (2k) − 2k ρU12 coth [2 (−1 + h) k] − k tanh (2k), (B5)
We
coth (2k) tanh (2k)
d12 = − + , (B6)
4k 4k
044104-22 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

e12 = −i coth (2k) + i tanh (2k), (B7)


f12 = k coth (2k) − k tanh (2k), (B8)
coth (2k) tanh (2k)
a22 = − , (B9)
4k 4k
b22 = i coth (2k) − i tanh (2k), (B10)
c22 = −k coth (2k) + k tanh (2k), (B11)
coth (2k) tanh (2k) ρ coth [(−1 + h) 2k]
d22 = − − − , (B12)
4k 4k 2k
e22 = −i coth (2k) − 2i ρU1 coth [(−1 + h) 2k] − i tanh (2k), (B13)
4k 2
f22 = −+ k coth (2k) + 2k ρU12 coth [2 (−1 + h) k] + k tanh (2k) . (B14)
We
Constants of second-order dispersion Eq. (52) are given as
M21 = a12 × d22 − d12 × a22 , (B15)
M22 = a12 × e22 + b12 × d22 − a22 × e12 − b22 × d12 , (B16)
M23 = a12 × f22 + c12 × d22 + b12 × e22 − a22 × f12 − d12 × c22 − b22 × e12 , (B17)
M24 = b12 × f22 + c12 × e22 − b22 × f12 − e12 × c22 , (B18)
M25 = c12 × f22 − f12 × c22 , (B19)
The constants of Eq. (53) are given as
f g f g
Pa | s=ω12 d22 ω12 2 + ie ω + f
22 12 22 − P s=ω
b d12 ω122 + ie ω + f
12 12 12
C1a = 12
, (B20)
(iω12 − iω̄12 ) (iω12 − iω22 ) (iω12 − iω̄22 )
f g f g
Pa | s=ω̄12 d22 ω̄12 2 + ie ω̄ + f
22 12
b
22 − P s=ω̄ d12 ω̄12
2 + ie ω̄ + f
12 12 12
C2a = 12
, (B21)
(iω̄12 − iω12 ) (iω̄12 − iω22 ) (iω̄12 − iω̄22 )
f g f g
Pa | s=ω22 d22 ω22 2 + ie ω + f
22 22 22 − P s=ω
b d12 ω222 + ie ω + f
12 22 12
C3a = 22
, (B22)
(iω22 − iω12 ) (iω22 − iω̄12 ) (iω22 − iω̄22 )
f g f g
Pa | s=ω̄22 d22 ω̄22 2 + ie ω̄ + f
22 22
b
22 − P s=ω̄ d12 ω̄22
2 + ie ω̄ + f
12 22 12
C4a = 22
, (B23)
(iω̄22 − iω12 ) (iω̄22 − iω̄12 ) (iω̄22 − iω22 )
f g f g
Pa | s=ω11 d22 ω11 2 + ie ω + f
22 11 22 − P s=ω
b d12 ω112 + ie ω + f
12 11 12
C5a = 11
, (B24)
(iω11 − iω12 ) (iω11 − iω̄12 ) (iω11 − iω22 ) (iω11 − iω̄22 )
f g f g
Pa | s=ω̄11 d22 ω̄11 2 + ie ω̄ + f
22 11
b
22 − P s=ω̄ d12 ω̄11
2 + ie ω̄ + f
12 11 12
C6a = 11
, (B25)
(iω̄11 − iω12 ) (iω̄11 − iω̄12 ) (iω̄11 − iω22 ) (iω̄11 − iω̄22 )
f g


 Pa s=(ω11 +ω̄11 ) d22 (ω11 + ω̄11 )2 + ie22 (ω11 + ω̄11 ) + f22  

 f g 
 − Pb

(ω11 +ω̄11 ) d 12 (ω 11 + ω̄ 11 ) 2
+ ie 12 (ω 11 + ω̄ 11 ) + f 12


C7a =   , (B26)
[i (ω11 + ω̄11 ) − iω12 ] [i (ω11 + ω̄11 ) − iω̄12 ] [i (ω11 + ω̄11 ) − iω22 ] [i (ω11 + ω̄11 ) − iω̄22 ]
f g f g
Pa | s=ω21 d22 ω21 2 + ie ω + f
22 21 22 − P s=ω
b d12 ω212 + ie ω + f
12 21 12
C8a = 21
, (B27)
(iω21 − iω12 ) (iω21 − iω̄12 ) (iω21 − iω22 ) (iω21 − iω̄22 )
f g f g
Pa | s=ω̄12 d22 ω̄21 2 + ie ω̄ + f
22 21
b
22 − P s=ω̄ d12 ω̄21
2 + ie ω̄ + f
12 21 12
C9a = 21
, (B28)
(iω̄21 − iω12 ) (iω̄21 − iω̄22 ) (iω̄21 − iω22 ) (iω̄21 − iω̄22 )
f g


 Pa s=(ω21 +ω̄21 ) d22 (ω21 + ω̄21 )2 + ie22 (ω21 + ω̄21 ) + f22  

 f g
 −P
 b d
(ω21 +ω̄21 ) 12 21 (ω + ω̄ 21 ) 2
+ ie 12 (ω 21 + ω̄ 21 ) + f 12 

C10a =   , (B29)
[i (ω21 + ω̄21 ) − iω12 ] [i (ω21 + ω̄21 ) − iω̄12 ] [i (ω21 + ω̄21 ) − iω22 ] [i (ω21 + ω̄21 ) − iω̄21 ]
f g


 Pa s=(ω11 +ω21 ) d22 (ω11 + ω21 )2 + ie22 (ω11 + ω21 ) + f22  

 f g
(ω11 +ω21 ) d12 (ω11 + ω21 ) + ie12 (ω11 + ω21 ) + f12 
 − Pb
 2 

C11a =  , (B30)
[i (ω11 + ω21 ) − iω12 ] [i (ω11 + ω21 ) − iω̄12 ] [i (ω11 + ω21 ) − iω22 ] [i (ω11 + ω21 ) − iω̄22 ]
044104-23 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

f g


 Pa s=(ω11 +ω̄21 ) d22 (ω11 + ω̄21 )2 + ie22 (ω11 + ω̄21 ) + f22  

 f g
s=(ω11 +ω̄21 ) d12 (ω11 + ω̄21 ) + ie12 (ω11 + ω̄21 ) + f12 
 −P
 b 2 

C12a =  , (B31)
[i (ω11 + ω̄21 ) − iω12 ] [i (ω11 + ω̄21 ) − iω̄21 ] [i (ω11 + ω̄21 ) − iω22 ] [i (ω11 + ω̄21 ) − iω̄22 ]
f g


 Pa s=(ω21 +ω̄11 ) d22 (ω21 + ω̄11 )2 + ie22 (ω21 + ω̄11 ) + f22  

 f g
s=(ω21 +ω̄11 ) d12 (ω21 + ω̄11 ) + ie12 (ω21 + ω̄11 ) + f12 
 −P
 b 2 

C13a =  , (B32)
[i (ω21 + ω̄11 ) − iω12 ] [i (ω21 + ω̄11 ) − iω̄12 ] [i (ω21 + ω̄11 ) − iω22 ] [i (ω21 + ω̄11 ) − iω̄22 ]
f g


 Pa s=(ω̄21 +ω̄11 ) d22 (ω̄21 + ω̄11 )2 + ie22 (ω̄21 + ω̄11 ) + f22  

 f g
s=(ω̄21 +ω̄11 ) d12 (ω̄21 + ω̄11 ) + ie12 (ω̄21 + ω̄11 ) + f12 
 −P
 b 2 

C14a =  . (B33)
[i (ω̄21 + ω̄11 ) − iω12 ] [i (ω̄21 + ω̄11 ) − iω̄12 ] [i (ω̄21 + ω̄11 ) − iω22 ] [i (ω̄21 + ω̄11 ) − iω̄22 ]

APPENDIX C: SOURCE TERMS CALCULATION OF MOMENTUM AND ENERGY BALANCE CONSTRAINT


EQUATIONS OF MEF
The source term for constraint equations in maximum entropy formulation is obtained as follows:
Momentum balance
Considering gas velocities at both the interfaces are higher than the liquid velocity, i.e., (U 1 > 1 and U 2 > 1), momentum
transfer takes place from gas to liquid. Thus,

" # " # X m X k π
1  2 1  2 
ml ul + ρg ug1 − ul l × b × Cf 1 + ρg ug2 − ul l × b × Cf 2 = ni,j di3 ρl vj , (C1)
2 2 i=1 j=1
6

where ρg = density of gas; ug1 and ug2 = velocity of gas at upper and lower interfaces, respectively; ρl = density of liquid;
ul = velocity of liquid; l = dimensional breakup length of the liquid sheet; b = width of the liquid sheet; C f 1 and C f 2 = drag
coefficient on the upper and lower interface, respectively; ni,j = number of droplets formed per unit time having diameter d i and
velocity vj .
Equation (C1) can be further simplified as
m P
k
π 3
 
6 di ρl
P  2   2 
ni,j vj
 
2 ρg ρ
1 1
i=1 j=1
ug1 − ul l × b × Cf 1 2 g g2u − ul l × b × Cf2
=1+ + , (C2)
π 3 ρl (2h × b) ul ul ρl (2h × b) ul ul
     
N 6 dm ρl ul

as ml ul = N π6 dm3 ρl ul = ρl (2h × b) ul ul .
   
So, Eq. (C2) can finally be written as
m X k
ni,j di 3 vj 1 ρg ug1 1 ρg ug2
! ! ! !2 ! ! !2 !
X l l
=1+ −1 Cf 1 + −1 Cf 2
i=1 j=1
N dm u l 4 ρ l u l h 4 ρ l u l h
m X
k
X   1 1
or Pi,j (Di )3 Vj = 1 + ρ(U1 − 1)2 LCf 1 + ρ(U2 − 1)2 LCf 2 , (C3)
i=1 j=1
4 4

where L = l/h (non-dimensional length); ρ = ρg /ρl (gas to liquid density ratio); U 1 = ug1 /ul and U 2 = ug2 /ul (non-dimensional
gas velocity); Di = d i /d m (non-dimensional droplet diameter); V j = vj /ul (non-dimensional droplet velocity); Pi,j = ni,j /N; C f 1
and C f 2 = drag force coefficient on liquid at upper and lower interface respectively.
The RHS of Eq. (C3) represents the momentum source term S 3 of Eq. (62) in the manuscript.
Energy balance
Considering gas velocities at both the interfaces are higher than the liquid velocity, i.e., (U 1 > 1 and U 2 > 1), transfer of
energy takes place from gas to liquid, so the energy balance equation can be written as
" #  "1  #
1 1  2 2 
2
ml u l + ρg ug1 − ul l × b × Cf 1 ug1 − ul + ρg ug2 − ul l × b × Cf 2 ug2 − ul
2 2 2
m X k m X k
X 1 π 3
  X
= ni,j di ρl v2j + ni,j πσdi2 . (C4)
i=1 j=1
2 6 i=1 j=1
044104-24 Dasgupta, Nath, and Bhanja Phys. Fluids 30, 044104 (2018)

Following a procedure similar to momentum source term calculation, after complex algebraic calculations, the final energy
balance equation is obtained as
m X
k m X
k
X X 1 1
B Pi,j (Di )2 + Pi,j (Di )3 Vj2 = 1 + ρ(U1 − 1)3 LCf 1 + ρ(U2 − 1)3 LCf 2 , (C5)
i=1 j=1 i=1 j=1
2 2

where B = 12σ/ρl ul2 dm ; σ = surface tension coefficient for liquid.


The RHS of Eq. (C5) represents the energy source term S 4 of Eq. (63) in the manuscript.

1 Asare, H. R., Takahashi, R. K., and Hoffman, M. A., “Liquid sheet jet 19 Marmottant, P. and Villermaux, E., “On spray formation,” J. Fluid Mech.
experiments: Comparison with linear theory,” J. Fluids Eng. 103, 595–603 498, 73–111 (2004).
(1981). 20 Matas, J. P., Marty, S., and Cartellier, A., “Experimental and analytical study
2 Avulapati, M. M. and Venkata, R. R., “Experimental studies on air- of the shear instability of a gas-liquid mixing layer,” Phys. Fluid 23, 094112
assisted impinging jet atomization,” Int. J. Multiphase Flow 57, 88–101 (2011).
(2013). 21 Mezhericher, M., Ladizhensky, I., and Etlin, I., “Atomization of liquids by
3 Chatterjee, S., Samanta, S., Mukhopadhyay, A., Ghosh, K., and Sen, S., disintegrating thin liquid films using gas jets,” Int. J. Multiphase Flow 88,
“Effect of a confined outer air stream on instability of an annular liquid 99–115 (2017).
sheet exposed to gas flow,” in ASME 2012 Gas Turbine India Conference 22 Mitra, S. K., Li, X., and Renksizbulut, M., “On the breakup of viscous liquid

(ASME, 2012), pp. 477–486. sheets by dual-mode linear analysis,” J. Propul. Power 17, 728–735 (2001).
4 Cousin, J. and Dumouchel, C., “Coupling of classical linear theory and max- 23 Mohanta, L., Cheung, F. B., and Bajorek, S. M., “Stability of coaxial jets

imum entropy formalism for prediction of drop size distribution in sprays: confined in a tube with heat and mass transfer,” Phys. A 443, 333–346
Application to pressure-swirl atomisers,” Atomization Sprays 6, 601–622 (2016).
(1996). 24 Nath, S., Datta, A., Mukhopadhyay, A., Sen, S., and Tharakan, T. J., “Pre-
5 Déjean, B., Berthoumieu, P., and Gajan, P., “Experimental study on the diction of size and velocity distributions in sprays formed by the breakup
influence of liquid and air boundary conditions on a planar air-blasted liq- of planar liquid sheets using maximum entropy formulation,” Atomization
uid sheet. Part I: Liquid and air thicknesses,” Int. J. Multiphase Flow 79, Sprays 21(6), 483–501 (2011).
202–213 (2016). 25 Nath, S., Mukhopadhyay, A., Datta, A., Sarkar, S., and Sen, S., “Effect of
6 Duke, D. J., Honnery, D., and Soria, J., “The growth of instabilities in annular confinement on breakup of planar liquid sheets sandwiched between two gas
liquid sheets,” Exp. Therm. Fluid Sci. 68, 89–99 (2015). streams and resulting spray characteristics,” Fluid Dyn. Res. 46(1), 015511
7 Fu, Q., Jia, B., and Yang, L., “Stability of a confined swirling annular liquid (2014).
layer with heat and mass transfer,” Int. J. Heat Mass Transfer 104, 644–649 26 Nath, S., Mukhopadhyay, A., Datta, A., and Sen, S., “Analysis of dis-

(2017). integration of planar liquid sheet sandwiched between gas streams with
8 Herrero, E. P., Martı́n, E. M., Valle, D., and Galán, M. A., “Instability unequal velocities and resulting spray formation,” in Triennial International
study of a swirling annular liquid sheet of polymer produced by air-blast Conference on Liquid Atomization and Spray Systems, 2012.
atomization,” Chem. Eng. J. 133, 69–77 (2007). 27 Nath, S., Mukhopadhyay, A., Datta, A., Sen, S., and Tharakan, T. J., “Influ-
9 Ibrahim, A. A. and Jog, M. A., “Nonlinear instability of an annular liq- ence of gas velocity on breakup of planar liquid sheets sandwiched between
uid sheet exposed to gas flow,” Int. J. Multiphase Flow 34, 647–664 two gas streams,” Atomization Sprays 20, 983–1003 (2010).
(2008). 28 Rayleigh, Lord, “On the instability of jets,” Proc. Lond. Math. Soc. 10, 4–13
10 Jazayeri, S. A. and Li, X., “Nonlinear instability of plane liquid sheets,” (1878).
J. Fluid Mech. 406, 281–308 (2000). 29 Rees, S. J. and Juniper, M. P., “The effect of surface tension on the stability
11 Jiang, D. J., Liu, H. F., Li, W. F., Xu, J. L., Wang, F. C., and Gong, X., of unconfined and confined planar jets and wakes,” J. Fluid Mech. 633,
“Modeling atomization of a round water jet by a high-speed annular air jet 71–97 (2009).
based on the self-similarity of droplet breakup,” Chem. Eng. Res. Des. 90, 30 Rostami, E., Ommi, F., Mirmohammadi, A., and Khodayari, H., “Effect of

185–192 (2012). gas-to-liquid density ratios at different liquid viscosity on the atomization
12 Juniper, M. P., “The effect of confinement on the stability of two- of the air-blast atomizer,” Aust. J. Mech. Eng. 15(2), 125–136 (2017).
dimensional shear flows,” J. Fluid Mech. 565, 171–195 (2006). 31 Shannon, C. E., “A mathematical theory of communication,” Bell Syst.
13 Kan, K. and Yoshinaga, T., “Instability of a planar liquid sheet with sur- Tech. J. 27, 379–423 (1948).
rounding fluids between two parallel walls,” Fluid Dyn. Res. 39, 389–412 32 Subramaniam, K. and Parthasarathy, R. N., “Effects of confinement on the

(2007). temporal instability of gas jets injected in viscous liquids,” Phys. Fluid 12,
14 Kim, W. T., Mitra, S. K., Li, X., Prociw, L. A., and Hu, T. C. J., “A pre- 89 (2000).
dictive model for the initial droplet size and velocity distributions in sprays 33 Warncke, K., Gepperth, S., Sauer, B., Sadiki, A., Janicka, J., Koch, R.,

and comparision with experiments,” Part. Part. Syst. Charact. 20, 135–149 and Bauer, H. J., “Experimental and numerical investigation of the primary
(2003). breakup of an airblasted liquid sheet,” Int. J. Multiphase Flow 91, 208–224
15 Leboucher, N., Roger, F., and Carreau, J. L., “Atomization characteristics of (2017).
an annular liquid sheet with inner and outer gas flows,” Atomization Sprays 34 Yan, K., Ning, Z., Lü, M., Sun, C., Fu, J., and Li, Y., “Interface instability

24, 1065–1088 (2014). mechanism of an annular viscous liquid sheet exposed to axially moving
16 Lefebvre, A. H., “Gas turbine combustion,” in Atomization and Sprays inner and outer gas,” Eur. J. Mech. - B/Fluids 52, 185–190 (2015).
(Taylor & Francis, 1989). 35 Yang, L., Chen, P., and Wang, C., “Effect of gas velocity on the weakly
17 Li, X., “On the instability of plane liquid sheets in two gas streams of unequal nonlinear instability of a planar viscous sheet,” Phys. Fluids 26, 074106
velocities,” Acta Mech. 106, 137–156 (1994). (2014).
18 Liu, H., Science and Engineering of Droplets: Fundamentals and Applica- 36 Yang, L., Wang, C., Fu, Q., Du, M., and Tong, M., “Weakly nonlinear

tions, 1st ed. (Noyes Publications, 2000). instability of planar viscous sheets,” J. Fluid Mech. 735, 249–287 (2013).

Vous aimerez peut-être aussi