You are on page 1of 16

Breakup of a liquid rivulet falling over an inclined plate: Identification of a critical

Weber number
Rajesh K. Singh, Janine E. Galvin, Greg A. Whyatt, and Xin Sun

Citation: Physics of Fluids 29, 052101 (2017); doi: 10.1063/1.4981920


View online: http://dx.doi.org/10.1063/1.4981920
View Table of Contents: http://aip.scitation.org/toc/phf/29/5
Published by the American Institute of Physics

Articles you may be interested in


Effects of heat sink and source and entropy generation on MHD mixed convection of a Cu-water nanofluid in
a lid-driven square porous enclosure with partial slip
Physics of Fluids 29, 052001052001 (2017); 10.1063/1.4981911

Many-body dissipative particle dynamics modeling of fluid flow in fine-grained nanoporous shales
Physics of Fluids 29, 056601056601 (2017); 10.1063/1.4981136

Influence of interfacial viscosity on the dielectrophoresis of drops


Physics of Fluids 29, 052002052002 (2017); 10.1063/1.4982662

The fate of pancake vortices


Physics of Fluids 29, 031701031701 (2017); 10.1063/1.4977975

Interactions of a co-rotating vortex pair at multiple offsets


Physics of Fluids 29, 057102057102 (2017); 10.1063/1.4982217

Characterizing shock waves in hydrogel using high speed imaging and a fiber-optic probe hydrophone
Physics of Fluids 29, 057101057101 (2017); 10.1063/1.4982062
PHYSICS OF FLUIDS 29, 052101 (2017)

Breakup of a liquid rivulet falling over an inclined plate:


Identification of a critical Weber number
Rajesh K. Singh,1,a) Janine E. Galvin,1 Greg A. Whyatt,2 and Xin Sun2
1 National Energy Technology Laboratory, Albany, Oregon 97321, USA
2 Pacific Northwest National Laboratory, Richland, Washington 99352, USA
(Received 10 November 2016; accepted 6 April 2017; published online 4 May 2017)

We have numerically investigated the breakup of a rivulet falling over a smooth inclined plate using
the volume of fluid method. Rivulet breakup is a complex phenomenon dictated by many factors, such
as physical properties (viscosity and surface tension), contact angle, inertia, and plate inclination. An
extensive simulation was conducted wherein these factors were systematically investigated. Regimes
for a stable rivulet and an unstable rivulet that leads to breakup are examined in terms of a critical
value of the Weber number (Wecr ) that delineates these regimes. A higher Wecr implies that a higher
flow rate is required to maintain a stable rivulet. The impact of liquid properties is characterized by
the Kapitza number (Ka). Variation of Wecr with Ka shows two trends depending on the Ka value
of the liquid. Liquids with lower Ka values, corresponding to high viscosities and/or low surface
tensions, show linear variation and smaller value of the critical Weber number. In other words, the
lower the liquid Ka value, the more stable the rivulet will tend to be with changes in liquid inertia. A
liquid having higher Ka value exhibits larger value of Wecr and quadratic variation of Wecr with Ka.
This behavior is more pronounced with increasing contact angle (γ). Higher contact angles promote
rivulet breakup so that inertia must be higher to suppress breakup, consequently Wecr increases with
increasing γ. The effect of plate inclination on breakup shows that Wecr decreases with increased
inclination angle (θ) owing to higher effective liquid inertia. However, the effect is negligible beyond
θ > 60◦ . The effect of the inlet size reveals that Wecr decreases with inlet cross-sectional area, but the
corresponding solvent flow rate for rivulet breakup remains unchanged. A phenomenological scaling
for the critical Weber number with the Kapitza number and contact angle is presented, which may offer
insight into rivulet breakup. Published by AIP Publishing. [http://dx.doi.org/10.1063/1.4981920]

I. INTRODUCTION understanding the hydrodynamics at the scale of the liquid


film. As shown previously,5,6 a film falling down an inclined
Fossil-fueled based power plants account for 40% of all plate exhibits different flow patterns such as a full film, rivulet,
CO2 emissions,1 and a need exists to mitigate these emissions and droplet. These flow features are dependent upon various
to control global warming. Post-combustion carbon capture parameters, such as solvent physical properties, solvent flow
by chemical absorption using countercurrent gas-liquid flow rates, plate surface texture and plate inclination. In this paper,
in structured packing is a popular and efficient technology for the volume of fluid method (VOF) is used to examine the tran-
reduction of CO2 emission from power plants.1 The structured sition from rivulet to droplet flow by breakup of the rivulet.
packed column provides a large surface area for mass trans- A better understanding of this flow regime transition and how
fer between the phases with minimum pressure drop across the various parameters play a role may be useful in the opera-
the column.2 These structured packed columns are several tion and design of a structured packing column where rivulets
meters in length (diameter: 5–10 m and height h: 20–30 m).3,4 may occur. It is also relevant to other areas where breakup of a
In contrast, the size of the structured packing unit is much rivulet into droplets plays a key role in design and performance,
smaller, typically around 20 cm, and the liquid-film on the such as in microfluidics and inkjet printings.
packing is typically less than a millimeter in thickness. Thus, Rivulets flowing down an inclined plate exhibit a wide
a wide disparity in length scales exists in the structured packed variety of instabilities, such as interfacial waves, braiding,
column. It is not computationally feasible to perform compu- meandering, breakup of a rivulet into droplets, and transition
tational fluid dynamics (CFD) simulations at the length scale of rivulet flow from laminar to turbulent.7 These instabilities
of the column size while accounting for the actual geometry are complex phenomenon dictated by many factors such as
of the internal packing and the local gas-liquid interactions at contact line motion (the point where the interface meets the
a film level. Accordingly, multiscale modelling must be used solid surface), contact angle, surface tension, gravity, viscos-
to resolve these differences in scale.4 Gravity-driven film flow ity, and inertia.8,9 Surface tension generally acts to destabilize
down an inclined plate provides a simple configuration for the rivulet whereas gravity and viscosity tend to stabilize the
rivulet.7 The classical Rayleigh–Plateau instability observed
a) Author to whom correspondence should be addressed. Electronic addresses: in freestanding liquid jets, wherein a falling stream breaks
rajesh-kumar.singh@netl.doe.gov and rajeshsingh.175@gmail.com up into smaller drops, is an example of a capillary instability

1070-6631/2017/29(5)/052101/15/$30.00 29, 052101-1 Published by AIP Publishing.


052101-2 Singh et al. Phys. Fluids 29, 052101 (2017)

which is driven by surface tension. In the case of a rivulet, the Rivulet breakup is dictated by a balance of various forces,
analysis becomes more complex due to the additional solid- such as viscous, capillary, gravity, inertia, and the force due to
liquid interaction.10 Over the past several decades, rivulet solid-liquid interaction.10 Two mechanisms for rivulet breakup
stability has been studied using a variety of techniques: ana- are recognized: capillary wave growth, a counterpart to the
lytically using linearized stability analysis,7,8,11,12 the energy Rayleigh-Plateau instability found in free surface liquid jets,
minimization method,13–15 or the perturbation method;16 and end-pinching.31 In capillary wave breakup, surface tension
experimentally17–20 and numerically.21,22 gives rise to a capillary pressure gradient that leads to thinner
The stability of a static rivulet on a horizontal surface in and thicker regions similar to the Rayleigh-Plateau instabil-
the absence of gravity at different contact line conditions (e.g., ity. This perturbation grows and eventually the rivulet may
fixed or moving contact lines with either fixed contact angles break up into small droplets. The size and spacing between
or smoothly varying contact angles) was investigated in the the droplets depend on the contact angle; higher contact
pioneering study of Davis.12 A condition for stability in terms angles (corresponding to hydrophobicity) lead to more closely
of a critical wavenumber was presented as a function of the spaced larger droplets.21,32 As noted earlier, such breakup
equilibrium contact angle. This study was further extended by phenomenon tends to occur in narrow rivulets having mobile
Young and Davis7 to explore the effects of contact line motion contact lines.7 It is worth noting that increasing the fluid inertia
and slip at the surface on the stability of a rivulet. The rivulet tends to mitigate this type of instability.7,9 End-pinching, the
size and contact-line motion strongly influence the stability of second mechanism, has been observed in breakup of a rivulet
a rivulet. A wider rivulet or a rivulet having immobile con- of finite length along a horizontal surface in the presence of
tact lines is prone to kinematic wave instabilities, whereas a gravity.32,33 The tip/end of the rivulet retracts because of sur-
narrower rivulet with moving contact lines is disposed to cap- face tension and the subsequent gathering of fluid into the tip
illary breakup. Both of these studies7,12 were conducted in the results in rapid bulb formation at the end. The bulbous end then
absence of gravity. The stability of a transverse rivulet (rivulet pinches off from a rivulet and a droplet appears. This instability
width is greater than its length in the flow direction) has been differs from the aforementioned capillary breakup instabil-
studied in the presence of gravity.23,24 In this case, gravity tends ity in that it is not a simultaneous occurrence along different
to stabilize the rivulet by suppressing the instability. Inertia is regions of the rivulet. Instead, perturbations at the ends of the
another important factor affecting the rivulet behavior and the domain are observed and propagate toward the uniform bulk
manner of instability. resulting in a smaller daughter droplet that appears because of
Inertia may give rise to kinematic instabilities (i.e., the pinch-off process. Recently, Wilson et al.34 investigated the
formation of surface waves at the interface),25 braiding,26 and breakup of a thin rivulet into daughter-rivulets (multiple small
meandering instabilities in the rivulet.27,28 When contact lines rivulets) using the lubrication approximation and a prescribed
are less mobile, surface tension (or capillary effects) inhibits a minimum value of rivulet semi-width above which breakup
the formation of kinematic instabilities whereas inertia effects occurs.
enhance it.7 Mertens et al.26 observed such an instability and The capillary wave breakup mechanism for a rivulet on a
referred to it as braiding (a stationary flow pattern that varies surface has been studied numerically using the finite element
in width and height). They also attributed it to a competition method21 and the phase-field method,31 analytically,10,13 and
between surface tension which tends to narrow the rivulet and experimentally.17 A good agreement with classical Rayleigh
fluid inertia which acts to widen the rivulet. When the contact prediction was found in the experimental study for the rupture
lines are less mobile, as in these cases, the capillary insta- of a polymer rivulet. Despite the knowledge gained by these
bility (discussed below) is suppressed. A meandering rivulet, studies, many have either simplifying assumptions or under-
wherein a rivulet destabilizes or deviates from a straight line, lying limitations in their analysis. For example, the numeri-
is generally observed upon increasing flow rate. The rivulet cal studies based on an Arbitrary Lagrangian-Eulerian (ALE)
shape is irregular in general and given sufficient time may technique by Ubal et al.21 were limited to investigating onset
become stationary or destabilize and break down depending on of breakup phenomenon. Since this method cannot follow
flow conditions.29 The mechanism by which the straight rivulet changes in domain topology, their simulations could not con-
destabilizes into a meandering rivulet is not fully understood. tinue beyond breakup. Thus the study of post break phenomena
Kim et al.28 report that the meandering instability is a competi- was not possible. The phase field simulation for rivulet breakup
tion between inertia and surface tension but is also related to the neglected inertia,31 while the analytic stability analyses that
tangential velocity difference across the interface and effects rely on a lubrication approximation fail to correctly predict
of the dynamic wetting. While Birner et al.27 have reported the rivulet behavior at large contact angles.23 Furthermore,
that the meandering instability arises from disturbances in the stability analysis becomes even more complex when trying
flow rate and that a rivulet will usually tend to meander unless to accommodate the effects of inertia as the simplifications
the flow rate is maintained at a highly constant value. More made possible by lubrication theory are no longer appropri-
recently Couvreur and Daerr 30 emphasize the role played by ate.25 Multiphase flow simulations using the volume of fluid
the geometry of the contact line (including the roughness of the (VOF) method provide a tool to fully explore these effects and
contact line and wetting hysteresis) on an initial critical flow are therefore useful for investigating rivulet breakup dynamics.
rate above which a straight rivulet destabilizes and begins to In this paper, we consider breakup of a rivulet on an
meander. While interesting, such instabilities are not the pri- inclined plate over a wide range of parameters including phys-
mary focus of the current effort which is to study breakup of ical properties (viscosity and surface tension), contact angle,
the rivulet. and plate inclination angle. We employ VOF formulation as
052101-3 Singh et al. Phys. Fluids 29, 052101 (2017)

an interface capturing technique for multiphase flow study. the specification of a static contact angle (γ). A dynamic
In Section II, a mathematical model and numerical formu- contact angle model, discussed next, was also invoked for
lation are presented. Section III discusses the problem setup comparative purposes in selective cases. In either case (static
including discretization and the different solvent properties or dynamic), the boundary condition is not imposed at the wall
examined. Simulation results on stability of the rivulet are itself, but instead the contact angle that the fluid is assumed to
presented in Section IV. A new correlation is developed for make with the wall is used to adjust the surface normal in cells
a critical Weber number demarcating the transition from a near the wall. Specifically, the normal vector at the interface
stable to an unstable rivulet that leads to breakup and sub- is adjusted near the wall by the following equation:
sequent droplet formation. The results of the simulations are n̂ = n̂wall cos γ + n̂t sin γ, (5)
then summarized.
where n̂wall and n̂t are unit vectors normal and tangential to the
II. MATHEMATICAL FORMULATION AND NUMERICAL wall, respectively. The vector n̂t lies in the wall and normal
METHOD to the contact line (where the interface meets the wall) and is
computed using the following expression:
The volume of fluid (VOF) multiphase method35 is used to
n̂t = ∇f |∇f |.

(6)
study the breakup of a rivulet flowing down a smooth inclined
plate Recall that unlike an interface tacking method, such as In practice, the contact angle may vary as the rivulet
the Arbitrary Lagrangian-Eulerian technique (ALE), the VOF evolves (e.g., as it spreads or wets the surface) so that a
method can overcome changes in the interface topology which dynamic contact angle, measured at leading and trailing edge
will occur during breakup of a rivulet.21 In the VOF method, of rivulet, may differ significantly from the static value.38 Still
the entire flow field is treated as a single phase. Therefore, the dynamic contact angle is not well understood and a com-
the conservation equations are solved for a single shared field. plete theory on contact angle hysteresis is still under develop-
While a brief sketch of the VOF is provided here, a more ment.39,40 Nevertheless, a dynamic contact angle model would
detailed explanation can be found in the Fluent theory guide.36 provide a more accurate description of the wetting and spread-
The governing equations involved in the simulation are shown ing process. In this effort, the correlation by Yokoi et al.40 for
∇ · u = 0, (1) dynamic contact angle was selected and implemented using a
User Defined Function (UDF). In this case, the dynamic con-
∂(ρu)   tact angle becomes a function of the velocity of the contact
+ ∇ · (ρuu) = −∇p + µ∇ · ∇u + (∇u)T + ρg + F. line (UCL ). Their work is based on Tanner’s law,41 and it is
∂t
(2) reproduced here as follows:

{
Here, ρ is the fluid density, p is the pressure, g is gravitational
! 1/3
Ca
 
acceleration, and µ is the fluid viscosity. The surface tension min γe + , γmDA  if UCL ≥ 0
 kA 
force, F, produces a jump in the normal traction across the γ (UCL ) = , (7)
interface. 
Ca
! 1/3 
The conservation equations (1) and (2) are solved by max γe + , γmDR  if UCL < 0
 kR 
a finite volume method. The surface tension force (F) is
expressed as a volume force through the continuous surface where Ca is the capillary number (Ca = µUCL /σ); kA and kR
force (CSF) model.37 This force is distributed across the thin are the empirical constants for advancing and receding cases,
interfacial layer respectively, that depend on the surface material; and γ is
ρκ∇f the equilibrium contact angle. The quantity γmDA refers to
F=σ  . (3) the maximum dynamic advancing angle, while γmDR refers
2 ρg + ρl
1
to the minimum receding angle. In measurements, these cor-
Here σ is the interfacial tension (which is constant in the CSF respond to limits in the contact angle as UCL increases and
model), κ is the local curvature of the interface, and ∇f is the decreases, respectively. Thus, the value of the dynamic contact
gradient in the liquid phase volume fraction (f ). angle lies within the interval γmDR ≤ γD ≤ γmDA . As evident,
The interface between the phases is captured by solving an this model requires several parameters for the characterization
additional transport equation (4) for the liquid phase volume of the dynamic contact angle: kA , kR , γmDA , γmDR , and γe . The
fraction (scalar f ). Accordingly, the value of f varies from 0 selection of these is discussed later.
to 1 (0 corresponding to cells with all gas and 1 corresponding The flow simulations were conducted in Fluent 14.036
to all liquid) based on the finite volume method using an implicit transient
∂f formulation. The conservation equations (Eqs. (1) and (2))
+ u . ∇f = 0. (4)
∂t were solved using a pressure-based solver. Specifically, the
In the case of a two-phase simulation, Equation (4) is only Pressure Implicit with Split of Operators (PISO) algorithm42
solved for the secondary phase, and volume fraction of the was employed for coupling between velocity and pressure.
n The second-order upwind scheme was used in the spatial dis-
primary phase is computed by satisfying constraint fi = 1.
P
i=1 cretization of all equations. For the volume fraction transport
The behavior of the interface between two fluids in con- equation, an explicit formulation was employed. An algebraic
tact with the wall is influenced by wall adhesion.37 In the VOF multigrid (AMG) method was used to accelerate the conver-
model used here, this behavior is, by default, dictated through gence of the solver. Interface capturing was achieved with the
052101-4 Singh et al. Phys. Fluids 29, 052101 (2017)

geometric reconstruction method using the piecewise linear TABLE I. Physical properties of the solvents at 25 ◦ C and 1 atm.
interface calculation (PLIC).43 The convergence of the solu- (MEA: monoethanolamine. MPZ: 1-methylepiperazine. MDEA: N-
methyldiethanolamine. %: Percentage by weight. x: mole fraction).
tion was assumed when the sum of the normalized residual
for each conservation equation was less than or equal to 10 5 . Solvent µ (mPa·s) ρ (Kg/m3 ) σ (mN/m) Ka
The stability of the transient simulation was controlled by the
Courant–Friedrichs–Lewy condition (CFL) condition with a Water 0.89 997 72.8 3969
value of the Courant number of 0.50. Therefore, a very small 20% MEA 1.18 99648 57.849 2173
30% MEA 2.52 98848 5549 750
time step (∆t), varying from 10−5 to 10−4 s, was required
40% MEA 3.71 97948 54.849 450
to satisfy this condition. Consequently, the simulations were
0.07x MPZ50 5.56 1005.3 54.42 258
computationally expensive. 48.8% MDEA 9.25 1016.6 47.56 117
0.10x MPZ50 10.75 1000.9 47.25 93
III. PROBLEM SETUP AND SOLVENT PROPERTIES 0.51x MPZ50 13.36 946.41 34.37 50
0.41x MPZ50 23.48 962.20 35.89 25
We investigate rivulet breakup and identification of flow 0.31x MPZ50 36.42 981.31 38.40 15
regimes for rivulet flow down a smooth inclined plate. A
schematic of the simulation setup is presented in Figure 1(a)
investigated. However, it should be noted that surface tension
showing the inlet and outlet boundaries, the side walls, and the
plays a relatively minor factor in changing the Ka number com-
top and bottom (plate) boundaries. The flow domain consists
pared to viscosity as the values of surface tension vary within
of a smooth plate (60 × 50 mm2 ) inclined 60◦ to horizontal
a relatively narrow range.
(i.e., inclination angle, θ = 60◦ ) with a height of 7 mm (see
The liquid was introduced through a small inlet of dimen-
Figure 1(b)). Air (density (ρ) = 1.185 kg/m3 and viscosity
sion 4 × 2 mm2 located in the middle of the inlet face (see
(µ) = 1.831 × 10 5 Pa s) and industrially used solvents for car-
Figure 1(a)). The inlet was defined by a uniform and constant
bon capture were used as working fluids (see Table I). Air was
velocity perpendicular to the boundary. The remaining part
considered as a stagnant phase.
of the inlet cross section was set as a pressure outlet bound-
A number of alkanolamine-based solutions are being used
ary with zero gauge pressure. The influence of inlet size was
as solvents for carbon capture processes.44,45 Therefore, var-
examined for a single case, which is discussed in more detail
ious aqueous alkanolamines at different concentrations were
below. In brief, however, the solvent flow rate corresponding
used to represent the liquid phase in order to cover a wide
to rivulet breakup was found to be insensitive to inlet size.
range of physical properties. Consistent with previous work,6
The solvent then exits through the outlet face in the presence
the fluid properties are studied in terms of the Kapitza num-
of gravity. The outlet and top boundaries were set as pressure
ber (Ka), which has a fixed value for each solvent and is
outlets with zero gauge pressure. Unless otherwise stated, the
independent of the flow rate
plate and side walls were set as no-slip walls with a static
ρ
! 1/3 contact angle (γ).
Ka = σ 4 . (8) The contact angle is a varying and complex quantity51
µ g
and the value corresponding to each solvent was not read-
The Kapitza number has been extensively used in the ily available in the literature. Note that a solid surface may
film flow community to investigate interfacial wave phe- exhibit different values of γ with change in σ (e.g., vary-
nomena.46,47 As evident, a wide range of Ka numbers are ing solvent).52 Similarly, for a given liquid, the contact angle

FIG. 1. (a) Prospective view of the computational flow


domain showing details of domain. The solvent is flowing
down from top of the inclined smooth plate. (b) Front
view of the flow domain showing the inclination angle
and direction of the gravity.
052101-5 Singh et al. Phys. Fluids 29, 052101 (2017)

may also vary with a change in solid surface. As a result, a (2Wh/(W + h)), where W and h are the width and height of the
static contact angle value of 70◦ , corresponding to water on inlet, respectively.
steel, was used in the preliminary simulations. However, the Meshing is a critical step in constructing a numerical
effect of varying contact angle (γ = 30◦ –80◦ ) on the rivulet simulation that will impact the convergence and accuracy of
breakup and the subsequent droplet formation were also inves- the simulation. The flow domain was discretized with a non-
tigated. As already noted, most cases were conducted using a uniform structured grid for efficient computations. The mesh
static contact angle (SCA). A dynamic contact angle (DCA; density inside the rivulet and near the interface is finer than
see Equation (7)) model was implemented but the simulation the region adjacent to the surrounding gas (see Figure 2(a)).
campaign could not be readily rerun due to the computational Accordingly, a very fine grid was used near the plate and along
expense involved. As noted earlier, however, a dynamic con- the center of the flow domain (see exploded view). A grid inde-
tact angle provides a more accurate description of spreading pendence study was also conducted to determine a reasonable
and wetting processes. In view of this, the impact of a DCA mesh while maintaining grid independent predictions. Only
was explored in selective cases for comparative purposes. It is the grid across the width of the inlet was varied, as opposed to
worth noting that the DCA model involves several empirical the whole domain, as breakup of the rivulet occurs proximate
parameters that are not readily available and so representative to the longitudinal centerline of the flow domain where the
values were selected. Ultimately, the SCA and DCA simu- inlet was located. The rivulet shape at the onset of breakup was
lations yielded similar results regarding the identification of evaluated for different grid resolutions using water (Ka = 3969)
a critical Weber number (Wecr ) that represents the transition with θ = 60◦ and γ = 70◦ and a Weber number (We) of 0.26. At
from rivulet to droplet flow. The critical Weber number based this Weber number, a water rivulet undergoes the breakup pro-
on the DCA was slightly lower. This aspect is discussed in cess. The grid resolution is varied from 20 grid points (GPs)
more detail in Sec. V. across the width of the inlet region to 40 GP. The shape of the
The Weber number is a representation of the fluid inertia rivulet at the onset of breakup does not substantially change
and it has been extensively used in the analyses of this type of with resolutions greater than 26 GP (see Figure 2(b)).
flow.53 Here it is computed according to the inlet conditions The value of the critical Weber number (Wecr ) that cor-
as follows: responds to the transition from rivulet to droplet flow was
. also examined for this setup. The predicted value of Wecr
We = ρU 2 DH σ, (9)
remains approximately the same following grid resolutions
where U is the inlet velocity and DH is the inlet hydraulic of 26 GP or higher (see Table II). Hence, a resolution con-
diameter. Recall that the hydraulic diameter is defined as 4A/P sisting of 26 GP across the inlet region was chosen as the
where A is the cross-sectional area and P is the wetted perime- optimum mesh. This number of GP corresponds to 1.26 M
ter of the cross section. For a rectangular inlet, this becomes cells over the entire simulation domain. It is worth noting that

FIG. 2. (a) Discretization of the com-


putational flow domain showing the
mesh. A very fine mesh was used both
near the plate to capture the film flow
field and in the center of the domain
to capture the rivulet flow dynamics.
(b) Shape of the rivulet at the onset
of breakup for varying grid resolution
using water (Ka = 3969) at We = 0.26,
γ = 70◦ , θ = 60◦ . After 26 GP, the shape
of the rivulet only marginally changes
with further resolution.
052101-6 Singh et al. Phys. Fluids 29, 052101 (2017)

TABLE II. Variation of the critical Weber number (Wecr ) with grid resolu-
tions for a water (Ka = 3969) rivulet at γ = 70◦ .

Grid points at inlet width Number of cells (M) Wecr

20 1.15 0.58
26 1.26 0.50
32 1.31 0.49
40 1.52 0.50

in the investigation of varying inlet size, the number of grid


points across the inlet is also varied to maintain a consistent
grid size according to these grid independent results. Thus, the
total number of cells in those cases will increase as the inlet
size is increased. Overall, mesh sizes ranging from 1.25 to
1.85 M cells were used, which falls within the range reported
by similar numerical studies.20,23

IV. EXPERIMENTAL SETUP


In this work, experiments were carried out for rivulet flow
down a smooth 1 in. thick polycarbonate plate inclined at a 60◦
angle to the horizontal (see Figure 3(a)). The results from the
experiments were used for validation purposes. Three liquids
were used in the experiments including water and two standard
silicon oils with ∼10 cP and 100 cP viscosity. The liquid was
introduced to the top surface of the plate by flowing over a
20 mm wide weir formed by cutting a square pocket into the
top of the polycarbonate plate (see Figure 3(b)). When testing
the water flow, the fluid entered at the back of the pocket via a
1/4 in. metal tube and then passed through two 30 mesh screens
to eliminate turbulence and provide a smooth fluid surface at
the weir. Testing of the ∼10 cP standard silicon oil required
a single 30 mesh screen, while the ∼100 cP fluid required no
screen as the fluid appeared smooth and calm at the top of FIG. 3. (a) Schematic diagram of the test rig of an inclined polycarbonate
the weir without it. The liquid flow rate was set by setting the plate inclined 60◦ to the horizontal, (b) experimental arrangement showing
speed on a gear pump. The fluid was diverted to a cup for a set the rivulet flow over the inclined plate. A 1-cm square grid was placed on the
back side of the plate for dimensional reference.
period and the weight gain determined to determine the flow.
The test was then run without adjusting the pump speed. The
temperature of the circulating fluid was measured using a type flask. Note, the density and surface tension were measured
K thermocouple and used to estimate the actual viscosity at after the addition of the Sudan III dye. The measured surface
test temperature. tension values for the viscosity standards were conducted after
To aid the visibility of the viscosity standards while flow- adding the dye. Data without the presence of the dye were not
ing down the plate, a dye (Sudan III) was added which provides taken to check the influence of the dye on surface tension. The
a red coloration to the flow. The camera was also angled to measured values re-close to what is expected based on litera-
catch the reflection of an overhead fluorescent light to better ture (i.e., 0.0183 vs. 0.0209 N/m (measured vs. literature) for
determine the position of the flow boundaries. Snapshots of 10 cP and 0.0185 vs. 0.0202 N/m (measured vs. literature)
the flow and a corresponding short video were recorded for for 100 cP). So the effect of the dye on the surface tension is
these liquids at different flow rates. A 1-cm square grid placed significant in this case.
on the back side of the plate for dimensional reference was
used to extract the rivulet width. The properties of the fluids
V. RESULTS AND DISCUSSIONS
were also measured. Surface tension of the silicon oils were
measured using a Kruss K-12 tensiometer. The instrument per- In this section, simulation results on the breakup of a
formance was checked before measurements using pure water rivulet on a smooth inclined plate are presented. A critical
and pure ethanol and then checked after measurements using Weber number (Wecr ) is identified which effectively repre-
pure water. The high viscosity of the standards interfered with sents the flow rate for which rivulet breakup occurs and coin-
the ability of the instrument to make measurements using its cides with a transition in the flow regime (from continuous
automated program requiring the measurements to be made rivulet to droplet). As already noted, rivulet dynamics over an
manually. The densities of the silicon oils were determined inclined plate are complex and dictated by many factors, such
by measuring the weight gain in filling a 25 ml volumetric as solvent properties (µ and σ), contact angle (γ), and plate
052101-7 Singh et al. Phys. Fluids 29, 052101 (2017)

inclination angle (θ).6 Accordingly, the simulation results are 0.5) rivulet matches well with the experiment for water (Ka
presented in terms of the following representative dimension- = 3969) at two flow rates 9.89 × 10−6 and 8.22 × 10−6 m3 /s
less parameters: Weber number (We), Kapitza number (Ka), (corresponding to Weber numbers of 0.60 and 0.41, respec-
and dimensionless time (t * ). The Weber number and Kapitza tively). In this case, predictions from both static (SCA) and
number were defined in Equations (8) and (9), respectively. To dynamic contact angle (DCA) simulations are shown. Both
nondimensionalize time, a time scale was defined as DH µ/σ experiment and simulations exhibit a similar stationary “braid-
by using the inlet hydraulic diameter (DH ) as the length scale ing” pattern in the flow26 where the rivulet varies in width and
and the capillary velocity (σ/µ) as the velocity scale.18 height. The predicted edge of the rivulet (rim) and the max-
imum height (hump) are approximately the same position as
A. Comparison with experiments observed in the experiment.
Comparing the rivulet shape obtained from the dynamic
In our previous study of a fully wetted film over a smooth
and static contact angle simulations shows that they are not
inclined plate,6 VOF predictions for wetted area and film thick-
significantly different for this case. Once the rivulets achieved
ness were found to agree well with the experimental results
pseudo-steady state, the contact lines are relatively immo-
of Hoffmann et al.54 and Nusselt theory,55 respectively. To
bile. So, it is unsurprising that the dynamic and static con-
further ensure the reliability of the current simulations, the
tact angle simulations provide similar predictions. In these
predicted results are qualitatively and quantitatively compared
cases, a static contact angle of 80◦ was specified in these
with those obtained from the experiments conducted as part of
cases. Recall that the dynamic contact angle model requires
this effort. CFD simulations were conducted for water as well
a number of empirical parameters wherein assumptions are
as for two highly viscous general purpose silicon oils: 10 cS
made for the evolution and hysteresis of the contact angle.39
and 100 cS. In this setup, the simulations were conducted for
Here values were selected as to achieve good agreement with
a sufficiently long time as to achieve pseudo-steady state for
experimental results resulting in kA = 9 × 10−9 , kR = 9 × 10−8 ,
wetted area of the plate and solvent mass flow at the exit. Once
γmDA = 74.8◦ , γmDR = 53.3◦ , and γe = 70◦ . These values were
the rivulets achieved pseudo-steady state, the rivulet width was
used in all cases employing a DCA model.
measured and compared with that from the experiment.
In cases of increasing flow rate, the rivulet can become
As shown in Figure 4, the predicted shape of the inter-
significantly diverted and a meandering rivulet may emerge
face (defined using an iso-surface at volume fraction (f ) of
(see the references of Introduction). This was observed exper-
imentally for the case of water and increasing flow rate. In
such cases of a moving contact line, the contact angle may
be expected to vary. Specification of a static contact angle
would provide an inaccurate description of the process and so
a dynamic contact model becomes necessary. Therefore, sim-
ulations with the DCA model were employed for comparative
purposes. The results are shown in Figure 5 for two flow rates
(a) 2.20 × 10 3 and (b) 9.30 × 10 4 m3 /s (corresponding to
Weber numbers of 0.03 and 0.005, respectively). Given the
dynamic nature of this phenomenon, the results presented cor-
respond to a single snapshot in time wherein the experimental
results and simulation predictions show similar behavior. As
evident, the flow rates reported in the meandering rivulet cases
(Figure 5) are lower than those of the braiding rivulet exam-
ples (Figure 4). This may be somewhat expected as inertia is
reported to reduce contact line motion (thereby reducing the
dynamic contact angle).56
The simulation predictions for the two general purpose
silicon oils, 10 cS (Ka = 43, ρ = 934 kg/m3 , µ = 9.98 mP, and
σ = 18.3 mN/m) and 100 cS (Ka = 2.10, ρ = 965 kg/m3 ,
µ = 98.2 mP, and σ = 18.5 mN/m), at varying flow rates were
also examined. For these viscous liquids, a stable rivulet (with-
out any undulation) was found at all flow rates investigated.
Variation in rivulet width with the Weber number (We) is pre-
sented in Figure 6. As in the experiment, the width of the
rivulet is measured at a location 60 mm down from the inlet
FIG. 4. Comparison of the CFD predicted rivulet shape using a dynamic as depicted by the red line in the inset of Figures 6(a) and
contact angle (DCA) and static contact angle (SCA) with that captured from 6(b). As shown, the predicted rivulet width matches well with
experiment for water (Ka = 3969) with θ = 60◦ at 2 Wb numbers: (a) We the experimental measurement at different Weber numbers for
= 0.41 and (b) We = 0.60 (corresponding to flow rates of (a) 8.22 × 10 3 and
both solvents. For 10 cS silicon oil, Figure 6(a), a small dis-
(b) 9.89 × 10 3 m3 /s). The hump and rim show the braiding rivulet due to
instability. The dynamic contact angle is defined by γA = 75◦ , γR = 53.3◦ , and crepancy in the results is seen at low flow rates; however, the
γe = 70◦ while the static contact angle (γ) is 80◦ . difference is ≤10%. An agreement between the CFD predicted
052101-8 Singh et al. Phys. Fluids 29, 052101 (2017)

FIG. 5. Comparison of the CFD predicted rivulet shape using dynamic con-
tact angle (DCA) simulation with that captured from experiment for water
(Ka = 3969) with γA = 75◦ ), γR = 53.3◦ and γe = 70◦ and θ = 60◦ at 2 Wb
numbers (a) We = 0.03 and (b) We = 0.005 (corresponding to low rates of (a)
2.20 × 10 3 and (b) 9.30 × 10 4 m3 /s). In both cases, a dynamic meandering
instability is observed.

rivulet shape and that obtained from experiments is also con-


firmed by the insets. Based on these and previous results, VOF
is considered as an appropriate method for the present multi-
phase flow study. Indeed, VOF has been widely used by others FIG. 6. Comparison between the predicted and measured rivulet width, (a)
10 cS (Ka = 43 and γ = 5◦ ) and (b) 100 cS (Ka = 2 and γ = 8◦ ) general purpose
for studying the different flow regimes exhibited by film flow- silicon oils with θ = 60◦ . Width of the rivulet was measured at 60 mm from
ing down an inclined plate: droplet, rivulet, and film (fully inlet. Insets show the predicted and observed shape of the interface from the
wetted plate).6,53,57,58 simulation and experiment, respectively, at a fixed We number with (a) We
= 1.02 and (b) We = 0.75.

B. Rivulet breakup
An earlier study on interfacial and wetted areas of a sta- resulting in rupture/breakup.7 Accordingly, simulations were
ble rivulet found that the interfacial area was greater than the initially conducted at a higher flow rate wherein a stable rivulet
wetted area and that this was due to the rivulet morphology is observed. The solvent flow rate was then gradually reduced
(curvature at the rim along the contact line and to slight rippling to identify the transition from rivulet to droplet. As earlier, the
at the surface).6 In the case of a structured packing column effects of the flow rate are presented in terms of the Weber
for solvent absorption, uniform films are thought to maxi- number.
mize mass transfer area, and therefore, breakup into droplets Figure 7 shows the interface shape for water (Ka = 3969)
would be considered unwanted phenomenon. Understanding at different flow rates. Recall that the interface is defined by an
the stability properties of the flow in such a system and the iso-surface at f = 0.50. As the flow rate is decreased the width
mechanisms involved in breakup and dewetting is of indus- of the rivulet becomes narrower (We = 0.82 and 0.53). A fur-
trial interest. In this view, rigorous simulations are conducted ther decrease in flow rate reveals smaller rivulets and droplets
to identify the rivulet and droplet flow regimes. (We = 0.37 and 0.13). Therefore, breakup occurs between We
As reported in the literature, transition of flow regimes = 0.53 and 0.37. The transition between flow regimes is shown
from a stable film to a rivulet and to a droplet is observed with to depend on a critical Weber number (Wecr ). The value of the
decreasing liquid flow rate.53 At low-flow rates, surface tension critical Weber number (Wecr ) was determined by observing
forces dominate over inertia, which acts to destabilize the film animations of the predicted interface and monitoring the exit
052101-9 Singh et al. Phys. Fluids 29, 052101 (2017)

FIG. 7. Shape of the water rivulet (Ka = 3969) at decreasing inlet Weber num-
bers (We) showing the rivulet breakup process and formation of a droplet with
decreasing flow rate. θ = 60◦ and γ = 70◦ .

mass flow rate for either intermittent or steady rivulet flow


(indicating droplet or rivulet flow, respectively) with iterative
refinement in the value of Weber number.
To show the breakup phenomenon more clearly, the
temporal evolution of the interface for water (Ka = 3969)
at We = 0.13 is depicted in Figure 8(a) where time is non-
dimensionalized as t ∗ = tσ/µDH . Here end-pinching can be
observed to lead to breakup of the rivulet. As evident, a bul-
bous end has developed at t * = 3988 along with the onset of
wrist formation. The tip of the rivulet begins to retract because
of surface tension and liquid gathers at the tip. Eventually, a FIG. 8. (a) Temporal evolution rivulet shape (f = 0.50) for water at a Weber
bulbous shape forms at the tip of rivulet that further grows number (We) of 0.13 wherein end pinching leads to the formation of the droplet
(see Figure 8(b)). The wrist formation proximate to the bul- with θ = 60◦ and γ = 70◦ . Time (t) is non-dimensionalized as t ∗ = tσ / DH µ.
(b) A color map of capillary pressure at the onset of breaking (t* = 6749)
bous end, observed clearly at t * = 5829, continuously retracts shows that the greatest pressure appears at the neck.
to become thinner with time. A neck appears at t * = 6749,
which is a precursor to end pinching.
To better explain this breakup mechanism, capillary pres-
sure at the interface was computed and is shown in Figure 8(b).
A local maximum negative capillary pressure exists at the
neck owing to its larger curvature. Due to the pressure gradi-
ent, fluid is accelerated at the neck and driven toward the end
of the rivulet. The accelerated fluid at the neck leads to fur-
ther reduction in the neck radius. Eventually breakup occurs
(t * = 6982) as the bulbous end pinches off and a small cor-
nered droplet (a drop that deviates from the classical spherical
shape) forms. Specifically, the advancing contact line at the
front of droplet remains rounded whereas the rear contact line
at the trailing edge transforms into corner. Thus the corner
at the trailing edge of the droplet develops due to contact
line motion.59 The breakup process repeats when the bul-
bous end reappears at the tip of the rivulet (see Figure 8(a) at
t * = 7670).
The effect of inlet size on the observed behavior of rivulet
breakup was also examined. Specifically, the height of the
inlet (h) was kept constant as 2 mm while the width (W ) was
varied. Figure 9 shows that the value of the critical Weber
number decreases with increasing inlet size. In contrast, the FIG. 9. The effect of inlet width on the critical Weber number (Wecr ) for
corresponding volumetric flow rate was found to remain nearly rivulet breakup using water (Ka = 3969) with γ = 70◦ and θ = 60◦ . Inset shows
constant for all cases (see the inset of Figure 9). This behavior that the corresponding solvent flow rate is insensitive to inlet size.
052101-10 Singh et al. Phys. Fluids 29, 052101 (2017)

is consistent with our previous findings concerning the dynam- of a stable rivulet and below the curve a droplet or breakup
ics of a rivulet falling over an inclined plate. In that case, inlet of rivulet (shape of the ensuing droplet varies depending on
size had no effect on the developed width of the rivulet for a the case). The region for droplet formation increases with
given flow rate and contact angle.6 The decrease in Wecr with increasing Kapitza number as shown in Figure 10(a). In other
increase in inlet width can be explained by considering the words, higher flow is required to maintain a stable rivulet with
definition of the We number (Equation (9)) and the change increasing Ka value as explained further below.
in inlet velocity with inlet size. In particular, as the inlet size A vertical line is drawn in Figure 10(a) that divides the
is increased the hydraulic diameter also increases, however, plot into two regions of variation in the critical Weber num-
the magnitude of the square in inlet velocity decreases more ber (Wecr ) with the Kapitza number. Below Ka (≤100), Fig-
significantly so that Wecr decreases. ure 10(b), Wecr linearly varies with the Kapitza number. This
is referred to as region I. Above Ka (>100), Figure 10(c),
C. Effect of the solvent properties
Wecr shows a quadratic variation with Ka, Wecr ∼ Ka2 . This
is referred to as region II. The trend in variation of Wecr is
The breakup of a rivulet into droplets is a complex process expressed here as follows:
dictated by many factors including the physical properties of
Wecr = AKa Ka ≤ 100
the fluid. As mentioned earlier the fluid properties are studied (10)
= AKa + BKa2 > 100,
in terms of the Kapitza number.
The breakup of a rivulet and subsequent regime transi- where A and B are the empirical constants. Recall that high
tion is first explored in We Ka space at a fixed value of the values of Ka imply high interfacial surface tension and/or
contact angle and inclination angle: γ = 70◦ and θ = 60◦ . The low viscosity. Higher values of the surface tension give rise
impact of these latter two parameters is studied subsequently. to enhanced capillary forces and capillary forces act as a
Figure 10(a) indicates the occurrence of two flow regimes and destabilizing effect that then requires higher inertia forces to
their transition as it depends on liquid inertia. Specifically, the stabilize the rivulet. In short, surface tension tends to have a
curve represents a critical Weber number (Wecr ) with values destabilizing effect, which inertia counteracts.7 On the other
of the Weber number above the curve reflecting the presence hand, low value of Ka corresponds to high viscosity and/or

FIG. 10. (a) Variation of the critical


Weber number (Wecr) with the Kapitza
number (Ka) for θ = 60◦ and γ = 70◦ .
The value of We above the red curve
shows rivulet and below falls in droplet
(♦: droplet, : rivulet, and —: Wecr ).
Plots of Wecr against Ka shows (b)
linear variation in Region–I and (c)
quadratic variation in Region–II. Inset
shows (b) a cornered droplet with a
pinned satellite droplet for a highly vis-
cous solvent (Ka = 50) and (c) only a
cornered droplet for water (Ka = 3969).
052101-11 Singh et al. Phys. Fluids 29, 052101 (2017)

low surface tension. Such solvents show enhanced wetting from 30◦ to 60◦ . This result is also consistent with previous
characteristics. At these Ka values, viscous forces tend to dom- observations based on linear stability analysis based on the
inate acting as a dissipative factor that partially suppresses lubrication approximation that reveals the stabilizing effect
the breakup instability thereby enhancing the stability of the of gravity.9,10 Similar to the variation of interfacial area with
rivulet.12 Therefore, a stable rivulet can be achieved at lower inclination angle,6 the critical Weber number also becomes
inertia. insensitive to changes in θ beyond 60◦ . For example, the value
The simulations also exhibit slightly different morpholo- of Wecr for 60◦ and 75◦ is almost identical in Figure 11. This
gies during the breakup process depending on the Ka value. is not surprising considering that (gsinθ) does not significantly
The end-pinching of highly viscous solvents (region 1) show vary in those limits.
a small satellite drop along with a bigger daughter droplet.
This behavior is shown in the inset of Figure 10(b) for Ka E. Effect of contact angle
= 50. The smaller satellite drop remains pinned to the plate, it So far the flow simulations were generally restricted to
is unable to move, because of its light weight.59 For higher a fixed contact angle (γ) of 70◦ . The contact angle between
Ka values (region II), a satellite drop is not predicted in the fluid and the surface is another important quantity that
the breakup of the rivulet (see the inset of Figure 10(c) for reflects the wetting properties of the surface. Smaller values
Ka = 3969). of the contact angle were found to lead to increased interfacial
area owing to the enhanced surface wettability.6 Note that the
D. Effect of inclination angle contact angle is a characteristic of a given solid-liquid sys-
Besides solvent physical properties, the effect of plate tem in specific environment.60 That is, a liquid can exhibit a
inclination angle on the breakup of a rivulet is also examined. different contact angle with a change of solid substrate. More-
Inclination angle has been shown to impact various flow fea- over, for a given solid surface, the value of cosγ vs. σ was
tures such as film thickness, wetted area, and interfacial area.6 observed to follow a linear trend for a homologous series
As the flow distribution in a structured packing is a key fac- of liquids. Lower values of σ correspond to smaller contact
tor in the absorption efficiency of the column, it is important angles.52
to understand how the inclination angle affects the breakup Contact angle will also play a role on the stability of the
of rivulet and subsequent flow regime transition. Accordingly, rivulet. The curvature of the interface increases with increasing
the inclination angle was varied from 30◦ to 75◦ (θ = 30◦ 75◦ ) contact angle, which leads to enhanced capillary pressure and
at a fixed contact angle (γ = 70◦ ) for all solvents studied. By converging of the flow that leads to breakup of a rivulet.31
changing the inclination angle, the gravitational effect (gsinθ) This tendency toward breakup may be exacerbated for solvents
on the rivulet also alters. In particular, an increased value of with large Ka (i.e., characterized in general by high surface
θ results in higher film velocity even at the same inlet flow tension). To better understand the effects of contact angle on
rate and, in turn, higher inertia. Recall that inertia acts to sta- rivulet breakup and identification of the critical Weber number
bilize the rivulet against opposing capillary forces that tend (Wecr ), extensive simulations are conducted for a wide range
to destabilize the rivulet.10 As expected, Wecr is reduced with of static γ values (30◦ − 80◦ ).
increasing θ as shown in Figure 11 for all Ka values studied. Variation of the critical Weber number (Wecr ) with
For example, Wecr decreases by ∼17% when θ is increased contact angle (γ) for different solvents at fixed inclination
angle (θ = 60◦ ) is plotted in Figure 12(a). As expected, Wecr
increases with increased value of γ for all solvents. Further-
more, variation of Wecr with γ is steeper for higher values
of Ka number. Recall that high Ka and γ values promote
breakup requiring enhanced inertia to moderate the insta-
bility. To examine this behavior more closely, the critical
Weber number is renormalized with the critical Weber number
at γ = 70◦ (We70cr ): Wecr = Wecr /Wecr . Two regions in varia-
* 70

tion of Wecr with γ are revealed (figure not shown). In both


regions, the critical Weber number shows a scaling relation as
We*cr ∼ (1 − cosγ)m . Furthermore, the value of the exponent m
is higher at lower values of Ka corresponding to lower values
of σ,
Wecr ∼ (1 − cos γ)m
m = 3, Ka ≤ 100 (11)
= 9 4, > 100.


The scaling relation shown Equation (11) is presented


in Figures 12(b) and 12(c). It is worth noting some simi-
larities between variation in the critical Weber number with
FIG. 11. Variation of the critical Weber number (Wecr ) with Kapitza numbers contact angle observed here and variation in interfacial area
greater than 100 (Ka > 100) at contact angle (γ) of 70◦ and four inclination with contact angle observed in some earlier studies. In par-
angles (θ). Wecr is not affected by θ beyond θ ≥ 60◦ . ticular, variation in interfacial area with contact angle also
052101-12 Singh et al. Phys. Fluids 29, 052101 (2017)

FIG. 12. (a) Variation of the critical


Weber number (Wecr ) with contact
angle (γ) for different solvents. Scaling
of Wecr with γ in (b) region/zone-I and
(c) region/zone-II. Insets show snap-
shots of the interface for (b) 10 MPZ
(Ka = 96) and (c) water (Ka = 3969) just
after the break-up of the rivulet.

exhibited two distinct regions for a given set of parameters (Equation (11)) are combined. The following phenomenolog-
influencing the rivulet flow.6,61 In the previous effort, a scal- ical correlation is obtained as shown in Figure 13,
ing relation between interfacial area with contact angle as
AIn ∼ (1 − cosγ)−n was shown, where AIn is the normalized Wecr ∼ Ka(1 − cos γ)3 Ka ≤ 100
interfacial area and n is an exponent that depends on Ka.6 (12)
∼ Ka2 (1 − cos γ)9/4 > 100.
Solvents having low Ka (<100) value showed steeper varia-
tion of AIn with γ than those having high Ka (>100). In the In this relationship, solvents with low values of surface
present case, however, the trends in variation of Wecr with tension (low values of Ka) show a linear variation in Wecr
contact angle were found to be opposite: solvents having high with Ka and reduced sensitivity to the value of the contact
Ka value showed stiffer variation of Wecr with γ value con- angle as the term (1 − cosγ)3 decays more quickly than the
tact angle than those having low Ka value. It is worth noting term (1 − cosγ)9/4 . In comparison, solvents with high val-
that the variation in the Wecr with contact angle and variation ues of surface tension (high values of Ka) show a quadratic
in AIn with contact angle are physically consistent. For high variation in Wecr with Ka and greater sensitivity to the value
Ka and γ values, the solvent will demonstrate less wettabil- of the contact angle. So solvents with low values of surface
ity and an enhanced tendency for instability. Therefore, higher tension (low values of Ka) correspond to lower values of
inertia is required to stabilize the rivulet, so eventually Wecr WeCr . A lower value of Wecr implies a more stable rivulet
increases. even at low flow rate. It is worth noting that low values
Earlier the critical Weber number representing rivulet of Ka generally correspond to highly viscous solvents, and
breakup was shown to vary as Wecr ∼ Kap , where p = 1 for that solvents with high viscosity are characterized by lower
Ka < 100 and p = 2 for Ka > 100 for a fixed contact angle values of surface tension62–64 and, in turn, smaller contact
γ = 70◦ (see Figures 10(b) and 10(c) and Equation (10)). This angles.
scaling relation was found to hold for all contact angles, as While this investigation employed a constant and static
shown in the inset of Figure 13. All curves collapse when value for contact angle, it is recognized that the contact angle
the expressions for the critical Weber number as a function is a varying and complex quantity.51 Still the dynamic contact
of the Kapitza number (Equation (10)) and the contact angle angle is not well understood and a complete theory on contact
052101-13 Singh et al. Phys. Fluids 29, 052101 (2017)

FIG. 14. Comparison of the rivulet shape at the onset of breakup for water
(Ka = 3969) and θ = 60◦ for (a) a static contact angle (SCA) defined by γ
= 70◦ corresponding to Wecr = 0.36 and (b) a dynamic contact angle (DCA)
defined by γA = 75◦ , γR = 53.3◦ and γe = 70◦ corresponding to We = 0.26.

VI. CONCLUSION
We conducted three dimensional multiphase flow simu-
lations with VOF interface capturing method for studying the
breakup of a rivulet flowing down a smooth inclined plate.
Understanding the factors that lead to rivulet breakup is rel-
evant to industries where the microscopic flow pattern plays
an important role, such as in packings that are employed in
separation and distillation processes. The CFD predictions
compared well with experimental results of a rivulet flowing
down an inclined plate using water as well as highly viscous
liquids. Specifically, the predicted morphology (shape) and
width of the rivulet matched well with those observed in the
experiments. Rivulet breakup is a complex phenomenon. This
process is influenced by many factors including solvent proper-
ties, flow rate (inertia), and inclination angle. Contact angle is
another important parameter representing the wetting charac-
teristics of the system (interaction of the solvent and plate). In
FIG. 13. Scaling for the normalized critical Weber number (Wecr ) with the this effort, rivulet breakup through end-pinching was examined
contact angle (γ) and Kapitza number (Ka) for (a) Region I and (b) Region II. wherein a rivulet tip retracts due to surface tension causing the
Insets shows the variation of Wecr with Ka at different values of γ with θ liquid to collect into a bulbous end that is eventually pinched
= 60◦ .
off.33 This long term instability leads to the breakup and forma-
tion of a droplet. An extensive CFD simulation campaign was
angle hysteresis is still in development.39,40 Therefore, a com- conducted to explore the role of the aforementioned factors on
parison between the static contact angle (SCA) and a dynamic the breakup phenomenon.
contact angle (DCA) model on rivulet breakup was conducted A critical value of the Weber number (Wecr ) that separates
for a single case involving water (Ka = 3969). The shape of the droplet and rivulet regimes was computed as a function of
rivulet at the onset of breakup is presented in Figure 14 corre- the Kapitza number. For We values lower than the Wecr , a
sponding to a SCA and DCA simulation at We = 0.26 (recall droplet is observed, and conversely a rivulet is observed for
that rivulet breakup is expected at this We number). A com- higher values of We. The results show that the critical Weber
parison reveals slight differences in the shape of the bulbous number increases with increased Kapitza number. This trend
end with the DCA case predicting a nearly spherical bulbous can be understood because high viscosity dampens growth
end and the SCA case resulting in a more cornered shape. In in the instability, subsequently hindering breakup. Recall that
addition, the length of the wrist adjacent to the bulbous end high values of solvent viscosity and/or low values of surface
is longer for the DCA setup than that of the SCA. Besides tension correspond to low Kapitza numbers. Therefore, low Ka
comparing rivulet shapes at the onset of breakup, the critical solvents will tend to exhibit rivulet flow even at low flow rates.
Weber number (Wecr ) was also compared. The value of Wecr The exact nature of the trend in the variation of Wecr with Ka
corresponding to the DCA model was found to be 10% lower depends on the Ka value. Solvents with lower Ka value show
than the corresponding one for the SCA condition. Finally, it Wecr varying linearly with Ka, while solvents with higher Ka
is worth noting that highly viscous solvents may exhibit small value show Wecr having a quadratic variation with Ka. The
contact angles so that variation in the advancing and receding effect of plate inclination was also explored. The critical Weber
contact angle will be not significant. number decreases with increased inclination angle (θ) owing
052101-14 Singh et al. Phys. Fluids 29, 052101 (2017)

to accelerated liquid velocity. However, the effect is negligible 19 L. Kondic et al., “Nanoparticle assembly via the dewetting of patterned thin

beyond θ > 60◦ . metal lines: Understanding the instability mechanisms,” Phys. Rev. E: Stat.,
Nonlinear, Soft Matter Phys. 79(2), 026302 (2009).
The role of contact angle (γ) on rivulet breakup was also 20 M. F. G. Johnson et al., “Experimental study of rivulet formation on
systematically examined. As the contact angle is increased an inclined plate by fluorescent imaging,” J. Fluid Mech. 394, 339–354
so too does Wecr (i.e., greater inertia is required to achieve (1999).
21 S. Ubal et al., “The influence of inertia and contact angle on the instability
stable rivulet flow). As with Wecr versus Ka, Wecr versus γ
of partially wetting liquid strips: A numerical analysis study,” Phys. Fluids
shows two trends for variation. Solvents with high Ka values 26(3), 032106 (2014).
show steeper variation of Wecr with contact angle than those 22 A. Y. Tong and Z. Wang, “Relaxation dynamics of a free elongated liquid

with low values of Ka (corresponding to low values of surface ligament,” Phys. Fluids 19(9), 092101 (2007).
23 J. A. Diez, A. G. González, and L. Kondic, “Stability of a finite-length
tension). By combining these analyses, a phenomenological
rivulet under partial wetting conditions,” J. Phys.: Conf. Ser. 166, 012009
scaling for the critical Weber number with the Kapitza number (2009).
and contact angle is presented. 24 J. A. Diez, A. G. González, and L. Kondic, “Instability of a transverse liquid

rivulet on an inclined plane,” Phys. Fluids 24(3), 032104 (2012).


25 L. M. Hocking and S. H. Davis, “Inertial effects in time-dependent motion

ACKNOWLEDGMENTS of thin films and drops,” J. Fluid Mech. 467, 1–17 (2002).
26 K. Mertens, V. Putkarzdze, and P. Vorobieff, “Morphology of a stream flow-

This research was supported in part by an appointment to ing down an inclined plane. Part 1. Braiding,” J. Fluid Mech. 531, 49–58
the National Energy Technology Laboratory Research Partici- (2005).
27 B. Birnir et al., “Morphology of a stream flowing down an inclined plane.
pation Program, sponsored by the Office of Fossil Energy, U.S. Part 2. Meandering,” J. Fluid Mech. 607, 401–411 (2008).
Department of Energy, through Carbon Capture Simulation 28 H.-Y. Kim, J.-H. Kim, and B. H. Kang, “Meandering instability of a rivulet,”

Initiative (CCSI) and administered by the Oak Ridge Institute J. Fluid Mech. 498, 245–256 (2004).
29 A. Daerr et al., “General mechanism for the meandering instability
for Science and Education. The CFD simulations were per-
of rivulets of Newtonian fluids,” Phys. Rev. Lett. 106(18), 184501
formed on NETL supercomputer “Joule.” The authors would (2011).
also like to thank Professor Sankaran Sundaresan, Prince- 30 S. Couvreur and A. Daerr, “The role of wetting heterogeneities in the

ton University, for his valuable inputs and discussion on the meandering instability of a partial wetting rivulet,” EPL 99(2), 24004
(2012).
project. 31 G. Ghigliotti, C. Zhou, and J. J. Feng, “Simulations of the breakup of liquid

filaments on a partially wetting solid substrate,” Phys. Fluids 25(7), 072102


1 L. Spiegel and W. Meier, “Distillation columns with structured packings in (2013).
the next decade,” Chem. Eng. Res. Des. 81(1), 39–47 (2003). 32 A. G. González et al., “Rupture of a fluid strip under partial wetting
2 J. Mackowiak, Fluid Dynamics of Packed Columns (Springer-Verlag, Berlin, conditions,” EPL 77(4), 44001 (2007).
2010). 33 J. Eggers, “Nonlinear dynamics and breakup of free-surface flows,” Rev.
3 L. Raynal, F. Ben Rayana, and A. Royon-Lebeaud, “Use of CFD for CO2 Mod. Phys. 69(3), 865–930 (1997).
absorbers optimum design: From local scale to large industrial scale,” 34 S. K. Wilson, J. M. Sullivan, and B. R. Duffy, “The energetics of the breakup

Energy Procedia 1(1), 917–924 (2009). of a sheet and of a rivulet on a vertical substrate in the presence of a uniform
4 L. Raynal and A. Royon-Lebeaud, “A multi-scale approach for CFD calcu- surface shear stress,” J. Fluid Mech. 674, 281–306 (2011).
lations of gas–liquid flow within large size column equipped with structured 35 C. W. Hirt and B. D. Nichols, “Volume of fluid (VOF) method for

packing,” Chem. Eng. Sci. 62(24), 7196–7204 (2007). the dynamics of free boundaries,” J. Comput. Phys. 39(1), 201–225
5 P. Schmuki and M. Laso, “On the stability of rivulet flow,” J. Fluid Mech. (1981).
215, 125–143 (1990). 36 ANSYS Fluent Theory Guide, 2015.
6 R. K. Singh, J. E. Galvin, and X. Sun, “Three-dimensional simulation of 37 J. U. Brackbill, D. B. Kothe, and C. Zemach, “A continuum method for

rivulet and film flows over an inclined plate: Effects of solvent properties modeling surface tension.,” J. Comput. Phys. 100(2), 335–354 (1992).
and contact angle,” Chem. Eng. Sci. 142, 244–257 (2016). 38 A. A. Keller, V. Broje, and K. Setty, “Effect of advancing velocity and fluid
7 G. W. Young and S. H. Davis, “Rivulet instabilities,” J. Fluid Mech. 176, viscosity on the dynamic contact angle of petroleum hydrocarbons,” J. Pet.
1–31 (1987). Sci. Eng. 58(1-2), 201–206 (2007).
8 M. A. Herrada et al., “Stability of a rivulet flowing in a microchannel,” Int. 39 B. Birnir et al., “Meandering fluid streams in the presence of flow-rate

J. Multiphase Flow 69, 1–7 (2015). fluctuations,” Phys. Rev. Lett. 101(11), 114501 (2008).
9 S. Mechkov, M. Rauscher, and S. Dietrich, “Stability of liquid ridges on 40 K. Yokoi et al., “Numerical studies of the influence of the dynamic contact

chemical micro- and nanostripes,” Phys. Rev. E: Stat., Nonlinear, Soft angle on a droplet impacting on a dry surface,” Phys. Fluids 21(7), 072102
Matter Phys. 77(6), 061605 (2008). (2009).
10 J. A. Diez, A. G. González, and L. Kondic, “On the breakup of fluid rivulets,” 41 L. H. Tanner, “The spreading of silicone oil drops on horizontal surfaces,”

Phys. Fluids 21(8), 082105 (2009). J. Phys. D: Appl. Phys. 12(9), 1473 (1979).
11 K. Sekimot, R. Oguma, and K. Kawasaki, “Morphological stability analysis 42 R. I. Issa, “Solution of the implicitly discretised fluid flow equations by

of partial wetting,” Ann. Phys. 176(2), 359–392 (1987). operator-splitting,” J. Comput. Phys. 62(1), 40–65 (1986).
12 S. H. Davis, “Moving contact lines and rivulet instabilities. Part 1. The static 43 D. L. Youngs, “Time-dependent multi-material flow with large fluid

rivulet,” J. Fluid Mech. 98(02), 225–242 (1980). distortion,” Numer. Methods Fluid Dyn. 24(2), 273–285 (1982).
13 H. H. Saber and M. S. El-Genk, “On the breakup of a thin liquid film subject 44 H. Yang et al., “Progress in carbon dioxide separation and capture: A

to interfacial shear,” J. Fluid Mech. 500, 113–133 (2004). review,” J. Environ. Sci. 20(1), 14–27 (2008).
14 M. S. El-Genk and H. H. Saber, “Minimum thickness of a flowing down 45 N. Razi, O. Bolland, and H. Svendsen, “Review of design correlations for

liquid film on a vertical surface,” Int. J. Heat Mass Transfer 44(15), 2809– CO2 absorption into MEA using structured packings,” Int. J. Greenhouse
2825 (2001). Gas Control 9(0), 193–219 (2012).
15 S. K. Wilson and B. R. Duffy, “When is it energetically favorable for a rivulet 46 H. Chang, “Wave evolution on a falling film,” Annu. Rev. Fluid Mech. 26(1),

of perfectly wetting fluid to split?,” Phys. Fluids 17(7), 078104 (2005). 103–136 (1994).
16 S. P. Lin, “Instability of a liquid film flowing down an inclined plane,” Phys. 47 C. E. Meza and V. Balakotaiah, “Modeling and experimental studies of

Fluids 10(2), 308 (1967). large amplitude waves on vertically falling films,” Chem. Eng. Sci. 63(19),
17 J.-y. Park et al., “Anisotropic rupture of polymer strips driven by Rayleigh 4704–4734 (2008).
instability,” J. Chem. Phys. 124(21), 214710 (2006). 48 Y. Maham et al., “Volumetric properties of aqueous solutions of
18 J. Petit et al., “Break-up dynamics of fluctuating liquid threads,” Proc. Natl. monoethanolamine, mono- and dimethylethanolamines at temperatures
Acad. Sci. U. S. A. 109(45), 18327–18331 (2012). from 5 to 80 ◦ C I,” Thermochim. Acta 386(2), 111–118 (2002).
052101-15 Singh et al. Phys. Fluids 29, 052101 (2017)

49 R. Tahery and H. Modarress, “A new and a simple model for surface tension 57 A. Ataki and H. J. Bart, “The use of the VOF-model to study the wetting of

prediction of water and organic liquid mixtures,” Iran. J. Sci. Technol. Trans. solid surfaces,” Chem. Eng. Technol. 27(10), 1109–1114 (2004).
B-Eng. 29(B5), 501–509 (2005). 58 Y. Xu et al., “Detailed investigations of the countercurrent multiphase (gas–
50 A. V. Rayer et al., “Physicochemical properties of {1-methyl piperazine (1) liquid and gas–liquid–liquid) flow behavior by three-dimensional computa-
+ water (2)} system at T = (298.15 to 343.15) K and atmospheric pressure,” tional fluid dynamics simulations,” Ind. Eng. Chem. Res. 53(18), 7797–7809
J. Chem. Thermodyn. 43(12), 1897–1905 (2011). (2014).
51 P. G. de Gennes, “Wetting: Statics and dynamics,” Rev. Mod. Phys. 57(3), 59 T. Podgorski, J. M. Flesselles, and L. Limat, “Corners, cusps, and pearls in

827–863 (1985). running drops,” Phys. Rev. Lett. 87(3), 036102 (2001).
52 W. A. Zisman, “Relation of the equilibrium contact angle to liquid and 60 J. H. Snoeijer and B. Andreotti, “A microscopic view on contact angle

solid constitution,” Contact Angle, Wettability, and Adhesion (American selection,” Phys. Fluids 20(5), 057101 (2008).
Chemical Society, 1964), pp. 1–51. 61 J. J. Gualito et al., “Design method for distillation columns filled with metal-
53 Y. Iso et al., “Numerical and experimental study on liquid film flows on lic, ceramic, or plastic structured packings,” Ind. Eng. Chem. Res. 36(5),
packing elements in absorbers for post-combustion CO2 capture,” Energy 1747–1757 (1997).
Procedia 37(0), 860–868 (2013). 62 A. H. Pelofsky, “Surface tension-viscosity relation for liquids,” J. Chem.
54 A. Hoffmann et al., “Fluid dynamics in multiphase distillation processes in Eng. Data 11(3), 394–397 (1966).
packed towers,” Comput. Chem. Eng. 29(6), 1433–1437 (2005). 63 A. J. Queimada et al., “Generalized relation between surface tension and
55 W. Nusselt, “Die Oberflächenkondesation des Wasserdampfes,” Z. Ver. viscosity: A study on pure and mixed n-alkanes,” Fluid Phase Equilib. 222–
Dtsch. Ing. 60(27), 541–546 (1916). 223, 161–168 (2004).
56 K. Stoev, E. Ramé, and S. Garoff, “Effects of inertia on the hydrodynamics 64 H. Schonhorn, “Surface tension-viscosity relationship for liquids,” J. Chem.

near moving contact lines,” Phys. Fluids 11(11), 3209–3216 (1999). Eng. Data 12(4), 524–525 (1967).

You might also like