You are on page 1of 18

Injection of a heavy fluid into a light fluid in

a closed-end pipe
Cite as: Phys. Fluids 32, 063302 (2020); https://doi.org/10.1063/5.0009102
Submitted: 27 March 2020 • Accepted: 22 May 2020 • Published Online: 09 June 2020

S. Akbari and S. M. Taghavi

ARTICLES YOU MAY BE INTERESTED IN

On respiratory droplets and face masks


Physics of Fluids 32, 063303 (2020); https://doi.org/10.1063/5.0015044

Likelihood of survival of coronavirus in a respiratory droplet deposited on a solid surface


Physics of Fluids 32, 061704 (2020); https://doi.org/10.1063/5.0012009

Direct numerical simulation of multiscale flow physics of binary droplet collision


Physics of Fluids 32, 062103 (2020); https://doi.org/10.1063/5.0006695

Phys. Fluids 32, 063302 (2020); https://doi.org/10.1063/5.0009102 32, 063302

© 2020 Author(s).
Physics of Fluids ARTICLE scitation.org/journal/phf

Injection of a heavy fluid into a light fluid


in a closed-end pipe
Cite as: Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102
Submitted: 27 March 2020 • Accepted: 22 May 2020 •
Published Online: 9 June 2020

S. Akbari and S. M. Taghavia)

AFFILIATIONS
Department of Chemical Engineering, Université Laval, Québec, Quebec G1V 0A6, Canada

a)
Author to whom correspondence should be addressed: Seyed-Mohammad.Taghavi@gch.ulaval.ca

ABSTRACT
We report experimental results on the injection of a heavy fluid into a light one in a closed-end pipe, inclined at intermediate angles. The
injection of the heavy fluid is made using an inner duct with a smaller diameter than that of the pipe in which the light fluid is placed. The
fluids used are miscible and Newtonian, and they have the same viscosity. Our observation shows that, during the removal/replacement of the
light fluid by the heavy fluid, at least four distinct flow stages can be identified: (i) initial buoyant jet of the heavy fluid, (ii) development of a
mixing region, (iii) slumping flow of the heavy fluid, and (iv) heavy fluid front reaching the pipe end and returning toward the mixing region.
Using high-speed camera images along with the ultrasound Doppler velocimetry and laser induced fluorescence data, the flow characteristics
in these flow stages are quantified, and they are described in detail vs the dimensionless groups that govern the flow dynamics, namely, the
Froude number (Fr), the Reynolds number (Re), the Archimedes number (Ar), and the pipe inclination angle (β). While our findings are
of fundamental importance, they can also be used to provide a fluid mechanics understanding of the dump bailing method in the plug and
abandonment (P&A) of oil and gas wells.
Published under license by AIP Publishing. https://doi.org/10.1063/5.0009102., s

I. INTRODUCTION processes, which usually include the injection of cement slurry into
drilling mud in certain areas of oil or gas well. This is to isolate the oil
Fluid injection into a flow geometry containing another fluid or gas reservoir and avoid reservoir fluids from migrating to upper
attracts research interests due to its prominent applications in vari- zones especially with underground water resources.2 For the pro-
ous industries, such as the plug and abandonment (P&A) of oil and cess to be successful, the in-place drilling mud must be displaced and
gas wells, transport of crude oil, oil recovery, food processing, and replaced by the injected cement slurry in a way that mixing between
channel cleaning.1 The objective of fluid injection is commonly to the two fluids is minimized. The process performance depends on
displace, remove, or replace an in situ fluid by an injected fluid. The several factors, e.g., the fluid injection method, the well configura-
behaviors of such a flow depend on buoyancy, e.g., when the two flu- tion, the flow geometry (e.g., circular casing or annulus), and the
ids have different densities, the injection rate of the heavy fluid, and well inclination.3 One of the common fluid injection methods in
the flow geometry configuration, to name a few. For fluid injection P&A processes is the dump bailing method4,5 in which the cement
into confined flow geometries, the flow can be studied as a func- slurry (heavy fluid) is continuously injected to remove the drilling
tion of the parameters that govern its dynamics, such as the phys- mud (light fluid) in a circular casing near the bottom of the well
ical properties of the fluids (e.g., densities and viscosities), the flow (i.e., a closed-end pipe), as schematically shown in Fig. 1. The fluid
geometry type (e.g., pipe, annulus, or channel), and the operational injection is made using a cylindrical shape bailer that has a diameter
conditions (e.g., injection velocity and geometry inclination). Based smaller than that of the closed-end pipe. Motivated by understand-
on the aforementioned parameters involved as well as the competi- ing fluid mechanics issues in this process, in this paper, we experi-
tion between the governing forces, various kinds of flow regimes are mentally investigate the problem of the injection of a heavy fluid into
expected to occur. a light fluid in an inclined closed-end pipe.
Our underlying motivation in studying the flow problem The injection flow problem may involve a jet-like behavior at
of fluid injection is to understand the fluid mechanics of P&A early stages, depending on the flow geometry, injection velocity, and

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-1


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 1. A sequence of schematic images of the cement slurry injection through the dump bailing method: (a) initial well condition, typically full of in situ fluid (drilling mud); (b)
a bridge plug is placed, and a dump bailer injects the cement slurry on the bridge plug; and (c) the in situ fluid is removed and replaced by the cement slurry. The shape of
the interface between the fluids is illustrative.

buoyancy. A series of previous studies have considered the effects experimentally studied the injection of a heavy fluid into a light
of the injection of a heavy fluid into a light fluid in confined flow fluid in a horizontal channel. They found that most mixing occurs
geometries, such as high-depth channels, vertical or inclined pipes, immediately downstream of the inlet and also near the head of the
and porous media, as summarized in Table I. Briefly, these stud- heavy fluid, with an increase in proportion of the entrainment occur-
ies have shown that the injection flow initially involves a buoyant ring in a mixing region, near the inflow, as the flow Froude num-
jet, for which it is possible to quantify volume and momentum flux ber increases. Their work revealed that the structure of the flow is
distributions, turbulent velocities (both axial and radial), and con- strongly controlled by mixing between the heavy and light fluids,
centration distributions. While some studies have been restricted near the inlet and also at the head of the heavy fluid, leading to hor-
to the case of bounded containers (e.g., the filling box problem for izontal and vertical stratifications of the density profile. Du et al.18
containers with height ≤ width), others have shown that the flow investigated the mixing of buoyant flows produced by continuously
becomes more complicated when the aspect ratio is very large (e.g., releasing dense brine into an inclined tank filled with fresh water.
height/width > 6).6 Using the light attenuation technique, they showed that the mix-
Several studies have aimed to shed light on the dynamics and ing between the fluids is strengthened by an increase in the volume
mixing of buoyant flows, typically focusing on relating the bulk flow rate of the heavy fluid, and the current downstream of the injec-
properties of the heavy fluid flow to its mixing properties and tion boundary becomes more mixed as the buoyancy force increases.
investigating the front velocity and mixing behind the front of the Other experimental works19–21 showed that when the flow is buoy-
heavy fluid and in the body of the flow.14–16 Sher and Woods17 ant, the density difference acts to segregate the heavy and light fluids

TABLE I. Summary of studies on buoyant jets in relation to the current work.

Flow conditions Analysis References

Buoyancy-driven turbulent flows Role of buoyancy in a plane Kotsovinos,7 Kotsovinos


in the presences of buoyant jets turbulent buoyant jet and List8
Turbulent vertical buoyant jets
Turbulent mixing in jets Papanicolaou and List9
over a wide range of initial flux
Filling box problem in a
Theoretical study of the flow Vauquelin et al.10
non-Boussinesq case
Injection of a light fluid in Characterization of mixing
Xue et al.11
a box of heavy fluid flow behaviors
Vertical turbulent buoyant jets Investigation of initial and fully
Milton-McGurk et al.12
into a tank developed stages of the flow
Turbulent offset jets into Characteristics of jets with
Assoudi et al.13
a rectangular box different offset heights and densities

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-2


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

in the transverse direction as the heavy fluid usually advances under- aspects. First, in most of the previous studies on fluid injection and
neath the light fluid. To study the effects of an injection flow on the exchange flows, the initial heights of the two fluids are equal,14,21,41
hydrodynamics of displacement flows, Taghavi et al.22 investigated while there are only a few works in the literature, which consider
the displacement flow of buoyant miscible fluids of equal viscosity in exchange or injection flows from a partial inlet.17,25,42,43 Second, we
a long nearly horizontal pipe. They observed that the flow becomes consider a closed-end pipe, implying that the inlet and outlet bound-
stable as the injection rate increases (also observed by Zheng et al.23 aries in our flow geometry configuration are on the same side, affect-
in an inclined pipe) and that the penetration front velocity grows ing the propagation of the heavy fluid due to the flow of the lighter
linearly with the injection rate. Using two-layer flow modeling and fluid. By considering this novel flow, our contribution is to quantify
direct numerical simulations, Hogg et al.24 analyzed the injection of different flow behaviors of the injection buoyant flow via analyz-
a heavy fluid into a horizontal channel initially containing a light ing velocities and mixing between the fluids and characterizing and
fluid, finding the conditions for which the heavy fluid can completely classifying different flow stages.
displace the light one. We note that the flows driven by an injection Section II introduces the experimental setup, as well as the
differ from those arising from releasing of a fixed volume of fluid experimental parameters and their ranges. Then, Sec. III defines the
behind a valve (e.g., lock-exchange flows25 ), in which the fluids inter- governing dimensionless numbers in this study. In presenting our
penetrate, only due to buoyancy forces, without the presence of an results, first Secs. IV A and IV B introduce the general observation
injection. and define the main observed flow stages in detail. Then, Sec. IV C
Table II summarizes some of the relevant studies on fluid injec- analyzes the mixing between the two fluids. Section IV D qualita-
tion and releasing into confined flow geometries, including channels tively examines the stability of the interface between the two fluids
and pipes, when the propagation of a heavy fluid removes a light in different conditions. Finally, Sec. V concludes the paper with a
fluid. In this table, a full inlet configuration refers to the situation brief summary.
where the heights of the heavy and light fluids on both sides of the
inlet boundary are equal, e.g., in a typical lock-exchange flow; how-
ever, in a partial inlet configuration, the heavy fluid propagates into
II. EXPERIMENTAL SETUP AND PROCEDURE
the light fluid through an inlet for which the height is a fraction of
the light fluid’s initial height. In addition, the terms open-end and To study the injection of a heavy fluid into a light fluid, an
closed-end refer to the situation where the flow geometry end is open experimental setup is designed. The setup consists of two transpar-
or closed, respectively. In the former case, the inlet and outlet bound- ent pipes: one as the main flow domain with the inner diameter (D̂)
aries are at the opposite sides of the flow geometry, while in the latter of 3.81 (cm) and a length of 2.46 (m), in which the light fluid is
case, the inlet and outlet boundaries are positioned on the same side. placed; the other pipe is for the injection of the heavy fluid with inner
Note that, for lock-exchange flows, both ends of the flow geometry and outer diameters of 1.27 (cm) and 2.54 (cm), respectively, and the
are typically closed. length of 2 (m). The schematic of the experimental setup is shown in
Our study is concerned with the injection of a heavy fluid from Fig. 2. The focus of this study is to analyze the injection flow of the
a partial inlet into a lighter fluid in a closed-end pipe. Here, the term heavy fluid from the inner pipe into the outer pipe, where the light
partial inlet implies that the inlet cross-sectional area, from which fluid is initially placed, over the distance between the inner pipe exit
the heavy fluid is injected, is smaller than the cross-sectional area of and the outer pipe end. Over this distance, the heavy fluid enters the
the main flow domain. Our work is novel in terms of at least two flow domain and displaces the lighter fluid. The domain of study is

TABLE II. Summary of studies on the fluid injection and releasing of a heavy fluid into a light fluid in confined flow geometries.

Initial conditions, flow constraints Geometry, inclination Method References

Lock-exchange and partial/full inlet Rectangular, horizontal Experiment and modeling Rottman and Simpson26
Lock-exchange and partial/full inlet Channel, horizontal Experiment and modeling Shin et al.,25 Chiapponi et al.27
Lock-exchange and full inlet Rectangular channel, horizontal Experiment and numerical Lowe et al.,28 Inghilesi et al.29
Dai,30,31 Dai and Huang,32
Lock-exchange and partial inlet Inclined surface in a channel Experiment and numerical
Steenhauer et al.33
Pipe or channel, horizontal Taghavi et al.,19,22 Alba et al.,34
Injection and full inlet, open-end Experiment and numerical
to vertical Amiri et al.,35 Zhou et al.36
Sher and Woods,17
Nourmohammadi et al.,37
Injection and partial inlet, open-end Channel, horizontal Experiment
Longo et al.,38 Yuan and
Horner-Devine39,40
Injection and partial inlet, closed-end Pipe, moderate inclinations Experiment Current study

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-3


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 2. Schematic view of the experi-


mental setup. The domain of study is
marked by dotted lines. The LIF instru-
ment and the UDV probe are also
depicted. Q̂ represents the injection flow
rate of the heavy fluid. The interface
shape and the laser light sheet are illus-
trative only.

marked in Fig. 2, i.e., with the length of 46 (cm) and the diameter they are used as references of light intensities for the Beer–Lambert
of 3.81 (cm); the injection inlet has a diameter of 1.27 (cm), i.e., the law.44 Using an in-house MATLAB program, post-processing of the
inner pipe diameter. The setup is mounted on a frame, which could recorded images is done to characterize the interfacial and mixing
be tilted from horizontal to vertical. behaviors.
In each experimental test, first the pipe is filled with the lighter To analyze the degree of mixing along the depth of the pipe, the
fluid (fluid L, water). Then, the heavier fluid (fluid H, salt-water two-dimensional planar Laser Induced Fluorescence (LIF) method
solution using NaCl as an additive) is injected via an elevated con- is used as a non-intrusive visualization technique. To apply this
tainer through the inner pipe into the outer pipe where the lighter method, the pipe is enclosed in a transparent rectangular box, and
fluid is placed. During the experiment, the lighter fluid exits the flow the volume between the pipe and the internal walls of the box is
domain, toward another container, as shown in Fig. 2. The fluid filled with deionized water, matching the fluids inside the pipe in
heights in both containers are nearly constant in order to keep a terms of the refractive index. The flow is seeded with a fluorescent
nearly constant hydrostatic pressure in all experiments. The injec- dye, namely, Rhodamine-B, at a concentration of 50 (mg/l),45 and it
tion flow rate is adjusted by a needle valve in each experiment, is illuminated by a laser sheet from a DPSS laser system (Laserglow
and it is measured by using a turbine flowmeter (FTB-421, Omega) technologies) with an excitation wavelength of 532 (nm). The sheet
with an accuracy of ±3%. The flow rate data are transferred to a optics arrangement is capable of producing a sheet of light. The scat-
computer through an acquisition card (model USB-6002, National tered light is recorded by using a camera positioned perpendicular to
Instruments). the laser light sheet, as illustrated in Fig. 2.
The densities of the fluids are measured using a high-accuracy To obtain information about the flow local dynamics in cer-
density meter (Anton Paar DMA 35), and their viscosities are quan- tain cases, Ultrasound Doppler Velocimetry (UDV) measurements
tified using a rheometer (Discovery HR-3, TA Instruments), both are performed. The UDV (DOP4000 from Signal Processing S.A.)
showing values close to those of pure water. is based on the principle of the reflection of sound waves from par-
The flow development is captured using a high speed cam- ticles moving with the flow. Both the location and the velocity of
era (Basler acA2040, with 4096 gray scale levels), covering the flow the particles can be derived from the time delay between the emit-
domain, as shown in Fig. 2. The field of view that is covered by the ted and received signals and the frequency shift due to the Doppler
camera is typically 3.81 × 46 (cm2 ). The images are recorded at a effect. The advantage of this method is the acquisition of velocity
specified frame rate (typically 20 frames/s). For visualization pur- profiles along the sound propagation line within very short time
poses, the heavier fluid is dyed with a little amount of ink (fountain intervals. Regarding the concentration of the UDV seeding particles
pen India black ink). In order to obtain images with uniform light- [copolyamide with the size distribution of 60 (μm)], around 0.2 (g/l)
ing across the pipe, light emitting diodes are used behind the pipe, is added to both fluids.46
and diffusive panels are put in between. Light absorption calibration The injection experiments are conducted for a wide range of
is performed in the usual fashion: before each experiment, an image injection rates and density differences. Table III shows the main
of the pipe containing the transparent (lighter) fluid and an image dimensional parameters of the flow. Note that some of the param-
of the pipe containing the darker (heavier) fluid are recorded, and eters, such as the viscosity and the hydrostatic pressure, are kept

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-4


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

TABLE III. Ranges and values of the dimensional parameters used in our experiments.

Parameter Name Range or value

D̂ Inner diameter of outer pipe (flow domain) 3.81 (cm)


D̂in,1 Inner diameter of inner pipe 1.27 (cm)
D̂in,2 Outer diameter of inner pipe 2.54 (cm)
ρ̂H Heavy fluid density 1000–1018 (kg/m3 )
ρ̂L Light fluid density 998 (kg/m3 )
ρ̂H + ρ̂L
ρ̄ˆ = Average density 999–1008 (kg/m3 )
2
μ̂ Fluid viscosity 0.001 (Pa s)
L̂ Length of flow domain 46 (cm)
V̂0 Injection velocity 68–280 (mm/s)
μ̂
v̂ Kinematic viscosity ( ) ∼ 10−6 (m2 /s)
ρ̄ˆ
D̂m Molecular diffusivity ∼ 2 × 10−9 (m2 /s)
ĝ Gravitational acceleration 9.81 (m/s2 )

constant in all experiments. In order to check the short and long Reynolds number, Re, the Froude number, Fr, and the pipe inclina-
term reproducibility, some experiments are repeated. tion angle from horizontal, β. We consider Re to vary in the range of
O(103 ) to O(104 ) and Fr to be in the range of O(10−1 ) to O(10). In
III. SCOPE OF EXPERIMENTS terms of the inclination angle, β, we consider low to moderate incli-
Let us define the scope of our experiments in terms of the nations angles for which some slumping patterns of the interface
dimensionless groups of the flow, the number of which based on between the fluids may be expected, at least at longer times.
Buckingham’s Pi theorem must be 8, which are given in Table IV. The results of our experiments can be presented vs three
However, some of these numbers are constant, and some others are parameters, e.g., Re, Fr, β, or their combinations. In this context,
not relevant in our work. First of all, although we consider misci- the parameter Re/Fr is useful because it is independent of the injec-
ble flows, we focus on the regimes of Pe ≫ 1 so that the molecular tion velocity. In our work, Re/Fr is closely related to the Archimedes
diffusion does not affect the flows studied in the time scale of the number, Ar,
experiments, i.e., the two fluids do not have time to mix under the ρ̄ˆ (ρ̂H − ρ̂L )ĝ D̂3 Re 2
Ar = 2
= 2( ) , (1)
effect of the molecular diffusion. We also focus on small density dif- μ̂ Fr
ferences, implying small Atwood numbers (At ≪ 1). In addition, which varies in the range of (1 − 11) × 106 and is preferred to
we consider fixed geometrical parameters: different aspect ratios of the present buoyant effects. The Archimedes number quantifies the
the flow geometry are fixed for all experiments. These are mainly motion of the fluids to the density difference, and its large values
DR1 , DR2 , and δ. The latter is a parameter of O(10), while the two indicate the dominance of buoyancy forces in the flow.
formers are the parameters of O(1). Considering these, as Table IV In order to better understand the physics behind the flow
shows, the main dimensionless numbers in our experiments are the dynamics in Sec. IV, let us introduce the critical characteristic veloc-
ities in our flow. First of all, the characteristic velocity in relation to
TABLE IV. Dimensionless parameters in our experiments. the injection imposed inertia is the mean injection velocity, V̂0 . The
characteristic buoyant-inertial velocity, V̂i , can also be obtained if
Parameter Name Definition Range the buoyant stresses (of order (ρ̂H − ρ̂L )ĝ D̂) and the resulting inertial
stresses (of order (ρ̂H + ρ̂L )Vi2 ) are balanced, i.e.,
δ Aspect ratio L̂/D̂ 12.07 √
DR1 Diameter ratio 1 D̂in,1 /D̂ 0.34 (ρ̂H − ρ̂L )ĝ D̂ ∼ (ρ̂H + ρ̂L )V̂i2 ⇒ V̂i = Atĝ D̂. (2)
DR2 Diameter ratio 2 D̂in,2 /D̂ 0.67
On the other hand, the characteristic buoyant-viscous velocity, V̂v ,
V̂0 D̂
Re Reynolds number 1000−11 200 can be defined if the buoyant stresses and the resulting viscous
v̂ μ̂V̂
ρ̂H − ρ̂L stresses (of order D̂v ) are balanced, i.e.,
At Atwood number 0.001, 0.005, 0.01
ρ̂H + ρ̂L
V̂0 μ̂V̂v (ρ̂H − ρ̂L )ĝ D̂2 2Atĝ D̂2
Fr Froude number √ 2−15 (ρ̂H − ρ̂L )ĝ D̂ ∼ ⇒ V̂v = = . (3)
D̂ μ̂ υ̂
Atĝ D̂
V̂0 D̂ One can expect that, in general, our flows must be governed by these
Pe Péclet number ∼ 105
D̂m three characteristic velocities, i.e., V̂0 , V̂i , and V̂v , or their combina-
β Inclination angle From horizontal 8, 15, 30, 45○ tions. Our dimensionless groups can also be simply written vs these
characteristic velocities, i.e.,

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-5


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

V̂0 D̂ V̂0 V̂v


Re = ≡ , (4)
υ̂ 2V̂i2

V̂0 V̂0
Fr = √ ≡ , (5)
Atĝ D̂ V̂i
2
ρ̄ˆ (ρ̂H − ρ̂L )ĝ D̂3 1 V̂v
Ar = ≡ ( ) . (6)
μ̂2 2 V̂i
To generalize our results, in Sec. IV, most of them are presented
using the main dimensionless numbers, i.e., Re, Ar, Fr, and β. The
variation of these parameters can also be viewed in the context of the
characteristic velocities explained, which may bring further insights
into the physics of the flow. For example, for a fixed Reynolds num-
ber (e.g., fixed V̂0 ), the variation in Fr implies that the characteristic
buoyant-inertial velocity (V̂i ) changes; on the other hand, increasing
Re for a fixed Ar suggests that the imposed inertia increases in the
flow, and increasing Ar for a fixed Fr implies that the characteris-
tic buoyant-viscous velocity (V̂v ) increases. In Sec. IV, the lengths,
velocities, and times are made dimensionless by dividing them with
D̂, V̂0 , and D̂/V̂0 , respectively.

IV. RESULTS AND DISCUSSION


We now present our experimental results, starting with intro-
ducing the general observations and then describing the main flow FIG. 3. Experimental snapshots of the flow development observed in a typical
experiment. The flow parameters are Re = 8700, Fr = 11.82, and β = 15○ . From
stages observed in our experiments. Then, we present some of the
top to bottom, t = [4.49, 32.31, 52.96, 104.13, 148.11, 157.98, 171.49, 204.66,
critical mixing behaviors. Finally, we clarify the effects of the flow and 320.47]. Unless otherwise stated, the dimensionless field of view is 1 × 12,
parameters on the stability of the interface and mixing between the and the flow direction is from right to left, here and elsewhere.
two fluids.

A. General observations B. Flow stages


Our general observations of the flow behavior of the two flu- In Sec. IV A, we observed qualitative results on the flow devel-
ids are presented in this section. Approximately 120 experiments are opment, which we can classify into four flow stages:
conducted to cover a wide range of experimental parameters, includ- Stage (i): initial jet— the heavier fluid enters the domain as a jet.
ing the density difference, the injection flow velocity, and the inclina- Stage (ii): mixing region— a mixing region is developed.
tion angle. In each experiment, as the gate valve is opened, the dark
heavy fluid is injected into the flow domain and penetrates into the
in-place transparent light fluid. Figure 3 depicts the image sequence
of a typical experiment, with time increasing from top to bottom.
Initially, by entering the heavy fluid into the domain, a buoyant jet
is created, and then, a mixing region is developed. Then, the heavy
fluid starts to slump, on the pipe lower wall, underneath the light
fluid, and it advances toward the pipe end. As soon as the front of
the heavy fluid reaches the pipe end, it hits the walls, rises upward,
and then moves in the reverse direction. The returning heavy fluid
moves toward the outlet boundary, while the injected heavy fluid
passes on the pipe lower wall, and the light fluid is pushed out of the
flow domain simultaneously.
Figure 4 shows the location of the heavy fluid front vs time
along the flow domain, from the inlet to the pipe end, for different
values of Fr. As can be seen for each value of Fr, first the heavy fluid
front accelerates and quickly moves over a considerable distance,
thanks to the initial jet strength; then, the front slows down and FIG. 4. Heavy fluid front position along the pipe vs time from the beginning of the
finally moves nearly steadily. Decreasing Fr (i.e., increasing buoy- flow until the heavy fluid front reaches the pipe end. Different symbols represent
Fr = 11.82 (▸), Fr = 5.07 (⧫), and Fr = 3.54 (●). The other flow parameters are
ancy) results in achieving the steady front velocity more rapidly and
β = 15○ and Re ≈ 8400.
reaching the pipe end faster.

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-6


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

The injected fluid has the initial momentum due to the imposed
velocity, but there is also a buoyant force due to the density differ-
ence between the fluids. As the heavy fluid enters the flow domain,
these sources of energy create a jet, known as a buoyant jet.
In Fig. 6, the early stage of the flow, i.e., the initial jet, is
depicted for different values of Re. From left to right in each fig-
ure, the value of Re increases, and as the value of Ar (representing
buoyancy) is fixed, this implies that the imposed injection inertia
increases. It may be a priori expected that a higher imposed inertia
results in a stronger jet, creating more Kelvin–Helmholtz type insta-
bilities48 and consequently more efficient mixing. For Re = 1200,
FIG. 5. Experimental snapshots of the different flow stages observed in our experi-
ments. From top to bottom, flow stages (i)–(iv) are shown at t = [3.78, 19.55, 90.24, as shown in Fig. 6(a), the jet flow is not completely disturbed, and
and 133.85], respectively. The flow parameters are Re = 7300, Fr = 9.93, and as the jet advances, the transverse mixing between the two fluids
β = 15○ . remains limited. At a higher Reynolds number shown in Fig. 6(b),
i.e., Re = 3000, the mixing between the fluids is stronger. For Fig. 6(c)
with Re = 6200, the flow is quite turbulent during the advancement
Stage (iii): slumping region— the heavier fluid slumps under- of the jet, and by increasing the Reynolds number to Re = 8700,
neath the light fluid on the pipe lower wall and moves the mixing and turbulency increase [see Fig. 6(d)]. Considering
toward the end of the pipe. these results, we can crudely consider Re ≈ 3000 for the transi-
Stage (iv): front reaching the pipe end and returning— the heavy tion to a highly turbulent jet, with efficient transverse mixing, in our
fluid front reaches the end of the pipe, hits the pipe experiments.
end wall, and then returns back toward the mixing Our findings in terms of the initial jet behaviors are qualita-
region. tively in agreement with those of previous studies. In most studies, it
Figure 5 shows the experimental snapshots of each flow stage. has been found that the jet transitions to the turbulent one when the
In what follows, we distinguish between these stages and describe flow Reynolds number surpasses 2000.6 Fischer et al.49 showed that
the flow behavior of each stage with the help of the dimensionless for the Reynolds number less than 4000, the turbulent jet may not be
parameters. fully developed.
Before we proceed to Sec. IV B 1, note that classifying the flow Due to the inner pipe thickness in our experiments, the initial
stages improves our understanding about the flow and facilitates jet has a small offset with respect to the pipe lower wall (as schemati-
quantifying the key flow features. This has also been done in previ- cally illustrated in Fig. 7). Thus, before touching the pipe lower wall,
ous studies, e.g., that by Kaloudis et al.47 who numerically studied the the heavier fluid has to travel a distance (with respect to the inlet),
problem of constant fluid injection into a confined space (i.e., a two- which we call Xa . After this distance, the heavy fluid is “attached” to
dimensional open channel). They analyzed the interaction of the the pipe lower wall while advancing into the light fluid. This distance
flow with the solid boundaries of the restricted propagation domain, (Xa ) can affect the removal efficiency and mixing between the fluids.
and they defined different observed flow stages. For buoyant flows Our results show that increasing the density difference or the incli-
on an inclined surface, Dai and Huang32 also observed three dis- nation angle results in a decrease in Xa , while changing the imposed
tinct flow stages after releasing a heavy fluid from behind a lock; the injection velocity does not much affect Xa . To illustrate √ this in a
heavy fluid moves in an acceleration phase, reaching a maximum dimensionless form, Fig. 8 shows the variation of Xa vs Ar sin β for
front velocity, followed by a deceleration phase. √ of Re superimposed using the color and size of the
different values
symbol. As Ar sin β increases (stronger buoyancy), Xa decreases,
1. Initial jet while changing Re √ has minor effects on Xa . The specific form of the
At the very first stage, the heavy fluid is injected into the flow control parameter Ar sin β can be better understood by rewriting
domain where the light fluid is placed, forming a jet-like behavior. it in the following form:

FIG. 6. Illustration of the buoyant jet using the snapshots of experiments at β = 15○ and Ar = 1 × 106 for (a) Re = 1200 and [ttop , tbottom ] = [1.18, 1.96], (b) Re = 3000 and
[ttop , tbottom ] = [2.81, 4.68], (c) Re = 6200 and [ttop , tbottom ] = [5.80, 9.66], and (d) Re = 8700 and [ttop , tbottom ] = [8.10, 13.50]. The dimensionless field of view in each snapshot
is 1 × 3.9, right below the inlet.

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-7


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 7. Schematic of the early phase of


the initial jet. Due to the inner pipe thick-
ness, the initial jet has an offset with
respect to the flow domain lower wall.


√ √ (V̂i sin β)D̂ Analyzing and interpreting the results available in the literature
Ar sin β ≡ 2 , (7) can help estimate the value of Xa for the corresponding iso-dense
υ̂
flows as ∼ 0.87,50 which in comparison with our results reveals that

which is basically a Reynolds number defined using V̂i sin β as the presence of the density difference reduces Xa . Thus, we may
the relevant velocity scale, which can be simply derived by balanc- simply assume that the iso-dense attachment distance is the leading
ing the characteristic longitudinal buoyant stress [(ρ̂H − ρ̂L )ĝ D̂ sin β] order term in the expansion of Xa , for example, in the form of
with the resulting inertial stress [see
√ Eq. (2) for a similar deriva- √ √ f ′′ (0)
tion]. To explain why increasing Ar sin β decreases Xa , consider Xa = f ( Ar sin β) ≈ f (0) + Ar sin βf ′ (0) + Ar sin β +⋯,
2
the lower interface near the inlet schematically depicted in Fig. 7, (8)
where the heavy fluid is initially placed above the light one near the in which f (0) ≈ 0.87, independent of Ar and β. With this ansatz, we
inlet: in this specific region (with the length of Xa ), the longitudi- use our experimental data to fit the coefficients f ′ (0) and f ′′ (0) in
nal component of the buoyant force is resistive to the expansion
√ of the equation above to arrive at
the interface, and the Reynolds number in the form of Ar sin β √
quantifies such a resistance. Note that this is in contrast to the other Xa ≈ 0.87 − 0.44 × 10−3 Ar sin β + 0.83 × 10−7 Ar sin β + ⋯ , (9)
flow regions where the heavy fluid is placed below the light one
showing an acceptable comparison with the data in Fig. 8. With-
and therefore helps the longitudinal expansion of the interface or
out specifying the value of f (0) in Eq. (8), the following fit can be
the mixing zone. As√shown in Fig. 8, increasing the Reynolds num-
obtained:
ber in the form of Ar sin β gives rise to the longitudinal buoyant √
force that locally resists the expansion of the interface, leading to a Xa ≈ 0.78 − 0.30 × 10−3 Ar sin β + 0.41 × 10−7 Ar sin β + ⋯ ,
decrease in Xa . (10)
using which the comparison with the data in Fig. 8 slightly improves.
As our buoyant jet enters a confined environment, its profile
depends on the flow parameters, as shown in Fig. 9. In Fig. 9(a),
by increasing Re, the height of the jet increases. This is because a
jet with a higher inertia results in stronger instabilities and more
efficient transverse mixing.51 In addition, by increasing Re, the jet
inertia along the pipe increases with respect to the transverse buoy-
ant force that pushes the heavy fluid toward the pipe lower wall and
limits the transverse mixing. These effects together lead to a rise in
the jet height. In terms of the effects of increasing the jet inertia on
the flow, a similar behavior was observed by Kaloudis et al.47 whose
work considers an injected heavy fluid propagating within a deep
rectangular storage tank. In Fig. 9(b), the effect of Ar is illustrated. As
Ar increases, the height of the jet decreases. Note that here the trans-
verse buoyant stress is of order (ρ̂H − ρ̂L )ĝ D̂ cos β, which experiences
a relative increase by increasing Ar at a fixed Re. Consequently, by
increasing the transverse buoyant stress, the heavy fluid jet is fur-
√ ther driven toward the pipe lower wall, leading to shortening the jet
FIG. 8. Variation of the attachment distance, Xa , vs Ar sin β. The symbol colors height.
and sizes represent the values of Re. The solid and dashed lines present the To obtain the velocity profiles, the UDV probe is placed near
results of Eqs. (9) and (10), respectively.
the inlet at an angle 75○ relative to the pipe axis, as shown in

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-8


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 9. Change in the jet profile: (a) effect of Re at constant Ar and (b) effect of Ar at constant Re. In both figures, β = 15○ . The dimensionless field of view in each snapshot
is 1 × 3.9, right below the inlet.

Fig. 10(a). The velocity contours along the pipe diameter are shown the inlet. The longitudinal extent (or length) of the mixing zone
in Fig. 10(b). This figure shows the variation of the velocity within depends on the flow parameters, as illustrated in Fig. 11, showing
both fluids. The position of the pseudo interface between the two flu- the length of the mixing region in different conditions. In Fig. 11(a),
ids is marked by the black line. The position of the zero axial velocity, the effect of Ar on the length of the mixing region is shown. As
identified via the contour of zero velocity, is marked by the red line. Ar increases, the mixing region extends, implying that increasing
Below the red line, the fluid velocity is positive, implying that the the buoyancy force enhances the mixing in the longitudinal direc-
fluid moves downward, and above the red line, the fluid velocity is tion. To justify this observation, first note that, in addition to the
negative, implying that the fluid advances upward (i.e., the zero net initial imposed inertia of the jet, the mixing region can expand
flux condition). along the pipe length due to the longitudinal buoyant stress, which
There are a few studies in the literature making observations is of order (ρ̂H − ρ̂L )ĝ D̂ sin β. Therefore, increasing Ar implies that
similar to ours in terms of the flux conservation explained above, the buoyancy force driving the flow in the lengthwise direction
an example of which is an injection flow study by Longo et al.38 increases, which in return longitudinally expands the mixing region.
They considered the problem of a heavy fluid injected into a con- Figure 11(b) shows that, on the other hand, the variation in Re does
fined horizontal channel of circular cross section (partial inlet and not seem to have a significant effect on the mixing region length,
open-end). By measuring the horizontal velocity profiles using an particularly for Re > 3000. This is because, at large Re, the effec-
Ultrasonic Velocity Profiler (UVP), they observed that a returning tive penetration length of the initial jet and the extent of the sub-
flow in the horizontal direction fulfills the mass conservation. sequent mixing zone become independent of Re. In fact, although
a larger Re initially induces a stronger jet, it also causes the fluids
2. Mixing region to mix more efficiently in the transverse direction, which counter-
The second stage of the flow is associated with the develop- balances the longitudinal expansion of the resulting mixing zone.
ment of a mixing region. After the initial jet stage, the fluids effec- In addition, as time grows, the longitudinal buoyant force starts
tively mix in the transverse and axial directions, and the advance- to take over the imposed inertia in governing the expansion of the
ment of the heavy fluid creates a mixing zone. This mixing zone mixing zone away from the inlet. These simultaneous effects lead
overshadows the initial jet, and it occupies a large area close to to a negligible impact of Re on the mixing zone length. Finally, as

FIG. 10. UDV measurements obtained


for Re = 4840, Fr = 2.06, and β = 30○ :
(a) experimental snapshots from top to
bottom at t = [6.00, 9.93, 19.86, 25.83,
29.80, and 38.00] and (b) contours of
near inlet velocity vs time along the pipe
diameter. The zero axial velocity bound-
ary is marked by the red line. The black
line shows the position where the nor-
malized concentration is equal to 0.5,
i.e., the pseudo interface position.

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-9


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 12. Heavy fluid unsteady front velocity, Vft , vs time from the beginning of the
flow until the heavy fluid front reaches the pipe end. Different symbols represent
Fr = 11.82 (▸), Fr = 5.07 (⧫), and Fr = 3.54 (●). The solid symbols are for
β = 15○ , and the hollow ones are for β = 45○ . Re ≈ 8400 for all experiments.

part of the fluid flow, quantifying its features, such as the heavy fluid
front velocity, is of importance. For example, the value of the heavy
fluid front velocity can represent the removal efficiency of the light
fluid by the heavy fluid.
Figure 12 illustrates the heavy fluid unsteady front velocity,
Vft , vs t. The front velocities are initially large, with values close
to Vf ≈ 1, implying that the front velocities have nearly the same
values as those of the initial jet. At longer times, however, when
the front enters the slumping region, the front velocities eventually
reach nearly constant values. This figure also shows that by increas-
ing the buoyancy (decreasing Fr), the front velocities in the slumping
region increase.
To quantify the dependency of the long-time front velocity on
the flow parameters, let us momentarily use the dimensional quanti-
FIG. 11. Change in the length of the mixing region vs different flow parameters: (a)
Re = 8780 and β = 45○ at different Ar, (b) Ar = 1 × 106 and β = 45○ at different
ties and plot Fig. 13, showing the variation of the dimensional heavy
Re, and (c) Ar = 1 × 106 and Re = 6700 at different β. The box roughly marks
the extent of the mixing region in each snapshot.

Fig. 11(c) illustrates, tilting the pipe toward the horizontal direc-
tion reduces the extent of the mixing region. This is simply because
decreasing the inclination angle directly reduces the characteristic
longitudinal buoyant stress, i.e., (ρ̂H − ρ̂L )ĝ D̂ sin β, thus significantly
limiting the flow driving force that expands the mixing zone, par-
ticularly away from the inlet. For example, by decreasing β from
45○ to 8○ in Fig. 11(c), the characteristic longitudinal buoyant stress
experiences a fivefold decrease, limiting the mixing zone lengthwise
expansion.
3. Slumping region
After developing the mixing region, the flow reaches the slump-
ing stage: the heavy fluid front slumps underneath the light fluid, it
advances on the pipe lower wall, and it moves toward the pipe end. FIG. 13. Variation of the heavy fluid front velocity, V̂f , vs the injection velocity, V̂0 .
Different marker colors represent At = 0.001 (green), At = 0.005 (red), and
In this stage, although the shape of the interface changes with time At = 0.01 (blue). Different marker shapes represent β = 45○ (◂), β = 30○ (▸),
and space, the heavy fluid front maintains a nearly constant velocity β = 15○ (⧫), and β = 8○ (●). The dotted lines show a plateau value for V̂f .
and almost a constant height. Since this flow stage forms the longest

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-10


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

fluid front velocity in the slumping region for different injection Note that these equations are only a function of Fr, independent of
velocities, density differences, and inclination angles. As can be seen, Re and β.
the injection velocity and the inclination angle do not seem to have Interestingly, Eq. (11) is in good agreement with the findings
a significant effect on the dimensional front velocity of the heavy of Ref. 52, which analyzed the interpenetration front dynamics of
fluid. However, by increasing the density difference (which improves the lock-exchange flows in a pipe, finding that the front veloc-
the buoyancy force), the dimensional front velocity increases. As ity follows V̂f = 0.7V̂i . This emphasizes the similarities between
the buoyancy force balanced by inertia mainly governs the front the front dynamics of the slumping stage of our flow with that
velocity in this flow stage, it is natural to use the characteristic of the lock-exchange flows, despite obvious differences in the flow
buoyant-inertial velocity, V̂i , to describe the flow, configuration.
V̂f ≈ f (V̂i ), 4. Front reaching pipe end and returning
At the end of the slumping stage, the heavy fluid eventually
which in the dimensionless form can be written as
reaches the pipe end and then returns back toward the mixing zone.
V̂f V̂i 1 The first question that may arise is when the front reaches the pipe
Vf = ≈ f( ) ≡ f ( ). end. As we have seen so far that the flow dynamics is mainly dom-
V̂0 V̂0 Fr
inated by the balance between buoyancy and inertia, for which V̂i
Based on this argument, Fig. 14 shows the variation of the long-time is relevant, it is natural to expect that the dimensional time of the
dimensionless heavy fluid front velocity, Vf , vs Fr for all experi- front reaching the pipe end (t̂e ) is a function of the characteristic
ments. The effects of the other parameters, i.e., Re and β, are also buoyant-inertial time defined as
included in this figure: the size and color of the symbols represent the
values of β and Re, respectively. This figure reveals that Fr is in fact t̂i = D̂/V̂i ,
the main parameter affecting the dynamics of the heavy fluid front leading to
in the slumping stage. As can be seen, Vf decreases by increasing Fr,
and it can be roughly predicted by the fitted line, t̂e ≈ f (t̂i ),

0.68 1 which in dimensionless form becomes


Vf ≈ + O( 2 ), (11)
Fr Fr t̂e ⎛ (D/V̂i ) ⎞ V̂0
te = ≈f ≡ f ( ) ≡ f (Fr).
while the agreement with the data can slightly and progressively (D/V̂0 ) ⎝ (D/V̂0 ) ⎠ V̂i
improve if the higher orders of 1/Fr are considered for fitting,
Figure 15 plots the variation of the dimensionless time of the
0.80 0.54 1 front reaching the pipe end (te ), with respect to the beginning of the
Vf ≈ − 2 + O( 3 ), (12)
Fr Fr Fr experiments, against Fr for different values of β and Re. The figure
shows that generally with increasing Fr (i.e., decreasing buoyancy),
1.05 2.71 4.21 1 it takes more time for the front to reach the pipe end. As can be
Vf ≈ − 2 + 3 + O( 4 ). (13) seen, all the data roughly coincide on a straight line with a slope
Fr Fr Fr Fr
of 14.58, i.e.,
te ≈ 14.58Fr. (14)

FIG. 14. Variation of the long-time front velocity (nearly steady), Vf , vs Fr shown
by the filled triangles. The symbol color represents the values of Re, and the
symbol size represents the values of β (e.g., the largest symbols correspond to
β = 45○ ). The black solid line, the red dotted line, and blue dashed line mark FIG. 15. Time of heavy fluid reaching pipe end, te , vs the Froude number, Fr.
the result of Eqs. (11)–(13), respectively. The backflow front velocities, Vf ,b , are The symbol color represents the values of Re, and the symbol size represents the
also superimposed using the hollow circles and thicker error bars, and they are values of β (e.g., the largest symbols correspond to β = 45○ ). The fitted line marks
discussed later in Sec. IV B 4. the result of Eq. (14).

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-11


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

To further analyze the flow behavior in the fourth stage, Fig. 17


plots the spatiotemporal diagrams of the depth-averaged concentra-
tions, in the plane of x and t, covering the flow from the first stage to
the last stage. The front is discernible on these diagrams by a sharp
boundary between the different comparative concentrations of the
two fluids. As can be observed, when the front turns back toward the
mixing region, it has a nearly constant propagation velocity, which is
inversely proportional to the slope of the lines on the spatiotempo-
ral diagrams. It can also be seen in Fig. 17 that by decreasing Fr, the
FIG. 16. Snapshots of the front propagation near the pipe end (t = 65), impact backflow velocity (Vf ,b ) increases. Figure 18 shows the same type of
with the pipe end wall (t = 76), and turning back (t = 89). The flow parameters
results as in Fig. 17 but for two different inclination angles, revealing
are Fr = 8.94, β = 8○ , and Re = 6587.
that changing β does not much affect Vf ,b as the values of Vf ,b have
only 4% difference, i.e., within the experimental error.
Similar to the slumping stage, in the last flow stage, the front
Figure 16 shows a typical example of the flow development near velocity has a major role in quantifying the removal of the in situ
the last stage of the flow for Fr = 8.94, Re = 6587, and β = 8○ , light fluid from the flow domain. Thus, it is of interest to quantify
showing that the heavy fluid front reaches the pipe end (t = 65), Vf ,b vs all the flow parameters. We have so far seen that Vf ,b is pri-
rises up on the pipe end wall (t = 76), and, after the impact, pene- marily a function of the buoyant force for which the characteristic
trates into the light fluid, on the upper wall of the pipe, as a backflow buoyant-inertial velocity, V̂i , or the Froude number becomes rele-
sets in (t = 89). As time progresses, the heavy fluid front goes back vant. Thus, it is natural to examine that if the relations developed
toward the mixing region (and eventually toward the inlet/outlet earlier for the velocity of the forward moving front are also valid for
boundaries). that of the backward moving front. For a large number of exper-
In relation to our work, Kaye and Hunt53 studied the impact iments, the values of Vf ,b are superimposed as hollow symbols on
of a heavy fluid on the flow geometry wall for different aspect ratios Fig. 14 in the plane of Vf ,b vs Fr. The size and color of the sym-
(width/height) of the filling box flow problem. They reported that, bols also represent the values of β and Re, respectively. In addition,
for small aspect ratios, the fluid front rises up and turns back when note that the error bars of Vf ,b are larger than those of Vf due to
reaching the wall; otherwise, the heavy fluid front slumps back with- the uncertainty in the front position due to mixing. Overall, this
out mixing with the light fluid, creating waves on the interface figure suggests a reasonable comparison of the backflow data with
between the fluids. Eqs. (11)–(13).

FIG. 17. Spatiotemporal images of the


depth-averaged concentrations for β
= 15○ and Re = 8000: (a) Fr = 10.65,
(b) Fr = 4.79, and (c) Fr = 2.89. The
blue dashed line in each figure indicates
the initial boundary between the heavy
and light fluids after the front reaches
the pipe end and turns back, with the
slope representing 1/Vf ,b . The color bar
represents the normalized concentration
values, here and elsewhere.

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-12


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 18. Spatiotemporal images of the


depth-averaged concentrations for Fr
= 8.90 and Re = 6600: (a) β = 8○
and (b) β = 45○ . The dashed blue
line in each figure indicates the boundary
between the heavy and light fluids after
the front reaches the pipe end and turns
back, with the slope representing 1/Vf ,b .

The results of Fig. 14 reveal that the forward and backward along the depth of the pipe spreads differently from what is observed
moving front velocities are not much different; this implies that as on the side view.
the forward moving front hits the wall at the end of the slump- Let us use the LIF technique to shed light on some of the mixing
ing region and returns toward the mixing zone, its velocity does flow features. For example, revisiting the spatiotemporal diagrams of
not much change initially. Of course, as the backward moving front the depth-averaged concentration in Figs. 17 and 18, we can notice
advances away from the end wall, the mixing between the two flu- “curvy streaks” corresponding to some flow features within the mix-
ids significantly increases, leading to a considerable decrease in the ing region near the inlet; these streaks can be recognized by the
effective buoyancy force that drives the flow, resulting in decreasing continuous dark and light alternations in the depth-averaged con-
the backflow front velocity. Much further away from the end wall, centration in the mixing region, forming thin, oblique, and slightly
as the backflow approaches the mixing zone, mixing dominates and curvy lines. Our hypothesis is that these streaks are in fact narrow
the flow becomes entirely diffusive, while the front fades away (see, mixing layers that move upward as their slopes are positive in the
for example, Fig. 18(a). plane of t and x. To better illustrate these mixing layers, Fig. 20
shows the concentration field spatiotemporal diagrams, obtained via
C. Perpendicular mixing the top view images of the LIF tests (upper sheet). To concentrate on
the regions near the inlet, the x-axis covers the pipe length only from
To analyze the perpendicular mixing between the two fluids,
the inlet to the middle of the pipe. In terms of their overall appear-
i.e., mixing in the planes perpendicular to the field of view, three
ance, the spatiotemporal diagrams based on the side and top view
different laser sheets are used, as illustrated in Fig. 19(a) (note that
images (respectively, corresponding to camera and LIF results) are
black-and-white images present the side view, and color images
similar (compare Fig. 20 with Fig. 18). However, it must be noted
present the LIF results throughout this study). These deliver the
that the LIF results capture and present the mixing layers obtained
concentration and mixing in three different depths of the pipe: one
over a thin perpendicular laser sheet.
coincides with the axis of the injection pipe, the second one is on the
As can be observed in Fig. 20, increasing the inclination angle
axis of the main pipe, and the last one is used in the upper region of
results in raising both the concentration values and the longitudi-
the main pipe. The images in this figure show the normalized con-
nal extent of the curvy streaks (or mixing layers) along the upper
centration field on the three laser sheets for the time when the front
section of the mixing region; however, the speed of the upward-
reaches the pipe end. In Fig. 19(b), the same experiment is illustrated
moving mixing layers (represented by the slopes of the dashed lines)
from the side view, based on the camera snapshot. As can be seen,
only slightly decreases by increasing β. The reason for the upward
when the heavy fluid propagates along the pipe, its concentration
motion of the mixing layers can be easily justified: as the flow geom-
etry end is closed, the zero net flux condition across each pipe cross
section must be respected. Thus, as the heavy fluid is injected into
the flow domain, loosely speaking, there must be an equal amount
of the light fluid or its mixture with the heavy fluid that moves in the
opposite direction; this is to compensate for the injected heavy fluid
coming into the flow domain. Therefore, the mixing layers move
upward to balance the injected volume of the heavy fluid. As the
injection is constant and continuous, the speed of the mixing lay-
ers near the inlet (represented by the superimposed oblique dashed
lines on Fig. 20) does not seem to much vary with time. Neverthe-
FIG. 19. (a) Normalized concentration field at three horizontal sheets at Fr = 2.70, less, away from the inlet and the intense mixing region, the mixing
Re = 4100, and β = 30○ . (b) Side view of an experiment with the same condition layers either disappear or their velocities become unsteady. Our find-
of (a).
ings show that, in general, the speed of these mixing layers seems to

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-13


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 20. Spatiotemporal diagrams of the


heavy fluid concentrations on the upper
sheet of the LIF tests, in the plane of
x and t, covering the flow from the first
stage to the middle of the flow domain at
Fr = 5.88 and Re = 9700: (a) β = 8○
and (b) β = 45○ . The dashed lines
represent the speed of the mixing layers.

increase with an increase in the imposed velocity and the density lead to efficient and rapid mixing. Some of the observed insta-
difference or a decrease in the inclination angle (results omitted for bilities originate from the high injection rate, and some are due
brevity). to the velocity gradients along with the density difference. While
Another slight observation can be made using Fig. 20. The lon- the stability of two-fluid interfaces has been quantitatively studied
gitudinal extent of the mixing region at smaller β is shorter than for many years, dating back to the efforts of Rayleigh, Helmholtz,
that at larger β. This was earlier observed in the side view images Kelvin, and Reynolds in the 19th century,54 here, let us have a
of Fig. 11(c), and the LIF results here also confirm that when β qualitative look on some of the interfacial instabilities seen in
decreases, the flow enters the slumping region at a smaller x. our experiments to provide a better understanding of the flow
Figure 21 shows the time-averaged, depth-averaged concen- dynamics.
tration field for the middle laser sheet vs the axial direction from In our experiments, the flow in regions near the inlet presents a
the beginning of the flow until the front reaches the pipe end. In completely mixed region, while in the slumping region, the interface
Fig. 21(a), the results are given for different inclinations. It can be between two fluids becomes visible due to the segregation between
seen that as the pipe is further inclined toward the vertical direction, two fluids. However, during the slumping stage, the interface still
the average concentration of the heavier fluid along the depth of the exhibits unstable behavior as interfacial perturbations grow remark-
pipe decreases. In Fig. 21(b), the results are given for different buoy- ably, resulting in partial transverse mixing between the two fluids.
ancy values (i.e., different values of Fr), showing a trend similar to To qualitatively analyze the instability and mixing between the flu-
that of β. ids, Fig. 22 illustrates images of the flow development at successive
times for different values of Fr. It can be seen that as Fr decreases, the
interface becomes destabilized to a greater extent. Since the heavier
D. Stability of interface fluid turns back after reaching the end of the pipe, the two flu-
Instabilities in our flows of interest usually appear in the form ids mix further, and the interface between them becomes gradually
of waves at the interface between the fluids, which may eventually dispersed.

¯
FIG. 21. (a) Time-averaged depth-averaged concentration field, C̄z,x,t , vs the streamwise location, x, from the beginning of the flow until the front reaches the pipe end,
measured at Fr = 6.53 and Re = 11000. Different markers represent β = 45○ (black filled left pointed triangle), β = 30○ (green filled right pointed triangle), β = 15○ (blue
filled diamond), and β = 8○ (red filled circle). The solid arrow shows the increase in β as a guide to the eye. (b) The same as (a) but at Re ≈ 7300 and β = 30○ . Different
markers represent Fr = 9.73 (green filled right pointed triangle), Fr = 4.67 (red filled diamond), and Fr = 3.02 (blue filled circle). The solid arrow shows the increase in Fr
as a guide to the eye.

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-14


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 22. Instability and mixing for different values of Fr at β = 45○ and Re = 4700. From top to bottom, Fr = 6.43, Fr = 3.09, and Fr = 2.05, respectively. In each row,
from right to left, t = [3.05, 18.27, and 33.55], and the last (left) image corresponds to the time when the heavy fluid front reaches the pipe end.

FIG. 23. Instabilities and mixing between


the fluids, at the moment when the heavy
fluid front reaches the pipe end, at dif-
ferent inclination angles for Fr = 3.25
and Re = 5500: (a) camera images (side
view) and (b) upper sheet view of the LIF
tests (top view).

In addition to buoyancy (quantified through Fr), the inclina- non-intrusive methods, such as high-speed imaging as well as UDV
tion of the pipe seems to also have a significant effect on the segre- and LIF techniques, are used to provide detailed characterizations of
gation and, consequently, on the mixing between the two fluids, as the flow.
illustrated in Fig. 23(a). Tilting the pipe away from the horizontal In general, during the flow development, four flow stages can
direction destabilizes the interface, creates Kelvin–Helmholtz insta- be identified: (i) initial jet, (ii) mixing region, (iii) slumping region,
bilities, and increases the mixing. In order to better show the effects and (iv) heavy fluid front reaching the pipe end and returning
of Kelvin–Helmholtz instabilities on the perpendicular mixing, the toward the mixing region. To have an overall understanding, these
LIF results corresponding to the upper horizontal laser sheet are flow stages are presented in a typical spatiotemporal diagram of
shown in Fig. 23(b). As expected, at near horizontal inclinations, the depth-averaged concentration field in Fig. 24. The flow features
we observe that Kelvin–Helmholtz instabilities are less effective than are quantified in this work for each flow stage, e.g., via character-
those at higher inclinations. izing the flow behaviors, heavy fluid front velocity, and mixing, vs
In relation to our work, the injection flow of two miscible fluids the governing dimensionless numbers, e.g., the Reynolds number
in inclined channels has been investigated by Sahu et al.55 Through (Re), the Archimedes number (Ar), the Froude number (Fr), and the
direct numerical simulations, they found that, in close to horizon- inclination angle (β).
tal inclinations, the fluids are distinctly segregated, even at notable As stated in the Introduction, one of our motivations for this
inertial and buoyant forces. In addition, by increasing the inclina- work is to better understand the fluid flows in the dump bailing
tion angle, inertial effects become stronger, and interfacial insta- method in P&A processes of oil and gas wells. From the indus-
bilities may grow. These findings are in qualitative agreement with trial perspective, reducing mixing between the process fluids while
ours. increasing the removal efficiency of the in-place fluid by the injected
fluid is desired. In this context, our experiments have revealed that
most mixing occurs at the early stages of the flow near the inlet
V. CONCLUSIONS boundary, the extent of which is reduced by a decrease in the incli-
Using an experimental approach, we have investigated the nation angle or buoyancy (the density difference). Comparatively,
problem of the injection of a heavy fluid into a light one in a closed- the minimum mixing between the fluids is seen in the slumping
end pipe. The injection of the heavy fluid is made via an inner stage, in which again decreasing the inclination angle and buoyancy
pipe with a diameter smaller than that of the closed-end pipe. The can reduce the transverse mixing. Moreover, the removal efficiency
two fluids are miscible, and the injection is continuous. Various of the in-place fluid by the injected one is particularly related to

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-15


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

7
N. E. Kotsovinos and E. J. List, “Plane turbulent buoyant jets. Part 1. Integral
properties,” J. Fluid Mech. 81(1), 25–44 (1977).
8
N. E. Kotsovinos, “Plane turbulent buoyant jets. Part 2. Turbulence structure,”
J. Fluid Mech. 81(1), 45–62 (1977).
9
P. N. Papanicolaou and E. J. List, “Investigations of round vertical turbulent
buoyant jets,” J. Fluid Mech. 195, 341–391 (1988).
10
O. Vauquelin, E. M. Koutaiba, E. Blanchard, and P. Fromy, “The discharge
plume parameter Γd and its implications for an emptying filling box,” J. Fluid
Mech. 817, 171–182 (2017).
11
N. Xue, S. Khodaparast, and H. A. Stone, “Fountain mixing in a filling box at
low Reynolds numbers,” Phys. Rev. Fluids 4(2), 024501 (2019).
12
L. Milton-McGurk, N. Williamson, S. W. Armfield, and M. P. Kirkpatrick,
“Experimental investigation into turbulent negatively buoyant jets using com-
bined PIV and PLIF measurements,” Int. J. Heat Fluid Flow 82, 108561 (2020).
13
A. Assoudi, A. Amamou, N. M. Saïd, and H. Bournot, “Characteristics and anal-
ysis of a turbulent offset jet including the effect of density and offset height,” Int.
FIG. 24. Spatiotemporal diagram of the depth-averaged concentration field, sum- J. Mech. Sci. 174, 105477 (2020).
14
marizing the four flow stages presented in this paper. The flow parameters are M. Van Reeuwijk, M. Holzner, and C. P. Caulfield, “Mixing and entrain-
Re = 10 300, Fr = 6.22, and β = 8○ . ment are suppressed in inclined gravity currents,” J. Fluid Mech. 873, 786–815
(2019).
15
A. Etrati and I. A. Frigaard, “A two-layer model for buoyant inertial displace-
ment flows in inclined pipes,” Phys. Fluids 30(2), 022107 (2018).
16
S. Balasubramanian and Q. Zhong, “Entrainment and mixing in lock-exchange
gravity currents using simultaneous velocity-density measurements,” Phys. Fluids
the heavy fluid front velocity (Vf ), which is mainly linked to the 30(5), 056601 (2018).
strength of the buoyancy force quantified via Fr (or the character- 17
D. Sher and A. W. Woods, “Mixing in continuous gravity currents,” J. Fluid
istic buoyant-inertial velocity, V̂i ); in this case, simple relations are Mech. 818, R4 (2017).
presented to predict Vf vs Fr. Finally, our results suggest that a simi- 18
T. Du, D. Yang, H. Wei, and Z. Zhang, “Experimental study on mixing and
lar buoyancy-inertia competition is largely responsible for governing stratification of buoyancy-driven flows produced by continuous buoyant source
the backflow velocity in the last identified flow stage. in narrow inclined tank,” Int. J. Heat Mass Transfer 121, 453–462 (2018).
19
Future directions of the current work can include analyzing S. M. Taghavi, T. Séon, K. Wielage-Burchard, D. M. Martinez, and I. A. Frigaard,
“Stationary residual layers in buoyant Newtonian displacement flows,” Phys.
the effects of large density differences (i.e., the non-Boussinesq
Fluids 23(4), 044105 (2011).
limit), the presence of a viscosity ratio between the two fluids, and 20
T. Séon, J. Hulin, D. Salin, B. Perrin, and E. Hinch, “Buoyant mixing of miscible
geometrical parameters. fluids in tilted tubes,” Phys. Fluids 16(12), 103–106 (2004).
21
S. M. Taghavi, K. Alba, T. Séon, K. Wielage-Burchard, D. M. Martinez, and I. A.
Frigaard, “Miscible displacement flows in near-horizontal ducts at low Atwood
ACKNOWLEDGMENTS number,” J. Fluid Mech. 696, 175–214 (2012).
22
This research has been carried out at the Univerité Laval. We S. M. Taghavi, T. Séon, D. M. Martinez, and I. A. Frigaard, “Influence of an
gratefully acknowledge the financial support provided by PTAC- imposed flow on the stability of a gravity current in a near horizontal duct,” Phys.
Fluids 22(3), 031702 (2010).
AUPRF via Grant No. PTAC-17-WARI-02 and from NSERC via 23
J. Zheng, W. Zhang, J. Jiang, and R. Guo, “CFD simulation and experimental
CRDPJ 516022-17 (“Plug and Abandon Strategies for Canada’s Oil study of water-oil displacement flow in an inclined pipe,” Int. J. Heat Technol.
& Gas Wells”). 35(3), 663–667 (2017).
24
A. J. Hogg, M. M. Nasr-Azadani, M. Ungarish, and E. Meiburg, “Sustained
gravity currents in a channel,” J. Fluid Mech. 798, 853–888 (2016).
DATA AVAILABILITY 25
J. O. Shin, S. B. Dalziel, and P. F. Linden, “Gravity currents produced by lock
exchange,” J. Fluid Mech. 521, 1–34 (2004).
The data that support the findings of this study are available 26
J. W. Rottman and J. E. Simpson, “Gravity currents produced by instantaneous
within the article. releases of a heavy fluid in a rectangular channel,” J. Fluid Mech. 135, 95–110
(1983).
27
REFERENCES L. Chiapponi, M. Ungarish, D. Petrolo, V. Di Federico, and S. Longo, “Non-
Boussinesq gravity currents and surface waves generated by lock release in a
1
A. Eslami and S. M. Taghavi, “Viscous fingering of yield stress fluids: The effects circular-section channel: Theoretical and experimental investigation,” J. Fluid
of wettability,” J. Non-Newtonian Fluid Mech. 264, 25–47 (2019). Mech. 869, 610–633 (2019).
2 28
M. J. Kaiser, “Rigless well abandonment remediation in the shallow water U.S. R. J. Lowe, J. W. Rottman, and P. F. Linden, “The non-Boussinesq lock-
Gulf of Mexico,” J. Pet. Sci. Technol. 151, 94–115 (2017). exchange problem. Part 1. Theory and experiments,” J. Fluid Mech. 537, 101–124
3
E. Trudel, M. Bizhani, M. Zare, and I. A. Frigaard, “Plug and abandonment prac- (2005).
29
tices and trends: A British Columbia perspective,” J. Pet. Sci. Technol. 183, 106417 R. Inghilesi, C. Adduce, V. Lombardi, F. Roman, and V. Armenio, “Axisymmet-
(2019). ric three-dimensional gravity currents generated by lock exchange,” J. Fluid Mech.
4
E. B. Nelson, Well Cementing (Newnes, 1990), Vol. 28. 851, 507–544 (2018).
5 30
M. Khalifeh and A. Saasen, Introduction to Permanent Plug and Abandonment of A. Dai, “Non-Boussinesq gravity currents propagating on different bottom
Wells (Springer, 2020). slopes,” J. Fluid Mech. 741, 658–680 (2014).
6 31
S. J. Barnett, “The dynamics of buoyant releases in confined spaces,” Ph.D. thesis, A. Dai, “High-resolution simulations of downslope gravity currents in the
University of Cambridge, 1992. acceleration phase,” Phys. Fluids 27(7), 076602 (2015).

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-16


Published under license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

32 43
A. Dai and Y.-l. Huang, “High-resolution simulations of non-Boussinesq S. Longo, M. Ungarish, V. Di Federico, L. Chiapponi, and F. Addona, “Gravity
downslope gravity currents in the acceleration phase,” Phys. Fluids 28(2), 026602 currents in a linearly stratified ambient fluid created by lock release and influx in
(2016). semi-circular and rectangular channels,” Phys. Fluids 28(9), 096602 (2016).
33 44
K. Steenhauer, T. Tokyay, and G. Constantinescu, “Dynamics and structure of T. Séon, “From turbulent mixing to gravity currents in confined geometry,”
planar gravity currents propagating down an inclined surface,” Phys. Fluids 29(3), Ph.D. thesis, Université Pierre et Marie Curie - Paris VI, 2006.
45
036604 (2017). T. Séon, J.-P. Hulin, D. Salin, B. Perrin, and E. J. Hinch, “Laser-induced fluo-
34
K. Alba, S. M. Taghavi, and I. A. Frigaard, “Miscible density-unstable displace- rescence measurements of buoyancy driven mixing in tilted tubes,” Phys. Fluids
ment flows in inclined tube,” Phys. Fluids 25(6), 067101 (2013). 18(4), 041701 (2006).
35 46
A. Amiri, F. Larachi, and S. M. Taghavi, “Buoyant miscible displacement flows S. M. Taghavi, “From displacement to mixing in a slightly inclined duct,” Ph.D.
in vertical pipe,” Phys. Fluids 28(10), 102105 (2016). thesis, University of British Columbia, 2011.
36 47
Z. Zhou, X. Yu, T.-J. Hsu, F. Shi, W. Rockwell Geyer, and J. T. Kirby, “On non- E. Kaloudis, D. G. E. Grigoriadis, and E. Papanicolaou, “Numerical simulations
hydrostatic coastal model simulations of shear instabilities in a stratified shear of constant-influx gravity currents in confined spaces: Application to thermal
flow at high Reynolds number,” J. Geophys. Res.: Oceans 122(4), 3081–3105, storage tanks,” Int. J. Therm. Sci. 108, 1–16 (2016).
48
https://doi.org/10.1002/2016jc012334 (2017). C. Staquet, “Two-dimensional secondary instabilities in a strongly stratified
37 shear layer,” J. Fluid Mech. 296, 73–126 (1995).
Z. Nourmohammadi, H. Afshin, and B. Firoozabadi, “Experimental observa-
49
tion of the flow structure of turbidity currents,” J. Hydraul. Res. 49(2), 168–177 H. B. Fischer, J. E. List, C. R. Koh, J. Imberger, and N. H. Brooks, Mixing in
(2011). Inland and Coastal Waters (Academic Press, New York, 1979).
38 50
S. Longo, M. Ungarish, V. Di Federico, L. Chiapponi, and F. Addona, “Gravity A. Nasr and J. C. S. Lai, “Comparison of flow characteristics in the near field
currents produced by constant and time varying inflow in a circular cross-section of two parallel plane jets and an offset plane jet,” Phys. Fluids 9(10), 2919–2931
channel: Experiments and theory,” Adv. Water Resour. 90, 10–23 (2016). (1997).
39 51
Y. Yuan and A. R. Horner-Devine, “Laboratory investigation of the impact of S. S. Aleyasin, “On the effects of initial conditions on statistical properties of
lateral spreading on buoyancy flux in a river plume,” J. Phys. Oceanogr. 43(12), single and twin turbulent jets,” Ph.D. thesis, University of Manitoba, 2018.
52
2588–2610 (2013). T. Séon, J.-P. Hulin, D. Salin, B. Perrin, and E. J. Hinch, “Buoyancy driven
40
Y. Yuan and A. R. Horner-Devine, “Experimental investigation of large-scale miscible front dynamics in tilted tubes,” Phys. Fluids 17(3), 031702 (2005).
53
vortices in a freely spreading gravity current,” Phys. Fluids 29(10), 106603 N. B. Kaye and G. R. Hunt, “Overturning in a filling box,” J. Fluid Mech. 576,
(2017). 297–323 (2007).
41 54
K. Bhaganagar, “Role of head of turbulent 3-D density currents in mixing during P. G. Drazin and W. H. Reid, Hydrodynamic Stability (Cambridge University
slumping regime,” Phys. Fluids 29(2), 020703 (2017). Press, 2004).
42 55
A. A. Alahyari and E. K. Longmire, “Development and structure of a gravity K. C. Sahu, H. Ding, P. Valluri, and O. K. Matar, “Pressure-driven miscible two-
current head,” Exp. Fluids 20(6), 410–416 (1996). fluid channel flow with density gradients,” Phys. Fluids 21(4), 043603 (2009).

Phys. Fluids 32, 063302 (2020); doi: 10.1063/5.0009102 32, 063302-17


Published under license by AIP Publishing

You might also like