ANRV400-FL42-06 ARI 17 November 2009 10:20

ANNUAL
REVIEWS Further Turbulent Dispersed
Click here for quick links to
Annual Reviews content online,
including:
Multiphase Flow
• Other articles in this volume
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

• Top cited articles

• Top downloaded articles
S. Balachandar1 and John K. Eaton2
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

• Our comprehensive search 1

Department of Mechanical and Aerospace Engineering, University of Florida, Gainesville,
Florida 32611; email: bala1s@ufl.edu
2
Department of Mechanical Engineering, Stanford University, Stanford, California 94305

Annu. Rev. Fluid Mech. 2010. 42:111–33 Key Words

First published online as a Review in Advance on turbulence modulation, preferential concentration, turbulent dispersion,
August 17, 2009
point-particle simulation, particulate and bubbly flows
The Annual Review of Fluid Mechanics is online at
fluid.annualreviews.org Abstract
This article’s doi: Turbulent dispersed multiphase flows are common in many engineering and
10.1146/annurev.fluid.010908.165243
environmental applications. The stochastic nature of both the carrier-phase
Copyright  c 2010 by Annual Reviews. turbulence and the dispersed-phase distribution makes the problem of tur-
All rights reserved
bulent dispersed multiphase flow far more complex than its single-phase
0066-4189/10/0115-0111$20.00 counterpart. In this article we first review the current state-of-the-art exper-
imental and computational techniques for turbulent dispersed multiphase
flows, their strengths and limitations, and opportunities for the future. The
review then focuses on three important aspects of turbulent dispersed mul-
tiphase flows: the preferential concentration of particles, droplets, and bub-
bles; the effect of turbulence on the coupling between the dispersed and
carrier phases; and modulation of carrier-phase turbulence due to the pres-
ence of particles and bubbles.

111
 ANRV400-FL42-06 ARI 17 November 2009 10:20

1. INTRODUCTION
Dispersed multiphase flows are common in many engineering and environmental applications,
and they are often turbulent. Examples of dispersed multiphase flows include particles suspended
in gas or liquid flow, the dispersion of droplets in a stream of gas, and bubbly flow. These flows
are thus characterized by a dispersed phase that is distributed within a carrier phase in the form
of particles, droplets, or bubbles. Dispersed multiphase flows are distinguished from other kinds
of multiphase flows, such as free-surface flows. In dispersed multiphase flows, the evolution of the
interface between the phases is considered of secondary importance. Processes such as droplet or
bubble break-up and agglomeration do indeed alter the interface between the phases. However,
in the context of dispersed multiphase flows, one accounts for the interface between the dispersed
and carrier phases in terms of particle-size spectra without considering the detailed evolution of
the interface. (Here and henceforth we use the term particles generically, and this term covers
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

drops and bubbles as well.)

Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

The fractional volume occupied by the dispersed phase (v ) and the mass loading (m ), defined
as the ratio of mass of the dispersed to carrier phase, are two critical parameters that determine the
level of interaction between the phases. When v and m are small, the dominant effect is that of
the turbulent carrier flow on the dynamics of the dispersed phase (i.e., one-way coupled). When
the mass of the dispersed phase is comparable with that of the carrier phase, the back influence
of the dispersed phase on the carrier-phase dynamics cannot be ignored (i.e., two-way coupled).
Finally, when v increases, interactions between particles (such as collision, agglomeration, and
break-up) become important, and this regime is described as four-way coupled (Elghobashi 1991,
1994). In the extreme limit of very large concentration, we encounter the granular flow regime,
where interparticle collision is the dominant mechanism and the effect of interstitial fluid becomes
less important. In this review, we focus on dilute dispersions in which the flow is either one-way
or two-way coupled.
Turbulence and multiphase flows are two of the most challenging topics in fluid mechanics,
and when combined they pose a formidable challenge, even in the dilute dispersed regime. The
inherent stochastic nature of the carrier-phase turbulence is further complicated by the random
distribution of the dispersed phase. The presence of the dispersed phase makes both experi-
mental measurements and numerical simulations of turbulent multiphase flows far more diffi-
cult than those of single-phase flows. Computational investigations using Eulerian-Eulerian and
Lagrangian-Eulerian techniques have provided valuable insight. However, in these approaches,
the multiphase flow is computed only at the macroscale, and the details of the flow at the mi-
croscale (of the order of a particle diameter) are incorporated through models. Similarly, fully
resolved measurements at the scale of the dispersed phase in turbulent flows still remain a chal-
lenge. In Sections 2 and 3, we review experimental and computational techniques for turbulent
multiphase flow.
Dispersed multiphase flows exhibit a variety of interesting phenomena that become important
in the context of a turbulent carrier-phase flow. One of the key features of particle distribution in
turbulent flows is the phenomenon of preferential accumulation (Elghobashi & Truesdell 1992,
Maxey 1987, Squires & Eaton 1991). It is now well accepted that even in isotropic turbulence,
particle distribution is not uniform. Heavier-than-fluid particles tend to accumulate in regions
of high strain rate and avoid regions of intense vorticity. In contrast, lighter-than-fluid particles
(or bubbles) tend to congregate in vortical regions. The physics of preferential accumulation are
reviewed in Section 4.
The effect of carrier-phase turbulence on interphase coupling can be significant. In turbulent
flows, the use of standard force laws and heat-transfer correlations, which are developed for

112 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

laminar flows, can result in significant error. Parameters such as particle-to-fluid length-scale and
timescale ratios can be expected to play a key role. Furthermore, when the particle Reynolds
number increases above a few hundred, the effect of vortex shedding introduces a stochastic
component to interphase coupling. The effect of turbulence on mean and fluctuating forces and
approaches to better modeling such forces are explored in Section 5.
Turbulence modulation by particles is an important aspect of two-phase flow research. In
a dilute suspension, there are several mechanisms that contribute to turbulence modulation:
(a) enhanced dissipation due to the presence of particles, (b) the transfer of kinetic energy to
the fluid from the particles, and (c) the formation of wakes and vortex shedding behind the par-
ticles. The relative importance of these mechanisms depends on parameters such as the particle-
to-turbulence length-scale ratio, particle Reynolds number, and particle-to-fluid density ratio.
In Section 6, we synthesize the large body of accumulated knowledge. Although there has been
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

substantial progress in the past few decades, this is still an open problem, ripe for future research.
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Finally, there have been a few earlier reviews of dispersed multiphase turbulence, but they
have been limited to either specific applications (Shaw 2003) or specific aspects of the particle-
turbulence interaction (Crowe et al. 1996, Gouesbet & Berlemont 1998, Guha 2008, Loth 2000,
Mashayek & Pandya 2003, Shirolkar et al. 1996). The present review is intended to complement
these and provide an up-to-date status of our understanding, and measurement and simulation
capabilities.

2. EXPERIMENTAL APPROACHES
Experimental research in dispersed flows has focused on measuring the mean velocity and Reynolds
stresses of the carrier phase, and the mean and root-mean-square velocity and concentration of
the particles. Higher-order quantities, including Lagrangian particle velocity correlations, the
carrier-phase turbulent dissipation rate, and two-particle and particle-fluid velocity correlations,
are also of interest.
Photographic techniques have been used to measure the particle velocity and concentration
distributions. Measurements of the carrier phase are more difficult because of particle interference.
A few workers have attempted intrusive measurements. Doig & Roper (1967) used a specially
designed pitot probe to measure mean gas velocities, and hot-wire anemometers have been used
to measure carrier-phase turbulence (Hetsroni & Sokolov 1971).
The widespread adoption of laser-based flow measurement techniques revolutionized the ex-
perimental study of turbulent dispersed flows. A standard laser-Doppler anemometer (LDA) mea-
sures the velocity of small tracer particles that follow the fluid motion. Larger, dispersed-phase
particles scatter more light and thus produce a stronger signal, making it straightforward to adapt
an LDA system to measure the particle phase. To measure the carrier phase, one adds fine tracer
particles, and the LDA signal processing must discriminate between Doppler bursts produced by
tracers and dispersed particles. In gas flows, tracers typically have diameters around 1 μm, whereas
the dispersed-phase particles have diameters of O(10–1000) μm.
To measure the particle velocities, one reduces the system sensitivity until only signals from
the particles are detected. This technique is tested by varying the tracer concentration and en-
suring that the LDA data rate does not change. To measure the fluid phase, it is customary to
use a high concentration of tracers and assume that the relatively few particles passing through
the measurement volume have a negligible effect on the measured carrier-phase statistics. This
technique only works for v  1. The effective LDA measurement volume size is much larger
for a large particle than for a tracer. Thus, the ratio between the particle and tracer sample rates
is higher than the ratio of the number densities.

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 113

 ANRV400-FL42-06 ARI 17 November 2009 10:20

LDA signal-processing schemes capable of discriminating between Doppler bursts produced

by particles and tracers are required to study flow modification by particles. Typically, the Doppler
signal amplitude is much larger for a particle than a tracer. To measure the velocity of both phases,
the photodetector output is fed to two separate processors. One processor is set to a low gain so
it only detects the particles. The other is set to a higher gain, and an amplitude threshold is set
so that Doppler bursts with high signal amplitude are rejected (Stock et al. 1975). This technique
is subject to cross talk in which some bursts originating from particles passing through the outer
edge of the LDA measurement volume are interpreted as tracers (Lee & Durst 1982). Modarress
et al. (1984) and Rogers & Eaton (1991) used special techniques to ensure that each validated
Doppler burst originated from a particle or tracer passing near the measurement volume center.
The ratio of the Doppler signal amplitude to the pedestal height is a better discriminator between
tracers and particles because a particle produces a much higher pedestal relative to the Doppler
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

amplitude as compared to a tracer. This technique has been used successfully for the measurement
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

of gas-phase turbulence of moderate v (Kulick et al. 1994, Muste et al. 1996, Sato et al. 1996).
The LDA phase-discrimination techniques discussed above are relatively complicated and often
require careful adjustment of optical or electronic processing systems to eliminate cross talk. The
best way to test the discrimination is to measure the probability density function of fluid velocity
in a region where the mean velocities of the two phases are widely separated. Cross talk appears
as a second peak in the carrier-phase probability density function.
A related measurement system is the phase-Doppler anemometer (Bachalo & Houser 1984),
which simultaneously measures the velocity and diameter of spherical liquid droplets. Kiger &
Lasheras (1995) used a phase-Doppler anemometer to measure gas-phase velocity in a droplet-
laden flow by assuming that the smallest droplets were flow tracers.
Recently, particle image velocimetry (PIV) has largely superseded the use of LDA in dispersed
two-phase flows because it allows the simultaneous measurement of the velocity of both phases
over an imaged area. Measurement of the particle-phase velocity field is simple for dilute flows
because the relatively large particles produce bright images using conventional lasers and cameras.
For systems with high number density and/or large imaging regions, each interrogation window
will contain several particles, and conventional digital cross-correlation PIV can be used. However,
for dilute flows or high-spatial-resolution systems, most interrogation windows contain at most
one particle, so particle tracking velocimetry is used. PIV images of large particles and bubbles
can appear as a dim circle with two bright glare points (Oakley et al. 1997). Kiger & Pan (2000)
developed a convolution method to find the particle centroid under these conditions.
As with LDA, the difficulty in PIV comes when attempting to measure the velocity of the carrier
phase. Good reviews of the various techniques are given by Kiger & Pan (2000) and Khalitov &
Longmire (2002). In liquid-solid or bubbly liquid flows, the tracer particles can be relatively large
and can be tagged with fluorescent molecules, providing unambiguous measurement of the carrier
phase (Hasan et al. 1993, Sridhar & Katz 1995). Particle-laden gas flows require very small tracer
particles, so the phase discrimination is based on the difference in image size or brightness between
the particles and tracers. Sakakibara et al. (1996) adjusted the camera so that scattered light from
the tracers never exceeded 70% of saturation, and then used thresholding to eliminate tracers from
the image. The resulting image was processed to give the particle phase velocity, and also was sub-
tracted from the original image to give a tracer-image field. A problem with this technique is that it
leaves low-intensity halos around the hole where each particle was eliminated. Paris & Eaton (1999)
modified the technique by first applying a low threshold to eliminate noise from the image. The
large particles were then identified using an object-growing algorithm that reduced the halo effect.
The median filtering technique described by Kiger & Pan (2000) effectively treats the tracer
images as noise in the particle image and eliminates them. Subtracting the median-filtered image

114 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

from the original image results in a tracer field image that yields little cross talk. Khalitov &
Longmire (2002) discriminated between tracers and particles using a size/brightness map in which
tracers were identified as small and dim objects and particles as large and bright.
One area of great experimental difficulty is the measurement of the carrier-phase velocity very
near the particle surface. Most phase-discrimination systems eliminate noise, including tracer
images near the particle surfaces. Tanaka & Eaton (2009) solved this problem by using a very-high-
resolution PIV system with an imaged area of only 3.7 mm by 4.7 mm. Phase discrimination based
only on particle image size resulted in essentially zero cross talk. However, the small measurement
area required a large number of image pairs to obtain converged turbulence statistics.
Two statistical quantities that arise in dispersed flow models are the particle-fluid velocity
correlation and the carrier-phase turbulent dissipation rate. The particle-fluid velocity correlation
is defined as the correlation between the particle and fluid velocities measured at the same point in
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

space and time. The correlation is ill-defined for practical systems with finite-size particles in which
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

the fluid and particle phases cannot be coincident. Sakakibara et al. (1996) correlated the particle
velocity with the average velocity of the nearest-neighbor, carrier-phase PIV measurement points.
However, it is easy to see that this process is dependent on the measurement resolution. The best
that one can hope to do is to obtain high-resolution measurements allowing the calculation of the
correlation with the flow velocity filtered over a range of scales.
Measurement of the turbulence-kinetic-energy dissipation rate is essential to understanding
turbulence modulation by particles. It is difficult to measure because it involves spatial derivatives
of all velocity components. However, in many flows, the turbulence is nearly isotropic, so the
dissipation can be calculated given precise measurements in a single plane. Paris & Eaton (1999)
attempted direct calculation of dissipation from PIV data but found that the measured dissipation
rate was sensitive to the PIV spatial resolution. Inadequate spatial resolution causes undermea-
surement of the dissipation rate. However, as one reduces the PIV spatial resolution below the
Kolmogorov scale, noise in the PIV measurements causes overmeasurement of the dissipation
(Saarenrinne & Piirto 2000). Various workers, including Sheng et al. (2000), have tried to account
for inadequate PIV spatial resolution by fitting the spatial spectrum measured over a limited wave-
number range to a model spectrum. These methods assume that the measured turbulence has a
similar spectrum to the model, and thus they are not general. In dispersed multiphase flow, there
is the additional problem that a significant fraction of the total dissipation occurs in the viscous
flow surrounding individual particles. Tanaka & Eaton (2007) have shown that both problems can
be addressed using very-high-spatial-resolution PIV measurements.
In summary, there have been rapid advances in the state of the art of simultaneous velocity
measurements for both phases in dispersed flows. However, measurement of the carrier-phase
velocity field near the particles remains a major problem that must be addressed separately for each
new experiment. High-spatial-resolution PIV techniques appear to be the best available option
for flows with low v , whereas further advances are needed before comparable measurements can
be acquired in all but very dilute bubbly flows.

3. COMPUTATIONAL APPROACHES
A range of complementary computational approaches is available for dispersed multiphase flows.
Figure 1 shows the range of applicability of the different computational approaches in terms
of relative particle size and v (Balachandar 2009; Elghobashi 1991, 1994). In discussing the
spectrum of available approaches, we specifically compare each approach with the prior ap-
proach. Past reviews of numerical methodologies for dispersed multiphase flows can be con-
sulted for additional information (Balachandar & Prosperetti 2004, Crowe 1982, Loth 2000,

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 115

 ANRV400-FL42-06 ARI 17 November 2009 10:20

Fully
resolved
0.1

DNS: (τp/τk) or LES: (τp/τξ)

DNS: (d/η) or LES: (d/ξ)
Lagrangian
point-particle
~1

Eulerian
0.2
Equilibrium
Eulerian
10–3
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

Dusty
gas
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Φv
10–6 10–3

Dilute suspension Dense suspension

one- or two-way coupled four-way coupled
Figure 1
Different approaches to turbulent multiphase flow. Their applicability is separated in terms of timescale and
length-scale ratios. DNS, direct numerical simulation; LES, large eddy simulation.

Prosperetti & Tryggvason 2007). Here we restrict attention to computational approaches for dilute
suspensions.

3.1. Dusty Gas and Equilibrium Eulerian Approaches

The dusty gas approach was advanced by Carrier (1958) and Marble (1970) primarily in the context
of particle-laden compressible flows. This approach assumes that the particles are sufficiently
small that they perfectly follow the local carrier phase. This allows the particle-laden flow to be
considered as a single fluid whose density depends on the local mass fraction of suspended particles.
This one-fluid approach also has been widely used in incompressible flows. The biggest advantage
of this approach is its simplicity. In addition to mass, momentum, and energy equations of the
mixture, only the concentration equation for the particulate phase needs to be solved. However,
this approach is only applicable for particles with a very small time constant.
The equilibrium Eulerian approach retains the computational simplicity of the dusty gas ap-
proach, but allows the particle velocity to be different from the surrounding carrier phase. It
assumes that the particles are sufficiently small that their motion is dictated only by the sur-
rounding fluid (Ferry & Balachandar 2001, Ferry et al. 2003). The particle velocity v can then be
explicitly expressed as an expansion in fluid velocity u with particle Stokes number (St = τ p /τk )
as the small parameter:
⎧
⎪ Du
⎪
⎪ −St(1 − β) for |w|  St
⎪
⎪ Dt
⎪
⎨ Du
v = u + w − St(1 − β) for |w| ∼ O(St) (1)
⎪
⎪  Dt 
⎪
⎪
⎪
⎪ Du
⎩w − St (1 − β) + w · ∇u for |w| ∼ O(1).
Dt
Here τ p = d 2 ρ/[18νφ(Re)] is the particle timescale, τ k is the Kolmogorov timescale, β = 3/(2ρ+1)
redefines the particle-to-fluid density ratio (ρ = ρ p /ρ f ), d is the particle diameter, and ν is the

116 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

kinematic viscosity of the fluid. The finite particle Reynolds number is accounted for with the
correction φ(Re) = 1 + 0.15Re 0.687 . The above expression is accurate to O(St), and higher-order
terms can be systematically derived (Ferry & Balachandar 2001). The particle settling velocity w
is also a key parameter, and the asymptotic form of v depends on the relative importance of w to
St. The velocities u, v, and w are normalized by the Kolmogorov velocity vk .
The advantage of the equilibrium Eulerian approach is that it captures the relative particle
motion more accurately and thereby enables important phenomena, such as preferential particle
accumulation and turbophoresis (Reeks 1983), to be captured faithfully. The accuracy of the
equilibrium approximation has been tested in a variety of turbulent flows. It appears that fairly
accurate results are obtained provided St  0.2 (Ferry & Balachandar 2002, Rani & Balachandar
2003, Shotorban & Balachandar 2006).
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

3.2. Eulerian Approach

Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

This two-fluid approach treats the carrier and the dispersed phases as interpenetrating fluid media,
and the particulate phase properties are given field representation (Crowe et al. 1998, Druzhinin
& Elghobashi 1998, Fevrier et al. 2005). In the dusty gas and equilibrium Eulerian approaches,
only the momentum and energy equations of the carrier phase are solved, along with the particle
concentration equation. The two-fluid formulation requires additional momentum and energy
equations for the particulate phase, with momentum and energy exchange between the phases
taken into account as source and sink terms.
The advantage of the two-fluid formulation over the equilibrium approximation is that the
restriction on St can be somewhat relaxed. Thus, the Eulerian approach is applicable for larger
particles. Furthermore, there are situations in which the equilibrium assumption is violated, such
as when particles are injected normal to the flow or downstream of a shock. Because additional
equations are solved, the Eulerian approach is computationally more expensive, especially for
polydisperse systems with a wide range of particle sizes. If the governing Eulerian equations are
derived consistently using the probability density function approach, one can handle polydispersity
by including particle size as one of the phase space variables (Fox et al. 2008, Pope 1985).

3.3. Lagrangian Point-Particle Approach

The Lagrangian point-particle approach perhaps has the longest history in turbulent dispersed
multiphase flow computations (Elghobashi 1991, Maxey 1987, Riley & Patterson 1974, Squires
& Eaton 1991). These early efforts were in the context of one-way coupled direct numerical
simulations (DNS) of the carrier phase. Later works have extended to two-way coupled simulations
(Elghobashi & Truesdell 1993, Ferrante & Elghobashi 2003, Squires & Eaton 1990), large eddy
simulation (LES) of the carrier phase (Portela & Oliemans 2002, Wang & Squires 1997), and four-
way coupled simulations that include particle-particle collisions (Vance et al. 2006, Yamamoto et al.
2001). This approach retains the Lagrangian description of the particles and solves the equations
of motion to track the position, mass, momentum, and energy of the particles (interphase coupling
approaches are discussed in Section 5). The Eulerian approaches assume the existence of unique
field representations for particle velocity and temperature, implicitly restricting the maximum
St that can be considered (Ferry & Balachandar 2001). If uniqueness is violated, then either a
probabilistic framework must be adopted or the Eulerian fields of particle properties must be
thought of as ensemble averages. Thus, the biggest advantage of the Lagrangian approach is that
there is no fundamental limitation on St due to the requirement of uniqueness. Furthermore, the
size of each particle is independent, and thus polydispersity can be handled easily.

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 117

 ANRV400-FL42-06 ARI 17 November 2009 10:20

The coupling of the Lagrangian particles back to the carrier phase poses interesting challenges.
Typically the hydrodynamic force on all the particles within a cell is distributed to the neighboring
grid points (see Equation 8 in Section 6). There needs to be enough particles in each cell to have a
smooth Eulerian representation of the feedback force from the particles. With fewer particles per
cell, smoothing of the feedback forcing is needed (Patankar & Joseph 2001, Sundaram & Collins
1996).

3.4. Fully Resolved Approach

All the above methods, either explicitly or implicitly, make the point-particle approximation and
thus are restricted to particles of size smaller than the Kolmogorov scale η, or the smallest resolved
eddies in the case of LES. The point-particle assumption is clear in the context of the Lagrangian
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

approach. In the context of the Eulerian approaches, the force and heat-transfer laws are invariably
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

based on the assumption that the particles are much smaller than the flow scales.
For particles of size comparable or larger than the smallest undisturbed flow scales of the
carrier phase, the ultimate option is to perform fully resolved DNS, in which all the scales of
ambient turbulence, and the flow scales introduced by the particles, are completely resolved. Such
simulations have been performed for a single particle (Bagchi & Balachandar 2003, 2004; Burton
& Eaton 2005; Merle et al. 2005; Zeng et al. 2008) and to a collection of up to O(1000) particles
in turbulent flows (Kajishima et al. 2001, Lu & Tryggvason 2006, Pan & Banerjee 1997, Ten
Cate et al. 2004, Uhlmann 2008). Most applications typically involve far more particles, and fully
resolved DNS of such systems is out of the question in the foreseeable future.

3.5. Appropriate Choice of Methodology

It can be argued that the point-particle Lagrangian approach is uniquely suited to handle large
particles (i.e., St > 1). The point-particle approach is clearly applicable even when St < 1. How-
ever, for such small particles, the Eulerian approaches may offer more computationally efficient
alternatives. In the limit St < 0.2, one may simply use the dusty gas or equilibrium Eulerian
approximation. Over the range 0.2  St  O(1), both the Eulerian two-fluid and Lagrangian
point-particle approaches become comparable.
The point-particle approach is on sound theoretical footing when d  η. However, in problems
involving millions of particles, even if d  η, the point-particle approach may be the only viable
option if St > 1 because a fully resolved simulation is not possible. The point-particle approach in
such cases where d  η can be appropriately termed the finite-size point-particle approach. From
the definitions of particle and Kolmogorov timescales, St can be expressed as

2ρ + 1 1 d 2
St = . (2)
36 φ(Re) η
As shown by Balachandar (2009), both Re and St can be expressed in terms of d/η, ρ, and the integral
to the Kolmogorov scale ratio of carrier-phase turbulence (L/η). Figure 2a shows St plotted against
d/η for varying ρ and L/η. The results are nearly independent of the intensity of carrier-phase
turbulence except for large ρ and d/η. If we restrict attention to d/η < 0.1, St is generally small
even for heavy particles in gas, and the dusty gas or the equilibrium Eulerian approach may suffice.
However, if we relax the particle-size restriction and consider d ∼ O(η), then for heavier-than-fluid
particles St can increase above unity and the point-particle approach becomes appropriate.
In case of LES of the carrier phase, the point-particle approach becomes the method of choice
provided τ p is larger than the timescale of the smallest resolved eddy [τξ = τk (ξ/η)2/3 , where ξ is

118 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

a DNS b LES ( ξ/η = 100)

104 104 Finite size for LES
Finite size for DNS
103 103

102 102
Lagrangian Lagrangian
101 point-particle 101 point-particle
τp/τk

τp/τξ
100 100
Eulerian Eulerian
10–1 10–1
Equilibrium Equilibrium
10–2 Eulerian 10–2 Eulerian
10–3 10–3
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

Dusty Dusty
10–4 10–4
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

gas gas
10–5 10–5
10–2 10–1 100 101 102 10–2 10–1 100 101 102
d/η d/η

Figure 2
The plot of timescale ratio versus nondimensional particle size, for varying density ratio and turbulence intensity. (a) τ p /τ k plotted for
direct numerical simulation (DNS) of the carrier phase. (b) τ p /τ ξ plotted for large eddy simulation (LES) of the carrier phase with the
cut-off length scale at 100η. The dashed gold line is for ρ = 0, the dashed-dotted purple line for ρ = 2.5, the dotted blue line for ρ =
25, and the solid red line for ρ = 1000. Lines with symbols represent L/η = 105 , whereas lines without symbols represent L/η = 103 .
It is arbitrarily assumed that d greater than 0.1 times the smallest eddy size will be considered to be large. Figure taken from
Balachandar 2009 and used with permission.

the smallest resolved LES length scale]. The timescale ratio as defined below is plotted in
Figure 2b for the particular case in which the LES filter is set at 100η:
 
τp 2ρ + 1 1 d 2 ξ 4/3
= . (3)
τξ 36 φ(Re) ξ η
Thus, even if we restrict attention to d/ξ < 0.1, St can be greater than 1, as long as the carrier-
phase Reynolds number is large enough that (ξ/η  1), making the point-particle approach the
method of choice. The point-particle LES approach can be pushed to consider larger particles (i.e.,
d ∼ ξ ), in which case even for bubbly flows, the Lagrangian point-particle approach may be the
only viable option.

4. PREFERENTIAL PARTICLE CONCENTRATION

Traditional theories of turbulent particle dispersion assume that the carrier-phase turbulence
applies random forcing to the particles, producing behavior similar to diffusion of a scalar con-
taminant. Accordingly, the concentration distribution evolves toward statistical uniformity in the
absence of external forces. We now know that particles may be drawn into dense clusters by tur-
bulence, and particle dispersion is often caused by deterministic eddy motions that produce highly
nonuniform concentration distributions (Eaton & Fessler 1994). This phenomenon of preferen-
tial concentration is also called inertial clustering because it is caused by the difference in inertia
between a particle and the fluid.
The ratio of particle-to-fluid inertia, expressed in terms of St, plays an important role in
preferential concentration. For very small St, the particles can follow all the motions of the turbu-
lence, and they disperse as fluid elements. The tendency for particles to locally accumulate with

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 119

 ANRV400-FL42-06 ARI 17 November 2009 10:20

increasing St can be seen by taking the divergence of Equation 1 and using incompressibility of
the carrier phase:
∇ · v = −St(1 − β) ||S||2 − || ||2 , (4)
where S and Ω are the local strain-rate and rotation-rate tensors, respectively. Heavier-than-
fluid particles (0 < β < 1) tend to accumulate (∇ · v < 0) in regions where the strain rate dominates
over vorticity, whereas lighter-than-fluid particles and bubbles (0 < β < 3) tend to accumulate
in regions of intense vorticity. The validity of the above expression, and the linear dependence
of preferential accumulation on St, applies only for small St. Large St particles are sluggish and
therefore have only a small response during the eddy’s lifetime. In this case, turbulence probably
can be modeled as providing small random impulses to the particles. At intermediate St, turbulent
eddies induce significant coherent motion of the particles, so preferential concentration is most
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

important when St ≈ 1.
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

4.1. Preferential Concentration in Gas Flows

Vortex structures can preferentially concentrate dense particles because the particle’s inertia pre-
vents them from following curved streamlines. Intense concentrations of particles may be formed
at convergence zones where multiple vortices interact to form a saddle-like flow. Early work on
preferential concentration examined free-shear flows. Experiments in plane mixing layers (Glawe
& Samimy 1993, Ishima et al. 1993, Kobayashi et al. 1988, Lazaro & Lasheras 1989, Wen et al.
1992) showed that large particles were flung away from the mixing layer by the roller vortices,
intermediate-size particles could become trapped in rings around vortices, and small particles
followed the flow. Numerous studies (starting from Crowe et al. 1985, Chien & Chung 1987,
and Ganan-Calvo & Lasheras 1991) showed that the main effects could be captured using two-
dimensional (2D) vortex models.
Preferential concentration has also been observed in axisymmetric jet flows. Chung & Troutt
(1988) represented a jet as a set of vortex rings and found that particles with St ≈ 1 based on
the vortex-passing frequency were the most strongly dispersed. At St = 10, so-called fingers of
particles were expelled from the jet core by the vortices. Both larger and smaller Stokes numbers
resulted in lower particle dispersion. Longmire & Eaton (1992) and Wicker & Eaton (2001)
experimentally studied acoustically forced round jets. Phase-locked particle concentration and
velocity measurements showed that the particles clustered in the high-strain region between the
vortex rings.
Preferential concentration is also important in flows with fully 3D vortex structure includ-
ing homogeneous turbulence. Maxey (1987) found inertial biasing of particle trajectories in 3D
isotropic turbulence using kinematic simulations. Squires & Eaton (1990) tracked inertial parti-
cles in a DNS of forced homogeneous turbulence, finding the strongest preferential concentration
for Stokes number (based on integral scale) of 0.15, with peak concentrations over 30 times the
mean. Conditional averages showed that the particles moved away from eddies and concentrated
in convergence zones (Squires & Eaton 1991).
Wang & Maxey (1993a) chose the Kolmogorov timescale to nondimensionalize the particle
timescale, surmising that preferential concentration is controlled by intense vortical structures
characteristic of small-scale turbulence. Subsequent simulations and experiments have shown this
to be the correct scaling, although it is untested at high Reynolds numbers.
Fessler et al. (1994) imaged the distribution of particles in the center plane of turbulent channel
flow and found the strongest preferential concentration for St ≈ 1. They divided each image
into a set of square boxes and calculated the distribution of the number of particles per box. A

120 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Figure 3
Laser-sheet image of St = 0.6 particles in homogeneous turbulence, illustrating preferential concentration.
Figure taken from Wood et al. 2005 and used with permission.

single parameter D = (σ − σ p )/μ was defined to indicate the deviation from a random (Poisson)
distribution. Here σ and σ p are the standard deviation of the measured and Poisson distributions
with the same mean μ, respectively. The measured D reached a maximum for St ≈ 1. Aliseda et al.
(2002) examined the preferential concentration of polydisperse water droplets in grid turbulence.
They found the strongest effect for St ≈ 1 and box sizes around 10η.
Sundaram & Collins (1997) and Wang et al. (2000) studied particle collisions in homogeneous
turbulence using Lagrangian/Eulerian DNS simulations. Sundaram & Collins (1997) used the
radial distribution function (RDF) as a measure of preferential concentration. The RDF is the
probability of finding a second particle at a given radial distance from any particle. It is useful
because the RDF evaluated at the particle radius appears in models for particle collision rates.
Wood et al. (2005) measured a 2D analog of the RDF in isotropic turbulence. Figure 3 shows
an image for St = 0.6, which had the strongest preferential concentration. The measured RDF
agreed qualitatively with the simulations, but the peak measured values were significantly lower
in the experiments. Holtzer & Collins (2002) identified an attenuation effect for 1D or 2D RDF
measurements that may have been responsible for the discrepancy. Saw et al. (2008) studied a
polydisperse spray in active-grid turbulence and measured a 1D RDF using a phase-Doppler
anemometer. In contrast to previous work, the results showed that the peak value of the RDF
increased monotonically with increasing St, but it is not clear if the investigators accounted for the
increasing size of the phase-Doppler anemometer measurement volume with increasing particle
size. Most recently, Salazar et al. (2008) measured the RDF using a 3D holographic imaging
system. This may be the future method of choice if resolution issues can be solved. There is now
broad consensus that in fully 3D turbulent flows, preferential concentration is strongest for St
near unity. The RDF is the most useful way to document preferential concentration in numerical
simulation results. However, further developments in experimental techniques are needed before
the RDF can be measured reliably.
Finally, we mention two areas where preferential concentration may play an important role in
the natural world. Shaw et al. (1998) describe how preferential concentration may affect both the
spatial distribution of water saturation and droplet-coalescence rates in clouds. Cuzzi et al. (2001)

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 121

 ANRV400-FL42-06 ARI 17 November 2009 10:20

pointed out that many meteorites comprise uniform-size particles and may have formed because
of preferential concentration in the protoplanetary nebula.

4.2. Preferential Concentration in Bubbly Flows

As discussed above, the inertial effect on bubbles is to draw them toward vortex cores. Videos of
captive dolphins at play offer excellent illustrations of both vortex dynamics and the preferential
concentration of bubbles in their cores (Marten et al. 1996). However, despite anecdotal evidence,
no consistent picture of preferential concentration has emerged over a range of flow types and
bubble sizes as it has for particle-laden flows.
Wang & Maxey (1993b) and Spelt & Biesheuvel (1997) simulated bubble motion in isotropic
turbulence and showed that bubbles collect in the intense vortex cores of Kolmogorov-scale
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

turbulence. Druzhinin & Elghobashi (1998) used a two-fluid DNS model of isotropic bubbly
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

turbulence but found only weak preferential concentration because the bubbles were very small. A
simple analysis indicated that the preferential concentration would increase linearly with increasing
Stokes number for values of St < 1. Vortex representations of mixing layers were used by Tio et al.
(1993) and Ruetsch & Meiburg (1993) to show that bubbles also can be trapped in the core of
shear-layer vortices.
There have been relatively few experimental studies of the preferential concentration of bub-
bles. The v of bubbles is much higher than for particles at equivalent number density, making
optical-imaging studies difficult. Rightley & Lasheras (2000) examined a mixing layer in which one
stream was laden with 50-μm microbubbles. The large-scale vortices produced a highly nonuni-
form bubble concentration field, but the bubbles were so small that preferential concentration
was insignificant. Aliseda & Lasheras (2006) examined a boundary layer laden with polydisperse
bubbles of d ≈ 200 μm. Laser-sheet illumination indicated preferential concentration with the
peak deviation from random distribution occurring for a counting box size of approximately 100
wall units. Calzavarini et al. (2008) examined grid turbulence laden with 200-μm bubbles and
interpreted spikes in a hot-wire signal as bubble impacts. The probability distribution of the
time between successive impacts suggested preferential concentration, but St for the bubbles was
approximately 0.005, far below the range at which inertial preferential concentration would be
expected.
In contrast to particle-laden flows, numerous questions remain about preferential concentration
in bubbly flows. For instance, the lift force in the bubble equation of motion appears to play a
significant role (Sridhar & Katz 1995). Unfortunately, there is not broad agreement on appropriate
models for the lift force. Moreover, typically St  1, unless the bubbles are larger than the scales of
turbulent motion (see Figure 2). Applying existing Lagrangian-Eulerian simulation techniques to
bubbles of d  η is not appropriate (see Section 5). Therefore, the importance of the preferential
concentration of bubbles due to fine-scale turbulence is an open research question.

5. EFFECT OF TURBULENCE ON INTERPHASE COUPLING

The key physics in dispersed multiphase flows are in the mass, momentum, and energy coupling
between the phases. Here we focus only on momentum coupling, and in both the Lagrangian
and the Eulerian treatments, the interfacial momentum coupling is taken into account through
drag and lift forces. In the limit of zero-particle Reynolds number, steady and unsteady Stokes
flow approximations have been used to obtain analytic expressions for the quasi-steady, pressure
gradient, added-mass, and Basset history components of the force (Crowe et al. 1998). At finite Re,
empirical corrections are in common use. These analytical and empirical expressions are strictly

122 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

applicable only to a spherical particle in a uniform cross flow. The purpose of this section is to
review current understanding of the effect of turbulence on force coupling between the particle
and the surrounding fluid.
Past experimental measurements on the effect of carrier-phase turbulence on the force on a
particle have been inconclusive, with some measuring drag increase, whereas others observed drag
reduction. For example, Uhlherr & Sinclair (1970) and Brucato et al. (1998) observed substantial
increase in the drag coefficient in a turbulent flow when compared with the standard drag. The
measurements of drag on 10–50-μm droplets in a turbulent field by Rudoff & Bachalo (1988),
however, showed substantial reduction in drag force due to ambient turbulence. In contrast,
Warnica et al. (1995) and Wu & Faeth (1994) concluded that ambient turbulence has no significant
influence on the mean drag force.
The following semiempirical relation is often used for the force (F) on a particle:
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.


Du Du d v
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

F = 3πμd (u − v)φ(Re) + m f + m f CM −
Dt Dt dt
3 2√ t
d (u − v)
+ d πρ f μ K (t, τ ) d τ + m p − m f g, (5)
2 −∞ dτ
where mp and mf are the mass of the particle and displaced fluid, respectively (Bagchi & Balachandar
2002, Magnaudet & Eames 2000). The acceleration due to gravity is g, and CM is the added mass
coefficient. The terms on the right are the quasi-steady standard drag, pressure gradient, added
mass, Basset history [K(t, τ ) is the kernel that weighs the past history of relative acceleration], and
gravitational forces, respectively. Additional forces, such as the lift and gravitational forces, can
be included as necessary. We note that D/Dt is the total derivative following the fluid velocity
and d/dt is the derivative following the particle velocity. In the above expression for force, u
and Du/Dt are defined as the undisturbed fluid velocity and total acceleration as seen by the
particle. Their interpretation is straightforward for d  η because the ambient flow appears nearly
uniform.
Equation 5, after ignoring the unsteady forces, can be applied to obtain an effective drag
coefficient for the case of particles falling through an isotropic turbulent field of zero mean as

24 1 d 0.687 1.687
CD,eff = ur + 0.15 ur , (6)
Re v | v | ν

where Re v is the mean Reynolds number based on the average particle velocity. The effective drag
coefficient can be higher or lower than the standard drag coefficient: (24/ Re v )(1+0.15 Re v 0.687 ).
Even if the ambient flow has zero mean velocity, the average fluid velocity seen by the parti-
cle need not be zero due to loitering, trapping, trajectory bias, and two-way coupling (Nielsen
1984, 1993). As a result, | ur | may be larger or smaller than | v |. Also, the effect of the non-
linear drag relation appears as | ur1.687 | = | ur |1.687 . Experiments (Aliseda et al. 2002, Yang &
Shy 2003) and simulations (Bosse et al. 2006, Wang & Maxey 1993a, Yang & Lei 1998) have
verified that the trajectory bias and two-way coupling effects generally enhance settling veloc-
ity, whereas the loitering, trapping, and nonlinear drag effects suppress it. These opposing in-
fluences of ambient turbulence on settling velocity are consistent with, and can partly explain,
the inconclusive experimental results on the influence of ambient turbulence on the mean drag
coefficient.
The point-particle approximation has been extended and employed for larger particles as well.
However, there are important challenges in the formulation and the use of a force formula of
the form given in Equation 5. Because there is no scale separation, it is hard to establish the
undisturbed ambient flow that would exist in the absence of the particle. Also, if the undisturbed

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 123

 ANRV400-FL42-06 ARI 17 November 2009 10:20

flow is known and denoted as u0 (x, t), it can be expanded about the location of the particle center
(xp ) as
u0 (x, t) = u0 (x p , t) + (x − x p ) · [∇u0 ](x p ,t) + (1/2)(x − x p )(x − x p ) : [∇∇u0 ](x p ,t) + · · · . (7)

For a small particle, the approximation u0 (x, t) ≈ u0 (x p , t) may be adequate. For larger particles
of size comparable with the carrier flow scales, leading-order information, such as u0 (x p , t) and
[∇u0 ](x p ,t) , alone may be insufficient, and the undisturbed ambient flow seen by the particle can
be far more complex.
The problem of a finite-sized particle of diameter larger than the Kolmogorov scale subjected
to isotropic turbulence has been considered by Bagchi & Balachandar (2003, 2004) and by Burton
& Eaton (2005). Merle et al. (2005) considered the problem of a stationary spherical bubble placed
along the centerline of a turbulent pipe flow, and Zeng et al. (2008, 2009) studied the corresponding
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

problem of a particle in a turbulent channel flow placed both in the buffer layer and in the channel
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

center. Their results can be summarized as follows. (a) As far as time-averaged mean drag force
on the particle, there is no systematic effect of ambient turbulence, and the standard drag law is
sufficient. (b) In case of d  η, Equation 5 accurately predicts the instantaneous time evolution
of the particle force. In fact, for such small particles, the dominant contribution to forces is
from the quasi-steady term. (c) For d  η, slow variations in particle force arising from scales
larger than d can be well described by Equation 5. However, rapid variations in the instantaneous
force arising from scales comparable to or smaller than d cannot be accurately captured by force
expressions of the form of Equation 5. (d ) Self-induced vortex shedding becomes important above
a certain critical particle Reynolds number, and the critical Reynolds number decreases with
increasing level of ambient turbulence. (e) Finally, self-induced vortex shedding introduces force
fluctuations, which are most pronounced in the lift component, and are not correlated to the
oncoming turbulent ambient flow.
Based on the above understanding, the following picture emerges. For small particles (d  η),
no special treatment is needed and Equation 5 is adequate to account for the turbulence effect on
particle motion. Such simplicity is possible because the range of turbulence scales in the carrier
phase is well separated from the particle-generated scales. Also, the Reynolds number of such
particles is typically quite low, and self-induced vortex shedding will not be present (Balachandar
2009). The particle-turbulence interaction becomes complex, and far more interesting, when
d  η. The best strategy here is to consider that the force on the particle comprises a deterministic
and a stochastic component. The contribution of smaller-scale eddies can be best accounted for
in terms of a stochastic component, and this is an area ripe for future research (Sawford 2001). If
the stochastic component is appropriately included, it can account for the dispersion of particles
that arise from the effect of the smaller scales of ambient motion. The above discussion of particle
force is equally well applicable to mass and energy exchange between the dispersed and the carrier
phases.

6. TURBULENCE MODULATION
Flows containing very small volume fractions (v  1) of very small particles (d < η) have
carrier-phase turbulence that is nearly the same as the equivalent single-phase flow. However,
at higher v or d ≥ η, extra mechanisms of turbulence production, distortion, and dissipation
become important, and the turbulent stresses can be either reduced or increased. This effect can
be dominant when the single-phase production rate is small.
Turbulence modulation is important because it can be so large as to qualitatively change the
behavior of natural or engineering systems. The mechanisms of turbulence modulation are poorly

124 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

understood because the wide range of relevant length scales from the particle diameter to the
size of the largest eddies causes problems for detailed simulations. Also, as discussed in Section
2, it is difficult to acquire accurate turbulence data for the carrier phase in particle-laden flows.
Finally, several different mechanisms can cause turbulence modulation, and sometimes multiple
mechanisms act simultaneously. Because of these competing factors, the present state of our
knowledge is incomplete, and many contradictory results have been published.

6.1. Observations of Turbulence Attenuation

There have been numerous experiments and numerical simulations that show turbulence modi-
fication in gas and liquid flows laden with solid particles. The results were reviewed recently by
Eaton (2006, 2009).
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

Gore & Crowe (1991) reviewed early work on jet and pipe flows and plotted the percentage
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

change in the turbulence intensity versus the ratio of the particle diameter to a turbulence integral
scale. They found that particles larger than approximately 1/10 of the integral scale augment
turbulence, whereas smaller particles attenuate it. This conclusion has been shown to be roughly
correct in subsequent experiments. However, the level of turbulence modification is dependent
on other factors, including at least St, Re, and m .
Kulick et al. (1994), Paris & Eaton (2001), and Kussin & Sommerfeld (2002) studied fully devel-
oped channel flow for a wide range of particle size and density. All cases with d  η had turbulence
attenuation, with the strongest reduction occurring near the channel center plane where produc-
tion is small. Small particles with St < 10 tracked the energy-containing eddies and showed little
attenuation. Maximum attenuation occurred for St ≈ 50. Paris & Eaton (2001) examined two dif-
ferent particle classes that had nearly identical St, but Reynolds numbers differing by a factor of two.
The attenuation was greater for the higher–Reynolds number particles, as seen in Figure 4, which
also shows that the attenuation increases monotonically with increasing m . Kussin & Sommerfeld
(2002) examined cases with particles larger than η and found significant turbulence augmentation
near the channel center plane where the particle Reynolds numbers were greater than 350.

0.8
TKE / unladen TKE

0.6
Copper

0.4
Glass

0.2

0
0 0.1 0.2 0.3 0.4
MLR
Figure 4
Turbulent-kinetic-energy (TKE) attenuation on the center plane of turbulent channel flow as a function of
mass loading ratio (MLR = m /ρ f ). 70-μm copper beads have St = 47 and Re = 10, and 150-μm glass
beads have St = 49 and Re = 20.

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 125

 ANRV400-FL42-06 ARI 17 November 2009 10:20

Most computational research in homogeneous turbulent flows used the Lagrangian point-
particle approach with interphase coupling between small particles and forced isotropic turbu-
lence (Boivin et al. 1998, Squires & Eaton 1990) or decaying isotropic turbulence (Elghobashi &
Truesdell 1993, Ferrante & Elghobashi 2003) (see Section 5). Substantial attenuation of turbu-
lence kinetic energy was found for 1  St  10, but at levels smaller than observed in pipe and
channel experiments.
Geiss et al. (2004) studied decaying grid turbulence in a downward airflow laden with light
loadings of large–Stokes number particles. No significant change in the turbulent kinetic energy
occurred for v  5 × 10−5 . At higher loadings, the smaller particles augmented the stream-
wise (gravity-aligned) turbulence and attenuated the transverse components, whereas the largest
particles augmented all turbulence components. Poelma et al. (2007) studied grid turbulence in
a downward water flow with d/η ranging from 0.5 to 1.5, m < 1%, and St from 0.6 to 0.48.
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

They observed significant turbulence modulation despite the small m and St. One of the pitfalls
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

of grid turbulence studies is that velocity fluctuations decay to a very low level, so even a small
increase or decrease of the turbulent fluctuations is a significant percentage change. Hwang &
Eaton (2006) studied forced isotropic turbulence at much higher Re and v . Solid particles with
dp /η ≈ 1, St ≈ 50, and Re ≈ 8 fell through the chamber. The measured attenuation is significantly
greater than the numerical simulations, even though the simulations do not include gravity effects,
which augment turbulence. Although the parameters are not matched, this difference provides a
strong indication that point-particle simulations underpredict turbulence attenuation. A similar
conclusion was reached by Yamamoto et al. (2001) and Eaton (2009), who compared channel-flow
simulations to experiments.

6.2. Mechanisms of Turbulence Modulation

In a dilute suspension, three mechanisms contributing to turbulence reduction can be identified:
(a) the enhanced inertia of the particle-laden flow, (b) increased dissipation arising from particle
drag, and (c) the enhanced effective viscosity of the particle-laden fluid. Conversely, turbulence can
be enhanced in two different ways: (d ) enhanced velocity fluctuation due to wake dynamics and
self-induced vortex shedding and (e) buoyancy-induced instabilities due to density variation arising
from preferential particle concentration. The length scale at which these different mechanisms are
active varies. Thus, compared to unladen turbulence, the suspended particles can simultaneously
augment and suppress over a different range of scales, and the overall modulation depends on the
relative strength of the different mechanisms.
The primary focus has been on mechanisms (b) and (d ) that arise from the imposition of the no-
slip and impermeability boundary conditions on the particles that are moving relative to the flow.
Large particles augment the turbulence by creating unsteady wakes above a critical Re. The critical
Reynolds number for the onset of vortex shedding somewhat decreases with ambient turbulence
(Bagchi & Balachandar 2004, Mittal 2000, Zeng et al. 2008). Turbulent kinetic energy increases
in the wake region by more than an order of magnitude partly because of unsteadiness arising
from ambient turbulence, but mostly because of vortex shedding (Bagchi & Balachandar 2004,
Zeng et al. 2009). Even in the absence of unsteadiness, the superposition of many laminar-like
wakes randomly positioned in space and time contributes to fluctuation (Chen & Faeth 2001,
Parthasarathy & Faeth 1990). The effect of superposed particle wakes is seen even in gas-particle
flows with low v of small particles. Simulations and experiments (see Rogers & Eaton 1991) have
shown an increase in the energy spectra at high wave number due to this effect.
The turbulence attenuation mechanism represented by most models is the extra dissipation
caused by unresponsive particles that apply forces opposing the turbulent fluctuations. Energy is

126 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

transferred from the turbulence to the small-scale flow around each particle where it is dissipated
rapidly. This effect can be analyzed with the particle-laden Navier-Stokes equation (Berlemont
et al. 1990, Rogers & Eaton 1991):
∂u∗ 1 1 ∗
+ u∗ · ∇u∗ = ∇ p ∗ + ν f ∇ 2 u∗ − f , (8)
∂t ∗ ρf ρf
where the asterisk indicates dimensional quantities, and f is the instantaneous force per unit volume
applied onto the particles by the fluid. If the particle diameter and spacing are smaller than η, f can
be assumed to be continuous. If we consider only the quasi-steady contribution (see Equation 5),
it can be represented by
m ∗
f∗ = (u − v∗ ) . (9)
τp
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

Assuming homogeneity, a transport equation for the turbulent kinetic energy (q2 /2) can be derived
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

as
D̄ q 2 m     1  
= P −− ui ui − ui vi − Ui − Vi m ui + m ui ui − m ui vi , (10)
Dt 2 ρ f τp ρ f τp
where capital letters indicate mean values, primes indicate fluctuating quantities, and overlines
indicate time or ensemble averaging. P and  are the single-phase turbulence production and
dissipation terms. The last term involving m is small at high St. Therefore, the equation is the
same as the single-phase equation, with the addition of a single term involving m , representing
the extra dissipation. Unfortunately, this extra-dissipation model does not capture the level of
turbulence attenuation or the experimentally observed dependence on the dimensionless particle
diameter.
Therefore, some other mechanism must significantly increase dissipation. Experiments show
that turbulence attenuation is strong when d  η, suggesting that local flow distortion around the
particles is important. Burton & Eaton (2005) performed highly resolved simulations of a single
particle interacting with homogeneous turbulence with d /η ≈ 2 and Re ≈ 20. They found strong
augmentation of the turbulence dissipation rate and reduction of the turbulent kinetic energy in
a sphere with a diameter of approximately 5d. The locally augmented dissipation is caused by the
distortion of turbulent eddies by the surface boundary conditions on the particles, showing that
the finite diameter of the particles plays a critical role in attenuating turbulence. Tanaka & Eaton
(2009) examined the local distortion of turbulence using high-resolution PIV in forced isotropic
turbulence. Ensemble averages of the dissipation field around moving particles showed that the
dissipation rate is augmented by up to a factor of three near the particle. As discussed in Section 3,
full resolution of a large number of particles is out of the question, and thus other Lagrangian and
Eulerian approaches must account for the effect of finite particle size. New modeling contributions
are needed in this area.
Recent efforts have sought new dimensionless parameters to collapse turbulence modulation
data. Poelma et al. (2007) postulated that the particle effects on turbulence are proportional to the
product of the number density and particle time constant made dimensionless using Kolmogorov
scales. They called this parameter the Stokes load:
v ρ p η3
 St = . (11)
3πνd τk
They found that the decay rate of grid turbulence varied linearly with this parameter. Tanaka &
Eaton (2008) nondimensionalized Equation 8 using an integral length scale (L) and the velocity of
large-scale turbulence to obtain the feedback force as −m f/(ρ f Pa St ), where the nondimensional

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 127

 ANRV400-FL42-06 ARI 17 November 2009 10:20

force f and the particle momentum number PaSt are defined as

τp ∗  η 3
f= f and Pa St = St Re 2L . (12)
m vk L
They compiled turbulence modulation data for pipes and channel flows and showed that turbulence
attenuation occurs for 3 × 103 ≤ Pa St ≤ 105 , whereas augmentation was found for lower and
higher PaSt . Their data compilation showed that there are no turbulence modulation experiments
with large-scale Reynolds number above 30,000, a glaring weakness in the experimental database.
There is ample evidence that bubbles can cause turbulence modulation in liquid flows (see
Druzhinin & Elghobashi 1998, Lance & Bataille 1991). However, there are open questions about
the accuracy of bubbly flow simulations, and experiments are difficult because of the high bubble
volume fraction. Therefore, we focus here on flows laden with solid particles, on which there is a
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

wealth of previous research.

Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

FUTURE DIRECTIONS
1. It is important to develop experimental techniques capable of yielding simultaneous fully
resolved measurements of both the carrier and dispersed phases.
2. While advancing the capabilities of fully resolved simulations, we must develop La-
grangian and two-fluid formulations capable of handling large particles of d  η.
3. The role of fine-scale turbulence in the preferential concentration of bubbles requires
further attention.
4. Interphase coupling models (drag, lift, and heat-transfer laws) must be extended to large
particles by including a stochastic component.
5. The mechanisms of turbulence modulation and their parametric dependence are poorly
understood and are wide open for fundamental investigation.

DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or financial holdings that
might be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS
S.B. acknowledges the support from the National Science Foundation (CBET 0639446 and EAR
0609712).

LITERATURE CITED
Aliseda A, Cartellier A, Hainaus F, Lasheras JC. 2002. Effect of preferential concentration on the settling
velocity of heavy particles in homogeneous isotropic turbulence. J. Fluid Mech. 468:77–105
Aliseda A, Lasheras JC. 2006. Effect of buoyancy on the dynamics of a turbulent boundary layer laden with
microbubles. J. Fluid Mech. 559:307–34
Bachalo WD, Houser MJ. 1984. Phase Doppler spray analyzer for simultaneous measurements of drop size
and velocity distribution. Opt. Eng. 23:583–90
Bagchi P, Balachandar S. 2002. Steady planar straining flow past a rigid sphere at moderate Reynolds number.
J. Fluid Mech. 466:365–407
Bagchi P, Balachandar S. 2003. Effect of turbulence on the drag and lift of a particle. Phys. Fluids 15:3496–513

128 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

Bagchi P, Balachandar S. 2004. Response of the wake of an isolated particle to an isotropic turbulent flow.
J. Fluid Mech. 518:95–123
Balachandar S. 2009. A scaling analysis for point-particle approaches to turbulent multiphase flows.
Int. J. Multiphase Flow 35:801–10
Balachandar S, Prosperetti A. 2004. IUTAM Symposium on Computational Approaches to Multiphase Flow. New
York: Springer
Berlemont A, Desjonqueres P, Gousebet G. 1990. Particle Lagrangian simulation in turbulent flows. Int.
J. Multiphase Flow 16:19–34
Boivin M, Simonin O, Squires KD. 1998. Direct numerical simulation of turbulence modulation by particles
in isotropic turbulence. J. Fluid Mech. 375:235–63
Bosse T, Kleiser L, Meiburg E. 2006. Small particles in homogeneous turbulence: settling velocity enhance-
ment by two-way coupling. Phys. Fluids 18:027102
Brucato A, Grisafi F, Montante G. 1998. Particle drag coefficients in turbulent fluids. Chem. Eng. Sci. 53:3295–
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

314
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Burton TM, Eaton JK. 2005. Fully resolved simulations of particle-turbulence interaction. J. Fluid Mech.
545:67–111
Calzavarini E, van den Berg TH, Toschi F, Lohse D. 2008. Quantifying microbubble clustering in turbulent
flow from single-point measurements. Phys. Fluids 20:040702
Carrier GF. 1958. Shock waves in a dusty gas. J. Fluid Mech. 4:376–82
Chen JH, Faeth GM. 2001. Continuous-phase properties of homogeneous particle-laden turbulent flows.
AIAA J. 39:180–83
Chien R, Chung JN. 1987. Effects of vortex pairing on particle-dispersion in turbulent shear flows. Int.
J. Multiphase Flow 13:785–802
Chung JN, Troutt TR. 1988. Simulation of particle dispersion in an axisymmetric jet. J. Fluid Mech. 186:199–
222
Crowe CT. 1982. Review: numerical models for dilute gas-particle flows. J. Fluid Eng. 104:297–303
Crowe CT, Gore RA, Troutt TR. 1985. Particle dispersion by coherent structures in free shear flows. Part.
Sci. Technol. 3:149–55
Crowe C, Sommerfeld M, Tsuji Y. 1998. Multiphase Flows with Droplets and Particles. Boca Raton, FL: CRC
Press
Crowe CT, Troutt TR, Chung JN. 1996. Numerical models for two-phase flows. Annu. Rev. Fluid Mech.
28:11–43
Cuzzi JN, Hogan RC, Paque JM, Dobrovolskis AR. 2001. Self-selective concentration of chondrules and other
small particles in protoplanetary nebula turbulence. Astrophys. J. 546:496–508
Doig ID, Roper GH. 1967. Air velocity profiles in the presence of concurrently transported particles. Ind.
Eng. Chem. Fund. 6:247–56
Druzhinin OA, Elghobashi S. 1998. Direct numerical simulations of bubble-laden turbulent flows using the
two-fluid formulation. Phys. Fluids 10:685–97
Eaton JK. 2006. Turbulence modulation by particles. In Multiphase Flow Handbook, ed. CT Crowe, pp. 86–98.
Boca Raton, FL: Taylor & Francis
Eaton JK. 2009. Two-way coupled turbulence simulations of gas-particle flows using point particle tracking.
Int. J. Multiphase Flow 35:792–800
Eaton JK, Fessler JR. 1994. Preferential concentration of particles by turbulence. Int. J. Multiphase Flow
20(Suppl.):169–209
Elghobashi S. 1991. Particle-laden turbulent flows: direct simulation and closure models. Appl. Sci. Res.
48:301–14
Elghobashi S. 1994. On predicting particle-laden turbulent flows. Appl. Sci. Res. 52:309–29
Elghobashi S, Truesdell GC. 1992. Direct simulation of particle dispersion in a decaying isotropic turbulence.
J. Fluid Mech. 242:655–700
Elghobashi S, Truesdell GC. 1993. On the two-way interaction between homogeneous turbulence and dis-
persed solid particles. I: Turbulence modification. Phys. Fluids A 5:1790–801
Ferrante A, Elghobashi S. 2003. On the physical mechanism of two-way coupling in particle-laden isotropic
turbulence. Phys. Fluids 15:315–29

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 129

 ANRV400-FL42-06 ARI 17 November 2009 10:20

Ferry J, Balachandar S. 2001. A fast Eulerian method for two-phase flow. Int. J. Multiphase Flow 27:1199–226
Ferry J, Balachandar S. 2002. Equilibrium expansion for the Eulerian velocity of small particles. Powder Technol.
125:131–39
Ferry J, Rani SL, Balachandar S. 2003. A locally implicit improvement of the equilibrium Eulerian method.
Int. J. Multiphase Flow 29:869–91
Fessler JR, Kulick JD, Eaton JK. 1994. Preferential concentration of heavy particles in a turbulent channel
flow. Phys. Fluids A 6:3742–49
Fevrier P, Simonin O, Squires KD. 2005. Partitioning of particle velocities in gas-solid turbulent flows into
a continuous field and a spatially uncorrelated random distribution: theoretical formalism and numerical
study. J. Fluid Mech. 533:1–46
Fox RO, Laurent F, Massot F. 2008. Numerical simulation of spray coalescence in an Eulerian framework:
direct quadrature method of moments and multi-fluid method. J. Comput. Phys. 277:3058–88
Ganan-Calvo AM, Lasheras JC. 1991. The dynamics and mixing of small spherical particles in a plane, free
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

shear layer. Phys. Fluids A 3:1207–17

Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Geiss S, Dreizler A, Stojanovic Z, Chrigui M, Sadiki A, Janicka A. 2004. Investigation of turbulence modifi-
cation in a non-reactive two-phase flow. Exp. Fluids 36:344–54
Glawe DD, Samimy M. 1993. Dispersion of solid particles in compressible mixing layers. J. Propuls. Power
9:83–89
Gore RA, Crowe CT. 1991. Modulation of turbulence by a dispersed phase. ASME J. Fluids Eng. 113:304–7
Gouesbet G, Berlemont A. 1998. Eulerian and Lagrangian approaches for predicting the behaviour of discrete
particles in turbulent flows. Prog. Energy Combust. Sci. 25:133–59
Guha A. 2008. Transport and deposition of particles in turbulent and laminar flows. Annu. Rev. Fluid Mech.
40:311–41
Hasan Y, Philip O, Schmidl W. 1993. Bubble collapse velocity measurements using a particle image velocimetry
technique with fluorescent tracers. ASME FED 172:85–92
Hetsroni G, Sokolov M. 1971. Distribution of mass, velocity and intensity of turbulence in a two-phase
turbulent jet. ASME J. Appl. Mech. 38:315–27
Holtzer GL, Collins LR. 2002. Relationship between the intrinsic radial distribution function for an isotropic
field of particles and lower dimensional measurements. J. Fluid Mech. 459:93–102
Hwang W, Eaton JK. 2006. Homogeneous and isotropic turbulence modulation by small heavy (st ∼ 50)
particles. J. Fluid Mech. 564:361–93
Ishima T, Hishida K, Maeda M. 1993. Effect of particle residence time on particle dispersion in a plane mixing
layer. ASME J. Fluids Eng. 115:751–59
Kajishima T, Takiguchi S, Hamasaki H, Miyake Y. 2001. Turbulence structure of particle-laden flow in a
vertical plane channel due to vortex shedding. JSME Int. J. Ser. B 44:526–35
Khalitov DA, Longmire EK. 2002. Simultaneous two-phase PIV by two-parameter phase discrimination. Exp.
Fluids 32:252–68
Kiger KT, Lasheras JC. 1995. The effect of vortex pairing on particle dispersion and kinetic energy transfer
in a two-phase turbulent shear layer. J. Fluid Mech. 302:149–78
Kiger KT, Pan C. 2000. PIV technique of the simultaneous measurement of dilute two-phase flows. ASME
J. Fluids Eng. 122:811–18
Kobayashi H, Masutani SM, Azuhata S, Arashi N, Hishinuma Y. 1988. Dispersed phase transport in a plane
mixing layer. In Transport Phenomena in Turbulent Flow, ed. M Hirata, N Kasagi, pp. 693–700. New York:
Hemisphere
Kulick JD, Fessler JR, Eaton JK. 1994. Particle response and turbulence modification in fully developed
channel flow. J. Fluid Mech. 277:109–34
Kussin J, Sommerfeld M. 2002. Experimental studies on particle behaviour and turbulence modification in
horizontal channel flow with different wall roughness. Exp. Fluids 33:143–59
Lance M, Bataille J. 1991. Turbulence in the liquid phase of a uniform bubbly air-water flow. J. Fluid Mech.
222:95–118
Lazaro BJ, Lasheras JC. 1989. Particle dispersion in a turbulent plane, free shear layer. Phys. Fluids A 1:1035–44
Lee SL, Durst F. 1982. On the motion of particles in turbulent duct flows. Int. J. Multiphase Flow 8:125–46

130 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

Longmire EK, Eaton JK. 1992. Structure of a particle-laden round jet. J. Fluid Mech. 236:217–57
Loth E. 2000. Numerical approaches for motion of dispersed particles, droplets and bubbles. Prog. Energy
Combust. Sci. 26:161–223
Lu JC, Tryggvason G. 2006. Numerical study of turbulent bubbly downflows in a vertical channel. Phys. Fluids
18:103302
Magnaudet J, Eames I. 2000. The motion of high Reynolds number bubbles in inhomogeneous flows. Annu.
Rev. Fluid Mech. 32:659–708
Marble FE. 1970. Dynamics of dusty gases. Annu. Rev. Fluid Mech. 2:397–446
Marten K, Shariff K, Psarakos S, White DJ. 1996. Ring bubbles of dolphins. Sci. Am. 275:82–87
Mashayek F, Pandya RVR. 2003. Analytical description of particle/droplet-laden turbulent flows. Prog. Energy
Combust. Sci. 29:329–78
Maxey MR. 1987. The gravitational settling of aerosol particles in homogeneous turbulence and random flow
fields. J. Fluid Mech. 174:441–65
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

Merle A, Legendre D, Magnaudet J. 2005. Forces on a high-Reynolds-number spherical bubble in a turbulent

Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

flow. J. Fluid Mech. 532:53–62

Mittal R. 2000. Response of the sphere wake to free-stream fluctuations. Theoret. Comput. Fluid Dyn. 13:397–
419
Modarress D, Tan H, Elghobashi S. 1984. Two-component LDA measurement in a two-phase turbulent jet.
AIAA J. 22:624–30
Muste M, Parthasarathy RN, Patel VC. 1996. Discriminator laser Doppler velocimetry for measurement of
liquid and particle velocities in sediment-laden flows. Exp. Fluids 22:45–56
Nielsen P. 1984. On the motion of suspended sand particles. J. Geophys. Res. 89:616–26
Nielsen P. 1993. Turbulence effects on the settling of suspended particles. J. Sediment. Petrol. 63:835–38
Oakley TR, Loth E, Adrian RJ. 1997. A two-phase cinematic PIV method for bubbly flows. ASME J. Fluids
Eng. 119:707–12
Pan Y, Banerjee S. 1997. Numerical investigation of the effects of large particles on wall turbulence. Phys.
Fluids 9:3786–807
Paris AD, Eaton JK. 1999. Measuring velocity gradients in a particle-laden channel flow. Proc. 3rd Int. Workshop
Part. Image Velocim., Santa Barbara, pp. 513–18
Paris AD, Eaton JK. 2001. Turbulence attenuation in a particle-laden channel flow. Rep. TSD-137, Dep. Mech.
Eng., Stanford Univ.
Parthasarathy RN, Faeth GM. 1990. Turbulence modulation in homogeneous dilute particle-laden flows.
J. Fluid Mech. 220:485–514
Patankar NA, Joseph DD. 2001. Lagrangian numerical simulation of particulate flows. Int. J. Multiphase Flow
27:1685–706
Poelma C, Westerweel J, Ooms G. 2007. Particle-fluid interaction in grid-generated turbulence. J. Fluid Mech.
589:315–51
Pope SB. 1985. PDF methods for turbulent reactive flows. Prog. Energy Combust. Sci. 11:119–92
Portela LM, Oliemans RVA. 2002. Direct and large eddy simulation of particle-laden flows using the
point-particle approach. In Direct and Large Eddy Simulations IV, ed. B Geurts, R Friedrich, O Metais,
pp. 453–60. New York: Springer
Prosperetti A, Tryggvason G. 2007. Computational Methods for Multiphase Flow. Cambridge, UK: Cambridge
Univ. Press
Rani SL, Balachandar S. 2003. Evaluation of the equilibrium Eulerian approach for the evolution of particle
concentration in isotropic turbulence. Int. J. Multiphase Flow 29:1793–816
Reeks MW. 1983. The transport of discrete particles in inhomogeneous turbulence. J. Aerosol Sci. 14:729–39
Rightley PM, Lasheras JC. 2000. Bubble dispersion and interphase coupling in a free-shear flow. J. Fluid Mech.
412:21–59
Riley JJ. Patterson GS. 1974. Diffusion experiments with numerically integrated isotropic turbulence. Phys.
Fluids 17:292–97
Rogers CB, Eaton JK. 1991. The effect of small particles on fluid turbulence in a flat-plate, turbulent boundary
layer in air. Phys. Fluids A 3:928–37

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 131

 ANRV400-FL42-06 ARI 17 November 2009 10:20

Rudoff RC, Bachalo WD. 1988. Measurement of droplet drag coefficients in polydispersed turbulent flow field.
Presented at Aerosp. Sci. Meet., 29th, Reno, NV, AIAA Pap. No. 88-0235
Ruetsch GR, Meiburg E. 1993. On the motion of small spherical bubbles in two-dimensional vortical flows.
Phys. Fluids A 5:2326–41
Saarenrinne P, Piirto M. 2000. Turbulent kinetic energy dissipation rate estimation for PIV vector fields. Exp.
Fluids 29(Suppl.):S300–7
Sakakibara J, Wicker RB, Eaton JK. 1996. Measurements of the particle-fluid velocity correlation and the
extra dissipation in a round jet. Int. J. Multiphase Flow 22:863–81
Salazar JPL, DeJong J, Cao L, Woodward SH, Meng H, Collins LR. 2008. Experimental and numerical
investigation of inertial particle clustering in isotropic turbulence. J. Fluid Mech. 600:245–56
Sato Y, Hishida K, Maeda M. 1996. Effect of dispersed phase on modification of turbulent flow in a wall jet.
Trans. ASME J. Fluids Eng. 118:307–15
Saw EW, Shaw RA, Ayyalasomayajula S, Chuang PY, Gylfason A. 2008. Inertial clustering of particles in
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

high-Reynolds-number turbulence. Phys. Rev. Lett. 100:214501

Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Sawford B. 2001. Turbulent relative dispersion. Annu. Rev. Fluid Mech. 33:289–317
Shaw RA. 2003. Particle-turbulence interactions in atmospheric clouds. Annu. Rev. Fluid Mech. 35:183–227
Shaw RA, Reade WC, Collins LR, Verlinde J. 1998. Preferential concentration of cloud droplets by turbulence:
effects on the early evolution of cumulus cloud droplet spectra. J. Atmos. Sci. 55:1965–76
Sheng J, Meng H, Fox RO. 2000. A large eddy PIV method for turbulence dissipation rate estimation. Chem.
Eng. Sci. 55:4423–34
Shirolkar JS, Coimbra CFM, McQuay MQ. 1996. Fundamental aspects of modeling turbulent particle dis-
persion in dilute flows. Prog. Energy Combust. Sci. 22:363–99
Shotorban B, Balachandar S. 2006. Particle concentration in homogeneous shear turbulence simulated via
Lagrangian and equilibrium Eulerian approaches. Phys. Fluids 18:065105
Spelt PDM, Biesheuvel A. 1997. On the motion of gas bubbles in homogeneous isotropic turbulence. J. Fluid
Mech. 336:221–44
Squires KD, Eaton JK. 1990. Particle response and turbulence modification in isotropic turbulence. Phys.
Fluids A 2:1191–203
Squires KD, Eaton JK. 1991. Preferential concentration of particles by turbulence. Phys. Fluids A 3:1169–79
Sridhar G, Katz J. 1995. Drag and lift forces on microscopic bubbles entrained by a vortex. Phys. Fluids 7:389–99
Stock DE, Jurewicz JT, Crowe CT, Eschbach JE. 1975. Measurement of both gas and particle velocity
in turbulent two-phase flow. In Turbulence in Liquids: Proc. 4th Biennial Symp., Rolla, MO, pp. 91–100.
Princeton, NJ: Science
Sundaram S, Collins LR. 1996. Numerical considerations in simulating a turbulent suspension of finite-volume
particles. J. Comp. Phys. 124:337–50
Sundaram S, Collins LR. 1997. Collision statistics in an isotropic particle-laden turbulent suspension. Part 1.
Direct numerical simulation. J. Fluid Mech. 335:75–109
Tanaka T, Eaton JK. 2007. A correction method for measuring turbulence kinetic energy dissipation rate by
PIV. Exp. Fluids 42:893–902
Tanaka T, Eaton JK. 2008. Classification of turbulence modification by dispersed spheres using a novel
dimensionless number. Phys. Rev. Lett. 101:114502
Tanaka T, Eaton JK. 2009. Sub-Kolmogorov resolution PIV measurements of particle-laden forced turbulence.
J. Fluid Mech. In press
Ten Cate A, Derksen JJ, Portela L, van den Akker HEA. 2004. Fully resolved simulations of colliding monodis-
perse spheres in forced isotropic turbulence. J. Fluid Mech. 519:233–71
Tio KK, Linan A, Lasheras JC, Ganan-Calvo AM. 1993. On the dynamics of buoyant and heavy particles in
a periodic Stuart vortex flow. J. Fluid Mech. 254:671–99
Uhlherr PHT, Sinclair CG. 1970. The effect of free-stream turbulence on the drag coefficients of spheres.
Proc. Chem. 1:1–13
Uhlmann M. 2008. Interface-resolved direct numerical simulation of vertical particulate channel flow in the
turbulent regime. Phys. Fluids 20:053305
Vance MW, Squires KD, Simonin O. 2006. Properties of the particle velocity field in gas-solid turbulent
channel flow. Phys. Fluids 18:063302

132 Balachandar · Eaton

 ANRV400-FL42-06 ARI 17 November 2009 10:20

Wang L-P, Maxey MR. 1993a. Settling velocity and concentration distribution of heavy particles in homoge-
neous isotropic turbulence. J. Fluid Mech. 256:27–68
Wang L-P, Maxey MR. 1993b. The motion of microbubbles in a forced isotropic and homogeneous turbulence.
Appl. Sci. Res. 51:291–96
Wang L-P, Wexler AS, Zhou Y. 2000. Statistical mechanical description and modeling of turbulent collision
of inertial particles. J. Fluid Mech. 415:117–53
Wang Q, Squires KD. 1997. Large eddy simulation of particle-laden turbulent channel flow. Phys. Fluids
8:1207–23
Warnica WD, Renksizbulut M, Strong AB. 1995. Drag coefficients of spherical liquid droplets. Part 2: Tur-
bulent gaseous fields. Exp. Fluids 18:265–76
Wen F, Kamalu N, Chung JN, Crowe CT, Troutt TR. 1992. Particle dispersion by vortex structures in plane
mixing layers. ASME J. Fluids Eng. 114:657–66
Wicker RB, Eaton JK. 2001. Structure of a swirling, recirculating coaxial free jet and its effect on particle
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

motion. Int. J. Multiphase Flow 27:949–70

Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Wood AM, Hwang W, Eaton JK. 2005. Preferential concentration of particles in homogeneous and isotropic
turbulence. Int. J. Multiphase Flow 31:1220–30
Wu JS, Faeth GM. 1994. Sphere wakes at moderate Reynolds numbers in a turbulent environment. AIAA J.
32:535–41
Yamamoto Y, Potthoff M, Tanaka T, Kajishima T, Tsuji Y. 2001. Large eddy simulation of turbulent gas-
particle flow in a vertical channel: effect of considering interparticle collisions. J. Fluid Mech. 442:303–34
Yang CY, Lei U. 1998. The role of the turbulent scales in the settling velocity of heavy particles in homogeneous
isotropic turbulence. J. Fluid Mech. 371:179–205
Yang TS, Shy SS. 2003. The settling velocity of heavy particles in an aqueous near-isotropic turbulence.
J. Fluid Mech. 526:171–216
Zeng L, Balachandar S, Fischer P, Najjar FM. 2008. Interactions of a stationary finite-sized particle with wall
turbulence. J. Fluid Mech. 594:271–305
Zeng L, Balachandar S, Najjar FM. 2009. Wake response of a stationary finite-sized particle in a turbulent
channel flow. Int. J. Multiphase Flow Under consideration

www.annualreviews.org • Turbulent Dispersed Multiphase Flows 133

 AR400-FM ARI 13 November 2009 15:33

Annual Review of

Contents Fluid Mechanics

Volume 42, 2010

Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

Singular Perturbation Theory: A Viscous Flow out of Göttingen

Robert E. O’Malley Jr. p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 1
Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Dynamics of Winds and Currents Coupled to Surface Waves

Peter P. Sullivan and James C. McWilliams p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p19
Fluvial Sedimentary Patterns
G. Seminara p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p43
Shear Bands in Matter with Granularity
Peter Schall and Martin van Hecke p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p67
Slip on Superhydrophobic Surfaces
Jonathan P. Rothstein p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p89
Turbulent Dispersed Multiphase Flow
S. Balachandar and John K. Eaton p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 111
Turbidity Currents and Their Deposits
Eckart Meiburg and Ben Kneller p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 135
Measurement of the Velocity Gradient Tensor in Turbulent Flows
James M. Wallace and Petar V. Vukoslavčević p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 157
Friction Drag Reduction of External Flows with Bubble and
Gas Injection
Steven L. Ceccio p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 183
Wave–Vortex Interactions in Fluids and Superfluids
Oliver Bühler p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 205
Laminar, Transitional, and Turbulent Flows in Rotor-Stator Cavities
Brian Launder, Sébastien Poncet, and Eric Serre p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 229
Scale-Dependent Models for Atmospheric Flows
Rupert Klein p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 249
Spike-Type Compressor Stall Inception, Detection, and Control
C.S. Tan, I. Day, S. Morris, and A. Wadia p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 275

vii
 AR400-FM ARI 13 November 2009 15:33

Airflow and Particle Transport in the Human Respiratory System

C. Kleinstreuer and Z. Zhang p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 301
Small-Scale Properties of Turbulent Rayleigh-Bénard Convection
Detlef Lohse and Ke-Qing Xia p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 335
Fluid Dynamics of Urban Atmospheres in Complex Terrain
H.J.S. Fernando p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 365
Turbulent Plumes in Nature
Andrew W. Woods p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 391
Access provided by Indian Institute of Science - Bangalore on 05/20/17. For personal use only.

Fluid Mechanics of Microrheology

Annu. Rev. Fluid Mech. 2010.42:111-133. Downloaded from www.annualreviews.org

Todd M. Squires and Thomas G. Mason p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 413

Lattice-Boltzmann Method for Complex Flows
Cyrus K. Aidun and Jonathan R. Clausen p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 439
Wavelet Methods in Computational Fluid Dynamics
Kai Schneider and Oleg V. Vasilyev p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 473
Dielectric Barrier Discharge Plasma Actuators for Flow Control
Thomas C. Corke, C. Lon Enloe, and Stephen P. Wilkinson p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 505
Applications of Holography in Fluid Mechanics and Particle Dynamics
Joseph Katz and Jian Sheng p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 531
Recent Advances in Micro-Particle Image Velocimetry
Steven T. Wereley and Carl D. Meinhart p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 557

Indexes

Cumulative Index of Contributing Authors, Volumes 1–42 p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 577

Cumulative Index of Chapter Titles, Volumes 1–42 p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 585

Errata

An online log of corrections to Annual Review of Fluid Mechanics articles may be found
at http://fluid.annualreviews.org/errata.shtml

viii Contents