Professional Documents
Culture Documents
Abstract
When bubbles are continuously released from a localised source at the bottom of a liquid layer, a plume is produced. As the bubble
plume rises due to its buoyancy, it entrains surrounding liquid, which is carried upward with the stream of bubbles. In the present
work, we investigate the motion of a plume of low-Reynolds number bubbles in a strati"ed liquid consisting of two homogeneous
layers of di!erent densities. The liquid environment is of "nite lateral extent. We develop a theoretical model for the #ow of the bubble
plume and the surrounding liquid. The full equations are solved numerically. The mixing at the interface is quanti"ed and the
time-evolution of the density pro"les in both layers is calculated. The model also predicts the rate of rise of the density interface. We
develop an analytical solution for the problem in the limit of strong strati"cations. Our theoretical predictions are compared with new
experimental results using plumes of small bubbles generated by electrolysis of an aqueous solution of sodium chloride and with
previous experimental results (McDougall (1978), Journal of Fluid Mechanics, 85, 655}672; Baines & Leitch (1992), Journal of Hydraulic
Engineering, 118(4), 559}577). 2000 Elsevier Science Ltd. All rights reserved.
0009-2509/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 9 9 ) 0 0 5 3 1 - X
2586 M. H. Chen, S. S. S. Cardoso / Chemical Engineering Science 55 (2000) 2585}2594
Finite size bubbles exhibit a slip velocity relative to the most studies have concentrated on these double-struc-
liquid rising in the plume and this leads to the formation ture plumes. However, for at least some of the larger-
of a bubble-rich core surrounded by bubble-depleted scale plumes found in industrial processes, the bubbles
rising liquid. The e!ect of this slip velocity will be greater are distributed widely across the plume. It is therefore of
for weak plumes i.e. for small volumetric #ow rates of gas. interest to study plumes of very small bubbles in the
A quantitative criterion determining the structure of laboratory.
a bubble plume may be obtained by comparing a scale Large bubbles possess wakes where liquid from deeper
for the velocity in the plume = with the slip velocity of levels in the environment may be carried and then de-
individual bubbles u . Dimensional arguments suggest trained at shallower levels. This is a relatively e!ective
@
transport mechanism. However, the work of Milgram
ol !o Q g (1983) suggests that mixing induced by small bubbles
= ol hl should be more e$cient than that by large bubbles.
& , (1) Indeed, a plume of small bubbles entrains a larger
u u
@ @ quantity of surrounding liquid and this e!ect dominates
where ol is the density of the liquid, o is the density of the over the wake transport mechanism exhibited by large
gas bubbles, Q is the volumetric #ow rate of gas at bubbles. Nevertheless, very little is known about the
the source, g is the acceleration of gravity and hl is the dynamics of plumes of low-Reynolds number bubbles,
typical vertical length scale of the plume. In small-scale whose ascent speed is very small compared to the convec-
experiments performed in the laboratory, the #ow rates tive velocities. It is therefore relevant to investigate the
are smaller than those found in industrial and natural mechanism by which such small bubbles drive mixing in
bubble plumes, whilst the bubble sizes remain approxim- a liquid and to assess its industrial potential. Some in-
ately the same, and hence =/u is typically smaller. This sight into the behaviour of plumes of small bubbles may
@
explains why most laboratory observations hitherto have be gained from previous studies of one-phase buoyant
revealed a double-structure plume. As a consequence, plumes (Morton, Taylor & Turner, 1956; Cardoso &
M. H. Chen, S. S. S. Cardoso / Chemical Engineering Science 55 (2000) 2585}2594 2587
Woods, 1993), but there are some fundamental di!er- density level. In an environment of "nite lateral extent,
ences, including bubble separation in strati"ed liquids the plume induces downward advection of the surround-
and the formation of sequences of bubble plumes in the ing liquid, and hence the density distribution in the
liquid layer. In the present work, we present new experi- environment evolves with time. The continuing plume,
mental results using bubbles of approximately 150 lm rising through the changing strati"cation, will in turn
diameter generated by electrolysis; these bubbles are at transport liquid to varying depths. This interaction be-
least one order of magnitude smaller than bubbles used tween the plume and the environment has been modelled
by previous researchers (see Table 1). numerically by Patterson and Imberger (1989), Leitch
A bubble plume rising in a density-strati"ed environ- and Baines (1989) and Schladow (1992). The e!ectiveness
ment will transport dense liquid upward from the deeper of bubble-driven mixing in such complex systems has
regions for some distance; a fraction of this liquid then been measured by the e$ciency of conversion of the
leaves the plume and spreads out horizontally at its own input bubble energy into potential energy of the strati"-
cation. The bubble energy at the source is measured by
the work required to produce the air #ow via a compres-
sor. Although peak e$ciencies of up to 15% have been
achieved for certain design and operating con"gurations
(Asaeda & Imberger, 1993; Schladow, 1992), the e$cien-
cy generally decreases with time as the strati"cation
weakens, and the average e$ciency of the complete mix-
ing process can be as low as 3}5%. However, comparison
with e$ciencies of mixing by one-phase plumes suggests
that there is scope for improvement of bubble plume
e$ciencies up to approximately 15% (Zilitinkevitch,
1991; Cardoso & Woods, 1993). In this paper, we discuss
how the e$ciency of mixing of a plume is a!ected by the
bubble size.
In the following sections, we develop a novel approach
to the modelling of the dynamics and mixing induced by
a bubble plume, appropriate in the limit of small bubbles.
We focus on the mixing driven by a continuous release
of bubbles in a strati"cation consisting of two layers of
liquids of di!erent densities. The liquid environment is of
"nite lateral extent. We calculate the evolution of the
density pro"le in both layers. Our theoretical predictions
are validated by comparison with both new and previous
experimental observations.
Fig. 1. Structure of a bubble plume rising in a homogeneous #uid: (a) The experiments were carried out in two Perspex
simple plume when =/u 1 and (b) double-core plume when =/u 1. tanks, tank A of square cross section 0.2 m;0.2 m and
@ @
Table 1
Comparison of the size and structure of the bubbles used in the present work and in previous studies
The bubble shape and the volume of the wake were obtained from Clift, Grace & Weber (1978).
2588 M. H. Chen, S. S. S. Cardoso / Chemical Engineering Science 55 (2000) 2585}2594
height 0.4 m and a larger tank, tank B, of 0.4 m;0.4 m Bubbles were generated at the bottom of the tank via
and height 0.6 m. A two-layer strati"cation was set up electrolysis of the NaCl solution. The electrodes consis-
with a lower layer of NaCl aqueous solution and a top ted of a mesh of "ne platinum wires (Fig. 3), with area
layer of fresh water; the two layers were separated by 1.5 cm;1.5 cm, connected to a d.c. power supply. Bub-
a sharp interface with approximately 0.5 cm thickness. bles of hydrogen were released at the cathode and of
The top layer was coloured with food dye to enhance the chlorine at the anode (Creighton & Koehler, 1944). The
visibility of the interface. A photograph of the experi- electrodes were vibrated magnetically to avoid coales-
mental system is shown in Fig. 2. cence of bubbles and hence ensure the detachment of
Fig. 2. Photograph of experimental set-up illustrating the structure of a plume of small bubbles in a two-layer strati"cation. White arrows indicate the
#ow of liquid.
Fig. 3. Electrodes of platinum used to generate bubbles of hydrogen at the cathode and chlorine at the anode.
M. H. Chen, S. S. S. Cardoso / Chemical Engineering Science 55 (2000) 2585}2594 2589
uniformly sized bubbles. The size of the bubbles was the liquid reaches zero; the detrained dense liquid mixes
determined in both layers using a laser analyser. The with upper layer #uid and then spreads radially, forming
bubble size distribution was found to be independent of a gravity current of intermediate density. The bubbles
height in the tank and approximately Gaussian. An aver- continue to rise through the upper layer, forming a new
age size of 150$50 lm was measured (Chen, 1998). Such bubble plume. The surrounding less dense liquid in the
a constant bubble size with height, indicates a low coales- upper layer is entrained and mixes with the residual
cence tendency, which is characteristic of bubbles in dense liquid in the plume. The mixture of the two liquids
concentrated salt solutions (Soo, 1967). Although small is carried to the free surface where the bubbles burst and
bubbles do tend to coalesce in fresh water, our laboratory a liquid current spreads radially. As the tank is of "nite
observations and bubble size measurements suggest that lateral extent, the continual entrainment of surrounding
coalescence was also negligible in the upper layer. We ambient #uid into the plume leads to a downward #ow in
believe that the thin "lm of NaCl solution surrounding the environment. Hence, a front of liquid, which has been
the bubbles moving into the upper layer hindered coales- in the plume in each layer, descends with time; this can be
cence. seen in the sequence of photographs in Fig. 4 as the liquid
Calibration of the #ow rate of gas generated versus the circulated through the plume becomes discoloured. The
intensity of the current was performed directly by sealing liquid above the front in the lower layer is less dense than
the top of the tank and measuring the gas #ow rate with the original liquid below due to the entrainment of less
a bubble soap meter. Although chlorine is relatively dense upper layer liquid at the interface. Therefore, the
soluble in water, our #ow rate measurements indicate plume will arrive at the interface even lighter. As time
a constant #ow rate for a given current intensity. This evolves, a stable density distribution develops, gradually
may be a consequence of either the mass of dissolved "lling up the original lower layer. In the upper layer,
chlorine being negligible compared to the total gas gener- the liquid detrained at the free surface is denser than
ated or of saturation being achieved very fast. Neverthe- the original upper layer liquid. It may therefore drive
less, since we have measured directly the #ow rate of gas buoyant convection and mixing in the upper layer.
generated, the dissolution of chlorine does not a!ect our Our measurements of the evolution of the density
experimental results. pro"le in the tank for a typical experiment are shown in
The position of the interface and the motion in the Fig. 5. These results suggest that very little lower layer
lower layer were monitored visually. The density distri- liquid is carried into the upper layer by the rising bub-
bution in the tank was determined by withdrawing small bles. This observation may be explained partially by the
samples of the liquid at di!erent levels with a syringe. The fact that small bubbles of diameter 150 lm do not have
density was measured by refractometry. The experi- wakes, as shown in Table 1, and hence are unable to trap
mental conditions for each run are summarised in Table 2. and carry lower layer liquid upward. In addition, the
buoyancy force associated with the bubbles moving
2.2. Qualitative observations across the interface is relatively large compared to the
viscous force that can drag lower layer liquid into the
The typical #ow pattern induced by a bubble plume upper layer. Therefore, the contribution of viscous en-
rising in a strati"ed liquid consisting of two homogene- trainment of lower layer liquid to the #ow rate and
ous layers of di!erent densities is shown in Fig. 4. The density of the plume in the upper layer is small. We
lower and upper-layer #uids have both been dyed for conclude then that the dominant mixing mechanism
visualisation. As the bubble plume rises across the den- at the interface is associated with the overshoot and
sity interface, its buoyancy decreases. As a result, a frac- collapse of the dense liquid transported in the plume. In
tion of the dense, lower layer liquid carried in the plume Section 3, we present a theoretical model that builds
is detrained above the interface, where the momentum of upon these experimental observations.
Table 2
The experimental conditions. Labels A and B refer to tank used (See Section 2.1)
Fig. 4. Sequence of photographs showing the evolution of the mixing induced by a bubble plume in a two-layer density strati"cation.
M. H. Chen, S. S. S. Cardoso / Chemical Engineering Science 55 (2000) 2585}2594 2591
d 1
aw "aD, (3) nawD"B . (10c)
dz 2
We require one more continuity condition, quantifying
d 1 *Dl the rate of entrainment of the upper layer liquid at the
awD "aw . (4)
dz 2 *z interface. We shall assume that a constant fraction f of
2592 M. H. Chen, S. S. S. Cardoso / Chemical Engineering Science 55 (2000) 2585}2594
the kinetic energy of the plume at the interface is con- 4. Results and discussion
verted into potential energy of the lower layer. This
energetic conversion is expressed by (Cardoso & Woods, Fig. 5 illustrates a typical evolution of the density
1993) pro"le with time. Several stages may be identi"ed: in the
"rst stage (t"0}10 mins), the lower layer becomes strat-
dhl f B
(Dl !D ) " . (11) i"ed, evolving into a region where the density increases
dt A with height. This is followed by the main stage
(t"10}110 mins) where the density di!erence between
This equation is equivalent to an entrainment law of the the upper and the lower layer is reduced and a homo-
type (1/w) dhl /dtJRi\, where Ri is the Richardson geneous density is eventually recorded at t"120 min.
number of the strati"cation, de"ned by The homogenisation of the solution occurs soon after the
(Dl !D )hl interface disappears, normally in the space of 1}2 min as
Ri" . (12) recorded in nearly all the experiments. It can be seen
w
from the graph that the density in the upper layer has
Cardoso and Woods (1993) have shown that such a mix- undergone only minor changes until the last stage. It can
ing law is valid for 0.2(Ri(3. The e$ciency of this therefore be concluded that the homogenisation of the
energetic conversion, f, has been found to lie in the range solution is mainly due to the increase of the depth of the
0.2}0.3 in penetrative convection studies (Zilitinkevich, lower layer and that only a very small amount of liquid
1991). from the lower layer is carried across the interface by the
Eqs. (2)}(4) and (7)}(9), (10a)}(10c) and (11) were solved bubbles. No strati"cation prevails in the lower layer soon
numerically using f"0.3; the theoretical results are com- after the "rst stage, showing that the plume drives good
pared with the experimental measurements in the next circulation in the "nite size tank.
section. Our theoretical predictions for the evolution of the
density pro"le have been plotted as solid lines in Fig. 6.
The agreement between the model predictions and the
3.2. A simplixed model for strong stratixcations
experimental results is good for the lower layer. For large
times, a small inversion in the density is observed at the
For strong strati"cations, the rate of rise of the inter-
top of the upper layer. The model does not capture this
face is very much smaller than the rate at which the
small e!ect as expected.
environmental #uid is circulated through the plume. We
Fig. 7 shows the evolution of the density pro"le in an
may therefore assume that after the descending front
experiment by Baines and Leitch (1992). The pro"les
reaches the bottom of the tank, the environment sur-
predicted by our new theoretical model, valid for small
rounding the plume is approximately homogeneous. In
Reynolds number bubbles, are also shown. The di!erence
this limit, the full model equations may be simpli"ed to
between the model and experimental results suggests that
an equation for the conservation of buoyancy in the
a plume of small bubbles is more e$cient at destratifying
lower layer,
the lower regions of the tank, whilst a plume of large
d bubbles is more e$cient at transporting lower layer
[hl (Dl !D )]"0 (13)
dt
d fB
hl (Dl !D )"! . (14)
dt A
fB t
hl "hl exp (15)
Ahl (Dl !D )
and the evolution of the density in the lower layer
!f B t
(Dl !D )"(Dl !D ) exp . (16)
Ahl (Dl !D )
Fig. 6. Evolution of the density pro"le: theoretical predictions (solid
Eqs. (15) and (16) are valid for large Richardson number lines) and experimental results for an initial Richardson number,
strati"cations, that is for Ri1. Ri "2.5.
M. H. Chen, S. S. S. Cardoso / Chemical Engineering Science 55 (2000) 2585}2594 2593
conversion is well within the range reported in penetra- on a large scale application, but alternative methods
tive convection studies (Zilitinkevich, 1991). Energy involving saturation of the liquid with a gas at high
losses are due to viscous dissipation and internal wave pressure followed by decompression can be explored.
radiation. From an engineering point of view, it is impor- Also, in non-aqueous systems, small bubbles may absorb
tant to know how much of the power supplied in the gas a signi"cant amount of vapour from the process liquid,
stream is converted into potential energy of the density and thus further recovery may be required. However, we
distribution. As the kinetic energy of the plume at the note that as long as the criterion =/u 1 is satis"ed, the
@
interface is equal to approximately half of the energy of plume will exhibit a structure similar to a plume of low-
the bubbles at the source (Cardoso & Woods, 1993), we Reynolds number bubbles. We conclude that, from the
may conclude that the e$ciency of conversion of the view point of energy e$ciency and absence of mechanical
bubble energy at the source into potential energy is moving parts, plumes of small bubbles have important
f /2+15%. engineering potential as a mixing technique.
5. Conclusions References
We have presented a new experimental and theoretical Asaeda, T., & Imberger, J. (1993). Structure of bubble plumes in linearly
study on the mixing induced by a plume of small bubbles strati"ed environments. Journal of Fluid Mechanics, 249, 35}57.
Baines, W. D. (1961). The principles of operation of bubbling systems,
in a two-layer density strati"cation. Our new experi- Proceedings of the Symposium on Air Bubbling, Ottawa.
mental results show that very little dense liquid from Baines, W. D., & Leitch, A. M. (1992). Destruction of strati"cation by
the lower layer is transported across the interface by the bubble plume. Journal of Hydraulic Engineering, 118(4), 559}577.
rising bubbles. The density in the upper layer therefore Cardoso, S. S. S., & Woods, A. W. (1993). Mixing by a turbulent plume
remains constant and equal to that at the beginning of in a con"ned strati"ed region. Journal of Fluid Mechanics, 250, 277}305.
Chen, M. H. (1998). Bubble-driven mixing (Internal Report), Dept. of
the mixing process. As a result, the mixing in the two- Chemical Engineering, University of Cambridge, UK.
layer system consists of turbulent entrainment of the Clift, R., Grace, J. R., & Weber, M. E. (1978). In Bubbles, drops and
liquid in the upper layer by the plume liquid detrained at particles (pp. 143}144). New York: Academic Press, Inc. (London) Ltd.
the interface, and the re-circulation of environmental Creighton, H. J., & Koehler, W. A. (1944). In Electrochemistry, vol. II:
liquid in the lower layer through the plume. Applications (pp. 275}279). New York: Wiley.
Jones, W. T. (1972). Air barriers as oil-spill containment devices. J. Soc.
A detailed numerical model has been developed. The Pet. Eng., 12, 126}142.
model predicts the position of the interface as a function Leitch, A. M., & Baines, W. D. (1989). Liquid volume #ux in a weak
of time and the evolution of the density pro"le in the bubble plume. Journal of Fluid Mechanics, 205, 77}98.
two-layer system. The agreement between our model Lemkert, C. J., & Imberger, J. (1993). Energetic bubble plumes in
predictions and experimental results is excellent. arbitrary strati"cation. Journal of Hydraulic Engineering, 119(6),
680}703.
We have also shown, that in the limit of strong strati"- Manins, P. C. (1979). Turbulent buoyant convection from a source in a
cations, the density pro"le in the lower layer remains con"ned region. Journal of Fluid Mechanics, 91, 765}781.
approximately uniform. We have developed a simple McDougall, T. J. (1978). Bubble plumes in strati"ed environments.
analytical solution for the evolution of the interface posi- Journal of Fluid Mechanics, 85, 655}672.
tion and the density di!erence across the interface. We Milgram, J. H. (1983). Mean #ow in round bubble plumes. Journal of
Fluid Mechanics, 133, 345}376.
have shown that this simpli"ed approach is valid for Morton, B. R., Taylor, G. I., & Turner, J. S. (1956). Turbulent gravi-
Ri *1.
tational convection from maintained and instantaneous sources.
Our study has also shown that a plume of small Proceedings of the Royal Society, A234, 1}23.
Reynolds number bubbles is e$cient at destratifying Patterson, J. C., & Imberger, J. (1989). Simulation of bubble plume
the lower regions of the tank, whilst a plume of large destrati"cation systems in reservoirs. Aquatic Sciences, 51(1), 3}18.
Schladow, S. G. (1992). Bubble plume dynamics in a strati"ed medium
bubbles is more e$cient at transporting lower layer and the implications for water quality amelioration in lakes. Water
liquid into the upper layer and thereby increasing the Resources Research, 28(2), 313}321.
density at the higher levels in the tank. The overall Soo, S. L. (1967). In Fluid dynamics of multiphase systems (p. 96). New
e$ciency of conversion of the energy released at the York: Blaisdell Publishing Company.
source of bubbles into potential energy is approximately Wilkinson, D. L. (1979). Two-dimensional bubble plumes. Journal of
Hydraulic Division ASCE, 105(2), 139}154.
15% for the plume of small bubbles used in our work. Wuest, A., Brooks, N., & Imboden, D. M. (1992). Bubble plume model-
This is signi"cantly larger than the average e$ciency of ling for lake restoration. Water resources research, 28(12),
mixing of a plume of large bubbles, which has been found 3235}3250.
by previous authors to be of order of 3}9% (Schladow, Taylor, G. I. (1955). The action of a surface current, used as a break-
1992). water. Proceedings of the Royal Society of London, A 231, 466}478.
Turner, J. S. (1979). In Buoyancy ewects in yuids (pp. 165}206).
There are some di$culties associated with the use of Cambridge: Cambridge University Press.
very small bubbles in industry. In particular, electrolysis Zilitinkevich, S. S. (1991). In Turbulent penetrative convection (pp. 1}77).
cannot be competitively used to produce small bubbles Avebury Technical.