UFL/COEL89/013
RESPONSE OF FINE SEDIMENTWATER INTERFACE
TO SHEAR FLOW
By
Rajesh Srinivas
1989
Thesis
RESPONSE OF FINE SEDIMENTWATER INTERFACE TO SHEAR FLOW
By
RAJESH SRINIVAS
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF MASTER OF SCIENCE
UNIVERSITY OF FLORIDA
1989
fOEstal Engineering Archives
University of florida
ACKNOWLEDGEMENTS
I would like to express my sincere gratitude to my advisor and chairman of
my graduate committee, Dr. Ashish J. Mehta, for his valuable and imaginative
guidance and ideas which have made this thesis possible. I am indebted to him
for going out of his way in acting like a mentor and guardian. My thanks also go
to Dr. R.G. Dean and Dr. D.M. Sheppard for serving on my committee. I am
also grateful to the personnel at the Coastal Engineering Laboratory, Roy Johnson,
Danny Brown and, especially, Vernon Sparkman for their help and suggestions in
building the flume and pump. Special thanks are also due to Shannon Smythe and
Barry Underwood for their excellent drafting work.
Finally, I would like to thank my parents for their unqualified support and faith
in me.
This study was supported by the U.S. Army Engineer Waterways Experiment
Station, Vicksburg, MS (contract DACW3989K0012) with project manager, Allen
M. Teeter.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ........
LIST OF FIGURES ............
LIST OF TABLES ............
LIST OF SYMBOLS ...........
ABSTRACT ................
CHAPTERS
1 INTRODUCTION ...........
1.1 Need for Study of Fluid Muds .
1.2 Some Observations of Fluid Mud
1.3 Approach to the Problem ....
1.4 Objectives .............
1.5 Plan of Study ...........
2 INSTABILITY MECHANISM ....
2.1 Discussion .............
2.2 KelvinHelmholtz Instability ...
2.2.1 Case of a Vortex Sheet..
2.2.2 Generalized Form of Kelvil
3 INSTABILITY OF STRATIFIED SHE
3.1 Background ............
3.2 Literature Review ........
3.2.1 Browand and Wang (1971)
3.2.2 Smyth, Klaassen and Pelti
.............
Entrainment . .
nHelmholtz Instability
.AR FLOWS .....
er (1987)........
3.2.3 Lawrence, Lasheras and Browand (1987) . . ... 33
3.2.4 Narimousa and Fernando (1987) . . ... 34
3.3 Conclusions ........... ......... ......... 38
4 ENTRAINMENT IN STRATIFIED SHEAR FLOWS ........... 40
4.1 General Aspects ........... .... .......... 40
4.2 Moore and Long (1971) ......................... 40
4.2.1 Results of Two Layer Steady State Experiments ...... 41
4.2.2 Results of Entrainment Experiments . . .... 44
4.2.3 Summary ................... .......... 44
4.3 Long (1974) .. ... ... ...... . . .. .. 45
4.4 Narimousa, Long and Kitaigorodskii (1986) . . .... 47
4.4.1 Deduction of u, ............. ........ 48
4.4.2 Entrainment Rates Based on u . . . ... 49
4.5 Wolanski, Asaeda and Imberger (1989) . . . .... 50
4.6 Conclusions ..................... .......... 51
5 METHODOLOGY ................. .......... 53
5.1 Apparatus .................... .. .......... 53
5.2 Procedure .............................. 61
6 RESULTS AND ANALYSIS ......................... 67
6.1 Definition of Richardson Number . . . .. .. 67
6.2 Initial Conditions ............................. 68
6.3 Evolution of Characteristic Profiles . . . ... 70
6.4 Shear Layer ................... ............. 74
6.5 Observations on the Interface ...................... 80
6.6 Entrainment Rate ................... ......... 84
6.7 Discussion in Terms of Equilibrium Peclet Number . ... 95
6.8 Comparison with Soft Bed Erosion . . . ... 98
. . . . . . 101
7.1 Summary .............
7.2 Conclusions ............
7.3 Recommendation for Further Wor]
APPENDICES
A TEST MATERIALS ..........
A.1 Kaolinite .............
A.2 Bentonite .............
B A NOTE ON RICHARDSON NUMBER
B.1 Introductory Note ........
B.2 Small Disturbances ........
B.3 Energy Considerations ......
BIBLIOGRAPHY .............
BIOGRAPHICAL SKETCH .......
. . . . . 101
. . . . . 101
k .. ............ .. 104
. . . . . 105
. . . . . 105
. . . . . 105
R .................. 107
. . . . . 107
. . . . . 107
. . . . . 108
. . . . . 110
. . . . . 114
7 SUMMARY AND CONCLUSIONS
LIST OF FIGURES
1.1 Definition sketch for fluid mud (source: Ross et al. 1988).... 3
1.2 Evolution of Suspended Sediment Concentration (source: Kirby
1986) . . . . . . . . .. 5
1.3 Internal waves produced by the passage of sailing vessels in
the Rotterdam Waterway (source: van Leussen and van Velzen
1989) . . . . . . . . 6
1.4 Field evidence of gravity driven underflows (source: Wright et
al. 1988) . . . . . . .. 7
2.1 Definition sketch of the flow for the case of a vortex sheet 16
3.1 Offset Velocity and Density Profiles . . . ... 28
3.2 Physical description of the complete flow configuration, with
density and velocity profiles (adapted from Narimousa and Fer
nando 1987. . . . . . . .. .....35
5.1 Recirculating flume of plexiglass used in the present investigation
(dimensions in centimeters) . . . . .. 55
5.2 Section AA of the flume, from Figure 5.1 (dimensions in cen
tim eters) ............................. 56
5.3 Section BB of the flume, from Figure 5.1 (dimensions in cen
tim eters) ............................. 57
5.4 Details of the disk pump system used in the present investigation
(dimensions in centimeters). . . . ..... 59
6.1 Sequence of concentration profiles of Run 9 with kaolinite de
picting the evolution of concentration with time. IF denotes
interface. ............................... 71
6.2 Evolution of the velocity profile in the mixedlayer for Run 6
with kaolinite. IF denotes interface. . . . ... 73
6.3 Change in the mixedlayer depth with time for Run 10 with
bentonite................... ............. 75
6.4 Rate of change of mixedlayer depth in Run 10 with bentonite. 76
6.5 Nondimensional shear layer thickness vs. Richardson number .78
6.6 Nondimensional shear layer thickness vs. Richardson number
on a loglog scale ........................... 79
6.7 Turbulent entrainment at t ~ 0.5 minute. Sediment kaolinite. 81
6.8 Interface at Ri, < 10. Sedimentkaolinite . . ... 81
6.9 Interface at Ri, < 10. Sedimentkaolinite. . . ... 82
6.10 Highly irregular interface at Ri, > 10. Sediment kaolinite. ... 82
6.11 Scour of growing crest at Ri, > 10. Sediment kaolinite. .... 83
6.12 Scour of grown crest at Ri, > 10. Sediment kaolinite. . 83
6.13 Subsiding crest at Ri, > 10. Sedimentkaolinite. . ... 85
6.14 Smokelike wisp being ejected from the tip of disturbances. Sediment
kaolinite. .... ............... .......... 85
6.15 Appearance of the interface at high Richardson numbers, Ri, >
25. Sedimentkaolinite... ............... ...... 86
6.16 Nondimensional buoyancy flux vs. Richardson number for all
the experiments. ........................... 93
6.17 Comparison of erosion rates of soft beds with the rates predicted
by equation (6.9). .......................... 100
LIST OF TABLES
6.1 Initial conditions of all Runs ................. .. 69
6.2 Relevant measured parameters for runs with kaolinite . 87
6.3 Relevant measured parameters for runs with bentonite . 88
6.4 Richardson numbers and entrainment rates for runs with kaoli
nite . . . . . . . . 89
6.5 Richardson numbers and entrainment rates for runs with ben
tonite . . . . . . . .. 90
6.6 Peclet numbers for equilibrium conditions . . .... 98
A.1 Chemical composition of kaolinite . . . .... 106
A.2 Chemical composition of bentonite . . .... 106
LIST OF SYMBOLS
b = buoyancy.
boo = buoyancy of unperturbed layer.
bi = rms buoyancy fluctuation.
C = concentration of the suspension.
C1 = mean concentration of the mixedlayer.
C2 = concentration of fluid mud at the level of the interface.
C2 = mean concentration of fluid mud.
c = disturbance wave speed.
c' = turbulent speed.
d = distance between the centers of the shear layer and the density interface.
dm = change in mass with time.
dt = time of the interval.
E = entrainment coefficient.
E = erosion rate.
Ef = floc erosion rate.
F = Froude number.
F, = vertical flux.
H = depth of the fluid mud layer.
H = total depth of the twolayered system.
i = V.
h = depth of the mixedlayer.
J = local Richardson number.
k = horizontal (xdirection) wave number of the perturbation.
K, = eddy diffusion coefficient.
k = resultant horizontal wave number of the perturbation.
L, = mixing length.
1 = horizontal (ydirection) wave number of the perturbation.
11 = length scale.
M2 = mass per unit area of the fluid mud.
N = buoyancy frequency.
n = Manning's resistance coefficient.
P = probability that a particle reaching the bed will deposit.
Pe = Peclet number.
p = pressure in the fluid.
p' = perturbation in the pressure due to the disturbance.
Q = nondimensional bouyancy flux.
q = bouyancy flux.
Ri = Richardson number.
Ricr = critical Richardson number.
Ri = minimum Richardson number.
Rio = overall Richardson number.
RiU = Richardson number based on the mean velocity of the mixed layer.
Ri. = Richardson number based on the friction velocity.
s = complex angular frequency of the disturbance.
T = surface tension.
Ta = advective time scale.
Td = diffusion time scale.
U = velocity of fluid.
u = representative velocity.
I
u' = perturbation in the horizontal (xdirection) velocity due to the disturbance,
or horizontal (xdirection) turbulent velocity.
= mean velocity of the mixedlayer.
S = rms turbulent horizontal velocity.
ue = entrainment velocity.
u, = friction velocity.
V = potential energy.
V1 = potential energy per unit mass.
v' = perturbation in the horizontal (xdirection) velocity due to the disturbance.
W = width of the sidewalls.
w' = perturbation in the vertical (zdirection) velocity due to the disturbance,
or vertical turbulent velocity.
w, = particle settling velocity.
wl = turbulent fluctuation of the vertical velocity.
w, = friction velocity of the sidewalls.
x = horizontal coordinate.
y = horizontal coordinate.
z = vertical coordinate.
a = horizontal (xdirection) of the perturbation.
a = a rate coefficient.
Pf = horizontal (ydirection) of the perturbation.
Ab = interfacial buoyancy jump.
Ap = interfacial density jump.
6 = thickness of the density interface.
6T = kinetic energy per unit volume of the flow.
6W = work done to overcome gravity.
6, = thickness of the shear layer.
6, = amplitude of the interfacial wave.
e = dissipation function.
r7 = displacement of the interface.
A = wavelength of the disturbance.
v = kinematic viscosity.
p = density of the fluid.
p = perturbation in the fluid density due to the disturbance.
pl = mean density of the mixedlayer.
= = shear stress.
= velocity potential.
= perturbation in the velocity potential due to the disturbance.
Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science
RESPONSE OF FINE SEDIMENTWATER INTERFACE TO SHEAR FLOW
By
RAJESH SRINIVAS
August 1989
Chairman: Ashish J. Mehta
Major Department: Coastal and Oceanographic Engineering
An experiment was conceived and executed to simulate the effects of turbulent
shear flow on fine sediment, specifically fluid mud. The tests were conducted in
a "racetrack" shaped recirculating flume with a disk pump. Experiments were
run with two types of fluid mud, consisting of kaolinite and bentonite in water.
Shear layer thickness and the nature of the interfacial instabilities were qualita
tively examined. Entrainment rates of fluid muds were examined as a function of
increasing Richardson number and an empirical relation was obtained between the
nondimensional buoyancy flux and the Richardson number. This relationship was
then compared with that obtained by previous experimenters for salt stratified sys
tems. This comparison made apparent the effect of sediment particles in causing
additional dissipation of turbulent kinetic energy at higher Richardson numbers as
the entrainment rate decreased substantially. Peclet number consideration showed
that the mixedlayer at these higher Richardson numbers appears to behave like a
suspension in equilibrium. The effect of varying the clay constituent of fluid mud on
the entrainment rate could not be fully investigated, although within the limits of
data no discernible trend differences could be clearly identified. A brief comparison
of the fluid mud entrainment rate, which is proportional to the cube of the flow
xiii
velocity, with soft bed erosion rate, which is proportional to the square of the flow
velocity, showed that fluid mud entrainment can dominate over bed erosion at low
current velocities.
CHAPTER 1
INTRODUCTION
1.1 Need for Study of Fluid Muds
A challenging aspect of many coastal and estuarine problems is the elucidation
of fine sediment transport behavior. The compelling factors for such investigations
are both economical and environmental. The last couple of decades have seen
extensive effort being applied to experimental and theoretical studies with a variety
of mathematical models developed for simulation of fine and cohesive sediment
transport. The common aspect in the modeling approach is a soil bed subject to
layer by layer or massive erosion. However, experimental observations verify the
existence of the sediment population in three distinct states: mobile, upper column
suspensions, high concentration nearbed suspensions, and settled muds (e.g., see
Kirby and Parker 1983). In mobile suspensions, the particles are dispersed and
stay in suspension by turbulent momentum exchange. Nearbed high concentration
suspensions or fluid muds, are partially supported by the fluid and partially by their
particle network while in settled muds the particles rest at the bottom supported
by their infrastructure (soil matrix).
The relatively high concentrations of fluid muds play a substantial role in hori
zontal transport to sedimentationprone areas. Indeed, in spite of low near bed ve
locities, the horizontal sediment mass flux can be considerable and can lead to "fluff"
accumulation in navigational channels. The movement of fluid muds has been cited
as the most likely cause of rapid sedimentation in ports located in muddy estuaries.
Obviously, ignoring fluid muds can lead to gross underestimation of sedimentation
rates. Almost totally neglected has been the issue of their upward turbulent en
2
trainment and mixing due to vorticity generation by shear flows (of current) above
them. Their loose structure permits fluid muds to entrain into the water column
easily and contribute substantially to degradation of water quality. This facet of
transport is evidently not simulated in solely considering erosion of cohesive beds
(which have a measurable shear strength).
Thus, one can assert that the consideration of the entrainment behavior of
this state of fluid muds is necessary to comprehensively simulate the mechanics
of fine sediment transport effectively. This implies that the prediction of fluid
mud behavior to hydrodynamic forcing by shear flows is necessary. Entrainment
rates need to be established and possible physical mechanisms causing this kind of
response need to be formulated, neither of which are presently widely available in
detail. These aspects are briefly examined in this experimental study.
1.2 Some Observations of Fluid Mud Entrainment
Typical variations in concentration and velocity with depth for muds, and the
related definition terminology are presented in Figure 1.1. Fluid muds are confined
to the region between the lutocline, i.e., the zone with a steep concentration gradi
ent, and the partially or fully consolidated bottom. The upper zone of fluid mud
may have both horizontal and vertical motion, while the lower zone may have some
vertical motion only. Using concentration as a measure, it is generally accepted
that these fluid muds fall in the range 20 to 320 g/1 (Ross et al. 1987). These
concentrations correspond to the bulk density range of 1.01 to 1.20 g/cm3, given a
sediment granular density of 2.65 g/cm3.
Fluid mud behavior is largely time dependent, varying with the physicochemical
properties of both sediment and water. The theological properties of fluid muds are
strongly affected by factors such as pH, salinity, mineralogical composition and
particle size. Their mechanical behavior is generally pseudoplastic while at very
high concentrations they resemble Bingham plastics (Bryant et al. 1980), as at
CONCENTRATION
1 102 103 104
(mgl ')
105
VELOCITY (msec 1)
Figure 1.1: Definition sketch for fluid mud (source: Ross et al. 1988).
10
or
106
4
high concentrations strong interparticle bonds provide an initial resistance to shear
deformation (when the applied stress is less than the yield stress, elastic deformation
is possible without any breakdown of structure leading to fluidization).
The dynamic behavior of fluid muds during a tidal cycle is well recognized by
presenting the sequence of concentration profiles recorded by Kirby (1986) (see Fig
ure 1.2). These are given for accelerating flow, while the reverse sequence prevails
for decelerating flow. Zone 1 is a very low concentration suspension, Zone 2 is the
lutocline layer, i.e., the zone with steep concentration gradients, while, Zone 3 is
high concentration suspension (similar to fluid mud). At slack water, the destabi
lizing shear forces are small compared to buoyancy stabilization and there is no
entrainment. The physical situation corresponds to a twophased system with fluid
mud separated from the overlying water by a distinct interface. As the velocity
picks up, the resulting turbulent kinetic energy becomes sufficient to overcome the
stable stratification of the fluid mud and there is subsequent entrainment.
High concentration ( 300 g/l) fluid mud layers of thicknesses more than a
meter have been observed in the Rotterdam Waterway (van Leussen and van Velzen
1989). The passage of sailing vessels over these layers produces internal waves (see
Figure 1.3) at their surface, in spite of the fact that the bottom stresses are quite
low.
Wright et al. (1988) made field measurements of dispersion of concentrated
sediment suspensions over the active delta front of the Yellow River in China. They
provided evidence of the existence of both hypopycnal (buoyant) plumes as well as
gravity driven hyperpycnal (near bottom) dispersal modes. Downslope advection
within the hyperpycnal plume of mixed, lower salinity water from the river mouth
caused vertical instability as regards the excess bulk density (including sediment
concentration, salinity and temperature). Once deposition began, tidal currents
contributing to vertical momentum exchange resulted in instability induced en
1
Zone 1
Conc.
Velocity 
L=Lutocline
Conc.

Conc.
Conc.
0,
4
Zone 1
Zone 2
L
Zone 3
Conc.
Sequence 15:Accelerating Phase
Figure 1.2: Evolution of Suspended Sediment Concentration (source: Kirby 1986)
Zone 3
Conc.
I. J
16'
A sr;;
Figure 1.3: Internal waves produced by the passage of sailing vessels in the Rotter
dam Waterway (source: van Leussen and van Velzen 1989).
E
I1
01
SE
5
Km
Figure 1.4: Field evidence of gravity driven underflows (source: Wright et al. 1988).
8
hanced mixing. They observed large amplitude high frequency internal waves at
close to the BruntVaisili frequency.
1.3 Approach to the Problem
In a most general sense, it can be asserted that shear flow in a stratified fluid is
a natural occurence and a crucial mechanism for turbulence production in the at
mosphere and oceans. A number of practical engineering problems, often associated
with a desire to thoroughly mix effluents entering the surroundings, also requires a
knowledge of the behavior of stratified shear flows.
There are numerous situations in nature where an understanding of the behavior
of velocitysheared density interfaces is important:
Wind generated waves in the ocean can be a manifestation of KelvinHelmholtz
type instabilities at the airwater interface.
The tangential stress which occurs when the wind blows over the ocean gener
ates a drift current in the upper layers of the ocean, which causes entrainment
of the stratified layers below. This has been cited as the mechanism responsi
ble for bringing deep sea nutrients into more accessible regions (Phillips 1977).
Substantial bearing on the world climate is attributed to drift currents in the
upper atmosphere causing growth of this mixed layer against previously stable
inversions.
The rising and subsequent spreading of methane gas in coal mines has an
important bearing on safety (Ellison and Turner 1959).
Gravity currents under a stratified layer over sloping bottoms are very com
mon in oceans.
In estuaries, the oceanic saltwater wedge penetrates upstream and under
lighter river water.
9
Finally, as mentioned before, shear flows can cause entrainment of underlying
fluid mud, which is the focal point of interest of this study.
Again, it can be stated in general terms that vorticty generation by shear flows
causes instabilities to appear at the density interface and these seem to be the
prime cause for mixing across this interface. A gamut of literature exists for the
same general kind of problem, with density stratification caused by salinity, or ther
mal effects, or both. These are analogous because of comparable density ranges
and statically stable arrangements. Salinity experiments have been conducted to
simulate oceanic situations which have velocity shear values similar to estuarine
environments, with resulting comparable values of the ratio of buoyancy to shear
forces. Interfacial instabilities and entrainment rates have been examined, theoret
ically as well as experimentally. However, a peculiar feature of these studies is the
fact that most investigators seem to arrive at quite different results, which they
then generally proceed to explain satisfactorally. So, relative newcomers are sad
dled with numerous and quite different relationships and explanations for observed
phenomena, without any explicit kind of unification. This is a potent indicator of
the fact that this process of production and dissipation of turbulent kinetic energy
which governs the buoyancy flux and generation, growth and collapse of instabilities
is a very complex process and far from being well understood.
Experiments considered here have additional complications due to nonNewtonian
rheology. Fluid muds are not autosuspensions. Settling is characteristic, and the
downward buoyancy flux due to particle fall velocity causes additional dissipation
of turbulence, which is obviously not the case for salinity and temperature stratified
experiments.
Defining, h as the the depth of the turbulent mixed layer, u, as a relevant
entrainment velocity = dh/dt (rate of propagation of the mixed layer), ul as the
turbulent velocity scale for the mixed layer, Ab as the buoyancy step across the
10
density interface = (gAp)/po, Ap as the interfacial density step, and po as a refer
ence density, the Buckingham7r theorem for dimensional analysis can be used for
determining the relevant nondimensional parameters governing the dynamics of
this situation. Intuitively, one can see that density and acceleration due to gravity
should be coupled as buoyancy. We can in fact identify the pertinent variables to
be Ab, ul, u and h; the fundamental dimensions being that of length, L, and time,
T (as mass becomes implicit in buoyancy). Choosing ul and Ab as our repeating
variables we can form the combinations Ab"ugh and Ab'u'ue. Now, we demand the
exponents of L and T to be zero in each combination. So, we obtain a = 1, f = 2,
7 = 0, and 6 = 1, giving us the nondimensional parameters A and , the
first of which is the Richardson number (Ri), whereas the second is an entrainment
coefficient (E). The dimensional analysis is completed by the statement f(Ri,E) =
0, or, further,
E = 7(Ri) (1.1)
The fact that such a functional relationship exists is borne out by the experi
mental results of many previous investigators, albeit in different forms.
This relationship between E and Ri represents interaction between mechanical
mixing energy and the potential energy stored in stratification that it is working
against. As entrainment is considered a turbulent process, effects of molecular
diffusion are largely ignored, although, some investigators have pointed out that
at high Ri, when turbulence is relatively weak, molecular diffusion does become
important for salinity and thermal types of experiments.
Experimenters have arrived at different power laws (of the form E oc Ri")
for subranges of Ri (for example, see Christodoulou 1986 and Narimousa et al.
1986). More complicated relationships have also been derived by evaluation of the
turbulent kinetic energy budget (Zemen and Tennekes 1977; Sherman et al. 1978;
Deardorff 1983; Atkinson 1988).
11
1.4 Objectives
With the proceeding discussion in mind, and after an indepth review of perti
nent literature regarding the mechanism of instabilities and the consequent entrain
ment, it was decided to run experiments to simulate entrainment of fluid muds by
turbulent velocityshear flows in a speciallydesigned flume. A 'racetrack' shaped
recirculating flume was constructed for this purpose in which a twolayered system
of fluid mud and water could be established. The flume was built of plexiglass, as
one of the prime objectives of the present investigation was to observe the nature of
interfacial instabilities. Shear flow was generated by using a specially designed disk
pump which is basically a system of interlocking plates on two parallel externally
driven shafts rotating in opposite directions. The horizontal velocity of the driven
fluid was constant over the depth of the diskpump. This diskpump was instrumen
tal in imparting horizontal homogeneity to the flow The velocity profile diverged
from the vertical at a distance from the level of the bottom disk of this pump, thus
producing flow with meanshear.
The ultimate objective of this investigation was to run a series of experiments
to simulate the effects of shear flow on the fluid mudwater interface and the re
sulting entrainment of relatively low to medium concentration fluid muds, and to
make phenomenological observations to obtain qualitative descriptions of interfacial
instabilities and quantitative expressions) for rates of entrainment by measuring
mass flux in relation to the destabilizing velocityshear. Another objective was to
determine the effect of varying the degree of cohesion of sediment on rates of en
trainment. This was done by using kaolinite and bentonite (see Appendix A), which
vary greatly in their degree of cohesion, since kaolinite is only weakly cohesive while
bentonite is cohesive and thixotropic.
12
1.5 Plan of Study
The following chapters document the investigation of the issue of entrainment
of fluid mud by shear flow to find a quantifiable relationship for this process, which,
as mentioned before, has hitherto remained largely unaddressed. Starting with
the justifiable surmise that fluid mud entrainment is a manifestation of interfacial
instability due to current shear, theoretical background for the production and
propagation of instabilities is first discussed, and thus the investigation begins in
Chapter 2 with a theoretical background of KelvinHelmholtz type of hydrodynamic
instability. The classic case of stability of a vortex sheet is discussed first in this
chapter, and this is followed by the more generalized version of KelvinHelmholtz
instability.
In Chapter 3, some of the more pertinent work of previous investigators on the
subject of instability of shear flows is reviewed. Considerable work has been done
in the area of numerical simulations of instabilities, but adequate support in the
form of accurately documented experimental evidence seems to be lacking. It must
be mentioned, however, that the recent work of Narimousa and Fernando (1987) is
both comprehensive as well as enlightening.
The question of entrainment rates due to shear flows of stably stratified fluids
is examined in Chapter 4. Again, the volume of work which has been done is
considerable, and only directly pertinent literature is considered for review.
Chapter 5 is devoted to the experimental methodology of the present investiga
tion. The details of the flume and the disk pump constructed for the present study,
the procedure of experimentation and methods of measurement are documented.
In Chapter 6, the results of the investigation are presented and analysed, while
Chapter 7 gives the main conclusions of the study.
In Appendix A a description of the constituent materials of fluid mud, namely
kaolinite and bentonite, prepared in the laboratory is included, while Appendix B
13
traces the history of the definition of the critical Richardson number for stability of
a stratified shear flow.
CHAPTER 2
INSTABILITY MECHANISM
2.1 Discussion
In general, instability occurs when there is an upset in the equilibrium of the
external, inertia and viscous forces in a fluid. Examples of external forces are
buoyancy in a fluid of variable density, surface tension, magnetohydrodynamic,
Coriolis and centrifugal forces. Surface tension and magnetic forces usually tend to
stabilize, while an interesting point to be noted regarding viscosity is that it can
both inhibit or amplify disturbances. An obvious effect is of dissipation of energy,
whence any flow is stable if viscosity is large enough. However, it's effect of diffusing
momentum may render flows unstable, as in parallel shear flows, which are stable
for the inviscid case.
The analysis is restricted to primarily steady flows, although tidal action in
estuaries is obviously unsteady. However, tidal flows may be considered to be steady
for the purpose at hand, since one is dealing with widely different time scales.
Analysis of unsteady flows is very complex in general. Boundaries of the flow are
an important factor, as well; the closer the boundary, the more efficient is the
constraining of disturbances, although boundary layer momentum diffusive effects
may serve to enhance instability.
Any flow is likely to be disturbed, at least slightly, by irregularities or vibrations
of the basic flow. This disturbance may die away, persist at the same magnitude,
or grow so much as to alter the very flow. Such flows are termed stable, neutrally
stable and unstable, respectively. Stability of parallel inviscid fluid flow has been
investigated since the latter half of the nineteenth century, when the instability
15
of homogeneous and nonhomogeneous flows were considered. Subsequent analy
ses have been with subtle modifications to this same basic problem, including for
compressible fluids, considerations for rotational systems, magnetohydrodynamic
effects, etc. A wide range of literature has emerged, of interest to specialized sec
tors in engineering. The consideration in this section will be for the most general
case, fluid dynamical, for studying this phenomenon of instability, rather than its
occurence or application.
2.2 KelvinHelmholtz Instability
2.2.1 Case of a Vortex Sheet
Formulation of the Problem
It has been understood since the nineteenth century that the dynamic insta
bility of a weakly stratified parallel shear flow leads to the formation of vortexlike
structures called Kelvin Helmholtz (KH) waves. Consider the basic flow of incom
pressible, inviscid fluids in two infinite horizontal streams of different velocities and
densities, one above the other (see Figure 2.1), and given by
S= 2 U= 2 P= 2 P = P P2gz (z > 0)
= 4 U =U1 P = P = P plgz (z < 0)
The interface has an elevation z = r7 (x,y,t), when the flow is disturbed.
The governing differential equation is
V2 4 = 0 (2.1)
i.e.,
V72 =O z >r
V201=0 z <
 U2
 U1
Figure 2.1: Definition sketch of the flow for the case of a vortex sheet
Boundary Conditions
(a) The initial disturbance is constrained to a finite region
V a ' U as z  foo (2.2)
(b) A particle at the interface moves with it, i.e.,
D[z tl(x, y,t) 0 (2.3)
Dt
(c) Pressure is continuous across the interface
p2(C2 2 ( 2)' gz) =
84d 1
Pi (C1 (V1)2 gz) at z = 7 (2.4)
at 2
by Bernoulli's theorem.
Solution
The above equations pose the nonlinear problem for instability of the basic
flow. For linear stability, we consider
02 = U2X + '2 (z > T) (2.5)
01 = Uzx + f'i (z < r7) (2.6)
Products of small increments '1, '2 and rj are neglected. There being no length
scale in the basic flow, it is difficult to justify linearization as regards r7. However, it
appears plausible assuming that the surface displacement and it's slopes are small,
and gl << U 2, U.
With these these assumptions, linearisation yields,
v22 = 0 Z>O (2.7)
v = 0 z<0 (2.8)
4, = 0 z  +00 (2.9)
18
V1 = 0 z_ oo (2.10)
a = a +U z = 0 (i = 1,2) (2.11)
az Wt ax
+Pt1 + g) p(U + + gr) z = 0 (2.12)
az at az at
We now use the method of normal modes, assuming that an arbitrary distur
bance can be resolved into independent modes of the form,
(q,, 'i, '2) = (i, 1, $2) exp[i(kx + ly) + st] (2.13)
[s = a + iw, thus, if a > 0, the mode is unstable, if a = 0, the mode is neutrally
stable and stable asymptoticallyy) for a < 0 ]
Thus, equations (2.7) and (2.8) yield,
~, = AieA + Bje.' where k = +12 (2.14)
From equations (2.9) and (2.10),
01 = Ae (2.15)
$2 = Ae*z (2.16)
The coefficients can be evaluated from equation (2.11) as
A, = (s + ikUi)/l (2.17)
As = ?i(s + ikU2)/k (2.18)
From equation (2.12), we can obtain,
pi(U1Alekzik +Alels + gl) =
P2(U2A2eIik + A2e's + gi) (2.19)
Thus, with the substitution of the coefficients,
pi{(s + ikUi) + kg}=
p2{(s + ikU2)' + kg} (2.20)
which can be written as
82(Pi + P2) + 2iks(piUi + p2U2)+
[g(PI P2) k2(plU + p2U,2)] = 0 (2.21)
This yields
ik(piUi + pU2) k2pip2(U U2)2 ) (2.22)
P1+P P2 (P +P2)2 PI +P2
Conclusions
Several conclusions are of interest here,
(1)If k = 0, then
S= gik(P P2) (2.23)
P1 + P2
i.e., perturbations transverse to the direction of streaming are unaffected by it's
presence.
(2)In every other direction, instability occurs for all wave numbers with
k > k( P ) (2.24)
plp2(U2l U2)
/
If the wave vector k is at an angle 0 to U, k = k cos 0, instability occurs for
> (2g(pp) .25)
PIP,(U2 Ua) cos2 0
For a given relative velocity of the layers, instability occurs for the minimum wave
number when the wave vector is in the direction of streaming, i.e.,
g(p2 P p)
kmi = 2  ) (2.26)
PxP2(U2 U2)
Instability occurs for k > k,,n.
This predicts the onset and development of instability, no matter how small
(U1 U2) may be. The presence of streaming overcomes the stability of the static
arrangement. This is the classic KelvinHelmholtz instability. Helmholtz (1868)
stated this as:
20
Every perfectly geometrically sharp edge by which a fluid flows must tear
it asunder and establish a surface of separation, however slowly the rest
of the fluid may move.
However, if the effects of surface tension are considered, stability is predicted if,
2g p2 2
(U1 U2)2 < 2 P(2.27)
kmin P1P2
where, kmin = minimum wave number for stability.
With this condition, we have stability for,
(U1 U2)2 < 2 Tg(p P) (2.28)
PIP2
where T is the surface tension.
2.2.2 Generalized Form of KelvinHelmholtz Instability
From the above discussion, for the case without surface tension, it can be in
ferred that the onset of KelvinHelmholtz instability is by the crinkling of the in
terface by shear, and this is independent of the magnitude of the relative velocity
of the two layers. A natural question to confront the reader is whether this result
is entirely fortuitous, due to the sudden discontinuity in the density and velocity
profiles, and not be true for continuous distributions. Thus, now, we take the case
of the stabilizing effect of gravity on a continuously stratified fluid and of the desta
bilizing influence of shear in a generalized form of Kelvin Helmholtz instability. We
start with a basic state in dynamic equilibrium,
u, = U,(z,) (2.29)
P. = P.(z.) (2.30)
P. = (Po). g p(z) dz (2.31)
for zl. < z, < z2., where, z. is the height and zl. and zz. are the horizontal
boundaries of the flow. The subscript indicates dimensional quantities. Taking L,
21
U and Po to be the characteristic length, velocity and density, respectively, of the
basic flow and further assuming the fluid to be inviscid and density to be convected
but not diffused, we nondimensionalize the equations of motion, incompressibility
and continuity to get,
u ,
p(t +u.Vu) = VPF2pk (2.32)
v.u = 0 (2.33)
ap
+u.VP = 0 (2.34)
at
where F = V/I/gL is a Froude number.
Perturbations are introduced into the flow,
u(z,t) = U(z)+u'(x,t) (2.35)
p(x,t) = P(z)+p'(x,t) (2.36)
p(, t) = Po F (z)dz' + p'(x, t) (2.37)
The form of the equations obviously permits us to take normal modes of the
form,
{u' (, t), p (, t), p(, t)} =
{u(z), (z), A(z)} exp[i(ax + 0y act)] (2.38)
where, the real part is understood. The fact that the solutions must remain bounded
as x, y + oo implies that a, must be real; but, the wave speed c may, in general,
be complex, i.e., c = c, + ici thus representing waves traveling in the direction
(a, /,0) with phase speed ac,/V/a2 + 2 and grow/decay in time as exp(acit). Thus,
aci > 0 implies instability, acj < 0 stability, while acc = 0 implies neutral stability.
Introducing these into equations (2.32) (2.34), and linearizing by neglecting
quadratic terms of the primed quantities and using equation (2.38) we obtain,
iap(U c)u + pU'ti = iap
(2.39)
22
iap(U c)0 = if# (2.40)
iaA(U c)b = D F2 (2.41)
iauc + if + Dtw = 0 (2.42)
ia(U c)A + p'w = 0 (2.43)
where differentiation with respect to z of a basic quantity is denoted by prime
whereas that of a perturbation by D.
Thus, from equations (2.39) and (2.40),
i A AU'^A
t = a (2.44)
ia,5(U c)
S= (2.45)
ap(U c)
Using these in conjunction with equation (2.42), we can obtain,
i' p U'8 iP2,
+ Dtb = 0 (2.46)
p(U c) ap(U c)
Eliminating ^ and 5, we finally arrive at,
(2 +I
(U c) D (a2 + )} (a + + {(U c)Do U'}
a2F2 (U c)p p
(2.47)
Yih (1955) applied Squire's transformation to the system to show that for a
threedimensional (3D) wave with wave number (a, #), there is a 2D wave with
the same complex velocity c, but wave number (Viai + 2, 0) and Froude number
aF/I1/xa +2, which thus has effectively reduced gravity but magnified growth rate
(a2 + P2)c, and thus is more unstable.
Equation (2.47) indicates that F2 occurs as a product of p'/p, so an overall
Richardson number is defined as
'R = gL2 dp.
pF2 V2 p, dz.
The BruntViisill frequency (or buoyancy frequency) N, is defined as
N(z.) = g I/. = RiN2(z)V/L2
Thus, we get,
RiN2/U' L2 g dA dU
V2 F. dz. dz
g /{ .(dUz)2}
dz dz.
as the local Richardson number, J, of the flow at each height z,, such that
2z dU,2
J= N(z)/( )2 (2.48)
dz.
In many applications, Fp (z,) varies more slowly with height than U, (z,) such that
p',/p < 1; whence Ri is of the order of magnitude unity as F < 1. Thus, as in the
Boussinesq approximation the last two terms of equation (2.47) are neglected; hence,
the effect of variation of density is neglected in inertia but retained in buoyancy.
With this approximation and considering only 2D waves we get,
2d 2 2U 1 RiN2
dz' dzz U c (U c) 2
which can be written as
(U c)(D2 a2)0 U"q + RiN20/(U c) = 0 (2.50)
with the corresponding boundary conditions at z = zl and z2, which is the Taylor
Goldstein equation, where
i = af/az (2.51)
w = iao(z) (2.52)
u = ak'/laz (2.53)
w = ao'/la (2.54)
0' = O(z)exp{ia(x ct)} (2.55)
Here, a > 0 can be assumed without any loss of generality, and also that each
unstable mode has a conjugate stable one.
Assuming ci 7 0, define
H = //VU c (2.56)
Substituting into equation (2.50) yields,
U" U"
D{(U c)DH} {2(U c) + +( RiN2)/(U c)}H = 0 (2.57)
2 4
Multiplying by the complex conjugate, H* and integrating,
Li 1 U'/4 RiNC
f{(U c){IDHI2 + a2 jH2} + UIH2 + U2/4 RiN2 H12} dz = 0 (2.58)
12 Uc
The imaginary part gives,
'\jDH\2 + a2JH2 + (RiN2 U/4) IH/1U cl2} dz = 0 (2.59)
Thus,
0 > I DHI2 dz
1
= f'{(RiN2 U'2/4) +a U c12}H2/IJU c2 dz (2.60)
(assuming ci $ 0). Thus, the local Ri has to satisfy RiN2/U" < 1/4 somewhere in
the field of flow for instability.
The same can also be established, although somewhat heuristically, by analyzing
the energy budget; the essential mechanism of instability being the conversion of
the available kinetic energy of the layers into kinetic energy of the disturbance,
overcoming the potential energy needed to raise or lower the fluid when d{./dz* < 0
everywhere. Consider two neighboring fluid particles of equal volumes at heights z.
and z. + 6z. being interchanged.
Thus, 6W = work per unit volume needed to overcome gravity = g6ps6z..
For horizontal momentum to be conserved, the particle at z, will have final
velocity (U, + k6U.)tand the particle at z. + 6z. have (U, + (1 k)6U,)ias it's final
velocity, where, k = some number between 0 and 1, and
SU. = ( )6z. (2.61)
dz,
25
Thus, the kinetic energy per unit volume released by the basic flow is,
1 1
6T = pU. + (.+ 6.)(U +6U.)2
2 2
AP,(U. + k6U,)2 (p + 6p,)(U, + (1 k)6U.)5 (2.62)
2 2
= k(1 k)p,(6U.)2 + U,6U,6p, (2.63)
< (6U,)2f + U.6U.6p. (2.64)
4
A necessary condition for this interchange, and consequently, instability is 6W <
6T, and therefore, somewhere in the field of flow,
dp. 1 dU. dU_.dp._
g < p.( )2 + U. (2.65)
dz. 4 dz. dz, dz,
i.e.,
_h 1
d < (2.66)
neglecting the inertial effects of the variation of density.
Miles (1961) stated that the sufficient condition for an inviscid, continuously
stratified flow to be stable to small disturbances is that the local Richardson num
ber should exceed 1 everywhere in the flow (a modified result is presented in Ap
pendix B). This does not imply that the flow becomes unstable if this falls below
somewhere. Counter examples have been found, for example, with a jetlike velocity
profile uoc sech2 z and an exponential density profile, in which case the flow can
become unstable if Ri,, < 0.214. Hazel (1972) has demonstrated the stabilizing ef
fect of rigid boundaries. One must consequently surmise that the entire profile (the
boundary conditions, viscosity, etc.) matters in determining the critical Richardson
number.
Thus, it is seen that the effect of velocityshear on statically stable stratification
can be to cause disturbances to appear at density interfaces which grow with time.
Intuitively, one can sense that after a period of sustained growth, the wave should
break, with the natural ramification being upward mixing of the denser fluid, i.e.
entrainment.
26
With the preceding background of the theory of velocityshear induced inter
facial instability, we now proceed to Chapter 3 where pertinent work on the same
phenomenon is reviewed. Some examples of numerical and laboratory simulations
are covered to give a feel for the magnitude as well as different facets of the problem.
CHAPTER 3
INSTABILITY OF STRATIFIED SHEAR FLOWS
3.1 Background
As noted in Chapter 1, shear induced instabilities are a very important factor
in the generation of turbulence and mixing in stratified flows. When 6, 6 and
d = 0 (see Figure 3.1), at sufficiently low Ri (= ~A), the primary instability
is of the KelvinHelmholtz (KH) type; however, the process of growth by pairing
becomes limited by the stabilizing effects of buoyancy (Corcos and Sherman 1976)
and a sufficiently large density difference will stabilize the flow.
As it is relevant in geophysical situations, the case of 6, > 6, with d = 0 was
studied by Holmboe (1962), who predicted a second mode of instability, now called
the Holmboe mode, which has been further studied by a number of researchers, for
example Hazel (1972). Theoretically, this comprises of two trains of growing interfa
cial waves traveling in opposite directions to the mean flow, eventually resulting in
a series of sharply cusped crests protruding alternately into each layer, with wisps
of fluid being ejected from these cusps (but, more often, experimental results indi
cate cusping only into the high speed layer which may possibly be attributed to the
selective vorticity concentrations in the high speed layer). Thus, when 6,/6 > 1,
theoretically, there is always a range of wavenumbers which is unstable, however
large Ri may be, with this second mode having maximum amplification rates at
nonzero Ri.
For small Ri transition to turbulence is by the first mode (i.e., KH) regardless
of 6,/6 values, with collapse by overturning due to the concentration of the available
vorticity into discrete lumps along the interface (Thorpe 1973). This results in finer
U2 P2
U(z) p(z)
U. 
IU PI
Figure 3.1: Offset Velocity and Density Profiles
29
scales of turbulence, and in a homogeneous fluid these lumps continue to pair with
the growth of the mixed layer. However, with stratification, entrainment of fluid into
the mixing layer degrades this vorticity in these lumps and this mixed layer growth
eventually stops, and if the initial Ri is small, turbulence grows till length scales
become large enough for buoyancy to play an important role, followed by collapse.
If 6,/6 > 1, this collapse is followed by mode 2 waves (Browand and Winant
1973). These seem to be like internal waves within the mixing layer, with nearly
horizontal wave crests and small wavelengths (Delisi and Corcos 1973); and, finally,
there is decay of the turbulence structure. Fernando (1988) mentions that turbulent
patches in stratified media may be generated by the mechanism of instability (by
wavebreaking and double diffusion).
Thus, stratification has this ability to destroy turbulence which may be a pos
sible explanation for it's intermittent character, as found in nature. McLean (1985)
observed longitudinal ripples on the bed while modeling deep ocean sediment trans
port, which he postulated to occur during deposition after high energy erosional
events due to helical circulation owing to a nonuniform turbulence field. This kind
of turbulence field can result because of lateral homogeneity of turbulence damping
by the aforementioned density stratification. Physically, this turbulent mixing layer
is destroyed by the stabilizing effect of gravitation on the largest scales of Ri.
When the initial Ri is large enough, say > 0.1, then turbulence production
depends strongly on the d/6 ratio, with initial instability of the mode 2 waves.
These decay by breaking at sharply peaked crests (Browand and Winant 1973),
with fluid ejected into the higher speed layer as thin wisps from these crests.
30
3.2 Literature Review
3.2.1 Browand and Wang (1971)
Background
A velocity shear interface of thickness 6, is considered between two horizontal
streams of velocities U1 and U2 and densities P1 and P2, with the density interface
of thickness 6. They define Ri = Abb,/(AU)2.
The velocity profiles agreed remarkably well with the hyperbolic function, often
used in stability analysis. The difference between the stability of a sheared layer
which is homogeneous and that which has a stable density interface was demon
strated.
Discussion
The effect of stratification on sheared layers is complex, with the mode unstable
in the absence of stratification, called Rayleigh waves, being stabilized while a new
one, the Holmboe mode is now unstable. The mode destabilized by gravity has
a nonzero wave speed when riding at the mean velocity (U1 + U2)/2. In these
coordinates, the disturbance is assumed to consist of one wave traveling upstream
and one traveling downstream, with the interface a standing wave of increasing
amplitude.
Disturbances in the case of a homogeneous shear layer can be thought of as
two almost independent distortions of the upper and lower boundaries of the con
stant vorticity region. Short wave length disturbances are totally independent. The
amplitude of the disturbance oscillates as the two distortions alternately reinforce
and obstruct. However, long wavelength disturbances influence each other to such
an extent that "slippage" of the upper and lower distortions can be stopped. The
relative phase is fixed in the position most favorable for growth (PFMMG) of the
perturbation. In the stratified case, additional vorticity is generated by the distor
31
tion of the central interface. This baroclinic vorticity is responsible for inhibiting
instability at low Richardson numbers (Rayleigh waves); however, at high Ri, strat
ification alters the slippage of the distortions such that the wave lingers more at the
PFMMG than in unfavorable regions (Holmboe waves).
In the regions of instability of Rayleigh waves, both Holmboe and Rayleigh
waves are indistinguishable, both being phase locked, and nonlinear growth is by
rollup or overturning. Previously well distributed vorticity is now concentrated
into discrete lumps along the interface and breaking is violent.
In regions where Holmboe waves are unstable, no rollup occurs. Interface
displacement simply grows in magnitude with each succeding oscillation, ultimately
breaking at the crests, which may be on both sides or not, according to as the
excitation is unforced or not, respectively.
3.2.2 Smyth, Klaassen and Peltier (1987)
These investigators performed numerical simulations of the evolution of Holm
boe waves. A series of simulations using progressively lower levels of stratification
led to Kelvin Helmholtz (KH) waves. The effect of strong statification on KH
waves depends on the ratio of the vertical distances over which the density and flow
velocity, i.e., 6 and 6, change.
If 6 > 6,, increasing stratification stabilizes the flow.
If 6 < 6,/2, increasing stratification causes the KH wave be replaced by Holm
boe type oscillatory waves.
From linear theory, the relationship between KH and Holmboe type instabilities
can be shown to be equivalent to a damped oscillator, governed by,
A"(t) + bA'(t) + cA(t) = 0
Stratification, represented by c, provides the restoring force. Shear, represented by
b, serves to transfer energy into or out of the oscillation.
32
Solutions are of the form A elt, where a = a, + ioi, subscripts denoting real
and imaginary parts respectively. If ai = 0, we have a monotonically growing distur
bance, i.e., KH waves. However, if c/b2, which is analogous to the bulk Richardson
number, grows beyond a certain value, this train gives way to oscillatory Holmboe
waves.
A linear analysis of the governing hydrodynamic equations was performed to
determine, for a given level of stratification and Ri (with 6, being the length scale),
the value of a, the wave number, which has maximum growth rate, a,, to determine
the horizontal wave length to impose on the nonlinear model.
The plot of a(a, Ri) showed that, for small values of Ri(< 0.3), the fastest
growing modes had ai = 0; while for higher Richarson numbers, ai had nonzero
values, i.e, Holmboe instability. Two points were taken from the Holmboe regime
and one from the KH regime for nonlinear analysis. By analysing the evolution of
the non dimensional perturbation kinetic energy for the three points they confirmed
the nature of the instabilities predicted by the linear analysis : slow exponential
growth coupled with fast oscillations characterising disturbances in the Holmboe
regime and monotonically growing waves in the KH regime.
Holmboe waves have two components, with equal growth rates and equal but
oppositely directed phase speeds. The position most favorable for growth (PMFFG)
is just before the "inphase" configuration in accordance with Holmboe (1962). In
the "in phase" configuration, the kinetic energy is maximized. The phase speed
is maximum just beyond this "inphase" position. This implies that as the level of
stratification decreases, the maximum phase speed increases relative to the cycle
averaged speed, resulting in a greater time spent in the PMFFG and thus effecting
increasing growth rates. When this level of stratification is further decreased, the
phase speed at the PMFFG should vanish, with phase locking of the two compo
nents. They should now rotate as a unit and grow into intertwined fingers of heavy
and light fluid as in KH waves.
With decreased stratification in the Holmboe regime, growth rates and oscil
lation frequency reduced as predicted, and also, the phase speed increased after
leaving the "inphase" position. With evolution, thin plumes of fluid were ejected
from the peaks of the waves, primarily after passing the "inphase" configuration.
The KH regime simulation, too, was in accordance with linear predictions.
3.2.3 Lawrence, Lasheras and Browand (1987)
Two layers of different velocities and densities were separated by interfaces of
thicknesses 6, and 6, respectively. The centers of the two interfaces were separated
by a distance d.
Theoretical Analysis
An eigenvalue relation was derived from the Taylor Goldstein equation and
stability diagrams are plotted of Ri vs. a, for different values of e, where, Ri =
Af5, a = k6, = instability wave number, e = 2d/6,, Ab = 12 AU = UIi Us2,
k = 27r/A, and A = wavelength.
With E = 0, there were two modes of instability : a nondispersive Kelvin 
Helmholtz type for Ri < 0.07 and a dispersive one, the Holmboe type, for all
(positive) Ri. In the overlap region, 0 < Ri < 0.07, KH had higher amplification
rates. For e > 0, the KH mode as well is dispersive and has higher growth rates.
For e > 1, the Holmboe mode disappeared.
Experimental Observations
For e > 0, concentrated spanwise vorticity was observed above the interface, in
the high speed layer only (and none in the lower low speed layer), causing inter
facial cusping into the upper layer. Initial instability was two dimensional. As Ri
decreased, the wavelength of the disturbances increased. At lower Ri, disturbances
developed considerable three dimensionality, with wave breaking, similar to KH bil
lows. This billowing was only in small wisps, demonstrating the inhibiting effect of
34
buoyancy. With increasing Ri, at fixed e, this tendency decreased and thin wisps
were lifted almost vertically into the upper layer. Instabilities were observed to pair
in the same manner as KH instabilities in unstratified fluid, with wisps ejected, just
after this pairing.
3.2.4 Narimousa and Fernando (1987)
The investigators discuss the effects of velocity induced shear at the density
interface of a twofluid system. One of their most important conclusions has been
regarding the entrainment Richardson number relationship : Eu oc (Ri;"), where,
Eu is an entrainment coefficient = Ue/u, u, = entrainment velocity, u = scaling
velocity, Ri = Richardson number = Abh/u2, Ab = interfacial buoyancy jump,
h = mixed layer depth, and n = a coefficient.
The investigators used a recirculating flume, which was free of the rotating
screen of the more popular annular flume experiments. Their twofluid system
consisted of initially fresh and salt water layers. The mixed layer (of initially fresh
water) was selectively driven over the heavier quiescent fluid by using a disk pump,
developed by Odell and Kovasznay (1971). The velocity of the mixed layer was
varied between 5 15 cm/s using variable pump rotation rates. Shear layer velocity
profile appeared linear while that in the viscous diffusive momentum layer resembled
Couette flow profiles.
For moderately high Richardson numbers, Ri, > 5, the density interface was
found to be topped by a thin layer of thickness 6,, with a weak density gradient
which had not yet got well mixed. This partially mixed fluid results owing to
the fact that energy of the eddies is not strong enough to entrain the fluid from
the stable interfacial layer, and mixing can only occur by wave breaking resulting
from the mixed layer turbulence at higher Richardson numbers, i.e., eddies assist
entrainment in two stages, from the interface to the intermediate layer and from
there into the mixed layer.
Mixed
Layer
NonTurbulent
Layer
Figure 3.2: Physical description of the complete flow configuration, with density
and velocity profiles (adapted from Narimousa and Fernando 1987.
P(z)
 SI
`_17
36
Fluid above this layer was homogeneous. At low Ri,, with high rates of en
trainment, the intermediate layer was absent. The entrainment interface consisted
of regularly spaced billows with high spatial density gradients within, with their
centers having small scale irregularities which could be the effect of local instability
regions due to the entrainment of heavy and light fluid into the core. However,
the final stage of mixing within these billows was fairly slow, with breakdowns into
regions containing small scale structures which may be due to the interaction of
two adjacent vortices. With increasing Ri, the frequency of billows progressively
decreased and entrainment was dominated by a wave breaking process, with wisps
of fluid being ejected into the upper layer. This kind of behavior was seen over
a whole range of Ri,(5 < Ri, < 20), with decreasing frequency as Ri, increased.
Also, large amplitude nonbreaking solitary waves were seen over Riu = 10 20.
The shear layer is very important as it is responsible for the turbulent kinetic
energy of entrainment and thereby controls the size of the energy containing eddies
at the interface. The investigators found that 6,/h was independent of Ri' (and
about 0.2) indicating that the size of the eddies should be scaled by h.
The average measured value of 6/h was also independent of Ri,, and around
0.040.08. This ratio was also confirmed by another interpretation of data as follows:
Observing that the buoyancy in the mixed layer and the gradient in the interface
are constant,
b(z) = bo + Ab(z h 6)/(6) for (h < z < h + 6) (3.1)
where, b(z) = mean buoyancy at elevation z, bo = buoyancy of lower unperturbed
layer, and z is positive down from the free surface.
Assuming horizontal homogeneity, Long (1978) integrated the buoyancy conser
vation equation,
Ob Oq
t = (3.2)
at 8z
37
where, q(z) = bw = buoyancy flux; b and w being the values of buoyancy and
vertical velocity fluctuations, respectively.
This yielded,
q(z) = q2z/h (0 < z < h) (3.3)
q2 = hd( (3.4)
dt
Sr2 d(Ab) Abr2 d6 Ab dh
q(z) = q2+( T) (h < z < + ) (3.5)
26 dt 62 dt 6 dt
where, q2 = buoyancy flux at the entrainment interface, and r = z h.
As q(h + 6) = 0, it is possible to obtain
d{Ab(h + 6/2)} 0 (3.6)
dt
By defining a characteristic velocity scale based on the initial buoyancy jump
and the depth of the initially homogenous layer, i.e.,
Vo2 = hoAbo
and defining 6 = ah one finally arrives at
h(1 + a/2) = Vo0/Ab (3.7)
Plotting this equation showed 6 ~ 0.06h.
Energy Budget Analysis
Analysis of the energy budget yielded the result that buoyancy flux, turbulent
energy production and dissipation terms were of the same order and that E ~ Ri1'.
Wave amplitudes at the interface, 6,, scaled by h were of the order of Ri1/2.
This may possibly be due to the energy containing eddies impinging on the interface.
The vertical kinetic energy of the eddies = w2 (where wi is the rms fluctuation of
the vertical velocity). Then the generated potential energy of the waves ~ N26.
Thus, 6, wi/N, where N = (Ab/6)1/2 = boundary frequency of the interfacial
I
layer.
6 ~ h (3.8)
wl ~ Au (3.9)
Hence,
S~ Ri1/2 (3.10)
Summary
(1)During entrainment, two layers, the density interfacial layer and the shear
layer, having direct bearing on the entrainment process developed and increased
linearly, independent of Ri,.
(2)Billows, formation and breakdown of large ordered vortices cause mixing at
low Ri,, while breaking waves cause it at higher Ri,.
(3)Wave amplitudes scaled well with the size of the energy containing eddies of
the size of the mixed layer.
(4)The rates of work done against buoyancy forces, kinetic energy dissipation
and shear production of turbulent kinetic energy were of the same order.
3.3 Conclusions
The preceding discussion documents some of the modes of interfacial instability
which are possible. The mode of instability is dependent on the stratification and
the ratio of the thicknesses of the shear layer and the density interface. When
6,/6 > 1, increasing stratification causes monotonically growing KelvinHelmholtz
waves to be replaced by the oscillatory Holmboe mode. The physical nature of the
modes differs as well, in that KelvinHelmholtz waves are associated with billowing
and lumping (and pairing) of vorticity near the interface, while Holmboe waves are
characterized by a series of nonlinearly crested waves cusping generally into the high
speed layer only. Billowing as well as cusping into the high speed layer were observed
in laboratory experiments by Narimousa and Fernando (1987) with the transition
39
in the mode of instability occurring with increasing Richardson number. Moore and
Long's (1971) experiments to determine entrainment rates in velocitysheared salt
stratified systems (see Chapter 4) also describe some of these phenomena in detail.
In effect, it can be concluded that velocityshear has a destabilizing effect on stable
stratification and can cause upward mixing of the heavier fluid. This effect of the
growth and breakdown of instabilities is examined in the next chapter.
CHAPTER 4
ENTRAINMENT IN STRATIFIED SHEAR FLOWS
4.1 General Aspects
The effect of interfacial instabilities in causing entrainment across the (stat
ically) stable density interface is considered in this chapter. As a considerable
amount of worthwhile and interesting work has been done on both shear flows and
flows without mean shear, a complete review is beyond the current scope. Thus,
only directly pertinent studies as regards shear flows are reviewed. Moore and Long
(1971) discuss their results with respect to those obtained by previous investiga
tors and Long (1974) theoretically examines many of these results, thereby making
this literature especially riveting. A recent experimental study by Narimousa, Long
and Kitaigorodskii (1986) is also reviewed. Not much published work is available
specifically regarding vertical entrainment of fluid muds, and thus the study using
kaolinite by Wolanski, Asaeda and Imberger (1989) is reviewed in spite of it being
for a meanshear free environment.
4.2 Moore and Long (1971)
The experiments were run in a racetrack shaped flume with a system of holes
and slits in the floor and in the ceiling, allowing fluid injection and withdrawal
to produce required steady state horizontally homogeneous shearing flows. Their
steady state was defined as keeping the level of the density inflexion point constant.
In the steady state twolayer experiments, the density and velocity profiles were
kept constant by adjusting the flow rates and replenishing salt to the lower saline
layer. This amount of salt per unit time, on dividing by the horizontal cross section
41
area of the flow tank, gave the salt mass flux.
In the entrainment experiments, the tank was filled with fluid with a linear
density gradient and then circulation of either fresh or salt water was started and
the density profile observed as a function of time.
4.2.1 Results of Two Layer Steady State Experiments
The investigators' overall Richardson number was defined as, Rio = ItAb/(2Au)
where, H = total depth, Ab = buoyancy difference between the top and bottom
layers of fluid, and 2AU = velocity difference between the top and bottom layers of
fluid. Also, q = buoyancy flux, and, Q = q/Ab(2AU) = non dimensional buoyancy
flux.
A layer of thickness 6,, with a velocity gradient, separating two homogeneous
layers of depths h each, developed. At low Rio, 6, was very large and decreased
with increasing Rio, until it ultimately became quite small.
For values of Rio greater than about three, turbulence in each homogeneous layer
caused erosion to a considerable extent of the layer over which the density gradient
initially manifested. The interface was clearly visible. The surface of the interface
was irregular in shape (with amplitudes ~ 0.5 cm, wavelength ~ 34 cm and width
~ 1 cm) with wisps of fluid being detached from the crests of disturbances, this
phenomenon being more observable for disturbances cusping into the lower density
layer. The speed of these waves was less than of the homogeneous layer above.
These grew in amplitude and then simply disappeared with a wisp of fluid ejected
from the tip, indicating that the original disturbance may well have been caused by
eddies scouring the interface, with it's "roller action" drawing dense fluid up into a
crest before it sharpened and was sheared off.
For values 1.5 < Rio < 3.0, the interface was less sharp and more diffuse
(with 6, increasing). The thickness of the region with the density gradient, 6,
also increased, as did the salt mass flux. Mixing now seemed to be more due to
42
internal wave breaking. For Rio < 1.0, very large eddies extended through the
diffused interface. For low values of Ri, 6, ~ 6 while for higher values of Rio,
6, > 6. Richardson number, Ri,,, defined using the average density gradient and
average velocity gradient over 6, had a value close to one.
Plotting the nondimensional buoyancy flux with Rio yielded the functional
relationship
Q = C1/Rio (4.1)
with C1, which may be weakly dependent on kinematic viscosity and diffusivity,
having a value ~ 8 x 10.
Other researchers have obtained relationships between E and Ri, where
E = u,/u (4.2)
with the entrainment velocity u, defined as the normal velocity of the interface, or for
steady flow experiments, the volume flow rate of the fluid being entrained divided
by the cross sectional area over which this is occurring, u = some representative
velocity and Ri = Richardson number computed for that particular experiment,
with Ap always representing the density jump between the turbulent homogeneous
layer and the fluid being entrained.
Rouse and Dodu (1955) used a two layer fluid system with turbulence being
generated by a mechanical agitator and pointed out that if the entrainment rate is
proportional to Ri1, the implication is that the rate of change of potential energy
due to entrainment is proportional to the rate of production of turbulent energy by
the agitator.
Ellison and Turner (1959) discussed entrainment rates of a layer of salt water
of thickness D flowing with velocity i under a layer of fresh water. Defining Ri =
AbD/U2, they obtained E ~ Ri1 for Ri < 1.
Lofquist (1960) got a similar relationship for Ri < 1, but his data were scattered
43
for Ri > 1, with a faster decrease in entrainment rates than is indicated by E ~
Ri1.
Turner (1968) studied mixing rates across a density interface with turbulence
being generated on either or both sides by a mechanical agitator and obtained
E ~ Ri1 for Ri < 1, but E ~ Ri3/2 for Ri > 1.
Kato and Phillips (1969) applied a constant shear stress r = pu2 at the upper
surface of a linearly stratified fluid and obtained E ~ Ri1, with values of Ri.
equivalent to Rio < 1. These investigators also demonstrated that the entrainment
coefficient E represented a time rate of change of potential energy per unit mass V1,
in non dimensional terms, i.e.,
2po dVI u,
2po dV1 E = KRi1 (4.3)
gApu, dt u,
with, Ri, = gh/u2 and K is some constant.
Moore and Long (1971) used this basis to compare their functional relationship
with other researchers and showed that the nondimensional flux is essentially the
same as an entrainment coefficient. Another way of showing this relationship is as
follows :
If the injectionwithdrawal system at the top is turned off and the interface
allowed to rise a distance dh = u, dt, then [mass(t + dt) mass(t)] = mass added
at the bottom = dm.
Letting lower density = pi + Api/2 and upper density = pi Api/2,
1 1
(Pi + 2Ap)(HT/2 + dh)A + (p Api)(H/2 dh)A
2 2
1 1
(Pl + 2API)(t/2)A (pi Apl)(A/2)A
= dm (4.4)
Thus,
Ue = dh/dt
= (1/ApiA)dm/dt (4.5)
Therefore,
UeAb = q (4.6)
If u, is defined thus for the steady state experiment, too, we get,
Q =E (4.7)
Thus, E ~ Q ~ Ri1 should be valid over 0 < Rio < 30, as evidenced by the
Moore and Long experiments. Lofquist's results maybe attributed to the horizontal
inhomogeniety of his experiments, while Turner's maybe due to the absense of a
mean velocity to his flow, his method of definition of the Richardson number, or
the absence of what he calls fine structure in his experiments.
These relationships were considered in terms of energy changes and it was shown
that the rate of change of potential energy of the system or the buoyancy flux and
the rate of dissipation of kinetic energy per unit volume were of the same order.
4.2.2 Results of Entrainment Experiments
The initially linearly stratified fluid was eroded and replaced by a homogeneous
layer of depth h(t), when the injection withdrawal system was applied to only one
side of the channel. The results showed that h3 oc t, similar to Kato and Phillips
(1969).
4.2.3 Summary
Over the range of Richardson numbers studied, results showed that the existence
of turbulent layers on either side of a region with a density gradient caused erosion
of this region to occur, with the formation of two homogeneous layers separated by
a layer with strong density and velocity gradients. The gradient Richardson number
of this transition layer tended to have a value of order one. The nondimensional
buoyancy flux Q was functionally related to the overall Richardson number, Rio,
by Q ~ Ri;o for 0 < Rio < 30. Entrainment experiments of an initially linearly
45
stratified fluid with the application of shear on one side resulted in the formation of a
homogeneous layer separated by an interface from the stratified layer,with h3(t) cc t.
4.3 Long (1974)
Long critically analyzed mixing processes across density interfaces including
cases without and with shear, which have been shown by previous investigators to
have different relationships with an overall Richardson number, Ri., based on the
buoyancy jump across the interface, the depth of the homogeneous layer and the
intensity of turbulence at the source.
At large Reynolds (Re) and Peclet (Pe) numbers, the fluxes of heat or salt and
the entrainment velocity appear to be proportional to minus one and minus three
halves powers of Ri. for flows with and without mean shear respectively, where the
higher entrainment rate for shear flows is attributed to the decrease of rms velocities
near the interface for increasing Ri. for cases of zero shear. Conforming to our area
of interest, this discussion will be restricted to the cases with mean shear.
Kato and Phillips (1969) applied a constant shear stress r = pu2 at the surface
of initially linearly stratified fluid in an annular flume using a rotating screen. This
resulted in the development of an upper homogeneous layer and lower stratified fluid
with an interfacial buoyancy jump Ab. Defining the rate of downward propagation
of this interface as u,, the investigators arrived at
Ue/u* = KiRi,1 (4.8)
with, Ri, = hAb/u*, h = depth of the homogeneous layer and K1 is some constant.
They also found that U/u, increased with time, where U is the speed of the
screen, with u, held constant. A simple analysis also reveals this quantity to be
independent of the Richardson number.
In Moore and Long's (1971) experiments in a race track shaped flume with salt
46
and fresh water, buoyancy flux q was measured at steady state, yielding
q = K2(Au)3/h (4.9)
with Au being the mean velocity difference of the two layers and K2 is some con
stant. Defining the entrainment velocity by, u,Ab = q, this yields equation (4.8) on
making the plausible assumption that Au/u. is independent of Ri., where pu2 is
the constant momentum flux in the tank.
The theory (Turner 1973) that erosion of the interface should depend on the
properties of turbulence near the interface (and not at the source), especially on the
rms velocity scale ul and the integral length scale 11 near the interface proposes a
relationship of the form
U,/u, = f(Ri) (4.10)
with, Ri = liAb/u2 assuming no dependence on any other quantities, and large Re
and Pe.
Now, we have,
ar a ((4.11)
8z at
with i as the mean horizontal velocity at depth z.
For the Moore and Long steady state experiments,
ar
=0
oz
As r = pu'w' and the correlation coefficient is of order one in the homogeneous
layers, thus, u. = r/ is proportional to ul and li ~ h. Thus,
u,/ul = KsRi' (4.12)
where Ks is a constant. The energy equation for these experiments is,
a /2) (4.13)
w(C /2) = [w (c r/2 + p'/po) + rn + q E(4.13)
where, c' = turbulent speed, p' = turbulent pressure, and e = dissipation function.
47
Now, the velocity difference is proporional to V/F and the two energy source
terms as well as the dissipation function are of order ul/h or u/l1 near the interface.
Assuming that q ~ uAb is of the same order, one again arrives at
Ue/ul ~ Ri1
as in equation (4.12).
The shearing experiments indicate that
q ~ ulh u./h (4.14)
In the homogeneous layer near the interface, q ulbl where bl is the rms
buoyancy fluctuation. With the assumption that this correlation is of order one,
we obtain that u2/(bih) 1, thus, the kinetic energy and the available potential
energy, b1hl, are of the same order.
Long thus interprets the experiments to indicate that turbulence causes poten
tial energy to increase at a rate proportional to the rate at which kinetic energy is
supplied to the region of the interface and not necessarily to generation at the source.
A plausible unifying argument leads to the conclusion that entrainment rates in
cases with or without shear are proportional to Ri' defined on the buoyancy jump
and velocities and lengths characteristic of turbulence near the interface.
4.4 Narimousa, Long and Kitaigorodskii (1986)
The flume and pump section in this study were the same as used by Narimousa
and Fernando (1987). Experiments were run with two kinds of systems : a linearly
stratified system and a two layered fluid system. We will confine our discussion to
the latter which comprises of fresh water over salt water.
During entrainment, interfacial Kelvin Helmholtz instabilities and wave break
ing were easily observed. At low Richardson numbers (Ri), turbulence caused the
disturbances to be highly irregular, however, this irregularity decreased as Ri grew
and finally internal waves developed and occupied the entire interfacial layer. These
48
instabilities (disturbances) were larger for higher pump speeds and smaller density
jumps.
These investigators attempted to find a relationship between E,, where the
entrainment velocity was scaled by the friction velocity u,, and Ri,, based on u, as
well, i.e.,
E. = ue/u, (4.15)
Ri, = hAb/u* (4.16)
Measurements were made of the mean mixed layer velocity, U, the mixed layer
depth, h, and the entrainment velocity u,, while the friction velocity u, was deduced
from the mean momentum balance equation for homogeneous turbulent shear flow.
Plots of h vs.t revealed that ue = dh/dt was constant.
4.4.1 Deduction of u,
The streamwise momentum equation for the mixed layer, the interface and a
thin layer below it where the velocity drops to zero, is
d(Uh)/dt = u w2hlW (4.17)
where u, is identified with the pressure gradient force and Reynolds stress force
accelerating the flow due to pump action and is the friction velocity of the pump;
while w, is the friction velocity of the side walls, W being the width of the side
walls. The second term accounts for the retarding action of the side walls.
The Blasius resistance formula for turbulent channel flow is,
U/w, = 8.74(Ww,/2v)1'/ (4.18)
i.e., w, = 0.15U7/8(2v/W)1/8 (4.19)
d(Uh) (Uh) ah
also, dt A at
SUdt (4.20)h
= Ucu, (4.20)
Thus,
u. = U,u, + {0.15U7/8( )1/8}2h/W (4.21)
As U,(= a ( h) was obtained from graphs, u, was easily calculated, and found to
increase very slowly with h.
The measured S = Uh, on being plotted against h had two distinct regions,
initially increasing linearly and then remaining constant, leading to the interesting
observation that U decreases with h, first slowly and then faster as the pump term
is balanced by the wall friction term. Experimentally, it was determined that U 
10.65u*.
4.4.2 Entrainment Rates Based on u,
The investigators start out with the assumption that hAb is constant, in slight
contrast to Narimousa and Fernando (1987) who had
hoAbo = hAb(1.03) (4.22)
which may be due to the fact that the region of the thin density interface was not
considered while employing the buoyancy conservation equation. This enabled them
to develop plots of E, vs. Ri,.
No simple unifying relationship was found over the entire range of Ri,, but for
subranges they found,
E, 0.65RiC1/2 15 < Ri, < 150 (4.23)
E, 7Ri1 150 < Ri, < 800 (4.24)
E, 5Ri;3/2 Ri, > 150 (4.25)
This can be attributed to the difference in the very nature of the entrainment
process over the three ranges. Initially, the mixed layer as well as the base of
the mixed layer are fully turbulent; the high turbulent shear and weak density
jumps result in the eddies of the mixed layer directly producing entrainment; in
50
the next range, the interface becomes less chaotic and Kelvin Helmholtz type of
instabilities occur which are less efficient in causing entrainment. Finally, due to
even lesser shear, entrainment decreases further, and this is similar to previous shear
free experiments (oscillating grid type), where Long (1978) obtained E ~ Ri7/4,
which is close to the one obtained here.
4.5 Wolanski, Asaeda and Imberger (1989)
Turbulence was generated in a plexiglass cylinder using oscillating grids along
it's walls. The cylinder was filled with a fluid mud mixture of kaolinite and tap water,
with initial concentrations always greater than 40 g/1. The grids were stopped after
fully mixing the fluid mud. A lutocline formed, separating the clear, upper layer
from a turbid bottom layer, and moved down with a constant velocity wfo, which
depends on the suspended sediment concentration. As the oscillation was started
again, the fall velocity wf reduced to less than wpo. There was no mixing for stroke
frequencies w < we, the critical frequency at which billowing and wisp formations
occurred. In this frequency range, wf/wf, decreased for increasing values of w, which
can be attributed to the breakup by turbulence of the clay flocs. At w = we, there
was active mixing across the interface, which eroded by moving downwards at a
velocity greater than wf/, thereby implying the presence of a higher intensity of
turbulence in the upper layer. However, very soon, the fall velocity decreased to
below wfo again as a balance prevailed the upward turbulent entrainment and the
downward gravity settling at the lutocline, with no more erosion of the lutocline.
For w > we, the lutocline was convoluted with large internal waves cusping into
the upper layer where the intensity of turbulence was lower due to sediment induced
dissipation. The height to which the fluid was entrained increased with increasing
stroke frequency. Also, the onset of turbulence occurred at almost the same value of
the Richardon number for all the experiments.
Thus, similar to heat and salt stratified experiments, buoyancy effects are dom
51
inant in inhibiting mixing across the lutocline. However, an additional feature
affecting the process is the extraction of turbulent kinetic energy by the sediment
to counteract the buoyancy flux due to sediment fall velocity, which, here, causes
a collapse of turbulence in the bottom layer with resultant erosion of the lutocline
only from the top. It must be mentioned that Wolanski and Brush (1975) found, in
oscillating grid type of experiments, that the entrainment rate decreased much faster
with increasing Richardson numbers than was the case with salt or heat stratified
experiments.
The dependence of the fall velocity on the suspended sediment concentration
served to limit the height of entrainment into the top layer and also stopped lutocline
erosion after an initial period of active mixing, which is what E and Hopfinger (1987)
as well had observed.
4.6 Conclusions
The preceding review of flows with mean shear show that many investigators
have found relations of the form E oc Ri' for saltstratified systems, although the
range of validity of this relation varies according to the method of defining the
Richardson number (see, for instance, Appendix B). Moore and Long (1971) re
lated the nondimensional buoyancy flux, Q, to the Richardson number according
to a similar (Q oc Ri') relation and showed that the entrainment coefficient, E
is equivalent to Q. Narimousa et al. (1986) arrived at different entrainment re
lations for (three) subranges of the Richardson number defined on the basis of
the friction velocity u,. The exponent in the E, oc Ri;" increased with increasing
Richardson number, thus resulting in decreasing rates of entrainment with increas
ing Richardson number, which they attributed to the difference in the very nature of
the entrainment processes over the subranges. Wolanski et al. (1989) demonstrated
that the behavior of sediment particles or aggregates is different from saltstratified
systems because of greater dissipation of turbulent kinetic energy to counteract the
52
sediment fall velocity thereby implying a lesser rate of entrainment for sediment
stratified systems, which is the focal point of the present investigation.
CHAPTER 5
METHODOLOGY
5.1 Apparatus
Turbulence generated by grid stirring has been the most popular mode of labo
ratory studies of dynamics of stratified systems. In a twolayered system separated
by a density interface, the grid is placed in either layer or there might be a system of
grids placed in vertical succession and extending into both layers. It can be shown
that the grid may be replaced by a virtual source of energy at a horizontal plane,
the "action" of the source being determined by a single "action parameter" (Long
1978) having the dimensions of viscosity and proportional to the constant eddy
viscosity in the turbulent fluid above the source. However, the issue of entrain
ment of fluid muds in estuarine situations is obviously the result of currentshear
induced turbulence. Thus, it appears more prudent and realistic to simulate this
phenomenon with a laboratory apparatus which can produce the required turbulent
kinetic energy for mixing by velocity shear.
In this respect, most previous experimenters have used flumes with annular ge
ometries, with a rotating screen applying shear stress at the surface of the stratified
fluid within it (Kato and Phillips 1969; Kantha, Phillips and Azad 1977; Deardorff
and Willis 1982). This annular flume has the advantage of being free of end walls
and thus avoids undesirable recirculating flow (as in some previous surface shear
free experiments with salt and fresh water of Ellison and Turner 1959, and Chu
and Baddour 1984). But, this kind of arrangement seems affected by secondary cir
culations in the radial direction (Scranton and Lindberg 1983) causing substantial
interfacial tilting. Deardorff and Yoon (1984), after an indepth study concluded
54
that the cause for this tilting lies in the uneven angular momentum distribution
across the annulus due to the solid body rotation of the screen. The fluid possses
a mean velocity towards the outside resulting in higher entrainment rates at the
outer wall relative to the inner wall.
The experiments were carried out in a specially designed recirculating flume (see
Figures 5.1, 5.2 and 5.3) which basically consists of two sections, the pump and
the observation sections, joined together by two semicircular annuli. This kind of
flume has been used by previous experimenters (for example, Moore and Long 1971;
Narimousa, Long and Kitaigorodskii 1986) and is free of the effects of end walls and
that of a rotating screen. Some secondary circulation is introduced in the process
of bending the flow; however, this is possibly to be minimized by the large radius
of curvature and the relatively long straight section used for observations. Here, it
might be noted that in some experiments of this general nature, secondary circula
tions are not undesirable (even though the geometric dimensions of the apparatus
will dictate their transverse length scales), since turbulent geophysical flows also
contain them (e.g., longitudinal rolls or Langmuir cells), albeit with independent
preferred wave numbers. Although the effect of streamline curvature is not too well
understood, the effects of variation in transverse length scales does not appear to
cause substantial variation in entrainment (Scranton and Lindberg 1983).
The flume was entirely made of plexiglass to enable visualization of the flow and
other desired parameters. Except for the walls of the two semicircular sections, the
plexiglass used was 1.25 cm thick everywhere, including the bed and the floor of the
flume. The walls of the semicircular section were 0.32 cm thick, this merely being
expedient to afford ease of bending to the design radius of curvature. The flume
was 61 cm over the floor throughout. The entire unit was placed on a specially built
table.
A 'bed' was constructed first and placed on the table. The floor of the flume
1.27cm Plexiglass
0.32cm /
Plexiglass
0.32cm
Plexiglass
1.27cm Plexiglass
PLAN
Figure 5.1: Recirculating flume of plexiglass used in the present investigation (di
mensions in centimeters).
Spur Gears (R=8.5)
Splitter
Plate
Flume
Wall
Drive
Mechanism
Support
Figure 5.2: Section AA of the flume, from Figure 5.1 (dimensions in centimeters).
Propeller Drive Motor
Flexible Shaft
Flume
Walls
Propeller
6 T' Splitter Plate
30.8
Floor
Bed
SECTION BB
Figure 5.3: Section BB of the flume, from Figure 5.1 (dimensions in centimeters).
58
was bolted and glued onto this. Next, the walls were cut, and bolted and sealed to
this floor. The walls were supported with 20 cm high and 5 cm wide sections at
periodic intervals and also connected with brackets at the top. Joints were sealed
with gussets and rubber to prevent any leakage.
The required turbulent shear flow was obtained by using a disk pump (see
Figure 5.4), first introduced by Odell and Kovasznay (1971), to selectively drive the
upper fresh water layer over the quiescent fluid mud.
The width of the flume was 10 cm everywhere, except at the pump section.
The disks of the pump were between walls 47.5 cm apart. To maintain a constant
crosssection of flow as much as possible, flow seperator sections (triangular in plan,
of dimensions 73.7 x 73.7 x 31.8 cm) were placed both up and downstream of the
pump, which created two channels of 5 cm width each on either side of the pump.
These channels guide and blend the flow into the semi circular section in front of
the pump, while upstream of the pump, these split and guide the flow onto the
pump. The flow seperators were of the same height as the flume, i.e., 61 cm. Also,
the two flow seperator sections were connected by a weighted wooden box of height
30.8 cm which served to maintain the required 5 cm width. The hole of the intake
valve for fluid mud was directly under a transverse slot (of dimensions 38 x 13 x 3
cm) at the bottom of this box. The entire pump section had vertical supports at
every 40 cm.
The radius of curvature of the curved section was 51 cm. Here, supports were
put at closely spaced intervals of 25 cm, to take into account the additional stresses
due to bending. The observation section was completely straight and 200 cm long.
Small holes were drilled into the outer wall of the observation section, near the
entrance (in the flow direction) to the observation section and thin, flexible pipes
were inserted into them, taking care that they did not intrude into the interior of
the flume. The other ends of these pipes were closed with metal clips. These holes
IV26 I=
0.325
0.s.
Nut Nut
All dimensions in centimeters
Vertical dimensions are greatly
exaggerated.
Figure 5.4: Details of the disk pump system used in the present investigation (di
mensions in centimeters).
60
were at closely spaced vertical intervals14 cm apart. This was done to extract
samples of fluid mud for the estimation of concentration profiles (as a function of
time). At the center of the outer wall of the observation section, a 38 x 56 cm size
grid of 2 cm mesh on a transparency was pasted to record the rate of progress of
injected dyelines. The entire flume had brackets at the top, every 10 cm, to tie the
walls together.
A Sony Betamax video recording system was set up about 1 m from the obser
vation section of the flume such that the line of vision of the camera was normal to
the sidewall of this section. The camera was focused onto the grid at the center of
the wall. Two powerful (1000 W) lamps were set behind the camera to provide the
requisite illumination for recording.
The reason for a disk pump was that it could produce quite homogeneous hor
izontal streaming of the flow. The pump imparted only a horizontal component of
velocity to the fluid. The disks of the disk pump were of two different diameters,
8 and 26 cm, which were alternately stacked on each of the shafts. The larger
disks were 0.325 cm thick, while the smaller ones were 0.65 cm thick. These shafts
were so positioned that a large disk of one shaft meshed with the smaller of the
other and so on. Thus, the two stacks meshed, leaving almost no space in between
them, but creating gaps at the outer edges, between the larger disks. When the
two shafts were driven in opposite directions by a 1/8th h.p. Dayton Permament
Magnet Gearmotor (F/L rpm 50, F/L torque 130 inch pounds) via a.chain and
sprockets arrangement, fluid was pulled around the outer channels by the viscous
drag of the larger disks and ejected as horizontal jets from within these gaps. The
disks were sandblasted to improve surface roughness to increase the efficiency of
the pump by increasing drag. With the disk pump in place, the bottommost disk
of the pump was just above the elevation of the splitter plate (described next).
Preliminary calibration tests were performed with a homogeneous fluid to test
61
the range of velocities obtainable with this pump. As the maximum mean velocity
obtained was only about 9 cm/s, it was decided to augment the velocity with the
assistance of a screw propeller. This propeller was placed (the axis of the propeller
was ~ 2 cm above the level of the splitter plate which is described below) in the
curved section downstream of the pump and before the entrance to the observation
section (Section BB, Figure 5.3), and driven by a motor placed outside via a flexible
shaft. The result was quite satisfactory with the maximum obtainable velocity with
the twolayered system of fluid mud and water in place being 14 cm/s. Also,
the horizontal homogeneity of the flow was not disturbed. To impart additional
horizontal homogeniety to the streamlines, a sidewalltosidewall thin metal splitter
plate was constructed in a horizontal plane 30.8 cms from the floor and this splitter
plate covered the entire pump section (which is the only region with a width greater
than 10 cm and having the flow seperators and the disk pump) and the entire curved
section downstream of the pump. This also served to prevent any suction effects
either due to the disk pump or the propeller (which together are referred as the
pump system hereafter) from affecting into the fluid mud below.
5.2 Procedure
Each experimental Run was divided into intervals of ~ 820 (generally ~ 10)
minutes each. At the end of each interval, the required measurements were "in
stantaneously" made, and these were considered to be the representative conditions
for that interval. It must be mentioned is that in the case of experiments with
saltstratified systems, the sole cause for the deepening of the mixedlayer is tur
bulent entrainment across the density interface. In a velocitysheared twolayered
system of fluid mud and water, the rate of propagation of the visual density inter
face (i.e., the rate of change of depth of the mixedlayer) is not the result of vertical
entrainment of fluid mud alone, but is also due to the settling characteristic of fluid
mud below the level of the interface. Thus, in this case, one cannot easily quantify
62
turbulent entrainment in terms of changes of the mixed layer depth. Therefore, a
more direct approach was adopted. The most basic effect of the turbulent kinetic
energy of the system is mass/buoyancy transfer across the interface. It therefore
appears to be the most logical quantity to measure and relate to a suitably defined
Richardson number.
From an estimate of the initial depth of the mixed layer, the flume was first filled
with the requisite predetermined height of tap water. Two types of test sediment
were consideredkaolinite and bentonite (see Appendix A), with the objective of de
termining the effect of varying degrees of cohesion on entrainment rates. Bentonite,
a montmorillonitic clay, is highly cohesive (and thixotropic) whereas kaolinite is not
as cohesive and properties of bentonite aggregates are not as uniform as that of
kaolinite. Sediment was wellmixed with tap water (for composition of tap water,
see Dixit 1982) in a vertical, steel cylinder of 77 cm diameter with the aid of a
IngersollRand twostage 10 h.p. air compressor with a maximum discharge rate of
14.0 kg/cm2. The compressed air was introduced into this vertical mixing tank at a
high flow rate through tiny holes in a Tshaped PVC pipe section placed at the bot
tom of the cylinder. This agitation was continued long enough until the fluid mud
was well mixed and quite homogeneous. In the case of kaolinite, the sedimentwater
mixture was thoroughly agitated for at least an hour which provided quite "homo
geneous" mixing, while bentonite was not as tractable in this respect. Bentonite,
which is highly thixotropic, formed lumps with a wide range of sizes (of upto 20
cm diameter) even when the sediment was introduced at a slow rate into agitated
tapwater. These lumps were dry inside although covered by a wet "skin". The
mixture was allowed to equilibrate for ~ 5 days with periodic agitation (upto ~ 6
hours every day) before a fairly uniform, workable mixture resulted.
The wellmixed fluid mud was instantaneously pumped into a horizontal cylin
drical feeder tank above the elevation of the flume. This fluid mud was then in
63
produced into the flume through the intake valve at the flume bottom. The mud
entered the flume with a vertical (upward) velocity at the position of the slot at
the bottom of the wooden box and on encountering the wooden obstruction turned
at right angles and flowed horizontally into the bottom of the flume, displacing the
lighter tap water upwards. With all stops open, the filling rate was about 2.5 cms
per minute per unit area of the flume. The time required to fill the flume with the
requisite volume of fluid mud was ~ 15 minutes. In the first three experimental
runs, the filling rate was slightly slower, while in the remaining runs the resulting
fluid mud layer underneath water was essentially homogeneous initially ( except
very near the bottom). The interface was always positioned so that it was just
under the level of the splitter plate such that the internal boundary to the diffusion
of momentum (Narimousa and Fernando 1987) formed by the interface, and the
physical boundary of the horizontal splitter plate would be almost continuous. In
all the runs, the method of filling fluid mud under water always resulted in the
formation of a diffuse intermediate layer (of thickness ~ 57 cms) just above the
interface. The density of this layer was found to be minimal (~ 105 g/cm3) and
this layer completely eroded within 11.5 minutes after starting the run.
As soon as the twolayered system was in place, the depth of the mixed layer, h,
was noted, and samples (N 10 cm3 each) of fluid mud, at discrete vertical intervals
(~ 5 cms), were withdrawn via the flexible tubes in the outer wall of the observation
section of the flume to obtain the initial concentration profile (across depth). These
samples were directly withdrawn into small (capacity ~ 60 cmS each), clean glass
bottles which were then tightly capped. The elevation of the position (from the
bottom of the flume) from which the sample had been taken was marked on the
corresponding bottle. The time was also noted. As the instantaneous concentra
tion profile was required, it was not considered expedient to spend more than 11.5
minutes for sampling, by which the number of samples was limited to a maximum
64
of six each time. The video recording system was turned on (to record the entire
experiment) and the experiment was begun by starting the pump system (the disk
pump and the propeller were started simultaneously) to rotate at a predetermined
rotation rate (which, in conjunction with varying buoyancy jumps across the inter
face, provided a wide range of Richardson numbers N 4 32). The rotation rate of
the disk pump was always at the maximum, while the rotation rate of the propeller
was adjusted such that the pump system could produce the desired predecided mean
initial velocity in the mixedlayer.
After the experimental run was in progress, with velocity shear causing fluid
mud entrainment across the density interface through massive undulations convo
luting the interface, sets of samples, for gravimetric analysis, were systematically
withdrawn at discrete time intervals (~ 10 minutes). Consecutively, dye lines were
also injected to get the corresponding velocity profile for that interval. The depth
of the mixed layer was noted.
Dyelines (of diluted rhodomine such that it would be almost neutrally buoy
ant in water) were injected into the flow and their movement across the grid was
recorded by the video camera. A syringe with a long needle (~ 35 cm) was used
for this purpose. The needle was introduced vertically into the observation section
through a slot in a bracket tying the sidewalls together at the top. The needle was
aligned with the upstream vertical edge of the grid and it's end was well within the
lower layer of fluid mud. The plunger was depressed and the spewing needle was
"instantaneously" pulled out leaving a clearly visible dyeline (which became diffuse
with downstream progress). The velocity profile could be easily determined by mea
suring the rate of downstream progress of this injected dyeline. For each interval,
dyelines were injected at least twice (and frequently three times) and averaged to
get a more accurate velocity profile. A problem which could not be circumvented
was that, below the level of the visual density interface, the turbidity of fluid mud
65
prevented visualization of the dyeline. However, Narimousa and Fernando (1987),
using a similar flume and pump system, found that the velocity rapidly decreased
to a very insignificant value at the interior of a density interface of finite thickness,
6 ~ 0.06x (depth of mixed layer). Thus, the contribution of this portion to the
mean overall velocity was assumed to be negligible. In the present investigation
as well, visual observation seemed to be in conformance with this argument. The
resulting velocity profile was integrated, and knowing the depth of the mixed layer,
the representative mean velocity for the interval could be obtained. Temerature
recordings of the mixedlayer were also made throughout the course of some of the
runs which showed that the increase of temperature of the mixedlayer by the end
of a run was not more than 2 OC (mean temperature was ~ 170C.
The recording of the experiment (on the video recording system) was played
back to obtain the rate of progress of injected dyelines. The representative velocity
distribution was thus obtained. This was drawn on a graphpaper to measure the
area which further gave the mean representative velocity for each interval. The
point of inflection of the velocity profiles (see Figure 3.2) were also noted as the
velocitygradient is responsible for the shear production causing entrainment. The
vertical distance of this point from the interface was designated 6,.
For the purpose of gravimetric analysis, Millipore Filtering System was used in
conjunction with a small, vacuum pump (which could produce a vacuum of upto
65 cms of Hg). Millispore filters (Filter Type HA, Pore Size 0.45 pm) were first
dried in an oven at a temperature of 50 oC for at least 3 hours. These were then
removed from the oven and allowed to equilibrate in a room (whose temperature
and relative humidity were monitored with an airconditioning unit) for a minimum
of 8 hours. These filters were then weighed in the same room on a Mettler balance
(Type H80) which was accurate upto 1 mg. These preweighed filters were then used
to dewater known volumes of sediment samples. In the case of bentonite, the sample
66
volume that could be used for this process of dewatering was only 0.5 cm3 as the
filters got clogged with the sediment particles for greater volumes of fluid mud. To
improve accuracy in obtaining concentration profiles for bentonite, this procedure of
dewatering was done for at least three subsamples for each base sample withdrawn
from any elevation of the flume at any time, and these were averaged.
The filtrate was allowed to remain on the paper which was then heated in the
oven again (at 50 OC for at least 6 hours) to remove the last vestiges of water.
The dried filter paper with dry sediment on it was again equilibrated in the same
monitored room and then weighed on the Mettler balance from which the mass of
sediment in a known volume of sample was easily obtained. This procedure was
carried out for all the samples, and, thus, the concentration profile of fluid mud was
known for each interval. Knowing the depth of the fluid mud, the mass flux (and
hence, the buoyancy flux) across the interface could be calculated.
CHAPTER 6
RESULTS AND ANALYSIS
6.1 Definition of Richardson Number
Vertical mixing across a density interface is dependent on the local Richardson
number (Turner 1986), e.g., the gradient Richardson number across the interface
in terms of the velocity and density differences across the interface. However, mea
surement of the local Richardson number is generally difficult (the thickness of the
interface needs to be determined) and a common procedure is to define an over
all Richardson number. The most suitable definition in the present case as well is
such an overall Richardson number in terms of the depth of the mixed layer and
the buoyancy jump across the interface. The depth of the mixed layer controls the
lengthscale of the energycontaining eddies, with the interface acting as an internal
boundary. Regarding the velocity scale, most of the previous researchers tend to
identify with the friction velocity, u,. In flume experiments without rotating screens,
it can be seen that most of the turbulence is produced at the density interface and
the sidewalls (Narimousa and Fernando 1987). However, in wall bounded flows,
most of the sidewall induced turbulence dissipates near the walls itself and only
a small portion diffuses outwards (Hinze 1975, p. 648). This is also confirmed by
Jones and Mulhearn (1983). Thus, most of the energy required for turbulent mixing
is a direct result of shear production at the interface and the most important scaling
velocity should be the velocity difference between the two layers, AU (e.g., Ellison
and Turner 1959, Lofquist 1960, Moore and Long 1971). In the present case, the
velocity of fluid mud at and below the level of the density interface was considered
negligible (although it could not be expressly measured, visual observations seemed
68
to confirm the fact) as in Narimousa and Fernado (1987). Thus, the mean velocity
of the mixedlayer was taken as the most representative velocity scale. With this,
the Richardson number is defined as
hAb
Ri, = g9hA (6.1)
with the interfacial buoyancy jump being
P2 P1
Ab = gP2 l (6.2)
PI
where #1 is the mean mixedlayer density and p2 is the density of fluid mud at the
level of the density interface.
6.2 Initial Conditions
The initial conditions for all the experimental Runs are listed in Table 6.1 (for
a physical description of the flow configuration, refer to Figure 3.2). The associated
terminology is as follows:
Ms = mass per unit area of fluid mud
h = depth of mixed layer
H = depth of fluid mud
C2 = mean concentration of fluid mud
C2 = concentration of fluid mud at the level of the interface
ii = mean velocity of the mixed layer
Ab = buoyancy step across the interface
The subscript 0 denotes initial conditions. It must be noted that in Runs 410,
the pump system was kept at some fixed (by not altering the speed controls) rotation
rate (the rate was tuned such that a pre determined mean velocity could be achieved
in the mixedlayer) throughout the course of the each run, the velocity profiles
being allowed to evolve with time, while the speed settings of the pump system
(specifically, only the propeller) was varied during the course of the experiment
for the remaining runs. The initial values of the mean velocity were in the range
7.413.1 cm/s. However, in the first three runs, mean velocity values even exceeded
these initial values as the rotation rate of the pump system was increased. In Runs
Table 6.1: Initial conditions of all Runs
RUN (M2)o ho Ho (C2)0 (C2)0 (U)o (Ab)o
NUMBER g/cm2 cm cm g/l g/l cm/s cm/s2
1 1.1685 32.2 19.8 59.0 29.0 7.4 17.7
2 3.0625 28.8 26.5 115.6 94.5 7.5 57.7
3 3.2375 26.0 25.0 129.5 110.0 11.8 67.2
4 2.5410 26.5 29.5 86.1 76.0 13.1 46.4
5 2.3450 25.3 27.7 84.7 76.0 11.9 46.4
6 1.6125 22.2 28.5 56.6 35.0 9.5 21.4
7 1.2450 28.0 27.0 46.1 30.0 11.0 18.3
8 1.9918 25.0 31.2 63.8 50.0 13.0 30.5
9 2.2350 23.6 28.2 79.3 62.0 9.6 37.9
10 1.0154 24.5 29.5 34.4 30.5 9.2 18.6
11 1.0800 23.6 28.0 38.6 28.0 9.9 17.1
70
19, kaolinite was the constituent sediment of fluid mud while the fluid mud was
of bentonite for Runs 10 and 11. The initial mean concentration of fluid mud was
in the range 45130 g/1 which corresponded to bulk density range of ~ 1.031.08
g/cm3. The upper limit of this range was imposed by the performance capabilities
of the pump system so as to obtain reasonable (for which entrainment was possible)
values of the Richardson number In the case of bentonite, higher values of mean
initial concentration could not be used because of difficulty in obtaining a fairly
uniform, wellmixed suspension.
6.3 Evolution of Characteristic Profiles
A typical timeevolution of the concentration profile below the level of the den
sity interface is shown in Figure 6.1. The data are for Run 9 with kaolinite as the
constituent sediment of fluid mud. Initially, i.e. at t = 0, the fluid mud was essen
tially quite wellmixed (with generally mild lutoclines) with obviously the steepest
gradient at the interface. Although the interface is shown to have an infinite gradi
ent, it is a wellknown fact that in similar and geophysical situations, the interface
is a region of thin but finite thickness (of the order of 1/20th the thickness of the
mixed layer) with a steep density gradient, see Narimousa and Fernando (1987).
The settling characteristic of the suspension caused a lutocline to develop for about
5 cm directly below this interface. The bottom 8 cm show a slightly steep lutocline
as well, which might be due to settling. With the passage of time, the interface
sharpens in the sense that the lutocline below it disappears. The concentration
of fluid mud at the level of the interface generally increases with time (except for
the profile at t = 21 minutes, which could be due to the local settling rate being
more than the rate of scour of the interface due to entrainment, temporarily). The
bulk concentration of the fluid mud always increased with time as the mud settled.
The major lutocline progressively steepened. It must also be noted that the mean
concentration of the mixed layer is simultaneously increasingly as well. However, as
50.0
Symbol t IF U
(minutes) (cm) (g/I)
0 0 28.2 79
o 9 23 95
S21 19.7 101
40.0 31 16.8 114
SA 44 14.8 121
E U 59 13 136
0
O.
0
1 30.0
OU
m
20.0
10.0
0.0 1 1 t li t I
40 60 80 100 120 140 160
CONCENTRATION (g/l)
Figure 6.1: Sequence of concentration profiles of Run 9 with kaolinite depicting the
evolution of concentration with time. IF denotes interface.
72
will be seen later, the buoyancy jump across the interface generally increased with
time (except for the second interval t = 9 to t = 21 minutes, when it appeared to
decrease). Entrainment progressed until the buoyancy jump became strong enough
to overcome the excess (after dissipation) turbulent kineticenergy which tended to
increase the potential energy of the system by causing entranment, at which point
the entrainment apparently decreased a lot.
Figure 6.2 shows the typical evolution of velocity profiles in the mixed layer for
Run 6 (with kaolinite). Initially, the profile was homogeneous without any gradient
at all in the mixedlayer, i.e. an apparent step velocity profile resulting in the case
of a vortex sheet discussed in Section 2.2.1. There was much entrainment at these
earliest times with massive convolutions covering the entire extent of the interface.
Closer examination revealed that the (thin, but finite) interface might itself be tur
bulent at these times. Initially, the mean velocity of the mixedlayer increased very
rapidly with time as the inertia of the system was being overcome. This generally
took between 3 to 4 minutes, by which time the mean velocity peaked. Next, the
mean velocity of the mixedlayer slowly decreased with time which might be due to
three reasons : (1) with the passage of time during the course of a run, with en
trainment (and settling), the elevation of the interface decreased, and progressively
more and more volume of fluid was being driven (considering the interface to act
as an internal boundary to the diffusion of momentum) by the pump system which
had a constant energy input; however, calculations to check conservation of mass
(ht) and momentum (h]2) for each run revealed discrepencies indicating that more
accurate measurements of velocity profiles need to be made if these quantities (mass
and momentum for each run) need to be accurately estimated, (2) sidewall friction
may not always be negligible, and (3) as the mixedlayer concentration increased
with time (due to mass flux into it), there was consequently increasing dissipation of
turbulent kinetic energy in the mixedlayer to counteract the downward buoyancy
60.0
Symbol t IF U
(minutes) (cm) (cm/s)
A 2 26 6.6
0 8 24 9.9 .
50.0 18 21.3 8.6
0 28 19.2 7.8
0 48 16.7 7.8
40.0
0
0
O
UL
> 30.0
0
.2 20.0
10.0
0.0 I
0.0 4.0 8.0 12.0
VELOCITY, (cm/s)
Figure 6.2: Evolution of the velocity profile in the mixedlayer for Run 6 with
kaolinite. IF denotes interface.
74
flux due to the sediment particle's fall velocity (see also Wolanski et al. 1989). This
can be considered in terms of the energy equation (see Abraham 1988)
dK , 9 g ,
dK= u7 wp E (6.3)
dt 8z p
with the primes denoting turbulent fluctuations, K the turbulent kinetic energy and
E the dissipation function. Assuming the turbulence to be in local balance, diffusive
transport is neglected and dK = 0. Thus,
uw' = p +e (6.4)
dz p
whence the production term of kinetic energy is balanced by the buoyancy term
(which is the conversion of input energy into the potential energy of the system) and
the dissipation function. Hence, at constant input of kinetic energy, as the buoyancy
term decreases, the dissipation of energy increases. Thus, with the passage of time,
at fixed input of energy due to the pump system, the available energy to effect
entrainment decreased. Visually, this resulted in decreased amplitudes of the waves
at the interface. Figure 6.3 shows that the mixedlayer depth (for Run 10 with
bentonite as the constituent sediment of fluid mud) increased very rapidly with
time initially, but slowed down after ~ 20 minutes. Figure 6.4 is more illustraive
as it plots the rate of change of the mixedlayer depth against time (obtained by
differentiating the curve fitted in Figure 6.3). As expected, the curve asymptotes
towards zero after about 25 minutes.
6.4 Shear Layer
The shear layer is obviously very important as it is directly responsible for
overcoming the static stability of the twolayer system (of fluid mud and water) and
causing mixing. The vertical distance, between the point where the velocity profile
deviates from the vertical in the mixed layer to the the level of the interface, was
taken to be the thickness of the shear layer, 6, (refer to Figure 3.2) (although the
velocity may decrease to zero inside the thin interface). At the start of each run
0 10 20 30 40 50
TIME (minutes)
60 70 80 90
Figure 6.3: Change in the mixedlayer depth with time for Run 10 with bentonite.
W
0
.
w
W
w
3.5 
3.0
2.5
S2.0
E
S 1.5
1.0
0.5
0 I I I
0 20 40 60 80 100
TIME (minutes)
Figure 6.4: Rate of change of mixedlayer depth in Run 10 with bentonite.
77
when, initially, the upper mixed layer was equilibrating to the energy input of the
pump system, the interface appeared to be in turbulent motion, but the velocity
below the level of the visual interface could not be determined. Above it, the mixed
layer was fully turbulent and dye injection showed an apparent step velocity profile
with the vortex sheet at the interface. Actually, as mentioned before, it might be
more realistic to assume the thin density interface (Narimousa and Fernando 1987)
to have a steep density gradient. This appears to be most plausible as the interface,
at those times, was convulsed by massive undulations of heights of the order
of ~ 68 cms causing much mixing. The effect of the splitter plate in causing
additional entrainment was also visible. Thus, the first interval of each run was not
considered while plotting data in Figure 6.5 where the data are from all the runs
(i.e., for both kaolinite and bentonite) have been included. Unfortunately, the data
are quite scattered for any definitive conclusions to be made. The values of 6, range
from 0.18 0.34 h. To avoid illusionary appearances on account of disparate
scales of the axes, the same was also plotted on a log log scale (see Figure 6.6)
which indicates that the 6, may be ~ 0.23h for 4 < Ri, < 20 with a slight increase
beyond 20. Long (1973) reported that Moore and Long's (1971) data indicated
6,/h ~ Ri'05 while Narimousa and Fernando (1987) found the nondimensional
shear layer thickness to be independent of Richardson number and about 0.2. In
the present case, the scatter of data may not actually be very surprising, as (1) the
diffusion of momentum into and maybe even below the level of the interface may
not always have been totally negligible, and (2) the conditions for two different runs
were not exactly duplicated for the same Richardson number, e.g., a value of Riu
= 10 may have been obtained in the second time interval of a run while it may
have occurred in the later intervals of the other. There will be greater dissipation
of kinetic energy in the second case in trying to counteract the settling tendency
of more sediment particles, as the concentration of the mixed layer increases with
0.35
O
6 0.30
0

* 0.25 0 0
0o o0
So00 o o0
S O 0 0
< O GD O O
S0.20 IO 0 0
00 O
C 0 0
Lu
I
2 0.15
Z
0
0.10 i i I I 'I t I I I 
0.0 10.0 20.0 30.0 40.0
RICHARDSON NUMBER, Riu
Figure 6.5: Nondimensional shear layer thickness vs. Richardson number
10
u6
U)
r)
UJi
w
0
uJ
1
cc
RICHARDSON NUMBER, Riu
Figure 6.6: Nondimensional shear layer thickness vs. Richardson number on a
loglog scale
10 102
80
time. Thus, it appears that the thickness of the shear layer increased with increasing
Richardson number, unlike in Moore and Long (1971). More accurate methods of
measurement are required for velocity profiles before any definitive conclusions can
be reached.
6.5 Observations on the Interface
When the pump section was turned on to start the experiment, the entrain
ment process started out with turbulent entrainment of the diffuse intermediate
layer which formed when the fluid mud was introduced under the water layer while
setting up the twolayer system. Figure 6.7 was taken within half a minute of start
ing the pump system. The grid squares are 2 times 2 cm. This intermediate layer
eroded completely within 11.5 minutes. Although the contribution to the density
of the mixed layer was minimal, there was a significant contribution to the turbidity
of the mixed layer, thereby rendering it opaque and obstructing visibility. Small
amounts of dye were injected and this dye stained the mixedlayer as it moved
around the flume and gave a color contrast with respect to the fluid mud layer. Ini
tially, the effect of the splitter plate in producing additional vorticity at the entrance
to the observation section was quite pronounced and was visible as deepening of the
interface there (this tilt was generally perceptible for about the first 810 minutes).
Mixing was caused by massive internal waves (upto 8 cm wave height in the up
stream portion of the observation section) breaking, as the steep velocity gradient
in the thin interface caused significant scour (of the interface). Figures 6.8 and 6.9
were taken backtoback in the same run at Ri, ~ 7. Wave heights were about
4 cm. The interface was highly irregular and action of eddies causing entrainment
is visible towards the left of the photographs. Entrainment due to internal wave
breaking seemed to cause most of the mass flux at Ri, < 10. Another mechanism of
entrainment, evident for Riu > 10, was as seen in the sequence of Figures 6.10, 6.11,
6.12 and 6.13. Figure 6.10 shows the highly irregular interface with eddies scour
t o
t t
Vt
I t 
t 
4
UPt)
Figure 6.7: Turbulent entrainment at t ~ 0.5 minute. Sediment kaolinite.
ItILM XLF
Figure 6.8: Interface at Ri, < 10. Sedimentkaolinite.
4 1'
Figure 6.9: Interface at Ri, < 10. Sedimentkaolinite.
Figure 6.10: Highly irregular interface at Riu > 10. Sediment kaolinite.
Figure 6.11: Scour of growing crest at Ri, > 10. Sediment kaolinite.
Figure 6.12: Scour of grown crest at Ri, > 10. Sediment kaolinite.
84
ing the interface. Some of the undulations formed pronounced crests which grew
in amplitude and sharpened with entrainment due to eddies mostly scouring their
backs and tips (Figures 6.11 and 6.12). The remaining portion of the crest then
subsided back towards the interface (Figure 6.13). When the same mechanism was
active at slightly lower Richardson numbers ( 1015), after the wave sharpened at
the crest, instead of breaking to form an eddy, this crest suddenly disappeared with
a thin of fluid being 'ejected' from the tip (see Figure 6.14). It appears possible
that the original undulations were caused by eddies from the mixed layer scouring
the interface, with it's 'roller action' causing crest growth and entrainment across
(the crest). When the eddy was strong enough, it could shear off the crest. These
phenomena of cusping into the upper (mixed) layer and appearance of 'smokelike
wisps' from these crests cusping into the faster layer seemed to indicate the ex
istence of Holmboe (mode 2) type of instabilities which was discussed earlier in
Section 3.1 and 3.2. Referring again to this review, this appears to be feasible as
6, > 6 and the levels of stratification attained were always quite high (as compared
to the lower, Ri < 3, similarly defined Richardson numbers obtained in experiments
with saltstratified systems). When the Richardson number still increased (beyond
~ 25), the interface was convoluted with smaller (less than 1 cm) disturbances (see
Figure 6.15) which appeared to be slightly more regular.
6.6 Entrainment Rate
Table 6.2 documents the parameters that were measured during the course of
each run for all the runs with kaolinite as the constituent of fluid mud (i.e., Runs
19). Table 6.3 does the same for runs with bentonite (i.e., Runs 10 and 11).
The listed parameters are the representative ones for each time interval of each
run. Table 6.4 contains the calculated parameters which further lead to the non
dimensional buoyancy flux and Richardson number for each interval of each run
with fluid mud of kaolinite. Reynolds number (Re) calculated according to Re =
Figure 6.13: Subsiding crest at Ri, > 10. Sedimentkaolinite.
Figure 6.14: Smokelike wisp being ejected from the tip of disturbances. Sedi
mentkaolinite.
