Exp Fluids (2014) 55:1773

DOI 10.1007/s00348-014-1773-x

REVIEW ARTICLE

Frictional drag reduction by bubble injection

Yuichi Murai

Received: 8 February 2014 / Revised: 15 June 2014 / Accepted: 17 June 2014 / Published online: 15 July 2014
Ó Springer-Verlag Berlin Heidelberg 2014

Abstract The injection of gas bubbles into a turbulent 1 Introduction

boundary layer of a liquid phase has multiple different
impacts on the original flow structure. Frictional drag The use of bubbles as a way of reducing skin frictional drag
reduction is a phenomenon resulting from their combined in turbulent flows has long been a focus for engineers in the
effects. This explains why a number of different void–drag expectation that it is applicable to ships and pipelines. The
reduction relationships have been reported to date, while first success was reported in the literature more than
early works pursued a simple universal mechanism. In the 40 years ago: McCormick and Bhattacharyya (1973)
last 15 years, a series of precisely designed experimenta- employed electrolysis to generate microbubbles in water
tions has led to the conclusion that the frictional drag and were able to demonstrate a significant degree of drag
reduction by bubble injection has multiple manifestations reduction for a submersible hull. If gas films are included
dependent on bubble size and flow speed. The phenomena in this topic, we could further go back to the paper of
are classified into several regimes of two-phase interaction Hirata and Nishiwaki (1963), who measured pressure loss
mechanisms. Each regime has inherent physics of bubbly in a horizontal channel flow containing gas in a film state.
liquid, highlighted by keywords such as bubbly mixture After the dawn of such technology, fluid engineering
rheology, the spectral response of bubbles in turbulence, researchers began discussing the basic mechanism, which
buoyancy-dominated bubble behavior, and gas cavity was conceptualized as a change in mixture properties such
breakup. Among the regimes, bubbles in some selected as the local average density and viscosity of the liquid
situations lose the drag reduction effect owing to extra containing bubbles. Legner (1984) suggested that shear
momentum transfer promoted by their active motions. This thickening owing to an increase in local effective viscosity
separates engineers into two communities: those studying reduces wall shear stress. Marie (1987) explained the
small bubbles for high-speed flow applications and those decrease in Reynolds shear stress with a local reduction in
studying large bubbles for low-speed flow applications. density. However, later experimental results did not para-
This article reviews the roles of bubbles in drag reduction, metrically support these hypotheses and rather indicated
which have been revealed from fundamental studies of the difficulty of elucidating the correct perspective of the
simplified flow geometries and from development of mechanism. On the one hand, this background encouraged
measurement techniques that resolve the inner layer full-dress laboratory research on a bubble-mixed boundary
structure of bubble-mixed turbulent boundary layers. layer. A series of laboratory experiments using water
channels and tunnels were conducted from the 1980s to
1990s, such as those of Madavan et al. (1985), Merkle and
Deutsch (1990, 1992), and Kato et al. (1999). On the other
hand, demands for practical use by vessels increased with
Y. Murai (&) changes in economical and environmental circumstances
Laboratory for Flow Control, Division of Energy and
such as jumps in the oil price and the implementation of
Environmental Systems, Faculty of Engineering, Hokkaido
University, N13W8, Sapporo 060-8628, Japan energy-saving strategies and international greenhouse gas
e-mail: murai@[Link] regulations. Since 2000, activity relating to this topic came

123
1773 Page 2 of 28 Exp Fluids (2014) 55:1773

to a hottest generation especially among maritime Taylor–Couette flow (van den Berg et al. 2005; Murai et al.
researchers, energetically challenging to reach a compre- 2008; van Gils et al. 2013). Furthermore, the lack of
hensive understanding of the drag reduction toward prac- reproducibility in drag reduction, which has long been a
tical use (Kodama et al. 2000; Moriguchi and Kato 2002). concern among experimentalists, has come to the fore. We
Experimentalists introduced state-of-the-art measurement are now familiar with the concept that reproducibility is
instrumentations into quantitative monitoring of the affected by contamination in water (Winkel et al. 2004;
boundary layer. They expanded measurement principles Takagi and Matsumoto 2011) and by naturally induced
established for single-phase flow to allow investigation of void waves that are generated by the bubble–drag time-lag
bubbly two-phase flow, such as for laser Doppler veloci- mechanism (Murai et al. 2007; Park et al. 2009). The
metry (LDV), particle image velocimetry (PIV), and present article reviews the basic concept, history, and
ultrasonic velocity profiling. Numerical researchers com- current discussion of bubble-induced drag reduction.
puted the two-way interaction between turbulence and
bubbles with direct numerical simulation (DNS)-like high-
resolution schemes (Xu et al. 2002; Ferrante and Elghob- 2 Drag reduction mechanism
ashi 2004; Pang et al. 2013). Although their simulations
were somewhat limited to sparse spherical bubbles, how Experimental correlation between the bubbling condition
bubbles attack surrounding turbulent eddies has been elu- and resultant drag reduction explains little about the
cidated. They left us a significant scientific perspective of underlying physics of drag reduction. The causal relation
the drag reduction mechanism, and also helped interpret comprises extremely complicated multivariable nonlinear
numerous past experimental data. functions. In fact, drag reduction mechanisms proposed by
In the last decade, there has been great progress in both researchers diverge seriously, which is confusing when
fundamental research and applications to industry. The attempting to design their practical applications.
progress was made possible by sophisticated measurement Table 1 lists the parameters we need to handle in
techniques for monitoring high-speed two-phase boundary experimentation for a single case of application. The target
layer flows (Kitagawa et al. 2005; Zhen and Hassan 2006; of interest is simply the wall shear stress, sw. Three primary
Murai et al. 2006a, b, c, 2009). Another impulse for the parameters are managed to modify the wall shear stress:
progress has been well-designed problem setting in flow velocity U, gas flow rate Q, and mean bubble size
research of multiphase fluid dynamics, such as the use of d. The performance of drag reduction itself is evaluated

Table 1 List of parameters in Target parameters

drag reduction by bubble
injection sw Wall shear stress
Apparent control parameters Boundary layer characterizing scales
U Liquid flow velocity d Boundary layer thickness
Q Bubble volume flow rate l Wall unit
d Bubble diameter h Air cavity equivalent gas layer thickness
Environmental parameters
P Hydrostatic pressure
g Acceleration of gravity
l Liquid viscosity
q Liquid density
Code Name Description Dominant in

Mechanism-governing
dimensionless parameters
Re Reynolds number Original flow structure Single-phase flow
Fr Froude number Buoyancy effect Air cavity and large bubble
regime
We Weber number Inertia deformation High turbulent regime
Ca Capillary number Viscous deformation Bubbles smaller than
coherent structure
Ma Mach number Compressibility High-speed flow
a Void fraction Impact to liquid All regimes

123
Exp Fluids (2014) 55:1773 Page 3 of 28 1773

according to the correlation among these explicitly man- methods can be categorized to the same approach in a
aged parameters. However, the internal physics of the broad sense and are commonly termed the gas layer
boundary layer is highly difficult to deduce as it is method in this paper, while gas–liquid interface takes dif-
described by so many parameters in the table; there are at ferent structures inside the boundary layer. In contrast,
least six dimensionless parameters involved when trying to there are techniques of producing and stabilizing gas cavity
solve the fluid dynamic mechanism. In this chapter, several such as by installing a stepwise stern in the upstream region
important perspectives in reading the experimental results of the target wall. The stern separates main stream from the
are offered, starting with the origin of the concept. wall so that a relatively stagnant gas cavity is maintained in
the vicinity of wall. This is termed the gas cavity method.
2.1 Fluid engineering basis The techniques rely on gravity and the geometry of the
target body and were comprehensively reviewed by Ceccio
In general formulation, frictional drag is described by (2010).
Summarizing above, we have to distinguish three pri-
1
D ¼ Cf   q  ðUw  Uf Þ2 A; ð1Þ mary types from gas supplying-based drag reduction
2 techniques as illustrated in Fig. 1. (a) Bubble drag reduc-
where D, Cf, q, Uw, Uf, and A are the frictional drag, tion (BDR) works with action of dispersed bubbles inside
friction coefficient, density of fluid, velocity of the moving the boundary layer. (b) Gas layer drag reduction (GLDR)
wall, velocity of the fluid outside the boundary layer, and relies on replacement of highly shearing liquid with gas in
area of fluid contact with the wall. In this aspect of fluid the form of froths or long gas films. (c) Gas cavity drag
engineering, the equation tells us that the drag can be reduction (GCDR) occurs when backward step provides a
reduced when any of the five variables on the right-hand large gas single-phase space. Elbing et al. (2008) and their
side of the equation is lowered. Apparently, reducing the group (Sanders et al. 2006; Mäkiharju et al. 2013a, b)
relative velocity, Uw - Uf, realizes drag reduction with its investigated the transition among these three types. By
squared impact, and this approach has been practically supplying air into water, they observed GLDR effect
applied in shipping to save fuel consumption per unit dis- between BDR and GCDR in a spatially developing two-
tance (Ronen 1982). Without the deceleration of the phase boundary layer. GLDR realizes with complex gas–
moving wall, the injection of bubbles around the wall liquid interfaces, but plays an important role in applications
reduces the local average density of the fluid, q, and where spatially developing turbulent boundary layers are
therefore reduces the drag. This is categorized as the inertia targeted such as for maritime vessels. Gas phase is
effect of drag reduction, which plays the major role in unlimited to water and seawater, but can be any suitable
turbulent flow states (Marie 1987). Maintaining large gas for liquid pipeline applications, such as, nitrogen,
bubbles close to the wall reduces A, the area of wall con- hydrogen, carbon dioxide, and gaseous fuel dependent on
tact. A 100 % replacement of liquid phase with air phase the combination to the liquid substances.
within the boundary layer removes almost all the drag, and Among the three types, BDR and GLDR accompany a
there only remains the friction of air, which is small decrease in the friction coefficient, Cf, in Eq. (1). In par-
enough compared with that of liquid. This replacement- ticular, BDR totally depends on how the friction coefficient
based drag reduction is termed the air layer method changes when bubbles are mixed into the boundary layer.
(Sanders et al. 2006). If the gas layer forms a thin, but We need to learn what happens to the inner layer of the
stable gas films at low shearing environment, it is termed turbulent boundary layer as the friction coefficient changes
the gas film method (Fukuda et al. 2000). These two (Kim 2003). If the friction coefficient increases largely, the

Fig. 1 Three types of frictional

drag reduction enabled by gas
injection. a Bubble drag
reduction. b Gas layer drag
reduction. c Gas cavity drag
reduction (Mäkiharju et al.
2013a, b)

123
1773 Page 4 of 28 Exp Fluids (2014) 55:1773

Fig. 2 Question how bubble

injection alters the inner
structure of a turbulent
boundary layer. a General view
of shear layer decomposition.
b The question of how bubbles
alter the original structure by
replacing with a bubbly two-
phase layer

decreasing effect of the fluid density or the area of contact role of turbulence (Jimenez 2012). Once bubbles are
would be canceled out. If Cf reduces together with the fluid mixed, such detailed discussion of the single-phase
density and area of contact, drag reduction would be boundary layer raises two major questions. The first
amplified according to their synergy effect. This is the question relates to where the bubbles tend to remain. Even
point of focus in this article. The change in friction coef- in flow without turbulence, bubbles have their own fluc-
ficient owing to the presence of bubbles clearly declares tuating nature owing to the time-lagging combination of
that the inner layer structure differs from that of single- transversal force components, such as the drag, lift, added
phase flow. Gabillet et al. (2002) found in a horizontal inertia, pressure gradient, and history forces. Deformability
turbulent channel flow with upward bubble injection that of bubbles at high shearing turbulence introduces further
the bubbles activated near-wall turbulence that increased stochastic behavior. The second question relates to how the
linearly with the void fraction. They concluded that the bubbles create a new layer that replaces the original layer.
bubbles remaining near the wall affect the boundary layer Fluid behavior within the new layer obeys the rule of
in a way similar to wall surface roughness, thus increasing bubbly two-phase flow dynamics, and hence it is not
drag. Non-spherical bubbles of intermediate size in the wall approximated by any single-phase flow models. Evidence
proximity potentially increase the friction coefficient, is available from pipe and channel flow experiments whose
which is analogous to the enhancement of heat and mass pressure drop characteristics of gas–liquid two-phase flow
transfer owing to the presence of such bubbles (Kitagawa are indescribable only with the mixture property of two
et al. 2008). As buoyancy acts on the boundary layer out- fluids (Lundin and McCready 2009). Well-known interfa-
ward, the boundary layer thickness expands so that drag cial patterns in a tube such as bubbly flow, plug flow, slug
can be reduced (Aliseda and Lasheras 2006). Recent papers flow, froth flow, and stratified flow can occur analogously
reported that dilute mixing of microbubbles comparable to in the two-phase shear layer sandwiched by the wall and
or smaller than turbulent eddy scales can sensitively reduce outer region. Historical studies on internal two-phase flows
the friction coefficient (e.g., Hara et al. 2011). For inertial remind us that drag reduction by bubble injection is at the
deformable bubbles at high Weber numbers, drag reduction heart of the issue of multiphase flow.
is restored owing to the collapse of coherent structures
(Huang et al. 2008). In this context, we should consider the 2.2 Mechanism transition diagram
multiple roles played by bubbles in a single case of
boundary layer flow. As bubbles of broad size are injected, Understanding of the mechanism of transition allows rea-
these roles overlap in the same flow field to be intertwined sonable design of drag reduction and improved perfor-
complicatedly. This issue is addressed below. mance. Unfortunately, the mechanism in use of bubbles is
Figure 2 schematically shows the influence of bubbles not explained by a couple of dimensionless parameters.
when they are mixed inside a turbulent boundary layer. What we see from data available today is a series of cor-
Initially, there are multiple different shear layers from the relations among the liquid flow speed, gas flow rate, mean
wall: a viscous sublayer, buffer layer, outer layer, and main bubble size (often lacking), and drag reduction ratio for a
flow region (Robinson 1991; Adrian 2007). The number of number of different flow configurations. Everything in the
layers depends on how precisely we observe the individual bubble–liquid interaction differs between internal and

123
Exp Fluids (2014) 55:1773 Page 5 of 28 1773

Fig. 3 Kaliroscope visualization of turbulent eddies in a horizontal direction so that streamwise vortices are mainly visualized. a Without
channel flow at 2 m/s. The main stream flows from left to right. Near- bubbles, b with large bubbles, c with small bubbles, d with
wall eddies are illuminated with a sheet of light in the spanwise microbubbles (Murai et al. 2006a, b, c)

external flows, between fully developed and spatially 2 m/s. The stripe pattern in the left photograph (a) indicates
developing flows, and between horizontal and vertical turbulent eddies close to the wall, visualized by Kalliro-
streams. There have been intensive studies to date on the scope [details of which were given by Dominguez-Lerma
turbulent flow characteristics of bubbly two-phase flow in a et al. (1985)]. Photograph (b) is a snapshot of the same flow
cylindrical tube (Lockhart and Martinelli 1949; Fujiwara when large bubbles are injected. The stripe pattern is
et al. 2004; Hosokawa and Tomiyama 2004, 2009; Ishii attenuated inside the area with bubbles, but remains in the
and Hibiki 2011). As the diameter of the tube changes, not area without bubbles. Photograph (c) shows the case of
only the Reynolds number but also all other dominant small bubbles, ranging from 1 to 5 mm in diameter. Since
dimensionless numbers change simultaneously such as the bubble size and the spacing of streamwise vortices are
Froude number, Weber number, capillary number, Ray- comparable to each other, the original stripe pattern dis-
leigh number, and Mach number. Buckingham’s P theo- appears. Photograph (d) is a snapshot when microbubbles
rem does not immediately benefit researchers in terms of of 50 lm in peak diameter are mixed into the channel flow
reducing the number of experimental parameters. Engi- at a volume fraction of 0.01 %. The stripe pattern is
neers accept dimensional consolidation of the measurement revived, but the spatio-temporal frequency changes. The
data since the physics is still only half understood. In photographs clearly show that the bubble–turbulent flow
application to vessels, there are further uncontrollable interaction has different mechanisms according to bubble
factors of the boundary layer, which make the causal size.
relationship between the fuel consumption rate and bub- Before reviewing the individual findings reported in past
bling flow rate unclear (Mizokami et al. 2010; Mäkiharju papers, the experimental conditions that each researcher
et al. 2012; Kumagai et al. 2010). The drag reduction employed are classified. Figure 4 plots the results of pub-
performance of vessels is affected by the specifics of each lished papers on a two-parameter domain, with the abscissa
vessel, meteorological factors, and seawater properties. giving the bubble size and the ordinate giving the flow
The overlap of fluid dynamic nonlinearity in multiphase speed. The plot includes results for horizontal channel
turbulent flow and the diversity of ship operation condi- flow, flow along horizontal flat plates, and model ships. It
tions mean that there is a severe lack of reproducibility in excludes results for the drag reduction of vertical pipe
ship drag reductions. With such problems in application, flows, vertical channel flows, and Taylor–Couette flows
engineers emphasize the importance of fundamental study because different flow configurations need another com-
that can classify the phenomenon into known and unknown parison on different plots. Numerical analysis is also
domains, or into reproducible and irreproducible regions of excluded. The plot shows that the success of drag reduction
parameter space. is roughly separated into two regions. One is the use of
Figure 3 shows the effect of bubbles on the turbulent relatively small bubbles at high flow speed (marked in M
flow structure, as observed from the top of a horizontal and S), and the other is the use of large bubbles at low flow
water channel flow (Murai et al. 2006a, b, c). Bubbles of speed (marked in L and V). Between these two regions,
three different sizes are injected into water that flows at few papers reported the success of drag reduction as

123
1773 Page 6 of 28 Exp Fluids (2014) 55:1773

Fig. 4 Distribution of technical

papers on the experimental
success of drag reduction
plotted on a two-parameter
domain. The central position
and diameter of each ellipse
indicate the average conditions
and the approximate range of
experimental tests in each report

indicated by gray regions. In the gray regions, the injection of ships were rarely published in journal articles. The
of bubbles rather increases the friction, whereas the mix- author is aware of a number of failures in ship drag
ture density of the boundary layer is invariably reduced. reduction via collaborations and private communications.
We have to avoid the condition falling in these regions to Accordingly, exact experimentations for fundamental flow
guarantee drag reduction. The upper gray region ‘‘Unre- configurations are strongly desired. In the following sub-
alizing’’ means that bubbles become unstable to keep their section, the performance and mechanism of drag reduction
initial size owing to shear stress. This region takes place in each region are elaborated.
only in the transition from small to large bubbles for coa-
lescence at high void fraction or large to small bubbles for 2.3 Gas cavity effect regime
fragmentation.
Table 2 lists all available reports on drag reduction Frictional resistance is reduced by having a gas cavity
using horizontal channel flows. The reports are roughly between a solid wall and outer flow. Fukuda et al. (2000)
classified by mean bubble size. Results from DNS are demonstrated that this effect is in proportion to the ratio of
recognized as numerical experiments and included here. In the area covered by gas to the whole area of the wall. The
the ‘‘Bubble injection into’’ column, ‘‘beneath’’ refers to mechanism is explained simply: The gas cavity cuts off the
bubble injection beneath the top wall of the channel, and contact between the liquid flow and the wall. However,
‘‘above’’ refers to bubble injection from the bottom wall of maintaining a stable gas cavity close to the wall requires
the channel. The column ‘‘Drag reduction %’’ presents the technical efforts. In case of external flow around a high-
maximum recorded drag reduction percentage, and ‘‘Gain speed moving body, blowing air from the front demands a
factor’’ is the ratio of the drag reduction per unit void gas flow rate larger than a critical value to keep the gas–
fraction [see Eq. (5)]. It is noted that these results cannot be liquid interface at the desired position. Cavitation-relevant
simply compared because of the very different conditions phenomena are coupled with the technique, implicitly or
applied in each assessment. Nevertheless, drag reduction explicitly (Callenaere et al. 2001). For slow flows below a
percentages of several tens have been reported. flat wall, the gas cavity naturally forms with buoyancy and
Table 3 summarizes reports on drag reduction in other can stably remain beneath the wall. In both cases, an
types of flow configurations. The use of a flat plate deals increase in flow speed results in a wavy gas–liquid inter-
with the bubble effect in a spatially developing boundary face owing to Kelvin–Helmholtz instability (Michel 1984).
layer, different from a fully developed channel flow. In In particular, the combination of high-speed liquid flow and
contrast, Taylor–Couette flow is preferably adopted to slow gas flow amplifies the instability so that the gas cavity
assess the fully developed state of two-way interaction is easily broken into an ensemble of bubbles as it migrates
between bubbles and turbulence. Application to ships or a long way downstream. Amromin and Mizine (2003)
model ships also recorded drag reduction in some cases; analyzed possibility of active flow control to keep partial
however, it should be noted that failures in drag reduction cavity stable. Even in the case of a gas cavity subject to

123
Exp Fluids (2014) 55:1773 Page 7 of 28 1773

Table 2 Reports on drag reduction for bubble injection into horizontal channel flows
Flow configuration Bubble size Year Investigators Bubble Drag Gain Central
injection reduction factor bubble
into % size

Horizontal channel flow Large bubbles 1963 Hirata and Nishiwaki Beneath 15 0.5 100 mm
2002 Gabillet, Colinm, and Fabre Above \0 \0 10 mm
2007 Murai, Fukuda, Oishi and Kodama Beneath 60 2 [10 mm
2009 Park, Tasaka, Murai, and Oishi Beneath 25 20 [40 mm
Small bubbles 1985 Madavan, Deusch, and Merkle Above 80 1.5 Broad
1995 Kim and Clever Beneath 65 NA NA
1996 Guin, Kato et al. Beneath 20 2 0.5 mm
1999 Kato, Iwashina et al. Beneath 60 4 0.5 mm
2000 Kodama, Kakugawa et al. Beneath 30 3.5 NA
2002 Moriguchi and Kato Beneath 40 4 0.5–2.5 mm
2005 Kitagawa, Hishida, and Kodama Beneath 2.5 5 0.5 mm
2006 Shen, Ceccio, and Perlin Beneath 35 3.5 0.05–0.4 mm
2006 Murai, Oishi, Takeda, Yamamoto Beneath 10 1.5 0.5 mm
2009 Huang, Murai, Yamamoto Beneath 5 2 1 mm
2009 Oishi, Murai, Tasaka, Takeda Beneath 10 4 1 mm
Microbubbles 2005 Hassan, Gutierrez Torres et al. Beneath 38 10 0.05 mm
2005 Murai, Oishi, Sasaki, and Kodama Beneath 20 1,000 0.03 mm
2005 Lu, Hamada, and Kato Beneath 25 800 0.03 mm
2011 Hara, Suzuki, and Yamamoto Beneath 20 1,100 0.03 mm
DNS 2002 Xu and Maxey Spherical 15 1.5 Spherical
2002 Kawamura and Kodama Deformable \0 -3 We = 37
2005 Lu, Fernandez, Tryggvason Deformable 25 1.3 We = 0.4

slow flows, a gravity wave forms as another factor of drag reduction as its thickness decreases, and how the gas
interfacial instability as analyzed by Matveev (2007). Once cavity maintains its function as its length in the main flow
the wave touches the wall, the cavity transforms to direction is shortened. In flow geometry of backward facing
dynamic two-phase flow similar to froth and churn flow. step, Mäkiharju et al. (2013a, b) found dependences of air
The main parameter that governs the gas cavity regime is cavity formation on Reynolds and Weber numbers which
the Froude number defined by correlate with gas shedding from the cavity. At enough high
U Reynolds numbers, their dependences are relaxed and high
Fr ¼ pffiffiffiffiffiffi ; ð2Þ drag reduction ratio up to 95 % was confirmed within the
gL
cavity closure (Lay et al. 2010). Their work supports design
where U, g, and L are the characteristic flow speed, of necessary ventilation flow rate to maintain the gas cavity
acceleration of gravity, and characteristic length of the flow drag reduction. Amromin et al. (2011) designed a ship hull
configuration. When Fr is sufficiently smaller than unity, with a bottom niche terminating in a cavity locker, which
the buoyancy force stabilizes the gas cavity. For Fr [1, suppresses cavity tail oscillations and reduces the escape of
waves are generated downstream to destroy the cavity. The gas from the cavity. They obtained approximately 25 %
following investigation on such application of an gas cavity drag reduction for 0.4 \ Fr \ 0.7 in a seaway and the
has been reported. power required to supply gas was less than 4 % of the gain
Katsui et al. (2003) measured the drag reduction ratio of in the required propulsion power.
a model ship in the cavity regime. They employed partitions
on the bottom of the ship to provide 12 independent air 2.4 Gas layer effect regime
cavities so that a certain tolerance to waves and ship
oscillation was secured for 0.1 \ Fr \ 0.2. In terms of In ordinary applications, gas cavity is provided artificially
power savings, the gas flow rate required for generating and such as by a stern (see Fig. 1c). When the name of gas
maintaining the air cavity needs to be minimized. This cavity changes to gas layer, it indicates that gas phase
raises two questions: How the gas cavity works properly for thickness is minimized necessary for isolating liquid from

123
1773 Page 8 of 28 Exp Fluids (2014) 55:1773

Table 3 Reports on drag reduction for bubble injection into various configurations of shear flows
Flow configuration Flow configuration Year Investigators Bubble Drag Gain Central
injection reduction factor bubble
into % size

Flat plate boundary layer 2004 Ferrante and Elghobashi DNS 20 10 0.06 mm
2006 Sanders, Winkel, Dowlingm et al. Beneath 90 3 0.3 mm
2007 Wu, Hsu, Lin Beneath 25 5 Broad
2008 Elbing, Winkel, Lay, and Ceccio Above 25 to 80 2 to 4 Broad
2010 Jacob, Olivieri, Miozzi et al. Beneath 9 250 0.15 mm
2013 Elbing, Makiharju, Wiggins et al. Above 95 1.5 Air layer
2013 Mäkiharju, Perlin, and Ceccio Beneath 90 1 Air Cavity
Taylor–Couette flow 2002 Rust and Manga Laminar 50 1.7 1 mm
2007 van den Berg, Luther et al. Turbulent 25 3 1 mm
2008 Murai, Oiwa, Takeda Turbulent 35 7 0.5 mm
2008 Sugiyama, Calzavarini, Lohse CFD 20 20 0.5 mm
2013 van Gils, Guzman, Sun, Lohse Turbulent 40 13 1 mm
2013 Watamura, Tasaka, Murai Transition 20 800 0.06 mm
2014 Maryami, Farahat et al. Turbulent 11 3 1 mm
Ship and model ship 1973 McCormick and Bhattacharyya Submerged 20 200 Microbubbles
1997 Latorre Ship 30 1 Air cavity
2000 Fukuda, Tokunaga et al. Model ship 20 1 Air cavity
2003 Latorre, Miller, and Philips Model ship 11 0.2 Small bubbles
2003 Takahashi, Kakugawa et al. Model ship 40 1 Small bubbles
2003 Katsui, Okamoto et al. Model ship 80 1 Air cavity
2010 Foeth, Eggers, and Quadvlieg Ship Total 0 0 NA
2010 Mizokami, Kawakita, Kodan et al. Ship Total 12 % NA NA
2011 Amromin, Karafiath, and Metcalf Ship Local 25 % NA Air cavity
2012 Mäkiharju, Perlin, and Ceccio Ship Total 14 % NA Air cavity
2010 Kumagai, Murai, and Takahashi Ship Total 14 % NA Small bubbles
Others
Pipe flow 1997 Liu Laminar \0 -6 2 mm
2005 Serizawa, Inui, and Eguchi Transition 75 150 0.04 mm
Sphere 2003 Cui, Fan, and Park Turbulent 15 NA 0.01 mm
2008 Murai and Oiwa Laminar -10 -3 2 mm
Spin shear 2013 Sakurai, Tasaka, and Murai Laminar -60 -30 2 mm
Gain factor is defined by drag reduction ratio divided by bulk void fraction which depends on flow geometry
Data in company’s website and unauthorized publications are excluded

the wall. Hence, gas layer is called so when the gas- proposed by Lee and Kim (2011). Once the wall structure
occupying thickness is less than the boundary layer thick- is allowed to change, stationary arrangement of bubbles in
ness. On this condition, we see two-phase flow patterns a desired pattern can support the drag reduction as reported
such as film flow, froth flow, and horizontally elongated by Kwon et al. (2014).
flat-bubble flow. Elbing et al. (2013) investigated how much gas flow rate
The wall coating of a hull for stabilizing a gas cavity is required to keep the gas layer drag reduction, not falling
was examined by Fukuda et al. (2000). They applied a into bubble drag reduction. They found its critical value
super water-repellent coating and achieved 80 % drag based on scaling for experimental data considering the lift
reduction at flow speeds from 4 to 8 m/s. Their result and buoyancy of dispersed bubbles. Their gas layer contained
implied that total replacement of water with gas is void fraction of 75 % within the boundary layer thickness.
unnecessary in the boundary layer as long as a thin gas film The relationship between the length of the gas bubble
remains on the wall surface. The concept of introducing a elongated in the main stream direction and local skin
super hydrophobic surface into the wall boundary layer was friction was measured by the author’s group (Murai et al.

123
Exp Fluids (2014) 55:1773 Page 9 of 28 1773

Fig. 5 Temporal fluctuation of the local wall shear stress owing to mean void fraction of 20 %. The projected void fraction is defined by
the passage of large bubbles in a horizontal air–water turbulent the ratio of the area occupied by the bubble image on the wall within
channel flow having a bulk mean liquid velocity of 1.0 m/s and bulk the circular area of the wall shear stress sensor (Murai et al. 2007)

2005a, b). Figure 5 shows two samples of the results viscosity of the liquid. Since turbulence has a broad
obtained by synchronized measurement of the local wall spectrum in terms of wavelength and frequency, bubble–
shear stress and local void fraction on the wall inside a turbulence interaction is deduced according to the bubble
horizontal channel flow. They found that bubbles longer size. While there are still deep discussions on the scaling
than about five times the boundary layer thickness reduce problem, the influence is classified at least into two cases.
the average drag around them. This tells that drag reduction One is the case that bubbles are smaller than tens of the
by gas cavity approach requires this length scale at least in wall unit of the boundary layer. Another is the use of
each unit. In contrast, shorter bubbles are not as effective in bubbles larger than those in the first case, but smaller than
reducing drag. Characteristics of such drag-neutral and the boundary layer thickness. The wall unit, l?, is defined
drag-increasing bubbles will be elaborated in the final part by
of this chapter. rffiffiffiffiffi rffiffiffiffiffi
While the gas cavity technique has not been linked to þ m sw 1 2 þ m 2
l ¼ ; us ¼ ; sw ¼ Cf qU ; ! l ¼ ;
turbulence modification, it obviously alters the local tur- us q 2 U Cf
bulence close to the gas–liquid interface. Velocity fluctu- ð4Þ
ations in the tangential direction of the interface are
conserved, but those in the normal direction are restricted where us and q are the friction velocity and density of the
so that turbulence travels in two dimensions (Ouellette liquid. sw and Cf are the wall shear stress and friction
2012). Hence, the scenario of drag reduction employing the coefficient. For an air–water combination with flow speed
gas cavity technique involves curious scientific phenomena of several meters per second, the wall unit becomes
such as the inverse cascade of two-dimensional turbulence roughly 10 lm. Hence, bubbles larger than 100 lm will
(e.g., Boffetta et al. 2000). As long as the coherent struc- directly attack the coherent structures in turbulence with
ture of turbulence is smaller than the length scale of the their volume effect. In contrast, bubbles smaller than
gas–liquid interface, long bubbles also provide local two- 100 lm alter the internal fluid properties inside individual
dimensionalization of turbulence around them as recently eddies. In both cases, the coherent structures that are a
visualized quantitatively by Oishi and Murai (2014) and source of friction can be modified in the following
Park et al. (2014). mechanism.
The effects of small particles and microbubbles on the
2.5 Microbubble regime: bubbles smaller than coherent turbulent boundary layer are commonly explained up to a
structures point. Small solid particles tend to stay in low-speed streak
regions (Narayanan and Lakehal 2003), and they alter the
As small bubbles are mixed into liquid, they interact with coherent structure around them so that drag is reduced
turbulence and the original turbulent structure inside the (Zhao et al. 2010, 2012). Such turbulent structure modified
boundary layer is modified. The Reynolds number by solid particles was summarized by Gore and Crowe
expresses the target flow field attacked by small bubbles; (1989, 1991) and Crowe et al. (1996). They clarified that
particles smaller than 1/10 of integral length scale of tur-
UL
Re ¼ ; ð3Þ bulence always relax the turbulent intensity. Motion of solid
m particles in turbulence was reviewed by Toschi and Bo-
where U, L and, m are the characteristic flow speed, char- denschartz (2009). The main difference of microbubbles
acteristic length of the flow configuration, and kinematic from solid particles is their own density relative to that of

123
1773 Page 10 of 28 Exp Fluids (2014) 55:1773

the continuous phase, which provides opposite Lagrangian Villafuerte (2003). They presented significant drag reduc-
acceleration relative to that of the continuous phase as the tion performance for a horizontal channel flow when
same pressure gradient acts on them. A number of numer- hydrogen microbubbles were mixed with water electroly-
ical research works have investigated this topic prior to sis. In their later papers (Hassan et al. 2005; Zhen and
experimental demonstrations. Felton and Loth (2001, 2002) Hassan 2006; Ortiz-Villafuerte and Hassan 2006), they
simulated the wall-perpendicular diffusion process of small reported that the action of microbubbles destroys coherent
spherical bubbles during downstream migration. The dif- structures. The author’s group also examined the effect of
fusion of bubbles is caused by a random diffusion in tur- hydrogen microbubbles in a 2-m/s channel flow (Murai
bulence, and it is thus promoted with a spatial gradient of et al. 2005a). They obtained 20 % drag reduction with only
turbulence intensity that decreases with distance from the a 0.02 % bubble volume fraction. The sensitivity of the
wall. The phenomenon is the same as that observed for solid drag reduction per unit void fraction can be evaluated by
particles and dye as can be simulated by Reynolds average  
DD=D 1 Cf
model equations. Xu et al. (2002) simulated the drag G ¼ 1 ð1  aÞ ; ð5Þ
a a Cf 0
reduction performance provided by spherical bubbles in a
DNS for a turbulent channel flow. They simulated bubble– where D, DD, and a are the original drag without bubbles,
liquid interaction at a nominal channel Reynolds number of reduced drag as bubbles are injected, and void fraction,
3,000 and bubble diameters of several tens of wall units. respectively. The right-hand side is the formula when
They found the realization of significant drag reduction in Eq. (1) is substituted into the definition. We refer to G as
the transient process of bubble diffusion, which ceased as the gain factor of drag reduction; it indicates the amplifi-
the void fraction reached a steady profile. Ferrante and cation of drag reduction relative to the inertia-originating
Elghobashi (2004, 2005) carried out a DNS for a spatially effect of drag reduction. When the friction coefficient is
developing turbulent boundary layer at a Reynolds number unmodified by bubbles, G becomes unity. Most gas cavity
of 1,400–3,000. Their interpretation of the result is that the methods have a value of G around unity.
motion of small bubbles provides positive divergence of the In the author’s first experiments on microbubbles, the
liquid velocity vector field close to the wall, and the bubbles value G was obtained surprisingly to be 1,000 at Re = 104
push the streamwise vortical structures away from the wall. (Murai et al. 2005a). In the experimental facility, water
It should be noted that in the field of numerical simu- electrolysis apparatus for generating both hydrogen and
lation, the term ‘‘microbubbles’’ is used when bubbles are oxygen microbubbles was flash mounted on the top surface
treated as spheres. Upon the spherical assumption, the of the channel so as not to affect on the liquid boundary
mathematical description of two-phase flow is dramatically layer downstream. The results of a succeeding experiment
simplified as described by the Eulerian–Lagrangian for- using the same facility was reported later by Hara et al.
mulation and point-source approximation based on Stoke- (2011) after careful checking of reproducibility in terms of
sian dynamics. Consequently, the computational load of the performance of bubble generation around the elec-
the DNS is lightened, and therefore, numerical research on trodes. Figure 6 shows their experimental data, confirming
microbubble drag reduction has progressed more quickly that the impact of the microbubble was on the same order;
than research on drag reduction by non-microbubbles. G = 1,100. According to their PTV measurement, Rey-
Experimentalists use the term ‘‘microbubbles’’ for a bubble nolds shear stress was reduced effectively only in the
size that is actually several tens of microns. In clean water, vicinity of the wall and that in the downstream region soon
the lower size limit of non-condensable microbubbles that recovered and rather increased outside the original
survive against their own surface tension is around 5 lm boundary layer. They attributed the dramatic drag reduc-
(e.g., Fujikawa et al. 2011). The upper limit may be sub- tion to the transient process of microbubble motion along
millimeter size, at which bubbles start to show non- the main stream. Lu et al. (2005b) constructed a clone of
spherical deformation in turbulence. The first experiment Hara’s bubble generation device for their different hori-
on microbubble drag reduction by McCormick and Bhat- zontal channel facility and obtained nearly the same degree
tacharyya (1973) employed water electrolysis. The mean of drag reduction; G = 800. For water flows beneath a
bubble diameter estimated using their setup is probably horizontal flat plate, a large effect of microbubbles of
tens of microns although they did not clearly state this. G = 250 was measured by Jacob et al. (2010). They found
Microbubble generation with water electrolysis did not by their PIV a decrease in coherency of the near-wall
come up in the literature again until 2003. Engineers structure. For different flow geometry, large gain factors
believed for 30 years that there was no significant differ- were obtained by Serizawa et al. (2005) for microbubble
ence between submillimeter bubbles and microbubbles. pipe flows (around G = 150) and by Watamura et al.
The first experiment performed this century on micro- (2013) for microbubble-added circular Couette flows
bubbles was that carried out by Hassan and Ortiz- (G = 800).

123
Exp Fluids (2014) 55:1773 Page 11 of 28 1773

Fig. 6 Sensitive drag reduction resulting from microbubbles profiles measured by PTV at 250 mm from the bubble injection point.
observed in a finite region from the bubble injection point. a Friction c The same profiles at 1,000 mm from the injection point (Hara et al.
coefficient relative to the single phase. b Reynolds shear stress 2011)

All the above experimental results are explained by the the critical Weber number (which is around 10) are
action of spherical bubbles on the coherent structure in unstable and fragment into multiple small bubbles and such
turbulence. As shown by numerical analysis of Maxey et al. large bubbles thus disappear downstream. The diameter of
(1996), microbubbles concentrate into strong vorticity the largest bubble that survives in turbulent shear flow is
regions, but low strain rate in Stokes regime of microbubble estimated by substituting the local shear rate into Eq. (6):
motion. L’vov et al. (2005) stated that the volumetric effect  2  
of microbubbles is lost as the microbubble diameter r du du 1
3
d ¼   Wec sw ¼ l  ¼ Cf qU 2 : ð7Þ
decreases. Their linearized theory inferred that drag q dy dy 2
reduction remains as observed in the single-phase turbulent Hence, the upper limit of the bubble diameter is
boundary layer that has wall-perpendicular distributions of obtained as
density and viscosity if clustering of microbubbles is (  )1=3
2
ignored. However, the mechanism in the microbubble 4r m
regime is hardly unified as a single mathematical model, d¼ Wec : ð8Þ
q Cf U 2
consequently. Particularly for gain factors obtained on the
order of hundreds, another scenario must be introduced, The above formula infers that the largest bubble in a
which will be picked up at the end of Sect. 2.8. 5-m/s air–water bubbly two-phase boundary layer has a
diameter of 500 lm, which is comparable with the length
2.6 Mesoscopic bubble regime: bubbles comparable scales of coherent structures such as the spanwise spacing
in size to coherent structures of streamwise vortices, sweeping flow into the wall, and
bursting eddies. In the region far from the wall, the local
The most difficult case for understanding the drag reduc- shear rate is lower and the upper limit of the diameter thus
tion mechanism may be that when the bubble size is higher. Equation (8) also indicates that the upper limit of
comparable to the length scale of coherent structures. This the diameter is a function of flow speed, U, to the power of
is so in both experimental and numerical studies. This -4/3. This fragmentation theory furthermore reminds us
condition is, however, the most frequently appearing in that a histogram of the bubble size will have a peak pop-
laboratory experiments and practical application because ulation close to the upper limit of the diameter at each
such bubbles are easily and naturally realized in the air– position from the wall. The bubbles having diameters
water combination of bubbly flow. This is explained with a around the upper limit always behave non-spherically and
Weber number defined by deform unsteadily in turbulence. Consequently, the mech-
  anism of drag reduction in this regime requires insights not
qUd2 d du
We ¼ ; Ud ¼  d; ð6Þ only into the volume-comparable effect with coherent
r dy
structures but also into the role of bubble deformation in
where Ud is the differential velocity of the liquid phase turbulence. This issue outweighs the density effect in
between two points on a single bubble surface, d is the maintaining steady drag reduction as explained later in
bubble diameter, and r is the surface tension of the bubble Sect. 3.1. Consequently, bubble breakup and bubble
interface. Bubbles that have a Weber number larger than deformation should be always considered in a couple.

123
1773 Page 12 of 28 Exp Fluids (2014) 55:1773

Fig. 7 Complex unsteady

bubble deformation in the wall
proximity of a horizontal
channel flow. a Eight
consecutive images taken over a
period of 2 ms. b Profiles of the
root mean square of the velocity
fluctuation in both phases,
indicating that the translational
velocity fluctuation of the
bubbles is relaxed significantly
by the deformability of bubbles
in strong turbulence (Kitagawa
et al. 2005)

Equation (8) just estimates it in semiempirical form at turbulence. This explains why deformable bubbles, no
dilute bubble situations. There are papers published on matter the precise control of bubble size, work robustly for
shear-triggered bubble breakup phenomena such as Hinze drag reduction.
(1955), and Hesketh et al. (1991). DNS of the mechanism of the interaction between tur-
Kato et al. (1999) and Moriguchi and Kato (2002) per- bulence and deformable bubbles or droplets in turbulent
formed horizontal channel flow experiments and found that shear flow was reported in several papers. Iwasaki et al.
the dependency on bubble size was insignificant in their (2001) found that droplets in turbulent Couette flow
tested range. Towing test experiments for a flat plate attenuated a near-wall streamwise vortex with their defor-
conducted by Takahashi et al. (2003) and for a catamaran mability. Kawamura and Kodama (2002) analyzed a sim-
conducted by Latorre et al. (2003) showed stable drag ilar flow field for air bubbles and found that bubble
reduction that was linear to the gas volume flow rate, deformation altered turbulence statistics. In their simula-
whereas the bubble size was not so carefully controlled. tion, there was an increase in drag owing to the DNS-
Shen et al. (2006) used surfactant to examine the effect of performable limit of the Weber number. That is, the vol-
the mean bubble size in a turbulent channel and concluded ume effect that enhances momentum transfer was stronger
that their drag reduction was insignificantly affected by the than the deformability effects. Iwasaki et al. (2001) con-
bubble size. The gain factors in these experiments ranged firmed the same tradeoff phenomenon between the volume
from 1.5 to 3.5; i.e., they were larger than unity. effect and deformability effect of immiscible droplets
To mechanism-pursuing researchers, the bubble size dispersed in a turbulent channel flow. Furthermore, Lu
insensitivity appears to be rather curious. Kitagawa et al. et al. (2005a) numerically confirmed in DNS that the
(2005) found a reason for the insensitivity. Their particle deformability of bubbles leads to a significant drag
tracking velocimetry for two-phase flow at 5 m/s and a reduction owing to suppression of streamwise vorticity
mean bubble diameter of 500 lm revealed that a bubble’s existing close to the wall, while bubbles that are less
soft deformation owing to surrounding turbulence absorbs deformed provide an additional shear rate near the viscous
Reynolds shear stress as shown in Fig. 7. The concept was sublayer to increase drag. The importance of the defor-
already known to Serizawa and Kataoka (1990), who mability is re-elaborated later in Sect. 3.1.
modeled bubble deformability in isotropic turbulence as a
temporal absorber of kinetic energy that then released the 2.7 Large bubble regime: bubbles larger than coherent
kinetic energy with a time lag. The time lag is on the order structures
of the resonance period of oscillation of the bubble shape
(Ryskin and Leal 1984). For the air–water combination, the Drag reduction is still provided by bubbles that are larger
resonance period of the upper limit of the diameter is much than coherent structures in the wall proximity, but suffi-
longer than the timescales of coherent structures so that ciently smaller than the boundary layer thickness. The
bubble deformability dampens the local acceleration of largeness of the bubbles results in high slip velocities

123
Exp Fluids (2014) 55:1773 Page 13 of 28 1773

Fig. 8 Profiles of turbulent

shear stress decomposed into
three terms measured in a
horizontal turbulent channel
flow. The bulk mean liquid
velocity is 2 m/s, and the mean
bubble diameter is 0.7 mm
relative to the channel height of
2 h = 10 mm. a Measured
profiles at 250 mm (i.e.,
x/h = 50) from the bubble
injection point. b Measured
profiles at 4 m or x/h = 800
(Murai et al. 2006b)

between the two phases. The slip velocity is governed in a where a denotes the local void fraction and l, q, u, and
complex manner, being affected by a combination of seven v are the viscosity, density, wall-parallel velocity and
force components, namely drag, lift, buoyancy, pressure wall-perpendicular velocity of the liquid phase, respec-
gradient, added inertia, and history forces. These forces act tively. The upper bars and primes, respectively, indicate
in different directions with time lags among them in tur- the time average and fluctuation. These profiles measured
bulence. For such a range of bubble size, Guin et al. (1996) at two locations are shown in Fig. 8. The first term is the
obtained 20 % drag reduction with a bulk mean void viscous shear stress, which plays a dominant role in the
fraction of 10 % in a horizontal turbulent channel flow. viscous sublayer, but is negligible outside the layer. The
They found dissimilarity of the void fraction profile in the second term is the Reynolds turbulent shear stress, which
wall-perpendicular direction as liquid and gas flow rates is relaxed by the average void fraction. This term implies
changed. This implies that drag reduction in this regime that drag reduction is simply proportional to the void
should be characterized by spatially developing bubbly fraction supplied in the boundary layer. This inertia effect
flow in the turbulent boundary layer. of drag reduction in turbulence has been widely confirmed
Inside the boundary layer, bubbles in this regime and has a gain factor around unity (Tsai and Chen 2011).
repetitively bounce along a wall. The case for a vertical flat However, the correlation of u0 v0 in the same term can also
wall was investigated by Tran-Cong et al. (2008). The be modified by the bubbles as the turbulence property
author’s group (Murai et al. 2006a) focused on such changes with them. Thus, the effect of the average void
behavior in a horizontal channel flow. Using bubbles with fraction can be amplified. The third term is the shear
mean diameter of 0.7 mm in flow traveling at 2 m/s, we stress induced by the correlation of a0 v0 , and it is inten-
observed active oscillation of the bubbles in the wall-per- sified by the mean streamwise velocity. If the flow is
pendicular direction during their migration downstream. perfectively homogeneous without slip between the two
The oscillation is attributed to the combination of four phases, this term disappears. The term takes a negative
dominant forces, namely drag, added inertia, lift, and value, contributing to drag reduction, when bubbles have
buoyancy, and could be simulated numerically only with wall-perpendicular oscillations interacting with vortical
the time-averaged liquid velocity profile. In particular, structures. A simplified interpretation of the term is that
bubbles close to the wall are rapidly decelerated by the such bubbles work as ‘‘virtual bursts’’ instead of liquid
increase in the drag coefficient (Masliyah et al. 1994) and a turbulence, replacing the sweep and ejection events of the
strong lift force acts on them in the next moment to sep- liquid phase with massless bubbles. The opposite effect is
arate them from the wall again. PTV of both phases by known in the case of a heavy particle mixture in boundary
Murai et al. (2006b) revealed that such a cyclic motion of layers (Kulick et al. 1994; Taniere et al. 1997). The data
bubbles within the turbulent boundary layer reduced local of Murai et al. (2006b) show that this term has stream-
turbulent shear stress. The mechanism was confirmed by wise persistency as long as bubbles have wall-perpen-
substituting PTV data of the liquid flow field into an dicular oscillation, and such a phenomenon stands out in
equation derived by Reynolds averaging the volume-aver- the case of relatively large bubbles. How the local tur-
aged conservation equation of bubbly two-phase flow: bulence modification in the buffer layer reflects on the
du wall skin friction was theoretically deduced by Fukagata
s¼l  qð1  aÞu0 v0 þ qa0 v0  u þ q  a0 u0 v0 ; ð9Þ et al. (2002).
dy

123
1773 Page 14 of 28 Exp Fluids (2014) 55:1773

2.8 Rheological effect regime viscosity decreases to values lower than the original liquid
viscosity at higher values of Ca. Rust and Manga (2002a)
Even without a slip between bubbles and liquid, the shear obtained the relationship between Ca and the bubble
stress is still affected by the presence of bubbles. As deformation ratio in simple shear flow and confirmed the
revealed in the earliest studies of Einstein (1906) and validity of the above formula by means of circular Couette
Batchelor (1967), the effective viscosity of a dilute sus- flow viscometry (Rust and Manga 2002b). In a typical case
pension is described by of drag reduction for a ship, Ca takes values that cross the
l l þ 5 l0 critical capillary number as the local shear rate is defined
¼ 1 þ 0 2 0 a; ð10Þ by the coherent structures in the boundary layer. It is thus
l0 l0 þ l
deduced that the shear-thinning property of the bubbly
where l , l0, l0 and a are the effective viscosity, original liquid affects turbulence. Zhen et al. (2013) simulated such
viscosity of the continuous phase, viscosity within the a response to hairpin vortices using a power-law model,
dispersion phase, and volume fraction of the dispersion. implying the hidden importance of the non-Newtonian
The formula is applicable to dilute spherical dispersion property of deformable bubbles in drag reduction.
(a \ 0.10) in simple shear flow. For solid particles, there is A further advanced question in regard to the rheo-
no chance to reduce effective viscosity because of l0 [ l0 : logical effect is the viscoelastic response of small bub-
Increasing of the volume fraction exponentially amplifies bles. Since the shear rate fluctuates at a high frequency
the viscosity (Stickel and Powell 2005). In the case of a around individual bubbles in the turbulent boundary
bubbly liquid, the viscosity of gas inside bubbles is suffi- layer, the effective viscosity formulated for steady shear
ciently small compared with that of the liquid and thus flow leads to misunderstanding. In general, the shear
stress of a multiphase fluid element with a high interfa-
l ¼ ð1 þ aÞl0 : ð11Þ
cial area concentration is described by tensor equations as
Hence, the effective viscosity of the spherical bubble presented by Doi and Ohta (1991). For a dispersed
mixture increases. Legner (1984) explained in a qualitative bubble system, Llewellin and Manga (2005) proposed
sense that the increase in effective viscosity thickens the introducing the dynamic capillary number to describe the
viscous sublayer and thus reduces the shear rate. However, ratio of elastic to viscous contributions to the local shear
the role of the effective viscosity in the entire turbulent stress. Murai and Oiwa (2008) confirmed with their
boundary layer is still unsolved. Sangani et al. (1997) falling-sphere viscometry that the effective viscosity
analyzed the rheological resistance of densely arranged increases drastically, departing from equilibrium defor-
spherical bubbles subject to a rapidly applied shear. They mation theory, as bubbles are subject to transient defor-
obtained an extra increase in the resistance over that given mation. The experiment performed by Cui et al. (2003)
by Eq. (11) owing to local bubble–bubble interaction. showed 15 % drag reduction for a falling sphere in a
L’vov et al. (2005) coupled the effective viscosity of highly turbulent regime of Re [ 104. Decoding of their
spherical bubbly liquid with Reynolds-averaged equations experimental condition matches the deformable bubble
and estimated how drag reduction is preserved in homo- regime, which alters the turbulent boundary layer sepa-
geneous bubbly flow situations when the void fraction has ration point on the sphere. For such viscoelasticity of
a wall-perpendicular profile. bubbly liquid originating from surface tension, it is sug-
The effective viscosity becomes a function of the cap- gested that there is a need for more fundamental studies
illary number as the bubble deforms significantly owing to referring to the elastic resonance of bubbles in temporally
an increase in the shear rate. Frankel and Acrivos (1970) fluctuating shear (Gao et al. 2011) and in a constraint
derived the formula environment (Prosperetti 2012). The momentum transfer
 
l 1  12
5 Ca
2
l0 c d du from a stepwise accelerating wall measured by Sakurai
¼1þ 6 2
a; Ca ¼ ; c ¼  ; ð12Þ et al. (2013) revealed a 60 % increment in effective
l0 1 þ 5 Ca r 2 dy
viscosity with a void fraction of only 2 % as Ca
which is the same as that derived by Schowalter et al. unsteadily changed across unity. Figure 9 shows their
(1968) when the void fraction is low and where r, c, Ca, sample data. The figure implies that such bubbles sup-
and d are the surface tension of the bubble surface, shear press turbulent eddies effectively; however, further
rate of the liquid around the bubbles, capillary number, and investigation is required for a more generalized formu-
sphere-equivalent bubble diameter. In this formula, the lation. Relevant to this rheological effect, we should
relative viscosity returns to unity at Ca2 = 5/12; i.e., the consider the phenomenological analogy to another type of
critical value of Ca at which the sign of the effect on drag reduction provided by a polymer surfactant solution
the void fraction changes is Cac = 0.65. The effective such as that measured by Li et al. (2008).

123
Exp Fluids (2014) 55:1773 Page 15 of 28 1773

Fig. 9 Increase in the effective viscosity of bubbly liquid subject to owing to rapid spinning of the cylinder. b Effective viscosity at a
transient bubble deformation. The two phases are air gas and highly point 10 mm from the wall, measured by ultrasound Doppler
viscous oil at room temperature, and fill a cylindrical container with a rheometry combined with a high-speed video camera system (Sakurai
diameter of 145 mm. a Top view of bubble deformation near the wall et al. 2013)

2.9 Fat bubbles comparable to the boundary layer as shown in Fig. 10. Four bubbles in the figure stably slide
thickness beneath the wall without temporal change to their shapes.
The shape-fixing effect is provided by gravity; i.e., Fr
Injection of fat bubbles into the liquid boundary layer number instead of the We number characterizes the shape.
beneath a horizontal wall greatly increases the friction The liquid film flow that remains between the wall and
coefficient. The resultant increases in wall shear stress bubble interface is governed by Ca: The top views of large
overshadow the effect of reduced fluid density. Fat bubbles bubbles accompany capillary waves. An important finding
referred to here are bubbles sufficiently larger than the from their PTV is that the velocity fluctuation correlation
coherent structure of turbulence as they occupy most of the u0 v0 takes a value one order of magnitude higher than that
boundary layer thickness with their single diameter. Unlike of the original Reynolds shear stress in a single-phase
an air cavity, the length in the streamwise direction is boundary layer. The result explains consistently the loss of
limited to a few times the boundary layer thickness. Their local skin friction reduction around large bubbles (Murai
Weber number defined by Eq. (6) ranges from 10 to 300 in et al. 2007).
the air–water case, which is much larger than the critical In the field of heat-transfer research, the role of such fat
Weber number for fragmentation in free shear flows. Such bubbles close to a wall has been investigated intensively.
an effect was measured in a vertical channel by Liu (1997). Fat bubbles, no matter of whether they comprise air or
The bubbles in his facility ranged between 5 and 15 mm in water vapor, promote turbulent heat transfer of the wall
diameter, which was comparable to the boundary layer unless the bubbles entirely cover the wall to form an adi-
thickness. In horizontal flow beneath a wall, the fatness is abatic sheet. A similar phenomenon for large bubbles is
maintained by buoyancy that keeps the bubbles stably observed in laminar heat transfer (Kitagawa and Murai
beneath the wall. Numerical analysis of Kawamura and 2013) and laminar mass transfer. In laminar flow states,
Kodama (2002) simulated the increase in wall shear stress bubbles larger than the thermal and concentration boundary
owing to the bubbles that mostly occupy the boundary layer thicknesses provide pseudo-turbulent transport of the
layer thickness. The finding of such a drag-increasing carrier phase. Such paradoxical experimental results
condition allows substantial improvement of drag reduc- between drag reduction and heat/mass transfer enhance-
tion performance. ment provide an analogous understanding of the role of
It is easily imaginable that fat bubbles moving along the large bubbles in the boundary layer.
wall force the ejection of liquid at their fronts and induce
an extra sweep in their rears. In the case that the bubbles 2.10 Transition diagram of the drag reduction
have significant slip velocity relative to the liquid, a similar mechanism
additional momentum exchange occurs owing to their large
volumes being subject to a steep velocity gradient. Oishi As mentioned throughout this chapter, there are multiple
and Murai (2014) measured bubble-induced secondary scenarios of the effects of bubbles in drag reduction. The
flows around such fat bubbles in a turbulent channel flow author does not explain them in complex detail, but each of

123
1773 Page 16 of 28 Exp Fluids (2014) 55:1773

Fig. 10 Secondary flow patterns induced by fat bubbles co-currently size, and c the spatial structure of the u0 v0 distribution formed around
migrating with liquid in a turbulent channel flow. a Typical top-view each bubble where the value is scaled by the friction velocity squared
images of fat bubbles of four different sizes, b streamlines of the (Oishi and Murai 2014)
secondary flow component measured by PTV corresponding to each

them physically exists and has been experimentally dem- diagram is approximately valid for horizontal turbulent
onstrated. Hence, we should remind ourselves that the boundary layers of water, but invalid for other flow con-
simplest term ‘‘air-lubrication,’’ which has been preferably figurations and other liquids such as highly viscous oil.
used in the field of drag reduction for ships, is a grand Words within each region highlight the phenomenon rele-
generic term for an extremely complex combination of vant to drag reduction, and the value of G indicates the
different mechanisms of drag reduction. Without such resultant mean gain factor [see Eq. (5)] realized in each
consideration, one would have an issue with the lack of region.
experimental reproducibility for the void–drag relationship, In the region ‘‘Spherical’’ at the bottom-left corner,
which is a primary problem in application. In parallel, we frictional drag is in proportion to the effective viscosity of
need to appreciate that multiphase flows have their inter- the spherical bubble mixture so that frictional drag
nally created deviations regardless of our strict control of increases in accordance with Eq. (11). The region
the flow geometry and inlet bubble injection conditions. ‘‘Yielding’’ on the right next is that for the use of highly
What we can do in this situation is to make a rough deformable bubbles, in which drag reduction is restored by
classification of the local dominant mechanism in param- the bubble yielding effect. The theoretical gain factor is
eter space by clustering similar results. This work helps in obtained as 5/3 as Ca goes to infinity in Eq. (12). Further
practical design straggled between the multiplicity of the large bubbles in slow two-phase flow transit to the region
drag reduction mechanism and the inevitable deviating of ‘‘Gas film’’ because of buoyancy and provide drag
nature of bubbly two-phase flow. On the basis of this reduction of G = 1.
concept, a rough sketch of a drag reduction mechanism The flow in the transition region from laminar to tur-
diagram is finally produced as shown in Fig. 11. The def- bulent states (0.1 \ U \ 1 m/s) is sensitively affected by
inition of the two-dimensional parameter space is the same the injection of bubbles since bubbles perturb the laminar
as that for Fig. 4. The lines that separate the domain into state and trigger the flow transition to turbulence (Huang
seven regions are determined from experimental data. et al. 2009). In most cases, therefore, bubbles rather acti-
Thus, the lines can be termed as transition lines of the vate the momentum transfer of the boundary layer so that
dominant drag reduction mechanism. It is noted that this drag increases. The same effect is known for free shear

123
Exp Fluids (2014) 55:1773 Page 17 of 28 1773

3 Current problems and ideas

Together with the progress in understanding the drag

reduction mechanism, limitations of the practical use of
drag reduction are gradually emerging. They are (1) the
short spatial persistency of the drag reduction effect for a
large target like a vessel and (2) management of the rich
unsteadiness that naturally occurs in the two-phase
boundary layer. In response to the former problem, Kim
and Cleaver (1995) proposed a simplified fitting function
with which to estimate the persistency of drag reduction as
a function of the gas flow rate. The function was such that
drag reduction is always and fully lost far downstream.
Fig. 11 Transition diagram of the drag reduction mechanism owing Main reason of the decline is bubbles’ turbulent diffusion
to bubble injection which will relax the peak void fraction near the wall.
Sanders et al. (2006) investigated the downstream transi-
flows and free-rising-bubble flows referred to as pseudo- tion of bubble distribution along a 11-m-flat plate. They
turbulence or ‘‘turbulence’’ (Lance and Bataille 1991). The found that small bubbles at far downstream region behaved
pseudoturbulence is decomposed into dipole-type potential similarly to a passive scalar turbulent diffusion process
disturbance and viscous wake disturbance. In either event, (Poreh and Cermak 1964). This turns out the fact that
flow transition is promoted. This is a delicate problem bubbles no longer affect liquid phase there. However, gas–
especially in numerical simulation since it cannot be liquid two-phase flow confined in a thin turbulent boundary
directly analyzed with a Eulerian formulation of volume- layer serves a further variety of flow patterns than we
averaged equations (Kitagawa et al. 2001). On bubble- expect. Experiments in them give us many more ideas for
induced turbulence close to the wall, Kobe university group the improvement, being challenged by many investigators.
collected data of vertical two-phase flow systems, which For instance, Elbing et al. (2008) found an abrupt transition
give partial analogous discussion to that observed in hori- from dispersed bubble state to gas layer structure, which
zontal systems. Hosokawa and Tomiyama (2013) found jumps up local drag reduction ratio from 25 to 80 %. Thus,
that Reynolds shear stress generated by bubble-induced we have ample scope for improvement if we carefully
pseudoturbulence was proportional to the ratio of void observe the nature of two-phase flow. This chapter intro-
fraction to shear rate. A paper from their group, Ojima duces and reviews ongoing research into the advances of
et al. (2014) clarified which terms in two-fluid formulation drag reduction.
turbulent flow model take dominancy in describing bubble-
induced turbulence. 3.1 Bubbly Taylor–Couette flows
Beyond the turbulent flow transition with an increase in
flow speed, drag reduction revives as shown by the region Shear flow generated in concentric annuli has been
‘‘Braking u0 v0 ’’ where the gain factor ranges from 1 to 5. employed by several experimentalists in the fundamental
Most cases of the so-called microbubble drag reduction for study of bubble-originating drag reduction. The flow field
a ship belong to this region. Use of larger bubbles at the is called Taylor–Couette flow (hereafter T–C flow). Its
same flow speed causes a transition to ‘‘Interfacial insta- fluid dynamic characteristics have been richly explored for
bility’’ before stabilizing as a gas layer along the wall. In single-phase Newtonian fluid (Andereck et al. 1986; Tak-
this region, bubbles subject to strong turbulent shear cannot eda et al. 1994). Historically, T–C flow has long been
keep their size due to active fragmentation and coales- investigated as a platform to elucidate scenarios of flow
cence. The region ‘‘Rheology’’ at the top-left corner in the transition between laminar circular Couette flow and highly
diagram has a high gain factor exceeding 100. Problems in turbulent Couette flow (La Porta et al. 2001) starting with
this region are the considerable power required for the onset of a Taylor vortex (Taylor 1923). Since the flow
microbubble generation and turbulent diffusion of micro- modes shift stepwise as the Reynolds number increases, the
bubbles from the inner layer of the turbulent boundary effect of bubble injection on the flow transition can be
layer. detected clearly. Another great advantage in using T–C
The transition lines in the diagram shown here are flow is that the flow is organized in a spatial closure
simply presented to inspire researchers in the field and are between two cylinders. Thereby, the momentum transfers
not to be taken as universal absolutes. It is anticipated that can be assessed in a steady state under an ensured global
the lines will be refined by further investigations. energy balance. One can expect that the influence of

123
1773 Page 18 of 28 Exp Fluids (2014) 55:1773

Fig. 12 Regularized bubble distribution and its effect on drag a Side view of bubble distributions at different Reynolds numbers,
reduction in Taylor–Couette flow. Silicone oil and air are used as showing random, toroidal, spiral, and turbulently diffused toroidal
the two phases. No base axial flow is given to the liquid, while modes. b Gain factor of drag reduction changing with the mode
bubbles are continuously injected from the bottom at a fixed flow rate. switching of the bubble distribution pattern (Murai et al. 2008)

bubbles is exactly judged in the confined fluid shear from reduction of 20–25 % with a bulk mean void fraction of
which no bubbles diffuse to escape. The idea contrasts the 4–8 % and Re from 105 to 106. They compared the drag
effect of bubbles with a spatially and temporally develop- reduction between small air bubbles and buoyant particles
ing two-phase boundary layer either in a channel flow or and ascertained that only air bubbles can achieve drag
from a flat wall. The comparison hints at how bubbles can reduction. This clearly demonstrated that bubble defor-
reduce drag in a fully developed state, while several papers mability plays a major role in drag reduction in the highly
reported the priority of largely enhanced drag reduction turbulent regime. The latest study of van Gils et al. (2013)
during spatial development of the two-phase boundary found a 40 % drag reduction with a 3 % void fraction by
layer (Xu et al. 2002). further increasing Re beyond 106. They witnessed that the
The flow structure of bubbly T–C flow is classified by how promoted drag reduction originated from a shift in the
the bubbles are mixed. Shiomi et al. (1993) measured bubble Weber number from around unity to the order of 10. This
distribution patterns varying with the flow rates of two axially scenario agrees with the work done by Kitagawa et al.
co-current phases. Atkhen et al. (2002) measured the phase (2005) for turbulent channel flows at a high Reynolds
velocity of the organized bubble distribution migrating with number. We are now reaching a conclusion that a steady
axial liquid flow. Air bubbles and cavitation bubbles were drag reduction effect for turbulent shear flow having high
compared by Djeridi et al. (2004). Chemical engineers Re number relies on bubble deformability.
reported the enhancement of the gas–liquid interfacial area For bubbly T–C flows at low Reynolds numbers, Murai
concentration by an array of vortical cells using horizontal- et al. (2008) obtained 35 % drag reduction at Re of about
axis-type T–C flow (Hubacz and Wronski 2004). A variety of 103 and a void fraction of 5 %. In this Re range, the Taylor
interfacial structures in the high gas volume fraction were also vortex conveys momentum between the two cylinders, and
reported (Wronski et al. 2005; Mehal et al. 2007). In these vortex spacing and its topology were investigated by
two-phase flows, there cause air spots owing to bubble coa- quantitative visualization of the inner structure. Their main
lescence promoted by Taylor vortices and the centrifugal finding was that bubbles increase the spacing between
force of the rotating system. For relatively small bubbles, wavy traveling Taylor vortices so that drag is reduced.
Climent et al. (2007) simulated preferential bubble accumu- Sugiyama et al. (2008) simulated numerically such an
lation to wavy vortical structures, which demonstrated the interaction and found a collapse of the waves owing to
organized action of small bubbles on coherent structures for spatially organized attack of spherical bubbles on the
dilute bubble injection. coherent structure. As the void fraction increases in this
The actual performance of drag reduction in bubbly T–C regime, the array of toroidal Taylor vortices shifts to spiral
flow was measured by two research groups: A group at the ones via emergence of the ellipsoidal instability of the
University of Twente investigated the highly turbulent Taylor vortex. The drag reduction performance per unit
regime, and a group at Hokkaido University focused on a bubble buoyancy is a maximum in this topological transi-
weak turbulent regime. In the former case, van den Berg tion regime as reported by Murai et al. (2008). Figure 12
et al. (2005) and van den Berg et al. (2007) obtained a drag shows a portion of their data. The internal two-phase flow

123
Exp Fluids (2014) 55:1773 Page 19 of 28 1773

Fig. 13 Mode transition

delayed by the dilute injection
of microbubbles in Taylor–
Couette flow. The top two
figures show the spatiotemporal
structure of the wavy Taylor
vortex visualized by a
Kalliroscope. The bottom two
show the corresponding power
spectra measured by ultrasound
Doppler velocimetry

in such a transition regime was measured by Yoshida et al. structures. For high-speed T–C flows, the effect of buoy-
(2009). They revealed global periodic switching between ancy becomes unimportant, and instead, bubble deforma-
the toroidal and spiral modes, which naturally arises from bility supports a large drag reduction since length scales of
the time lag between drag reduction and the formation of coherent structures are shortened to less than the bubble
the corresponding bubble distribution pattern. Their group size.
applied this interpretation to microbubbly T–C flow
recently. As represented in Fig. 13, Watamura et al. (2013) 3.2 Effect of void waves and clustering
found a delay in mode transition from the wavy Taylor
vortex to the modulated one with the dilute injection of Bubbly two-phase flow more or less shows naturally
microbubbles. Their drag reduction in this regime reached induced fluctuation in the void fraction except in the cases
10 % with a void fraction of only 0.01 %. The bubble size of ideal inertia-less Stokes and Poiseuille flows. The
dependent various trajectories in T–C flow were analyzed question is how the spatiotemporal scale and the amplitude
by Chouippe et al. (2014) based on Lagrangian approach. of such fluctuation affect the average drag reduction per-
They elucidated the condition to cause a statistic prefer- formance. The fluctuation originates from the relative
ential accumulation of bubbles subject to 3D turbulent motion of bubbles to the surrounding liquid. The term void
shear flow. A curious data of using 3D bubble behavior was wave expresses the propagation of the void fraction at a
reported by Maryami et al. (2014). They observed com- specific speed different from the bubble migration velocity.
bined improvement of drag reduction when microbubbles Compressibility of gas bubbles creates void waves together
and axial flow were cooperatively imposed into vertical with pressure waves (Biesheuvel and van Wijngaaden
T–C flow. 1984; Zhang and Prosperetti 1994). Since the speed of
Summarizing the drag reduction ascertained using T–C sound in bubbly liquid is dramatically lowered (roughly
flow, three roles of bubbles are confirmed. In slow T–C 30 m/s at a void fraction of 20 %), the bulk compressibility
flows, the buoyancy of bubbles alters the original T–C flow of bubbly mixture cannot be ignored in application to high-
structure to totally different structures as gas–liquid two- speed bubbly flows. For flow traveling faster than 10 m/s,
phase flow. However, a dilute injection of bubbles works the possibility of cavitation should also be considered
sufficiently well for drag reduction because the bubbles (Ceccio 2010). The compressibility-induced void wave has
automatically structure the concentration spatially in an oscillatory wave front originating from the volumetric
accordance with vortical structures. This is known as the pulsation of the bubble (Kameda and Matsumoto 1996). As
preferential concentration effect, which intensifies the the resonance frequency of the pulsation falls in the band of
bubble–liquid interaction in particular for unsteady vortical the turbulent coherent structure, there is interaction

123
1773 Page 20 of 28 Exp Fluids (2014) 55:1773

between them. The resonance frequency is known to be of acting on bubbles (Takagi et al. 2009). In short, the sur-
the order of 100 kHz for a 1-mm air bubble in water at factant determines the emergence of bubble clustering near
atmospheric pressure. It is, however, noted that the spher- the wall in such a vertical system, and thereby affects the
ical volumetric pulsation of individual bubbles produces turbulent boundary layer (Takagi and Matsumoto 2011).
only irrotational velocity fluctuation in the liquid phase, This is one reason why measurements of drag reduction are
and there has been no report of the importance of the sometimes diverse without the control of contamination.
pulsation in drag reduction, up to present. The author’s group is currently focusing on bubble–
Another factor triggering a void wave is the bubble’s bubble interaction seen in horizontal wall boundary layers.
translational motion. The equation of motion of a single Figure 14 shows a variety of bubble distribution patterns
spherical bubble comprises seven force components. photographed from the top of a horizontal channel flow.
Details were reviewed by Michaelides (1997) and Mag- The ordinate indicates a rough estimate of the projection of
naudet and Eames (2000). In monodispersed bubbly flow, the void fraction defined by the wall-occupying ratio of
the combination of these forces easily produces a spatially bubbles. While this mapping of bubble clustering is still
structured distribution of bubbles as a void wave. In the incomplete, it is immediately understandable that the
Stokes drag regime of bubble motion, bubbles in liquid bubble–turbulence interaction cannot be studied according
turbulence tend to accumulate into the region where the to the motion of a single bubble. A random distribution can
second invariant of velocity gradient tensor is negative, only be seen for the low void fraction in laminar flow. We
while particles heavier than liquid accumulate into the observed lateral waves, streamwise chains, longitudinal
positive region (Kitagawa et al. 2001). In sheared turbu- clusters, slugging waves, and V-shaped clusters. Some
lence, a curious example of such intrinsic behaviors was migrate steadily and others show the dynamic and active
analyzed by Tanaka (2013). He found that the Eulerian exchange of bubbles between clusters. We also observed a
phase velocity of bubbles takes the inverse sign to the uniform distribution for the high void fraction in a laminar
shearing direction. In bounded flows, the author’s group flow state. The uniform distribution, which should be dis-
found strong standing and pulsatile void waves in the tinguished from a random distribution, means that bubbles
buoyant rise of a bubble swarm close to a solid wall (Ki- strongly interact with repulsion forces for each neighbor.
tagawa et al. 2004; Kitagawa and Murai 2013, 2014). In a This is relevant to the observation made by Timkin and
horizontal flow configuration, a similar integration of mi- Gorelik (2010). They found homogeneously sliding bub-
crobubbles to pulsatile migration was reported by Wu et al. bles in a vertical tube, which multiply the pipe wall friction
(2007). They noted this effect as experimental fact when in the transitional regime from laminar to turbulence. In
significant drag reduction was confirmed. Their following turbulent flows, the photographs show streamwisely elon-
paper (Wu et al. 2008) analyzed which combination of gated void stripes that appear similar to what we know of
parameters the most importantly dominates average drag the wall turbulence structure. A surprising feature of
reduction. Fig. 14 is that there is less coalescence of bubbles inside
Bubble–bubble interaction is the third factor to take into the bubble cluster in highly turbulent regimes. This implies
account in the generation of void waves. This effect cannot that the repulsive force intensifies during the rapid con-
be described by any treatment of the equation of motion for gregation of bubbles. Huang et al. (2009) measured mod-
a single bubble. That is, the local liquid flows close to the ifications of sweep and ejection events through PIV and
bubble interface interact with each other as the distance clarified that a bubble swarm more strongly suppresses
between two bubbles decreases. In unbounded space, the actions than isolated bubbles. Oishi et al. (2009) measured
interaction was analyzed by Sangani and Didwania (1993) the correlation between the passage of void waves and the
and Seo et al. (2010) and measured by Brücker (1999) and local skin friction in a horizontal channel flow. Typical
Murai et al. (2006c). In vertical channel flows, local bubble data are shown in Fig. 15. They found a time lag in the
clustering and its effect on near-wall turbulence was ana- fluctuation between the two and ascertained that the time-
lyzed by Zenit et al. (2001) for spherical bubbles, and by averaged drag reduction was promoted as the void wave
Bunner and Tyggvason (2003) and Lu and Tryggvason had large amplitude. This demonstrates that there is an
(2007) for softly deformable bubbles. As plenty number of intrinsic dynamic two-way interaction enhanced by the
bubbles slide up along a vertical wall, bubble–bubble void waves along the wall. That is, a swarm of bubbles
interaction was clearly observed in two dimensions (Ki- reduces local frictional drag around it, and the drag
tagawa et al. 2004). In upward co-current bubbly channel reduction affects the migration of the swarm. Park et al.
flows, So et al. (2002) found intermittent swarms of sliding (2009) examined the same effect using artificially gener-
bubbles along the vertical walls as bubbles were nearly ated void waves produced by the intermittent injection of
monodispersed. The generation of such sliding bubbles bubbles. They obtained a gain of the drag reduction higher
depends on the water-in surfactant which changes lift force than that in the case without the intermittency, and their

123
Exp Fluids (2014) 55:1773 Page 21 of 28 1773

Fig. 14 Bubble distribution patterns photographed from the top of a height of the channel and bulk mean velocity of the liquid. Image
horizontal turbulent channel flow. Silicone oil (5cSt) is used to avoid sizes vary and are 50 mm for local imaging and 160 mm in the largest
the contamination effect. The Reynolds number is defined by the half case

Fig. 15 Synchronized measurement of the local wall shear stress and

local void fraction on the top wall of a horizontal bubbly channel
flow. Both sensing areas are defined by a circle having a diameter of
20 mm, where bubbles of 0.5–1.5 mm pass by at about 1 m/s (Oishi
et al. 2009)
Fig. 16 Three kinds of two-way interaction considered in turbulent
following paper reports how turbulent eddies are modified bubbly flow analysis
by each single passage of the void waves (Park et al. 2014).
Macroscopically, a certain part of the void wave effect is overlapping regions of two circles indicate two-way
explained by the discrepancy from the linearity of the drag interaction between the two elements. Overlap ‘‘A’’ can be
reduction versus the average void fraction. The extra effect analyzed employing a Eulerian–Eulerian two-phase flow
of the void waves could be attributed analogously to the approach such as the use of a two-fluid model, while the
compliant wall surface (Kramer 1960) and accelerating overlap ‘‘B’’ can be analyzed by taking a Eulerian–
boundary layer (Piomelli and Yuan 2013) since void waves Lagrangian approach. Overlap ‘‘C’’ indicates bubble–bub-
make the turbulent boundary layer thicker and thinner ble interaction that takes place when bubbles have a slip
unsteadily. velocity relative to the liquid phase. The triple overlap of
To summarize this section, the analytical methodology the three circles ‘‘D’’ is the future target of research toward
for investigating the drag reduction mechanism is illus- a comprehensive understanding of the drag reduction
trated in Fig. 16. Three circles are drawn in the figure to mechanism, which also leads to an improvement in drag
define explicitly the role of bubble–bubble interaction. The reduction performance.

123
1773 Page 22 of 28 Exp Fluids (2014) 55:1773

3.3 Advances in measurement techniques Digital image processing is widely employed in bubbly
flow measurement in two dimensions. Moriguchi and Kato
In the field of ship research, the overall measure of drag (2002) analyzed a bubble interfacial image illuminated by a
reduction performance is the reduction of the fuel con- laser sheet to measure the void fraction profile close to the
sumption rate. For large vessels such as tankers and con- wall in a horizontal channel flow. Backlighting of bubbles
tainer carriers, turbulent frictional drag accounts for more provides clear shadow images so that the shape information
than 70 % of total drag, and hence, a change owing to of bubbles is simultaneously obtained as applied by Fu-
bubble injection is reflected in the shaft power for pro- jiwara et al. (2004) for a vertical bubbly channel flow.
pulsion (Latorre 1997; Kumagai et al. 2010). In a towing While a single view of backlight bubble imaging is limited
experiment for a model ship and a flat plate, the towing to dilute bubbly flow, stereoscopic imaging allows mea-
force becomes a direct measure (Watanabe et al. 1998). In surement of the three-dimensional void fraction. Statistic
such research, the general relationship between the air and deterministic methods were proposed by Murai et al.
injection flow rate and the resultant drag reduction was (2001), and Luo et al. (2002), respectively. For bubbles in
obtained for each type of ship hull as the cruising speed laser illumination, Kawaguchi et al. (2002) proposed the
changes. The thickness of bubbly two-phase layer on the use of interference fringes converted to a one-dimensional
hull cannot be estimated accurately from these quantities stripe pattern for bubble detection and size measurement.
while only the equivalent thickness of a gas film is inferred Its combined use with PIV was realized for spraying phe-
assuming that all the gas migrates along the wall at the nomena (Hardalupas et al. 2010), which is expected to be
same speed as the liquid phase. Even the mean void frac- applied to microbubbly turbulent flows. For vertical bubbly
tion in the boundary layer can hardly be approximated pipe flows, Lelouvetel et al. (2014) has successfully
because it ordinarily has a steep profile as a function of the extracted turbulent cascade process altered by bubbles by
distance from the wall. Foeth et al. (2010) reported little means of PTV combined with bubble-shape projection
effect of drag reduction in a series of full-scale vessel imaging technique.
experiments using small bubbles. They explained this by For gas–liquid two-phase flow of highly concentrated
bubbles escaping the boundary layer. To solve such an gas–liquid interfaces, radiation type visualization works
issue in application, a variety of measurement techniques instead of the above-mentioned optical approach. X-ray
have been developed to elucidate the two-phase flow measurement instrumentation is designable since it is
characteristics as introduced below. The advance in mea- attenuated in liquid phase, but insignificant in gas phase.
surement technique serves to bridge our fundamental Stutz and Legoupil (2003) sandwiched a high-speed Ven-
understandings of the drag reduction mechanism and how turi throat by the source and the detector of X-ray and
practical applications should be designed. successfully measured oscillating cloud-state cavitation
Measurement of the void fraction is the most funda- bubbles at 1 kHz in sampling frequency. For two-phase
mental requirement for gas–liquid two-phase flow in any flow in a circular tube, many pairs of the X-ray sources and
flow geometry. Optical probing is an existing technique detectors can be arranged in the angular direction to realize
that allows point-wise measurement of the void fraction. computed tomographic visualization as presented by
For turbulent channel flow, Guin et al. (1996) and Gabillet Fischer and Hampel (2010). The application to a partial
et al. (2002) adopted an optical fiber probe that traversed to cavity flow was recently reported by Mäkiharju et al.
measure the streamwise development of the vertical void (2013a, b). They constructed a two-dimensional X-ray
fraction profile. Its advanced version realizes bubble- densitometry system which can acquire the void image at
counting and bubble size measurement by mounting mul- 1 kHz and successfully explored the internal unsteady void
tiple apertures on its sensing head and combining with profile as a function of distance from the wall.
time-resolved signal processing (Luther et al. 2004; Hig- Ultrasound senses bubbles from the differential acoustic
uchi and Saito 2010). For bubbly flows in a pipe, electrical impedances of two phases and has long been an alternative
capacitance tomography is recognized as a useful tech- to optical techniques. Amplitude modification of reflected
nology with which to approximately measure the cross- ultrasound is used in medical instruments where encapsu-
sectional void fraction distribution (e.g., Warsito and Fan lated microbubbles are injected into blood vessels as an
2001). In contrast, a wire-mesh sensor installed inside a ultrasonic contrast agent. For the detection of bubbles in a
pipe allows direct measurement of the phase distribution flowing state, the ultrasonic Doppler shift (thus frequency
(Prasser et al. 2001). Its extension to bubbly channel flow modification) also provides the interfacial location as
was presented by Richter et al. (2002). They succeeded in developed by the author’s group (Murai et al. 2006a, 2010).
simultaneous measurements of the transient void fluctua- In its application, void fraction profiles of a bubbly
tion, bubble size, and bubble velocity by means of a well- turbulent boundary layer were successfully measured.
devised spatial arrangement of the wire mesh. Figure 17 shows sample data from Murai et al. (2009).

123
Exp Fluids (2014) 55:1773 Page 23 of 28 1773

Fig. 17 Ultrasound Doppler

sensing of bubbles applied to
the turbulent boundary layer
formed beneath a long flat-
towed plate with the injection of
bubbles. a Detected bubbles
marked in black in
spatiotemporal Doppler shift
information. b Wall-
perpendicular profile of detected
bubbles b(y) and reconstructed
void fraction profile a(y) at a
towing speed of U = 4.0 m/s.
The measurement section is
25-m downstream of the bubble
injector. c The same profiles at
U = 5.1 m/s (Murai et al. 2009)

The signal of the bubble interface (‘‘DB’’ in the figure) on deformable bubbles, Huang et al. (2008) proposed a shal-
the map of the Doppler shift frequency is spatially inte- lowly set depth-of-focus (DOF) for backlighting PTV.
grated with a convolution function to obtain the short-time They measured bubble-induced velocity fluctuation within
average void fraction profile as a function of distance from a DOF-defined layer close to the wall. Figure 18 shows
the wall. The profiles are those beneath a long horizontal their samples, depicted by in-plane divergence of the
flat plate towed in stationary water at the National Mari- measured velocity vector field within 30 wall units from
time Research Institute (NMRI-Japan). The data show a the wall. The sink and source are recognized clearly in the
shift in the void peak toward the wall as the towing speed downstream and upstream regions of bubbles, respectively.
increases. Recent progress of the ultrasonic Doppler Since the pair of sink and source corresponds with ejection
method for fluid flow measurement was summarized in and sweep events along the wall, it was ascertained with
Springer’s book edited by Takeda (2012). spectral analysis that only small bubbles modify the
Measurement of liquid velocity distributions in bubbly coherent structures. The DOF-narrowing technique was
turbulent boundary layers might be the most laborious extended by Murai et al. (2006b) for flow in the wall-
work. Gabillet et al. (2002) developed hot-wire anemom- perpendicular direction. With a very different approach
etry for measurement of the liquid-phase velocity. They from the above, Hosokawa et al. (2009) realized the near-
raised the sampling frequency to 20 kHz so that each wall turbulent flow characterization of bubbly flow in a
passage of bubbles of a few millimeters was detected from square duct by means of molecular tagging velocimetry.
the sudden drop in the output voltage owing to the lower They performed photo-bleaching reactions in liquid phase
thermal conductivity inside bubbles. They introduced two on a microscale and successfully measured the velocity
thresholds into data processing to discriminate bubbles gradient tensors in a turbulent bubbly flow. Their following
from the velocity fluctuation of turbulent flow. Another paper (Hosokawa and Tomiyama (2013)) succeeded in
discrimination for hot-wire anemometry was proposed by validating Reynolds-averaged model for bubbly flows with
Rensen et al. (2005a) and applied to a bubbly flow (Rensen the data obtained by the same technique.
et al. 2005b). LDV applicable to bubbly turbulent flow was Summarizing this chapter, the author himself relearned
proposed by So et al. (2002). They modified the optical that the ideas for measurement tools lead to discoveries
geometry of a commercial two-color LDV system so that relating to the drag reduction mechanism. In opposing
laser beams detour about bubbles that accumulated inside direction, our continuous and enthusiastic effort to under-
turbulent boundary layers. stand the fundamental physics propels the development of
PIV requires image separation into two phases, prior to measurement techniques. As mentioned in Sects. 3.1 and
or during the digital image processing. For bubbly turbu- 3.2, advances in understanding provide the next questions
lent channel flows, infrared shadow imaging for bubbles to answer. The main feature of Sect. 3.3 is that the spa-
was combined with PIV by Fujiwara et al. (2004), and the tiotemporal resolving of measurements of the bubbly tur-
same with particle tracking velocimetry (PTV) by Kitaga- bulent boundary layer demonstrates the ample scope for
wa et al. (2005). To prevent irregular light scattering about improvement of drag reduction performance. At the same

123
1773 Page 24 of 28 Exp Fluids (2014) 55:1773

Fig. 18 Liquid velocity fluctuation induced by bubbles of various 0.5 mm. The two-dimensional divergence of velocity vector field at
sizes in the wall proximity of a horizontal turbulent channel flow. The 1.50 ± 0.15 mm from the wall is visualized in color for three
bulk mean liquid velocity is U = 2 m/s (Re = 8,800), flowing from measurement sections: a 250 mm, b 1 m, c 4 m from the bubble
left to right in each figure (100 9 100 mm). The bubble size has a injection point (Huang et al. 2008)
Gaussian distribution centered at 1.0 mm with a standard deviation of

time, the lack of reproducibility, which has long been a since we already learned that gas–liquid two-phase flow just
concern to experimentalists, is close to the solution. Con- in a circular pipe comprises several very different flow
sequently, the author can say that it is finally time to test patterns. In contrast, some experimentalists discovered
our new ideas in application devices based on historically highly organized interaction between two phases as the
integrated knowledge stated in this article. bubbles are controlled to be nearly spherical and monodi-
spersed. The use of such bubbles quite often induces natu-
rally forming void waves owing to the interphase interaction
4 Conclusions regularized in the time–space domain. The authors expect
that these phenomena will be potentially contributable to the
Drag reduction by bubble injection was reviewed. In con- next advance in drag reduction through bubble injection.
trast to past review papers such as that by Merkle and
Deutsch (1990), the present review highlights the great Acknowledgments The author would like to thank Prof. Koichi
Hishida and Dr. Yoshihiko Oishi for their assistance in preparing this
progress made in regard to ‘‘understanding’’ the internal manuscript, and also Prof. Hiroharu Kato and Dr. Yuji Tasaka for
mechanism of the drag reduction. The progress relies on their long support relating to this topic. The work of the author’s
recent advances in measurement techniques that have been group in the manuscript was supported by the Ministry of Education,
established to resolve complex bubbly two-phase flow Science, Sports and Culture, Japan, a Grant-in-Aid for Scientific
Research (KAKENHI Grant Numbers 24246033 and 21360077), and
structures subject to strong turbulence. In particular, the also financially supported by the New Energy Development Organi-
importance placed on fundamental research using channel zation, Japan (NEDO Project Number 08B36002d). The author
flows and circular Couette flows has led to a hot discussion expresses thanks for this support.
on the interpretation of the roles of a bubble’s slip velocity
and the interfacial deformability in turbulence. In a spatially
developing boundary layer, the slip plays the principal role References
in modifying the turbulent shear stress field. The deforma-
bility maintains drag reduction in fully developed turbu- Adrian RJ (2007) Hairpin vortex organization in wall turbulence.
lence and contributes to the insensitivity to the bubble size Phys Fluids 19:041301
Aliseda A, Lasheras JC (2006) Effect of buoyancy on the dynamics of
once bubbles start to deform. The rheology of bubbly liquid a turbulent boundary layer laden with microbubbles. J Fluid
in a wide dynamic range of the shear rate showed strong Mech 559:307–334
non-Newtonian properties affecting all regimes from lami- Amromin E, Mizine I (2003) Partial cavitation as drag reduction
nar to highly turbulent states. Namely, the mechanism of technique and problem of active flow control. Mar Technol
40:181–188
drag reduction is little explained by a single universal Amromin E, Karafiath G, Metcalf B (2011) Ship drag reduction by air
hypothesis, but has been classified into multiple regions in bottom ventilated cavitation in calm water and in waves. J Ship
the parameter space of our interest. This is not unfortunate Res 55:196–207

123
Exp Fluids (2014) 55:1773 Page 25 of 28 1773

Andereck CD, Liu SS, Swinny HL (1986) Flow regimes in a circular conf. ship drag reduction (SMOOTH-SHIPS), Instanbul, Turkey.
Couette system with independently rotating cylinders. J Fluid Paper No. 12, pp 9
Mech 164:155–183 Frankel NA, Acrivos A (1970) The constitutive equation for a dilute
Atkhen K, Fontaine J, Wesfreid JE (2002) Highly turbulent Couette– emulsion. J Fluid Mech 44:65–78
Taylor bubbly flow patterns. J Fluid Mech 422:55–68 Fujikawa S, Yano Y, Watanabe M (2011) Vapor–liquid interfaces,
Batchelor GK (1967) Effective viscosity of dilute dispersion: an bubbles and droplets: fundamentals and applications. Series of
introduction to fluid dynamics. Cambridge University Press, heat and mass transfer. Springer, Berlin
Cambridge, pp 246–255 Fujiwara A, Minato D, Hishida K (2004) Effect of bubble diameter on
Biesheuvel A, van Wijngaaden L (1984) Two-phase flow equations modification of turbulence in an upward pipe flow. Int J Heat
for a dilute dispersion of gas bubbles in liquid. J Fluid Mech Fluid Flow 25:481–488
148:301–318 Fukagata K, Iwamoto K, Kasagi N (2002) Contribution of Reynolds
Boffetta G, Celani A, Vergassola M (2000) Inverse energy cascade in stress distribution to the skin friction in wall-bounded flows.
two-dimensional turbulence: deviations from Gaussian behavior. Phys Fluids 14:L73–L76
Phys Rev E 61:29–32 Fukuda K, Tokunaga J, Nobunaga T, Nakatani T, Iwasaki T (2000)
Brücker C (1999) Structure and dynamics of the wake of bubbles and Frictional drag reduction with air lubricant over a super-water-
its relevance for bubble interaction. Phys Fluids 11:1781–1796 repellent surface. J Mar Sci Technol 5:123–130
Bunner B, Tryggvason G (2003) Effect of bubble deformation on the Gabillet C, Colin C, Fabre J (2002) Experimental study of bubble
properties of bubbly flow. J Fluid Mech 495:77–118 injection in a turbulent boundary layer. Int J Multiph Flow
Callenaere M, Franc JP, Michel JM, Riondet M (2001) The cavitation 28:553–578
instability induced by the development of a re-entrant jet. J Fluid Gao T, Hu HH, Castaneda PP (2011) Rheology of a suspension of
Mech 444:223–256 elastic particles in a viscous shear flow. J Fluid Mech
Ceccio S (2010) Frictional drag reduction of external flows with 687:209–237
bubble and gas injection. Annu Rev Fluid Mech 42:183–203 Gore RA, Crowe CT (1989) Effect of particle size on modulating
Chouippe A, Climent E, Legendre D, Gabillet C (2014) Numerical turbulent intensity. Int J Multiph Flow 15:279–285
simulation of bubble dispersion in turbulent Taylor–Couette Gore RA, Crowe CT (1991) Modulation of turbulence by a dispersed
flow. Phys Fluids 26:043304 phase. J Fluids Eng 113:304–307
Climent E, Simonnet M, Magnaudet J (2007) Preferential accumu- Guin MM, Kato H, Yamaguchi H, Maeda M, Miyanaga M (1996)
lation of bubbles in Couette–Taylor flow patterns. Phys Fluids Reduction of skin friction by microbubbles and its relation with
19:083301 near wall bubble concentration in a channel. J Mar Sci Technol
Crowe CT, Troutt TR, Chung JN (1996) Numerical models for two- 1:241–254
phase turbulent flows. Annu Rev Fluid Mech 28:11–43 Hara K, Suzuki T, Yamamoto F (2011) Image analysis applied to
Cui Z, Fan JM, Park AH (2003) Drag coefficients for a settling sphere study on frictional drag reduction by electrolytic microbubbles in
with microbubble drag reduction effects. Power Technol a turbulent channel flow. Exp Fluids 50:715–727
138:132–134 Hardalupas A, Sahu S, Taylor AMKP, Zarogoulidis K (2010)
Djeridi H, Gabillet C, Billard Y (2004) Two-phase Couette–Taylor Simultaneous planer measurement of droplet velocity and size
flow: arrangement of the dispersed phase and effect on the flow with gas phase velocities in a spray by combined ILIDS and PIV
structure. Phys Fluids 16:128–139 techniques. Exp Fluids 49:417–434
Doi M, Ohta T (1991) Dynamics and rheology of complex interfaces Hassan YA, Ortiz-Villafuerte J (2003) Investigation of microbub-
I. J Chem Phys 95(2):1242–1247 ble boundary layer using particle image velocimetry. In:
Dominguez-Lerma MA, Ahlers G, Channell DS (1985) Effects of Proceedings of ASME FEDSM’03 -45639 [CD-ROM], Fourth
Kalliroscope flow visualization particles on rotating Couette– ASME-JSME Joint Fluids Engineering Conference, Honolulu,
Taylor flow. Phys Fluids 28:1204–1206 HI
Einstein A (1906) Eine neue Bestimmung der Molekuldimensionen. Hassan YA, Gutierrez Torres CC, Jimenez-Bernal JA (2005) Tem-
Ann Phys 19:289–306 poral correlation modification by microbubbles injection in a
Elbing BR, Winkel ES, Lay KA, Ceccio SL, Dowling DR, Perlin M channel flow. Int Commun Heat Mass Transf 32:1009–1015
(2008) Bubble-induced skin-friction drag reduction and the Hesketh RP, Etchells AW, Russell TWF (1991) Bubble breakage in
abrupt transition to air-layer drag reduction. J Fluid Mech pipeline flow. Chem Eng Sci 46:1–9
612:201–236 Higuchi M, Saito T (2010) Quantitative characterizations of long-
Elbing BR, Mäkiharju S, Wiggins A, Perlin M, Dowling DR, Ceccio period fluctuations in a large-diameter bubble column based
SL (2013) On the scaling of air layer drag reduction. J Fluid on point-wise void fraction measurements. Chem Eng J
Mech 717:484–513 160:284–292
Felton K, Loth E (2001) Spherical bubble motion in a turbulent Hinze JO (1955) Fundamentals of the hydrodynamic mechanism of
boundary layer. Phys Fluids 13:2564–2577 splitting in dispersion processes. AIChE J 1:289–295
Felton K, Loth E (2002) Diffusion of spherical bubbles in a turbulent Hirata M, Nishiwaki N (1963) Skin friction and heat transfer for
boundary layer. Int J Multiph Flow 28:69–92 liquid flow over a porous wall with gas injection. Int J Heat Mass
Ferrante A, Elghobashi S (2004) On the physical mechanism of drag Transf 6:941–949
reduction in a spatially developing turbulent boundary layer Hosokawa S, Tomiyama A (2004) Turbulence modification in gas–
laden with microbubbles. J Fluid Mech 503:345–355 liquid and solid–liquid dispersed two-phase pipe flows. Int J
Ferrante A, Elghobashi S (2005) Reynolds number effect of drag Heat Fluid Flow 25:489–498
reduction in a microbubble-laden spatially developing turbulent Hosokawa S, Tomiyama A (2009) Multi-fluid simulation of turbulent
boundary layer. J Fluid Mech 543:93–106 bubbly pipe flows. Chem Eng Sci 64:5308–5318
Fischer F, Hampel U (2010) Ultra fast electron beam X-ray computed Hosokawa S, Tomiyama A (2013) Bubble-induced pseudo turbulence
tomography for two-phase flow measurement. Nuclear Eng Des in laminar pipe flows. Int J Heat Fluid Flow 40:97–105
240(9):2254–2259 Hosokawa S, Fukunaga T, Tomiyama (2009) Application of photo-
Foeth EJ, Eggers R, Quadvlieg EHHA (2010) The efficiency of air- bleaching molecular tagging velocimetry to turbulent bubbly
bubble lubrication for decreasing friction resistance. Prof. int. flow in a square duct. Exp Fluids 47:745–754

123
1773 Page 26 of 28 Exp Fluids (2014) 55:1773

Huang J, Murai Y, Yamamoto F (2008) Shallow DOF-based particle Kulick JD, Fessler JR, Eaton JK (1994) Particle response and
tracking velocimetry applied to horizontal bubbly wall turbu- turbulence modification in fully developed channel flow. J Fluid
lence. Flow Meas Instrum 19:93–105 Mech 277:109–134
Huang J, Murai Y, Yamamoto F (2009) Quadrant analysis of bubble Kumagai, I, Nakamura N, Murai Y, Tasaka Y, Takeda Y, Takahashi
induced velocity fluctuation in a transitional boundary layer. Y (2010) A new power-saving device for air bubble generation:
J Hydrodyn 21:93–99 hydrofoil air pump for ship drag reduction. In: Proceedings of
Hubacz R, Wronski S (2004) Horizontal Couette–Taylor flow in a international conference on ship drag reduction, Istanbul
two-phase gas–liquid system: flow patters. Exp Therm Fluid Sci (Smooth), pp 95–102
28:457–472 Kwon BH, Kim HH, Jeon HJ, Kim MC, Lee I, Chun S, Go JS (2014)
Ishii M, Hibiki T (2011) Drift-flux model: thermo-fluid dynamics of Experimental study on the reduction of skin frictional drag in
two-phase flow. Springer, Berlin pipe flow by using convex air bubbles. Exp Fluids 55:1772
Iwasaki T, Nishimura K, Tanaka M, Hagiwara Y (2001) Direct L’vov VS, Pomyalov A, Procaccia I, Tiberkevich V (2005) Drag
numerical simulation of turbulent Couette flow with immiscible reduction by microbubbles in turbulent flows: the limit of minute
droplets. Int J Heat Fluid Flow 22:332–342 bubbles. Phys Rev Let 94:174502
Jacob B, Olivieri A, Miozzi M, Campana EF, Piva R (2010) Drag La Porta A, Voth GA, Crawford AM, Alexander J, Bodenschatz E
reduction by microbubbles in a turbulent boundary layer. Phys (2001) Fluid particle accelerations in fully developed turbulence.
Fluids 22:115104 Nature 409:1017–1019
Jimenez J (2012) Cascades in wall-bounded turbulence. Annu Rev Lance M, Bataille J (1991) Turbulence in the liquid phase of a
Fluid Mech 44:27–45 uniform bubbly air–water flow. J Fliud Mech 222:95–118
Kameda M, Matsumoto Y (1996) Shock waves in a liquid containing Latorre R (1997) Ship hull drag reduction using bottom air injection.
small gas bubbles. Phys Fluids 8:322–335 Ocean Eng 24:161–175
Kato H, Iwashina T, Miyanaga M, Yamaguchi H (1999) Effect of Latorre R, Miller A, Philips R (2003) Micro-bubble resistance
microbubbles on the structure of turbulence in a turbulent reduction on a model SES catamaran. Ocean Eng 30:2297–2309
boundary layer. J Mar Sci Technol 4:115–162 Lay KA, Yakushiji R, Makiharju S, Perlin M, Ceccio SL (2010)
Katsui T, Okamoto Y, Kasahara Y, Shimoyama N, Iwasaki Y, Partial cavity drag reduction at high Reynolds number. J Ship
Soejima S (2003) A study of air lubrication method to reduce Res 54:109–119
frictional resistance of ship: experimental investigation by tanker Lee CY, Kim CJ (2011) Underwater restoration and retention of gases
form model ship and estimation of full scale ship performance. on superhyrdophibic surfaces for drag reduction. Phys Rev Lett
J Kansai Soc Nav Archit Jpn 239:45–53 (in Japanese) 106:014502
Kawaguchi T, Akasaka Y, Maeda M (2002) Size measurement of Legner HH (1984) Simple model for gas bubble drag reduction. Phys
droplets and bubbles by advanced interferometric laser imaging Fluids 27:2788–2790
technique. Meas Sci Technol 13:308 Lelouvetel J, Tanaka T, Sato Y, Hishida K (2014) Transport
Kawamura T, Kodama Y (2002) Numerical simulation method to mechanisms of the turbulent energy cascade in upward/down-
resolve interactions between bubbles and turbulence. Int J Heat ward bubbly flows. J Fluid Mech 741:514–542
Fluid Flow 23:627–638 Li FC, Kawaguchi Y, Yu B, Wei JJ, Hishida K (2008) Experimental
Kim J (2003) Control of turbulent boundary layers. Phys Fluids study of drag-reduction mechanism for a dilute surfactant
15:1093–1106 solution flow. Int J Heat Mass Transf 51:835–843
Kim SY, Cleaver JW (1995) The persistence of drag reduction Liu TJ (1997) Investigation of the wall shear stress in vertical bubbly
following the injection of microbubbles into a turbulent bound- flow under different bubble size conditions. Int J Multiph Flow
ary layer. Int Commun Heat Mass Transf 22:353–357 23:1085–1109
Kitagawa A, Murai Y (2013) Natural convection heat transfer from a Llewellin EW, Manga M (2005) Bubble suspension rheology and
vertical heated plate in water with microbubble injection. Chem implications for conduit flow. J Volcanol Geotherm Res
Eng Sci 99:215–224 143:205–217
Kitagawa A, Murai Y (2014) Pulsatory rise of microbubble swarm Lockhart RW, Martinelli RC (1949) Proposed correlation of data for
along a vertical wall. Chem Eng Sci 116:694–703 isothermal two-phase two-component flow in pipes. Chem Eng
Kitagawa A, Murai Y, Yamamoto F (2001) Two-way coupling of Process 45:39–48
Eulerian–Lagrangian model for dispersed multiphase flows using Lu J, Tryggvason G (2007) Effect of bubble size in turbulent
filtering functions. Int J Multiph Flow 27:2129–2153 bubbly downflow in a vertical channel. Chem Eng Sci
Kitagawa A, Sugiyama K, Murai Y (2004) Experimental detection of 62:3008–3018
bubble–bubble interactions in a wall-sliding bubble swarm. Int J Lu J, Fernandez A, Tryggvason G (2005a) The effect of bubbles on
Multiph Flow 30:1213–1234 the wall drag in a turbulent channel flow. Phys Fluids
Kitagawa A, Hishida K, Kodama Y (2005) Flow structure of 17(095102):1–12
microbubble-laden turbulent channel flow measured by PIV Lu X, Hamada M, Kato H (2005b) Effect of the turbulent frictional
combined with the shadow image technique. Exp Fluids drag reduction of microbubbles: experiments by bubbles of air
38:466–475 and hydrogen. In: Proc. fluid eng. conf. of Japan soc. mech. eng.
Kitagawa A, Kosuge K, Uchida K, Hagiwara Y (2008) Heat transfer (JSME), Paper No. 509:69–70
enhancement for laminar natural convection along a vertical Lundin MD, McCready MJ (2009) Modeling of bubble coalescence in
plate due to sub-millimeter-bubble injection. Exp Fluids bubbly co-current flows restricted by confined geometry. Chem
45:473–484 Eng Sci 64:4060–6067
Kodama Y, Kakugawa A, Takahashi T, Kawashima H (2000) Luo R, Song Q, Yang XY, Wang Z (2002) A three-dimensional
Experimental study on microbubbles and their applicability to photographic method for measurement of phase distribution in
ships for skin friction reduction. Int J Heat Fluid Flow dilute bubble flow. Exp Fluids 32:116–120
21:582–588 Luther S, Rensen J, Guet S (2004) Bubble aspect ratio and velocity
Kramer MO (1960) Boundary layer stabilization by distributed measurement using a four-point fiber-optical probe. Exp Fluids
damping. J Am Soc Nav Eng 72:25–34 36:326–333

123
Exp Fluids (2014) 55:1773 Page 27 of 28 1773

Madavan NK, Deutsch S, Merkle CL (1985) Measurements of local Murai Y, Fujii H, Tasaka Y, Takeda Y (2006a) Turbulent bubbly
skin friction in microbubble-modified turbulent boundary layer. channel flow investigated by ultrasound velocity profiler. J Fluid
J Fluid Mech 156:237–256 Sci Technol 1:12–23
Magnaudet J, Eames I (2000) The motion of high-Reynolds-number Murai Y, Oishi Y, Takeda Y, Yamamoto F (2006b) Turbulent shear
bubbles in inhomogeneous flows. Annu Rev Fluid Mech stress profiles in a bubbly channel flow assessed by particle
32:659–708 tracking velocimetry. Exp Fluids 41:343–352
Mäkiharju SA, Perlin M, Ceccio SL (2012) On the energy economics Murai Y, Qu JW, Yamamoto F (2006c) Three dimensional interaction
of air lubrication drag reduction. Int J Nav Archit Ocean Eng of bubbles at intermediate Reynolds numbers. Multiph Sci
4(4):412–422 Technol 18:175–197
Mäkiharju SA, Elbing BR, Wiggins A, Schinasi S, Vanden-Broeck Murai Y, Fukuda H, Oishi Y, Kodama Y, Yamamoto F (2007) Skin
JM, Perlin M, Ceccio SL (2013a) On the scaling of air friction reduction by large air bubbles in a horizontal channel
entrainment from a ventilated partial cavity. J Fluid Mech flow. Int J Multiph Flow 33:147–163
732:47–76 Murai Y, Oiwa H, Takeda Y (2008) Frictional drag reduction in
Mäkiharju SA, Gabillet C, Paik BG, Chang NA, Perlin M, Ceccio SL bubbly Couette-Taylor flow. Phys Fluids 20:034101
(2013b) Time-resolved two-dimensional X-ray densitometry of a Murai Y, Ohta S, Shigetomi A, Tasaka Y, Takeda Y (2009)
two-phase flow downstream of a ventilated cavity. Exp Fluids Development of an ultrasonic void fraction profiler. Meas Sci
54(7):1561 Technol 20:114003
Marie JL (1987) A simple analytical formulation for microbubble Murai Y, Tasaka Y, Nambu Y, Takeda Y, Gonzalez SR (2010)
drag reduction. J PhysicoChem Hydrodyn 13:213–220 Ultrasonic detection of moving interfaces in gas–liquid two-
Maryami R, Farahat S, Poor MJ, Mayam MHS (2014) Bubbly drag phase flow. Flow Meas Instrum 21:356–366
reduction in a vertical Couette–Taylor system with superimposed Narayanan C, Lakehal D (2003) Mechanism of particle deposition in
axial flow. Fluid Dyn Res 46:055504. doi:10.1088/0169-5983/ a fully developed turbulent open channel flow. Phys Fluids
46/5/05504 15:763–775
Masliyah J, Jauhari R, Gray M (1994) Drag coefficient for air bubbles Oishi Y, Murai Y (2014) Horizontal turbulent channel flow interacted
rising along an inclined surface. Chem Eng Sci 49:1905–1911 by a single large bubble. Exp Therm Fluid Sci 55:128–139
Matveev KI (2007) Three dimensional wave patterns in long air Oishi Y, Murai Y, Tasaka Y, Takeda Y (2009) Frictional drag
cavities on a horizontal plane. Ocean Eng 34:1882–1891 reduction by wavy advection of deformable bubbles. J Phys Conf
Maxey MR, Chang EJ, Wang LP (1996) Interaction of particles and Ser 147:012020
microbubbles with turbulence. Exp Therm Fluid Sci 12:417–425 Ojima S, Hayashi K, Hosokawa S, Tomiyama A (2014) Distribution
McCormick M, Bhattacharyya R (1973) Drag reduction of a of void fraction and liquid velocity in air-water bubble column.
submersible hull by electrolysis. Nav Eng J 85:11–16 Int J Multiph Flow 1–11. doi:10.1016/[Link].2014.
Mehel A, Gabillet C, Djeridi H (2007) Analysis of the flow pattern 05.008
modifications in a bubbly Couette–Taylor flow. Phys Fluids Ortiz-Villafuerte J, Hassan YA (2006) Investigation of microbubble
19:118101 boundary layer using particle tracking velocimetry. J Fluids Eng
Merkle CL, Deutsch S (1990) Drag reduction in liquid boundary 128:507–519
layers by gas injection. Prog Astronaut Aeronaut 123:351–411 Ouellette NT (2012) Turbulence in two dimensions. Phys Today
Merkle CL, Deutsch S (1992) Microbubble drag reduction in 68–69
liquid turbulent boundary layers. ASME Appl Mech Rev Pang MJ, Wei JJ, Yu B (2013) Numerical study on modulation of
45:103–127 microbubbles on turbulence frictional drag in a horizontal
Michaelides EE (1997) The transient equation of motion for particles, channel. Ocean Eng 81:58–68
bubbles, and droplets. J Fluids Eng 119:233–247 Park HJ, Oishi Y, Tasaka Y, Murai Y, Takeda Y (2009) Turbulent
Michel JM (1984) Some features of water flows with ventilated shear control with oscillatory bubble injection. J Phys Conf Ser
cavities. J Fluid Eng 106(3):319–326 147:012037
Mizokami S, Kawakita C, Kodan Y, Takano S, Higasa S, Shigenaga Park HJ, Tasaka Y, Murai Y, Oishi Y (2014) Vortical structures swept
R (2010) Experimental study of air lubrication method and by a bubble swarm in turbulent boundary layers. Chem Eng Sci
verification of effects on actual hull by means of sea trial. 116:486–496
Mitsubishi Heavy Ind Techn Rev 47(3):41–47 Piomelli U, Yuan J (2013) Numerical simulation of spatially
Moctezuma MF, Lima-Ochoterena R, Zenit R (2005) Velocity developing, accelerating boundary layer. Phys Fluids 25:101304
fluctuations resulting from the interaction of a bubble with a Poreh M, Cermak JE (1964) Study of diffusion from a line source in a
vertical wall. Phys Fluids 17:098106 turbulent boundary layer. Int J Heat Mass Transf 7:1083–1095
Moriguchi Y, Kato H (2002) Influence of microbubble diameter and Prasser HM, Scholz D, Zippe C (2001) Bubble size measurement
distribution on frictional resistance reduction. J Mar Sci Technol using wire-mesh sensors. Flow Meas Instrum 12:299–312
7:79–85 Prosperetti A (2012) Linear oscillations of constrained drops, bubbles,
Murai Y, Oiwa H (2008) Increase of effective viscosity in bubbly and plane liquid surfaces. Phys Fluids 24:032109
liquids from transient bubble deformation. Fluid Dyn Res Rensen J, Luther S, de Vries J, Lohse D (2005a) Hot-film
40:565–575 anemometry in bubbly flow I: bubble–probe interaction. Int J
Murai Y, Matsumoto Y, Yamamoto F (2001) Three-dimensional Multiph Flow 31:285–301
measurement of void fraction in a bubble plume using statistic Rensen J, Luther S, Lohse D (2005b) The effect of bubbles on
stereoscopic image processing. Exp Fluids 30:11–21 developed turbulence. J Fluid Mech 538:153–187
Murai Y, Oishi Y, Sasaki T, Kodama Y, Yamamoto F (2005a) Richter S, Aritomi M, Prasser HM, Humpel R (2002) Approach
Turbulent shear stress profiles in a horizontal bubbly channel towards spatial phase reconstruction in transient bubbly flow
flow. In: Proceedings of 6th international symposium on smart using a wire-mesh sensor. Int J Heat Mass Transf 45:1063–1075
control of turbulence 2005, Tokyo, 289–295 Robinson SK (1991) Coherent motions in the turbulent boundary
Murai Y, Sasaki T, Ishikawa M, Yamamoto F (2005b) Bubble-driven layer. Annu Rev Fluid Mech 23:601–639
convection around cylinders confined in a channel. J Fluids Eng Ronen D (1982) The effect of oil price on the optimal speed of ships.
127:117–123 J Oper Res 33:1035–1040

123
1773 Page 28 of 28 Exp Fluids (2014) 55:1773

Rust AC, Manga M (2002a) Effects of bubble deformation in the Taniere A, Oesterle B, Monnier JC (1997) On the behavior of solid
viscosity of dilute suspensions. J Non-Newton Fluid Mech particles in a horizontal boundary layer with turbulence and
104:53–63 saltation effects. Exp Fluids 23:463–471
Rust AC, Manga M (2002b) Bubble shapes and orientations in low Re Taylor GI (1923) Stability of a viscous liquid contained between two
simple shear flow. J Colloid Interface Sci 249:476–480 rotating cylinders. Philos Trans R Soc Lond Ser A 223:289–343
Ryskin G, Leal LG (1984) Numerical solution of free-boundary Timkin LS, Gorelik RS (2010) Specificity of laminar-turbulent
problems in fluid mechanics: part 1 the finite-difference transition un upward monodispersed microbubbly flow. Tech
technique. J Fluid Mech 148:1–17 Phys Lett 36:493–495
Sakurai K, Tasaka Y, Murai Y (2013) Modification of effective Toschi F, Bodenschartz E (2009) Lagrangian properties of particles in
viscosity on bubbly flows due to transient bubble deformation. turbulence. Annu Rev Fluid Mech 41:375–404
Trans. Japan Soc. Mech. Eng., Ser. B, 79: 1–11 (in Japanese). Tran-Cong S, Marie JL, Perkins RJ (2008) Bubble migration in a
English version of similar contents: Murai Y, Tasaka Y, Sakurai turbulent boundary layer. Int J Multiph Flow 34:786–807
K, Oyama K, Takeda Y (2010) Ultrasound Doppler rheometry Tsai JF, Chen CC (2011) Boundary layer mixture model for a
from spin response of viscoelastic and bubbly Liquids. In: microbubble drag reduction technique. Int Sch Res Netw
Proceedings 7th international symposium on ultrasonic Doppler 2011:405701
methods, Gothenburg, Sweden, 9–12 van den Berg TH, Luther S, Lathrop DP, Lohse D (2005) Drag
Sanders WC, Winkel ES, Dowling DR, Perlin M, Ceccio SL (2006) reduction in bubbly Taylor–Couette turbulence. Phys Rev Lett
Bubble friction drag reduction in a high-Reynolds-number flat- 94:044501
plate turbulent boundary layer. J Fluid Mech 552:353–380 van den Berg TH, Luther S, Lathrop D, Lohse D (2007) Bubbly
Sangani AS, Didwania AK (1993) Dynamic simulations of flows of turbulent drag reduction is a boundary effect. Phys Rev Lett
bubbly liquids at large Reynolds numbers. J Fluid Mech 98:084501
250:307–337 van Gils DPM, Guzman DN, Sun C, Lohse D (2013) The importance
Sangani AS, Kang SY, Tsao HK, Koch DL (1997) Rheology of dense of bubble deformability for strong drag reduction in bubbly
bubble suspensions. Phys Fluids 9(6):1540–1561 turbulent Taylor–Couette flow. J Fluid Mech 722:317–347
Schowalter WR, Chaffey CE, Brenner H (1968) Rheological behavior Warsito W, Fan LS (2001) Measurement of real-time flow structures
of a dilute emulsion. J Colloid Interface Sci 26:152–160 in gas–liquid and gas–liquid–solid flow systems using electrical
Seo JH, Lele SK, Tryggvason G (2010) Investigation and modeling of capacitance tomography (ECT). Chem Eng Sci 56:6455–6462
bubble–bubble interaction effect in homogeneous bubbly flows. Watamura T, Tasaka Y, Murai Y (2013) Intensified and attenuated waves
Phys Fluids 22:063302 in a microbubble Taylor-Couette flow. Phys Fluids 25:054107
Serizawa A, Kataoka I (1990) Turbulence suppression in bubbly two- Watanabe O, Masuko A, Shirose Y (1998) Measurements of drag
phase flow. Nuclear Eng Des 122:1–16 reduction by microbubbles using very long ship models. J Soc
Serizawa A, Inui T, Eguchi T (2005) Flow characteristics and pseudo- Nav Archit Jpn 183:53–63
laminarization of vertically upward air–water milky bubbly flow Winkel ES, Ceccio SL, Dowling DR, Perlin M (2004) Bubble-size
with micro bubbles in a pipe. Jpn J Multiph Flow 19:335–340 (in distributions produced by wall injection of air into flowing fresh
Japanese) water, saltwater and surfactant solutions. Exp Fluids 37:802–810
Shen X, Ceccio S, Perlin M (2006) Influence of bubble size on micro- Wronski S, Hubacz R, Ryszczuk T (2005) Interfacial area in a reactor
bubble drag reduction. Exp Fluids 41:415–424 with helicoidal flow for the two-phase gas–liquid system. Chem
Shiomi Y, Kutsuna H, Akagawa K, Ozawa M (1993) Two-phase flow Eng J 105:71–79
in an annulus with a rotating inner cylinder (flow pattern in Wu SJ, Hsu CH, Lin TT (2007) Model test of the surface and
bubbly flow region). Nuclear Eng Des 141:27–34 submerged vehicles with the micro-bubble drag reduction.
So S, Morikita H, Takagi S, Matsumoto Y (2002) Laser Doppler Ocean Eng 34:83–93
velocimetry measurement of turbulent bubbly channel flow. Exp Wu SJ, Ouyang K, Shiah SW (2008) Robust design of microbubble
Fluids 33:135–142 drag reduction in a channel flow using the Taguchi method.
Stickel J, Powell RL (2005) Fluid mechanics and rheology of dense Ocean Eng 35:856–863
suspensions. Annu Rev Fluid Mech 37:129–149 Xu J, Maxey ML, Karniadakis GE (2002) Numerical simulation of
Stutz B, Legoupil S (2003) X-ray measurements within unsteady turbulent drag reduction using micro-bubbles. J Fluid Mech
cavitation. Exp Fluids 35(2):130–138 468:271–281
Sugiyama K, Calzavarini E, Lohse D (2008) Microbubbly drag Yoshida K, Tasaka Y, Murai Y, Takeda Y (2009) Mode transition in
reduction in Taylor–Couette flow in wavy vortex regime. J Fluid bubbly Taylor–Couette flow measured by PTV. J Phys Conf Ser
Mech 608:21–41 147:012013
Takagi S, Matsumoto Y (2011) Surfactant effects on bubble motion Zenit R, Koch D, Sangani AS (2001) Measurements of the average
and bubbly flow. Annu Rev Fluid Mech 43:615–636 properties of a suspension of bubbles rising in a vertical channel.
Takagi S, Ogasawara T, Fukuta M, Matsumoto Y (2009) Surfactant J Fluid Mech 429:307–342
effect on the bubble motions and bubbly flow structures in a Zhang DZ, Prosperetti A (1994) Averaged equations for inviscid
vertical channel. Fluid Dyn Res 41:065003 disperse two-phase flow. J Fluid Mech 267:185–219
Takahashi T, Kakugawa A, Makino M, Kodama Y (2003) Experi- Zhao LH, Andersson HI, Gillissen JJJ (2010) Turbulence modulation
mental study on scale effect of drag reduction by microbubbles and drag reduction by spherical particles. Phys Fluids 22:081702
using very large flat plate ships. J Kansai Soc Nav Archit Jpn Zhao LH, Marchioli C, Andersson HI (2012) Stokes number effects
239:11–20 (in Japanese) on particle slip velocity in wall-bounded turbulence and
Takeda Y (2012) Ultrasonic Doppler velocity profiler for fluid flow. implications for dispersion models. Phys Fluids 24:021705
Fluid mechanics and its applications, Ser. 101, Springer, Berlin Zhen L, Hassan YA (2006) Wavelet autocorrelation identification of
Takeda Y, Fischer WE, Sakakibara J (1994) Decomposition of the the turbulent flow multi-scales for drag reduction process in
modulated waves in a rotating Couette system. Science microbubbly flows. Chem Eng Sci 61:7107–7114
263:502–505 Zhen N, Handler RA, Zhang Q, Oeth C (2013) Evolution of a hairpin
Tanaka M (2013) Inverse transverse migration of small bubbles in vortex in a shear-thinning fluid governed by a power-law model.
turbulence. J Phys Soc Jpn 82:044401 Phys Fluids 25:110703

123