Smoothed Particle Diverse Application

You might also like

You are on page 1of 26

FL44CH14-Monaghan ARI 1 December 2011 8:2

ANNUAL
REVIEWS Further Smoothed Particle
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

Click here for quick links to


Annual Reviews content online,
including:
Hydrodynamics and Its
• Other articles in this volume
• Top cited articles
Diverse Applications
• Top downloaded articles
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

• Our comprehensive search


J.J. Monaghan
School of Mathematical Sciences, Monash University, Melbourne, Vic 3800 Australia;
email: joe.monaghan@monash.edu

Annu. Rev. Fluid Mech. 2012. 44:323–46 Keywords


The Annual Review of Fluid Mechanics is online at particle methods, fluids, elastic bodies, impact, surface tension, turbulence
fluid.annualreviews.org

This article’s doi: Abstract


10.1146/annurev-fluid-120710-101220
This review focuses on the applications of smoothed particle hydrodynamics
Copyright  c 2012 by Annual Reviews. (SPH) to incompressible or nearly incompressible flow. In the past 17 years,
All rights reserved
the range of applications has increased as researchers have realized the abil-
0066-4189/12/0115-0323$20.00 ity of SPH algorithms to handle complex physical problems. These include
the disruption of free surfaces when a wave hits a rocky beach, multifluid
problems that may involve the motion of rigid and elastic bodies, non-
Newtonian fluids, virtual surgery, and chemical precipitation from fluids
moving through fractured media. SPH provides a fascinating tool that has
some of the properties of molecular dynamics while retaining the attributes
of the macroscopic equations of continuum mechanics.

323
FL44CH14-Monaghan ARI 1 December 2011 8:2

1. INTRODUCTION
Smoothed particle hydrodynamics (SPH) is a method for obtaining approximate numerical solu-
tions of the equations of fluid dynamics by replacing the fluid with a set of particles (Gingold &
Monaghan 1977). The equations of motion and properties of these particles are determined from
the continuum equations of fluid dynamics by interpolation from the particles. The interpolant
can be constructed using analytical functions, and spatial derivatives of the interpolated quantities
can then be found using ordinary calculus. There is no need to use a grid, and the description of
free surfaces, however complicated, is trivial. The precise way the interpolant is constructed has
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

been discussed in previous reviews (Monaghan 1992, 2005). The present review concentrates on
the applications of SPH other than those in astrophysics as the latter have recently been reviewed
by Springel (2010). However, before considering the applications, it is useful to become familiar
with the equations of SPH.
Let us consider a set of SPH particles such that particle b has mass mb , density ρ b , and position
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

rb . The interpolation formula for any scalar or tensor quantity A(r) is an integral interpolant of
the form
  mb A(rb )
A(r) = A(r )W (r − r , h)dr  W (r − rb , h), (1)
b
ρb
where dr  denotes a volume element, and the summation over particles is an approximation to
the integral. The function W(q, h) is a smoothing kernel that is a function of |q| and tends to a
delta function as h → 0. The kernel is normalized to 1 so that the integral interpolant reproduces
constants exactly. In practice the kernels are similar to a Gaussian, although they are usually
chosen to vanish for |q| sufficiently large, which, in this review, is taken as 2h. As a consequence,
although the summations are formally over all the particles, the only particles b that make a
contribution to the density of particle a are those for which |ra − rb | ≤ 2h. For further details, the
reader is referred to a discussion of kernels given elsewhere (Monaghan 2005). Finally, we note
the important point that if the gradient of quantity A is required, we can write from Equation 1
  mb A (rb )
∇ A(r) = A (r )∇W (r − r , h) dr  ∇W (r − rb , h) . (2)
b
ρb

Equation 1 can be used to calculate the density by replacing A by the density ρ and by replacing
r by ra . This gives

ρa = mb W (ra − rb , h). (3)
b

An alternative way to calculate the density is to convert the continuity equation to SPH form. To
do this, we write the continuity equation in the form

= −ρ∇ · v = −∇ · (ρv) + v · ∇ρ, (4)
dt
and note from Equation 1 that the two terms on the right-hand side can be written in SPH form
as

∇ · (ρv) → mb vb · ∇W (r − rb , h) (5)
b

and

v · ∇ρ → v · mb ∇W (r − rb , h) . (6)
b

If these expressions are evaluated at particle a (that is, r and v are replaced by ra and va , respectively)
and the gradient is taken with respect to the coordinates of particle a, the rate of change of density

324 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

at particle a is given by

dρa 
= mb (va − vb ) · ∇a W (ra − rb , h) , (7)
dt b

where ∇ a denotes a gradient taken with respect to the coordinates of particle a. The nondissipative
SPH acceleration equation for a fluid can be derived from the continuum equations or by using a
Lagrangian (Gingold & Monaghan 1982, Bonet & Lok 1999, Monaghan 2005). Either way, the
acceleration of particle a in a single fluid without dissipation is given by
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

  
d va Pa Pb
=− mb + 2 ∇a Wab (h) + ga , (8)
dt b
ρa2 ρb

where Pa is the pressure at particle a, ga is an external force/mass, and Wab (h) denotes W (ra −rb , h).
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

If Wab (h) is a function of |ra − rb |, it is easy to show that the SPH acceleration equation conserves
linear and angular momentum when it should. In the absence of dissipation, it can be shown
from the SPH Lagrangian that the SPH equations conserve a discrete form of the circulation
(Monaghan 2005).
The acceleration depends on the pressure, which, for a weakly compressible fluid, is usually
specified by an equation of state of the form
 γ 
ρ0a c a2 ρa
Pa = −1 , (9)
γ ρ0a

where ρ0a is the reference density of the fluid and ca its speed of sound (MacDonald 1966). In
the calculations to be described here, γ = 7. The common description of Equation 9 as the Tait
equation is incorrect. Other equations of state for liquids are discussed by Hayward (1967), but
in the pressure-density domain we consider, all these equations of state give similar results. An
approximate estimate of the relative variation in density δρ/ρ of a compressible fluid in a flow with
Mach number M is ∼M 2 . If the speed of sound ca is ∼10V, where V is the maximum speed of the
fluid relative to the boundaries, we can expect δρ/ρ ∼ 0.01, and this estimate proves reasonable
for
 most problems. For example, in a dam break with dam height D, this speed can be estimated as
2g D. In some problems there are no boundaries, for example, in the two-dimensional inviscid
evolution of a circular drop with initial radius R to an extended ellipse. If the x and y components of
the velocity are B(t)x and −B(t)y, respectively, good results are found when the speed is estimated
as |R B(0)|. In either case, the estimate of the maximum speed can be checked by monitoring the
variations in the density. For explicit time stepping, usually based on a second-order symplectic
method, the time step is normally controlled by the CFL (Courant, Friedrichs, Lewy) condition,
but for highly viscous flows, it is controlled by the viscosity.
Viscous fluids can be simulated by adding a term ab to the pressure terms in Equation 8. A
typical form for ab is
α c¯ab vab · rab
ab = − , (10)
ρ̄ab |rab |
where the notation c¯ab = 12 (c a + c b ) for scalars, and for vectors vab = va − vb , has been used
(Monaghan 1997). The continuum limit of this viscosity (Monaghan 2005) shows that, for the
Wendland kernel, the kinematic viscosity ν is

1
ν= αc h. (11)
8

www.annualreviews.org • SPH and Its Diverse Applications 325


FL44CH14-Monaghan ARI 1 December 2011 8:2

The expression in Equation 10 can therefore be written using ν. To take account of large differences
in viscosity, the following form of  is preferable:
16νa νb vab · rab
ab = − . (12)
(νa ρa + νb ρb ) h̄ ab |rab |
This form of the viscous term is similar to the heat-conduction term when there are discontinuous
changes in the thermal conductivity (Cleary & Monaghan 1999).
The motion of nearly incompressible fluids usually takes place in regions with rigid boundaries,
which may be fixed or moving. To include these boundaries in an SPH simulation, researchers use
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

various techniques. These include ghost particles, which are images of the fluid particles, layers
of fixed fluid particles, boundary force particles (Monaghan & Kajtar 2009), and forces calculated
from a Lagrangian formulation allowing for the kernel to be incomplete near a rigid boundary
(Feldman & Bonet 2007). In the case of ghost particles, the images may be obtained by reflection
across a boundary (with a suitable change of velocity to produce no-slip or free-slip boundaries)
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

or by reflection in a set of particles fixed on the boundary (Ferrari et al. 2009). An approach in the
graphics industry (Ihmsen et al. 2010) involves boundary particles that are used to adjust the veloc-
ity of fluid particles near the wall according to friction and collision rules combined with a pressure.
These are calculated as if the boundary particles were fluid particles. This technique has not been
tested as carefully as other methods because, in the graphics industry, appearance is more im-
portant than accuracy. For simple two-dimensional simulations, the ghost particles can give good
results, but when the boundaries change sharply, the placement of ghost particles must ensure that
their density does not become artificially large or small. Although this may be relatively easy in two
dimensions, it is much more difficult to handle in three dimensions with complicated surfaces. A
related problem occurs when there are two fluids and the ghost particles of the fluids become mixed
unphysically. The common belief that ghost particles are problem independent is wrong because
the way they are placed varies with the geometry of the boundaries, and whether the liquid is ho-
mogeneous. By comparison, the boundary force particles can be applied easily, but they can lead to
more noise than the ghost particles, and they involve a parameter that is problem dependent. The
method of Feldman & Bonet (2007) involves a correction arising from the kernel estimates being
incomplete when a particle has a nearby boundary. Although the method is elegant, its use becomes
complicated in three dimensions when there are moving rigid bodies of arbitrary shape in the fluid.
Other methods (Becker et al. 2009) make use of special time-stepping rules near the boundary,
but these have not been fully tested. The conclusion is that no method is entirely satisfactory.
If boundary force particles are used to define the rigid boundaries, the SPH equation of motion
takes the form
   
d va Pa Pb

=− mb + +  ab ∇a W ab + fa j − m j a j ∇a W a j , (13)
dt b
ρ 2
a ρ 2
b j

where j denotes the label of a boundary force particle, and faj denotes a boundary force per unit
mass on particle a due to particle j. This force is along their line of centers. This equation is
an SPH equivalent of the Sirovich (1967) formulation of the Navier-Stokes equation in which
forces are introduced as an alternative to specifying boundary conditions on the rigid bodies. The
Navier-Stokes equation is then
d vi 1 ∂σi j 1
= + σi j n j δ(s ), (14)
dt ρ ∂x j ρ
where σi j is the stress tensor, the function δ(s ) is a one-dimensional delta function, and s is the
perpendicular distance from the surface to the position where the fluid acceleration is required.
The unit vector nj is directed from the surface into the fluid. The last term of Equation 13 is the

326 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

SPH equivalent of the last term of Equation 14, which is also closely related to the immersed
boundary method of Peskin (1977). It is characteristic of the boundary force–particle approach
that the usual boundary conditions are not needed. The condition that the normal component
of the velocity must vanish at a rigid surface is taken care of by the boundary force faj , and the
condition that the tangential velocity vanishes for no-slip boundaries is taken care of by the viscous
interaction between the fluid and boundary particles. An alternative SPH acceleration equation
(Equation 16) is discussed in Section 3.
The reader will have noticed that Equation 13 is similar to the equations of molecular dynamics.
Recognizing this, but for a different system of particle-like equations, von Neumann (1963 [1944])
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

remarked, “The particle method is not only an approximation of the continuum fluid equations,
but also gives the rigorous equations for a particle system which approximates the molecular system
underlying, and more fundamental than the continuum equations.” One elementary example of
this behavior is an SPH shock simulation without viscosity. The SPH particles will generate
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

disordered motion as they pass through the shock in order to provide the equivalent of a jump in
the thermal energy. A more subtle example may occur in the case of fluid dynamics with different
fluids in which diffusion is allowed between the fluids. The complete equation for the flux of
matter involves not only the gradient of the chemical potential, but also the temperature gradient
(Landau & Lifshitz 1993, p. 231). Similarly, the complete equation for the thermal flux involves
not only the temperature gradient, but also the flux of matter. In the latter case, the equation
of heat transfer is often approximated by neglecting the flux of matter, but the SPH equations
approximating it may mimic the neglected term.

2. DAM BREAKS AND PLUNGING WAVES


A dam break was simulated in the first paper describing the application of SPH to weakly com-
pressible fluids (Monaghan 1994). This problem has since been simulated by many authors (e.g.,
Colicchio et al. 2002, Colagrossi  & Landrini 2003, Colagrossi 2004). In the present case, for a
dam of height H, we take V = g H , where g is the gravitational acceleration, and H is also the
initial height of the water in the dam. For this simulation, the particles were placed initially on
the vertices of a lattice of squares and given a mass m = ρ0 (d p)2 , where dp is the initial separation
of the particles.
Figure
 1 shows the initial stages of a dam break calculated for a Reynolds number R =
0.5D g H /ν = 200. The SPH particles were initially placed on a grid of squares. The length

a b
High

Speed

Low

Figure 1
Dam break in a two-dimensional tank 1.5 m long. The fluid initially occupies the region 1.0 < x < 1.5 and
0 < y < 0.5. The color coding is for local speed.

www.annualreviews.org • SPH and Its Diverse Applications 327


FL44CH14-Monaghan ARI 1 December 2011 8:2

a b
High

Speed
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

Low

Figure 2
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

A continuation of the flow in Figure 1. The fluid runs up the wall, disrupting and subsequently returning as
a plunging wave.

and height of the dam are equal, and the number of particles is 75 × 75. The calculation uses
a Wendland (1995) cubic kernel for two dimensions, with h = 1.5d p, where dp is the initial
particle spacing. In the absence of the left-hand wall, the flow continues and can be compared
with the experiments of Martin & Noyce (1952). The agreement (Monaghan 1994) is good, and
the differences that do exist appear to largely result from timing errors in the experiment. The
flow hits the left wall of the tank, as shown in Figure 2, and the upward-moving stream breaks
up, returns, and forms a plunging wave, which bounces from the nearly horizontal incoming
fluid. The breakup of the upward stream is seen in experiments (Issa 2004), although the details
require the inclusion of surface tension, which has been neglected here. The subsequent bouncing
of the plunging wave is more chaotic in experiments because the flow up the wall is affected by
the corners, which are absent from this two-dimensional simulation. A further difference between
the experiments and this simulation is the effect of air pressure on the plunging wave. Many SPH
simulations have been carried out for waves and related phenomena (see, e.g., Dalrymple & Rogers
2006 and the bore calculations of Landrini et al. 2007). In engineering simulations, a quantity of
interest is the pressure exerted on the walls by the fluid. If boundary force particles are used, the
pressure is the component of the force on the boundary particles normal to the boundary. These
pressures show high-amplitude, short-frequency oscillations, which can be reduced by averaging
over neighboring particles. The oscillations are reduced if a fully incompressible method is used.
Reduction of the oscillations by averaging over a sequence of time steps has not been studied. A
related problem common in industry is the simulation of injection molding using liquid metals.
This problem has been tackled successfully using SPH (Cleary et al. 2002).

3. GRAVITY CURRENTS AND OTHER MULTIFLUID PHENOMENA


In many problems, two or more fluids flow together. In the case of gravity currents, in which
the flow is driven by the density difference between the fluids, the density ratio is usually close
to 1, although one exception is the experimental work of Gröbelbauer et al. (1993), who studied
gaseous gravity currents with density ratios up to 26.5, and another is the case of breaking waves
in which the two fluids are water and air with a typical density ratio of 700. If the density ratio is
≤2, the equations described above can be used. Each fluid can be given its own equation of state,
although the simplest approach is to keep the equation of state the same and change the speed of
sound and the reference density.

328 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

If the density ratio is 2, then standard SPH, such as that described above, gives poor results,
and various, often complicated, changes have been made to the fundamental equations to get a
viable algorithm. For example, Colagrossi & Landrini (2003) and Colagrossi (2004) found that for
stable SPH quasi-incompressible simulations of two fluids with large density ratios, it was necessary
to introduce the following changes to their standard SPH algorithm: density renormalization (a
process designed to correct the density obtained through the continuity equation), a large surface
tension term in the low-density phase, a smoothing of the velocity field, and a very large difference
in the speed of sound in the two liquids. The first of these increases the computing time, although
whether it is significant depends on the number of time steps between renormalizations. The
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

second is unphysical because surface tension in a liquid is larger than in a gas. Velocity smoothing
has a stabilizing effect, but it also increases the computing time. The ratio of the speed of sound in
the low-density fluid (subscript g) to that in the high-density liquid (subscript
) was approximately

ρ
γg /(ρg γ
). For a density ratio 1,000:1, and taking the values of γ used by Colagrossi (2004),
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

this factor is ∼14, so the speed of sound in the gas is much greater than that in the liquid, which is
already a factor of 10 greater than the maximum speed of flow to keep the density fluctuations small.
Although this is unphysical, because the speed of sound in water is greater than that in air at room
temperature and pressure, it must be kept in mind that, in these weakly compressible calculations,
the speed of sound assigned to the liquid is chosen to ensure that the density fluctuations are
sufficiently low. For typical simulations, the speed of sound required for this is a factor of ∼50 less
than the true speed of sound in the liquid. In any case, Colagrossi & Landrini (2003) found that
the higher speed of sound in the gas was necessary to stabilize their algorithm. A consequence of
this high speed of sound is that the CFL time-step condition requires very small time steps and
very long calculation times. However, with these various devices, the results for dam breaks and
sloshing with two fluids were quite good.
Because of the discontinuity in the density in multifluid simulations, Hu & Adams (2006)
rewrote the SPH equations in terms of particle number density, which can easily be made contin-
uous across a density discontinuity. They applied the resulting algorithm to problems involving
surface tension and obtained good results, but their method was not applicable to free-surface
problems. In later papers, Hu & Adams (2007) applied a projection method to handle a system
with one or more incompressible fluids. This method was successful for incompressible fluids
without free surfaces, but it is not applicable to problems such as an expanding buoyant bubble
interacting with a free surface. In a recent paper, Grenier et al. (2009) proposed a new SPH-based
method for multiphase problems. The algorithm requires sweeps over the particles to determine
the particle volume distribution, the density, the rate of the change of volume (continuity equa-
tion), and the acceleration. The time-stepping algorithm is a fourth-order Runge-Kutta. The
algorithm therefore makes heavy demands on computer time. The following rather simple algo-
rithm ( J.J. Monaghan & A. Rafiee, manuscript submitted) is capable of handling the density ratios
that normally occur. First, the continuity equation is changed to

dρa  mη
= −ρa (vη − va ) · ∇a W (ra − rη , h) (15)
dt η
ρη

because the term multiplying −ρa on the right-hand side of Equation 15 is an SPH expression
for ∇ · v, which does not change if the ratio of mass to density of the contributing particles is
unchanged. Thus, if one of the fluids were replaced by a fluid with the same velocity field but
with double the density, the SPH particles would have double the mass, and the estimate of ∇ ·v
would be unchanged, which is correct. If boundary force particles are used, the summation in
Equation 15 is over the fluid and the boundary particles.

www.annualreviews.org • SPH and Its Diverse Applications 329


FL44CH14-Monaghan ARI 1 December 2011 8:2

1.0

0.8

0.6
y
0.4

0.2

0
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

0 1 2 3
x

Figure 3
The positions of the SPH particles for a gravity current of density 1,000 (blue) moving into a fluid of density
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

400 (red ). The head of the light current is different from that of the denser current, as seen in experiments
and predicted by theory.

Second, if the equations of motion are derived from a Lagrangian, with the continuity equation
as a constraint, the acceleration equation becomes
   
d va Pa + Pη
=− mη + aη + Raη ∇a W aη (h) + fa j + ga . (16)
dt η
ρa ρη j

The fact that the pressure terms in the acceleration equation should vary with the form of the
continuity equation is also a feature of finite-difference methods. The term involving faj is required
only if the fluid interacts with rigid boundaries defined by boundary force particles. The summation
over j is over the boundary particles.
The term Raη , equivalent to an extra pressure between the fluids, is required to handle a very
large density ratio and is due to Grenier et al. (2009). It applies only when the two particles
interacting come from different fluids and therefore acts in a band of width 2h on either side of
the interface between the fluids. The following form of Raη is due to J.J. Monaghan & A. Rafiee
(manuscript submitted) and is a variation of that used by Grenier et al. (2009):
 
ρd − ρ
Pa + Pη
Raη = 0.08 , (17)
ρd + ρ
ρa ρη
where ρ d is the reference density of the denser fluid, and ρ
is the reference density of the lighter
fluid. The factor 0.08 was obtained from numerical experiments, but it should be possible to
determine it by a stability analysis of the SPH equations.
Figure 3 shows the particle positions in an SPH simulation of a gravity current similar to one
of the simulations of Birman et al. (2005). The density ratio is 2.5, the tank is 3.0 m long, and the
fluid is initially 0.5 m deep. The number of particles used was 600 × 100, but reasonable results
are found with 300 × 50 particles. The denser fluid initially occupied the left half of the tank
and the low-density fluid the right-hand side. The characteristic appearance of the fronts of the
denser and less dense fluids, including the fluctuations in the velocity field, agrees with experiment
(Birman et al. 2005).
Figure 4 shows the velocity of the head of a dense current for a range of density ratios. The SPH
and Birman et al. (2005) simulation results are for a Reynolds number of 4,000. The experiments
are from Gröbelbauer et al. (1993) with Reynolds  numbers ranging from 10,000 to 100,000. The
particle velocities are scaled using the velocity g  H , where H is the initial depth of the fluid,
and g  = g(1 − ρ
/ρd ) is the effective gravity. The SPH simulation uses an initial particle spacing
of 0.01 of the depth, whereas the simulation of Birman et al. (2005) uses a cell width of 0.005

330 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

1.5

1.0

Velocity

0.5
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

0
0 0.2 0.4 0.6 0.8 1.0
Density ratio
Figure 4
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

The scaled speed of the head of a dense current against density ratio. The blue symbols and line denote the
SPH results. The red stars denote results from the simulation of Birman et al. (2005), and the open purple
squares are results from the experiments of Gröbelbauer et al. (1993).

of the depth. On this basis, and because it uses a combination of a spectral method and a high-
order finite-difference scheme, the results might be expected to be more accurate than the SPH
simulation with lower resolution, but this is not the case. Both sets of simulations give higher
speeds than observed in the experiments, which may result from the viscous interaction with the
walls of the laboratory tank. We can conclude from these results that SPH is capable of giving
good results for two fluid problems.
Experiments involving three currents with different densities can be simulated easily using
SPH. All that is required is to specify the reference density of each fluid and choose an equation of
state. These problems are relevant to the generation of tsunamis by pyroclastic flows (Monaghan
et al. 1999) and to the situation in which rivers from northern Russia drain into the Arctic Sea
(Wells & Wettlaufer 2005). An example of such a flow is shown in Figure 5, which is similar to
figures 18 and 19 of Monaghan et al. (1999) in which dense fluid from a lock runs down a ramp
into a stratified tank. The fluid in the lock has a density of 1,210, that in the upper tank has density
of 1,000, and that in the lower tank has density of 1,070. A simulation of this kind can be applied

1.0
Density 1,000
0.8
Density 1,210
0.6
y
0.4

0.2
Density 1,070
0
0 1 2 3
x

Figure 5
A three-fluid SPH simulation with an upper fluid (density of 1,000), lower fluid (density of 1,070), and lock
fluid (density of 1,210). This type of simulation is required for problems involving rivers, or other currents,
injecting fluid into a stratified sea.

www.annualreviews.org • SPH and Its Diverse Applications 331


FL44CH14-Monaghan ARI 1 December 2011 8:2

to mixing in the basin-filling experiments of Wells & Wettlaufer (2005) in which the Reynolds
number is large and to the analysis of mixing in highly viscous fluids (Robinson et al. 2008).
The graphics industry simulates bubbles in fluids using particle methods producing realistic
multifluid motion (Ihmsen et al. 2010, 2011). Although these methods have not been subjected to
careful tests, the ideas are interesting and may have applications in scientific simulations.

4. BODIES MOVING IN FLUIDS


Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

4.1. Landslides and Bodies Impacting Fluids


Some of the most fascinating problems in fluid mechanics occur when the fluid interacts with a
moving body. SPH has been used for many configurations involving moving bodies and fluids.
These include sloshing tanks (Delorme et al. 2009, Rafiee et al. 2010), ship-hull impact (Veen
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

2010), and fish-like swimming (Kajtar & Monaghan 2008). All these problems involve coupled
rigid body and fluid motion. Here we consider motion in two dimensions and use Cartesian
coordinates. The equation of motion of the fluid, which is assumed incompressible, is identical to
Equation 16 if the boundaries are defined by boundary force particles.
The motion of the center of mass R(k) of solid body k [with mass M(k)] is given by

d 2 Ri (k)
M (k) = − σ i j n j (k)dA(k). (18)
d t2
The rotation of rigid body k [with moment of inertia I(k)] is given by

d 2 θk
I (k) 2 = (dk × b)dA(k), (19)
dt
where dk is a vector from the center of mass of body k to the element of area dAk on the body’s
surface, and b is the force per unit area. In the following, to simplify the notation, the subscript k
always indicates the body k. Thus, for example, R(k) is replaced by Rk .
The SPH rigid-body equations can be constructed easily. The fluid exerts forces on the body,
which are determined by the boundary forces. The nonviscous force on boundary particle j due
to all fluid particles is
(nv)

fj = mj f ja , (20)
a

where f j a is the force per unit mass on boundary particle j due to particle a. The viscous force is
(v)
 
f j = −m j ma a j ∇ j W a j = m j ma a j ∇a W a j , (21)
a a

where we use the fact that ∇ j W a j = −∇a W a j . The total force on particle j is
(nv) (v)
fj = fj + fj . (22)

The equation for the center of mass motion of body k is then


d Vk 
Mk = fj, (23)
dt j ∈S k

where the labels j are from the set of boundary force particles defining the surface of body k. The
torque equation is
d k 
Ik = (r j − Rk ) × f j . (24)
dt j ∈Sk

332 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

High

Speed

Low

Figure 6
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

SPH particles showing the violent disruption of the body of fluid due to the impact of a box-like body. Some
fluid particles were forced under the box. The color coding is for speed. This type of simulation has been
used in an analysis of experiments of the kind shown in Figure 7.
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

The summations are over those values of boundary particle labels j belonging to body k. The
motion of a boundary particle can be determined from the motion of the center of mass and the
rotation about the center of mass. Thus, for particle j on body k,
drj
= Vk + k ẑ × (r j − Rk ), (25)
dt
where, for the two-dimensional examples, the rotation is around the z axis, which is perpendicular
to the plane of the motion.
A simple example is shown in Figures 6 and 7 in which a rigid box slides down a ramp into
a fluid. The setup is similar to the experiments of Monaghan et al. (2003). The SPH simulation
uses a Reynolds number of 1,000 and 22,540 particles. The depth is equal to 100 particle spacings.
After an initial phase, the impact generates a solitary wave with height that can be related to the
impact speed of the block. In the simulation and the experiment, fluid can force its way under the
block, causing it to rotate slightly.

4.2. Fish-Like Swimming


SPH has been applied to the fish-like motion of linked bodies moving in a fluid. The first simula-
tions (Kajtar & Monaghan 2008, 2010) involved three similar, linked ellipses moving in an infinite
fluid. These were compared with the calculations of Kanso et al. (2005) for an ideal fluid and
those of Eldredge (2008) for moderate Reynolds number. As in the calculations of the previous
section, SPH fluid and boundary force particles were used to determine the forces in the fluid and
on the bodies, and the dynamics of the entire system can then be determined from a Lagrangian.

2.5-cm
grid

Figure 7
A photograph from an experiment in which a weighted box slides down a ramp into a tank (Monaghan et al.
2003). Note the dramatic plunging wave, which is also seen in Figure 6.

www.annualreviews.org • SPH and Its Diverse Applications 333


FL44CH14-Monaghan ARI 1 December 2011 8:2

High

Speed

Low

Figure 8
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

An eel-like swimmer, consisting of three rigid ellipses connected by an elastic membrane, swimming near a
free surface. Although the swimmer disrupts the surface, the efficiency of the motion is higher than that at
greater depth. The color coding denotes speed.

However, in the case of the linked bodies, there are constraints that Kajtar & Monaghan (2008)
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

included via Lagrange multipliers. The results of the SPH simulations are in agreement with
scaling relations for the speed and the power generated, and the results agree with those of Kanso
et al. (2005) and Eldredge (2008). A characteristic feature of SPH is that it can accommodate
complicated physics, and in the present case, the bodies can be connected by an elastic skin using
SPH skin particles, which are attached to themselves and the bodies by linear elastic forces. The
simulation therefore involves fluid SPH particles, boundary force SPH particles, and elastic-skin
SPH particles. Other types of particles can be added as required.
The SPH algorithm can be applied to fish-like swimming near and through a free surface
(Kajtar & Monaghan 2010), as shown in Figure 8 in which an eel-like swimmer, built from three
identical ellipses connected by an elastic skin, moves close to a surface. In this simulation, the
upper surface is free, the lower boundary is rigid, and the left and right boundaries are periodic.
In Figure 9, the swimmer is the same, but the fluid is stratified. In this case, the swimmer leaves
behind a trail of vortices, and the interface between the two fluids is disrupted. More complicated
problems such as the dynamics of a predator attacking prey, or a fish following a scent, can also
be simulated easily using SPH, although in these cases models must be introduced to mimic how
the fish responds to sight and smell.

High

Speed

Low

Figure 9
The same eel-like swimmer as in Figure 8, but now the swimmer is in the dense lower layer of a stratified
fluid. The color coding is as in Figure 8 except that the darker green has been overlaid to show the position
of the interface between the two fluids. Note the trail of vortices behind the swimmer and the disruption to
the interface.

334 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

5. NON-NEWTONIAN FLUIDS
Fluids with properties similar to mud or molten chocolate can be described by the Navier-Stokes
equation with a viscosity dependent on the shear. Defining the shear-strain rate tensor Dij by
 
1 ∂vi ∂v j
Di j = + , (26)
2 ∂xj ∂ xi
the shear viscosity coefficient μ is a function of the invariant
 
D= Di j Di j . (27)
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

i j

The velocity gradients required for Dij can be calculated using the following SPH formula:
 i  mb (v i − v i ) ∂ W ab
∂v
= b a
.
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

(28)
∂x a
j
b
ρ b ∂ xa
j

A typical example is the Cross (1965) viscosity, which has the form μ = μ0 /(1 + K Dm/n ), where
m/n equals 2/3 or 4/5 (the rational fraction must have an odd denominator and even numerator
so that Dm/n is an even function of D). The Bingham viscosity requires the shear to vanish when
the stress is below a critical value τ c . Shao & Lo (2003) and Hosseini et al. (2007) have simulated
several flows with the Bingham and related viscosities using SPH. Instead of using the condition
for the normal Bingham viscosity, they use the following prescription: When D < τc /(2αμ∞ ), the
viscous stress is 2αμ∞ Di j , where α ∼ 100. Otherwise, τi j = (τc /D + 2μ∞ )Di j . These problems can
be easily solved using the weakly compressible model (Capone 2009). When there are two fluids
with very different viscosities, it is necessary to write the viscous term in the form of Equation 12.

6. SURFACE TENSION
There are at least three ways surface tension can be included within an SPH algorithm, and all
three have been tried. First, the surface force per unit mass given by
σ κn δ(s )
(29)
ρ
can be approximated. In this expression, σ is the surface-tension coefficient; n is the unit outward
normal; κ = −∇ ·n is the average curvature; and δ(s ) is a one-dimensional delta function, which
is a function of the distance s perpendicular to the surface. The approximation consists of finding
the normal to the surface of interest by assigning a color C to each fluid (Brackbill et al. 1992) and
finding the gradient of the color function from which n = ∇C/|∇C|. For example,
 mb (Cb − Ca )
∇Ca = ∇a Wab (h). (30)
b
ρb

The curvature can then be calculated by an SPH form of the divergence of n. Although reasonable
results have been obtained with this method (Morris 2000), high accuracy appears to require a
better algorithm rather than more particles. Hu & Adams (2006) used the number density instead
of the mass density to write the pressure and other gradient terms, and they used alternative forms
of the color gradient (Hu & Adams 2006, Adami et al. 2010). In this way, they obtained good
results for drops that either oscillate or are in a shear flow.
Second, the surface-tension equations follow by including an extra energy term in the particle
Lagrangian for SPH flows (Monaghan 1995, Morris 2000). Thus Monaghan (1995) writes the

www.annualreviews.org • SPH and Its Diverse Applications 335


FL44CH14-Monaghan ARI 1 December 2011 8:2

Lagrangian in the form


  
1 2 σ  mb
L= mb vb − u(ρb ) − |∇C|, (31)
b
2 C b ρb
where, as above, C is the color variable. Taking the above form for ∇C and, for convenience, C =
1, one can show that the acceleration equation will then contain an extra force/mass for particle a:
σm 
− 2 [nab Fab + (nab · rab )∇a Fab ]. (32)
ρa b

Here, rF (r) = ∇W (r, h). This force term conserves linear and angular momentum and reproduces
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

Equation 29 when the continuous limit is taken.


Third, a unique SPH approach is to think of the SPH particles as real physical particles and
introduce forces between them to mimic the effects of surface tension directly (Hunter 1992,
Tartakovsky & Meakin 2005a,b, Becker & Teschner 2007). For example, Tartakovsky & Meakin
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

(2005a) added the following force/mass between particles separated by a distance r:


 
3πr r
F(r) = s cos , (33)
2h r
for r/ h ≤ 1 and zero otherwise. Note that Tartakosky & Meakin (2005a,b) scaled the kernel so that
its nonzero domain is r ≤ h. F is repulsive for r/ h < 1/3 and attractive for 1/3 < r/ h < 1. The
parameter s is different for different fluids and is expected to be proportional to the surface tension.
Tartakosky & Meakin determined s by numerical experiments, e.g., the oscillation frequency of a
drop, or the pressure inside a drop. In principle, the relation between s and the surface tension could
be calculated by an integration over the fluid as in the original Laplace theory (see Rayleigh 1964).
The dependence on h could then be determined easily so that the surface tension could be specified
independently of resolution. Tartakosky & Meakin applied their surface-tension model to a series
of problems involving drops, and fluid moving through fractures, and obtained satisfactory results.

7. DIFFUSION AND PRECIPITATION


The efficient solution of diffusion equations such as the heat-conduction equation is fundamental
for dissipative processes. An advantage of the SPH equations for these dissipative problems is that
they can be written so that they mimic fundamental properties of the system and allow complicated
physics, such as phase changes, to be handled in a straightforward way. Appropriate forms of these
equations have been derived (Brookshaw 1985, Cleary & Monaghan 1999, Monaghan et al. 2005)
and applied to a wide variety of heat-conduction problems, including the Stefan problem and the
freezing of alloy solutions (Monaghan et al. 2005), and to problems involving radiative transfer in
the diffusion approximation (Whitehouse & Bate 2004).
A convenient form of the heat-conduction equation without heat sources or sinks is
dT 1
cp = ∇(κ∇T ), (34)
dt ρ
where T is the absolute temperature, cp the heat capacity per unit mass at constant pressure, ρ the
density, κ the coefficient of thermal conductivity, and d/dt the derivative following the motion.
Cleary & Monaghan (1999) derived the following SPH form of the heat-conduction equation:
d Ta  mb 4κa κb
c p,a = (T a − T b )Fab . (35)
dt b
ρa ρb (κa + κb )

Cleary & Monaghan (1999) showed that this SPH form of the heat-conduction equation had
similar accuracy to finite-difference methods and was not sensitive to the particle disorder that

336 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

occurs in some SPH calculations. In addition, heat-conduction problems with discontinuous κ, and
with κ varying with T, were accurately integrated. A slightly different κ term based on similar ideas
gives satisfactory results for jumps in κ by a factor of 109 (Parshikov & Medin 2002). Whitehouse &
Bate (2004) formulated the SPH transfer of radiation in the diffusion approximation and obtained
accurate results for test problems.
If the particles are thermally isolated (so they can only exchange heat among themselves), then
Equation 35 shows (noting Fab = Fba ) that the total heat content

ma c p,a T a (36)
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

is constant. When the system contains point sources or sinks of heat or matter, SPH allows a
simple effective formulation (Monaghan et al. 2005). The SPH heat-conduction and the salt-
diffusion equations predict an increase in entropy as they should, and they conserve heat and
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

matter (Monaghan et al. 2005).


Boundary conditions usually can be satisfied in a straightforward way if boundary particles
are used. If all the boundaries are adiabatic, the fluid particles exchange heat among themselves,
but not with boundary particles, and the symmetry of the SPH conduction equation ensures
that the system conserves its thermal energy. If one or more boundaries have fixed tempera-
tures, the SPH particles on those boundaries are included in the heat-conduction equation so
that the heat transferred to the boundary during a time step can be calculated. After this is
done, the temperatures of the boundary particles are set back to the specified boundary tem-
peratures for the next time step. If there are different fluids exchanging heat, the SPH particles
satisfy the previous equations, and the summations in Equation 35 are over all fluid particles.
The SPH treatment of the freezing is simple. Initially the SPH particles are assumed to be
liquid, tagged with an integer to denote liquid particles, and given the material properties of the
liquid. As heat is conducted from the liquid, some particles reach the solidification temperature
Tm . The heat per unit mass q lost by these particles after this time is then stored, and their
temperatures are kept at Tm . If particle a is in this condition then, when qa reaches the latent heat,
the integer tag is changed to that for the solid phase, and the properties of this phase (thermal
conductivity and heat capacity) are assigned to this particle. Between the solid particles and the
liquid particles, there is a region where the particles have reached the solidification temperature
but have not had their latent heat fully extracted. This is the mushy layer. An example of the SPH
solution of a one-dimensional Stefan problem is shown in Figure 10 (Monaghan et al. 2005).
The agreement between the SPH results and the exact result is excellent. The SPH simulation of
the freezing of a solution involves coupling the thermodynamics of the equilibrium state with the
heat-conduction equation and fluid motion. The SPH results are in satisfactory agreement with
experiment (Monaghan et al. 2005).
In related calculations, Tartakovsky et al. (2007) simulated precipitation using SPH. The typical
initial setup includes solid regions (represented by sets of SPH particles tagged as solid) and other
SPH particles representing the solution. The SPH equation for solute diffusion is supplemented by
a term representing the rate of precipitation. The results compare well with those from analytical
solutions and those from a lattice Boltzmann method.

8. ELASTICITY, FRACTURE, AND SOFT TISSUE


The elastic acceleration equation is
d vi 1 ∂σ i j
= , (37)
dt ρ ∂x j

www.annualreviews.org • SPH and Its Diverse Applications 337


FL44CH14-Monaghan ARI 1 December 2011 8:2

1.0

Temperature
0.5
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

0
0 0.2 0.4 0.6 0.8 1.0
y
Figure 10
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

Temperature plotted against distance from the cooling boundary for a two-dimensional Stefan problem.
The exact results are shown by the solid line and the SPH results by the solid diamonds. The change of slope
shows the interface between solid ice and liquid.

where i and j refer to Cartesian components, and σ i j = −Pδ i j + τ i j + S i j is the stress tensor
including the viscous τ ij and elastic Sij terms. If T i j = −Pδ i j + Si j , the acceleration equation can
be written in SPH form as

d vai  T ai j Tb
ij
∂ W ab
= mb + 2 + ab . (38)
dt b
ρ a
2 ρ b ∂ xi

The equations of elastic dynamics require an equation for S ij , which, in the simplest case in which
Hooke’s law holds, takes the form of a differential equation that is first order in time. These
equations were first applied by Libersky & Petschek (1990). They were subsequently used by
Benz & Asphaug (1994, 1995) together with a Grady-Kipp fracture model to study the impact and
fracture of asteroids. Benz & Asphaug showed that SPH gave outstanding results for the shattering
of a sphere of basalt struck by a projectile. Gray et al. (2001) applied the elastic equations together
with a technique to eliminate a tension instability (Monaghan 2000) to simulate bouncing balls
and oscillating beams (see also Bonet & Kulasegaram 2001, Vidal et al. 2007). This was extended
to the case of damage by Gray & Monaghan (2004) to study fracture around magma chambers.
These authors showed that the SPH algorithm is able to simulate the breakup of damaged beams,
the bouncing of rubber balls, and the oscillation of beams. These simulations emphasize the
advantages of SPH for this type of problem. First, the particles carry their damage with them,
whereas a finite-difference scheme would have to advect the damage. Second, the SPH code allows
a seamless passage from a continuous material to the thousands of fragments produced by impact.
As a result, the SPH simulations of this problem are the most accurate available. SPH has also
been used to simulate the collision of porous bodies in the early solar system ( Jutzi et al. 2008,
2009).
If the density varies significantly, the stress term should be changed according to
ij
σai j σij σ i j + σb
+ a2 → a . (39)
ρa2 ρa ρa ρb
There have been many applications of SPH to industrial problems involving elastic materials (see,
e.g., Cleary et al. 2006, Das & Cleary 2010). The SPH simulation of a fluid interacting with an
elastic body has also been studied by Antoci et al. (2007), who found that the simulations and
laboratory experiments are in good agreement. A related area of SPH application is the simulation

338 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

of the soft tissue that occurs in physiological problems (Hieber & Koumoutsakos 2008). In these
calculations, the linear Hooke’s law is replaced by a nonlinear function with parameters determined
from experiments with a liver. The application of SPH to other physiological problems is in
progress. One of these is the failure of lung function after long periods on artificial respirators.
The time-dependent behavior of a model lung is being studied by Xu, Adams, and Adami at the
Technical University, Munich. It is useful to note that researchers in the graphics simulation
industry treat elastic bodies by a number of techniques, one of which is the use of elastic rods
(Spillmann & Teschner 2009) that may be combined with SPH.
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

9. DUST, GRANULAR FLOW, AND GRANICLES


To simulate a dust storm or a volcanic pyroclastic flow, one must include particulate matter in
the SPH code. One way to do this is to follow Harlow & Amsden (1975) and treat the particulate
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

matter as a fluid that interacts with air or water by means of pressure and drag terms. This was
done by Monaghan & Kocharyan (1995), and the method has now been applied to the simulation
of dusty disks around stars (Barriére-Fouchet et al. 2005). The basic idea is to replace the dust
fluid by a set of dust SPH particles, each of which represents a large number of dust grains that,
together with the SPH particles for the gas or liquid, and SPH boundary particles for the terrain,
specify the system.
SPH can simulate the sedimentation of several dust species in a stratified fluid. Each fluid is
given its own set of particles with reference density and other parameters, as described above for
multifluid calculations. The different dust fluids involve different sets of SPH particles, and their
drag terms are different.
If it is desired to simulate a collapsing column of small grains, none of the previous algorithms
should be used because they do not include grain friction. A pile of grains simulated in this way
will collapse like a fluid. Some authors have treated soil-water interactions by giving the grain fluid
a very high viscosity (Ulrich & Rung 2010) with reasonable results. A uniquely SPH approach,
related to the discrete element method, is to replace each SPH particle by a set of particles that
move as a group that translates and rotates like a rigid body. They are called granicles (Capone
et al. 2008) and are related to the algorithms that use nonspherical particles (Pöschel & Buchholtz
1993). The shape of the granicles is irregular because it is determined by the linked set of particles,
which may be at the corners of a cube or any other shape. Combined with moderate viscosity,
granicles will form piles with a slope that is typical of granular material and varies with the shape
of the granicles. An example is shown in Figure 11 in which the granicles form a pile on a rough
boundary. A mixture of water SPH particles and granicles can mimic a Bingham fluid (Capone
et al. 2008). Further study of the behavior of granicles, especially how their properties relate to
those of actual granular material, would be interesting.

10. TURBULENCE
Although any of the standard turbulence models, such as the k −  model, can be added to the
SPH algorithm (Dalrymple & Rogers 2006, Violeau & Issa 2007), it seems preferable to construct
a turbulence model within the framework of SPH in such a way that general principles such as
conservation of energy, momentum, and circulation are satisfied. These aims can be achieved
by following the ideas of the Lagrangian-averaged Navier-Stokes alpha model (Holm 1999; for
further references, see Graham et al. 2007). A first attempt involved implicit smoothing (Monaghan
2002), but a variant with explicit smoothing is much more efficient (Monaghan 2011). Results from
the simulation of two-dimensional turbulence show that it reproduces the vorticity, energy decay,

www.annualreviews.org • SPH and Its Diverse Applications 339


FL44CH14-Monaghan ARI 1 December 2011 8:2

0.3

0.2

y
0.1
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

0
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

–0.2 0 0.2
x

Figure 11
An example of a pile of SPH granicles, each of which is a rigid body consisting of five particles, with the
central one marked by a circle. Note that the granicles have rotated relative to each other.

and velocity correlations of high-resolution calculations but with a factor-of-16 improvement in


speed.
The SPH turbulence model begins with smoothing. The smoothed velocity v̂a and the un-
smoothed velocity va for an SPH liquid particle a are related by
 mb
v̂a = va +  (vb − va )K (|ra − rb |,
), (40)
b
M

where mb is the mass of particle b, M is a mass closely related to the typical mass of a particle,
and 0 ≤  < 1 (Monaghan 2011). The function K is a smoothing function with typical length
scale
. The integral of K over the space of the simulation is equal to
d , where d is the number
of dimensions. The reader familiar with SPH will recognize this smoothing as the XSPH variant
of SPH (Monaghan 1989). It can be shown that, if the velocities are expanded in a Fourier series,
the values of coefficients with wave number greater than ∼ 1/
are reduced by a factor of |1 − |
by the smoothing (Monaghan 2011). Typical calculations use  = 0.8, so the coefficients of the
high-wave-number modes are reduced by a factor of five, and this occurs every time step. The
Lagrangian leading to the Euler equations is
  
1
L= mb v̂b · vb − u(ρb , s b ) , (41)
b
2

where ρ b is the density of particle b, u is the thermal energy/mass, and sb is its entropy. Particle
b moves with velocity v̂b . Writing L in terms of v̂ and r for each particle, it is straightforward,
although lengthy, to show that the Lagrange equations take the form
  
d vc Pc Pb   mb
=− mb + ∇c W c b + (vc − vb )2 ∇c K c b , (42)
dt b
ρ 2
c ρ 2
b 2 b
M

where P is the pressure, and the equation is the SPH equivalent of the Euler equations. To sim-
ulate the Navier-Stokes equations, one must add a standard SPH viscosity term. The continuum

340 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

1.0 1.0
a b

0.8 0.8

0.6 0.6

y y

0.4 0.4
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

0.2 0.2
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
x x

Figure 12
(a) The unsmoothed velocity field v and (b) the smoothed velocity field v̂ field for  = 0.9. The smoothing sharpens the velocity
contrast because it reduces noise.

equivalent of Equation 42 is the following Euler equation with an extra term due to the smoothing:


∂v 1  2
+ (v̂ · ∇) v = − ∇ P + ρ v (r) − v (r ) ∇ K (r − r ,
) dr . (43)
∂t ρ 2M

If the last term is approximated by expanding the velocity difference in a Taylor series and inte-
grated by parts to remove the derivative of K, the x component can be written

 
Fρ ∂v x 2 x ∂v y 2 y ∂
∇ v + ∇ v + [∇v x · ∇v x + ∇v y · ∇v y ] , (44)
2M ∂v ∂y ∂x

where

(x − x  ) K dr = (y − y  ) K dr = σ
4 ,
2 2
F= (45)

and σ is a nondimensional constant that depends on the smoothing kernel. We have assumed that
∇ ·v = 0, and the point r is sufficiently far from the boundaries that the integrations can be extended
over the entire domain for which K is nonzero. This term involving  is the equivalent of the
derivative of the turbulence stress tensor in LES models of turbulence and has the same structure
as the velocity derivative terms in the Lagrangian-averaged Navier-Stokes alpha equations (Holm
1999).
In recent calculations (Monaghan 2011), this model has been applied to two-dimensional
turbulence experiments and computations (see van Heijst et al. 2006 for extensive references).
Typical results of the smoothing with  = 0.9 are shown in Figure 12 5 s after the turbulence
was initiated from an initial state. The smoothing sharpens the velocity field by removing short
length-scale variations. Unpublished calculations explore the spin-up of fluid in a fixed box with
no-slip conditions and the changes that occur when the box is allowed to rotate freely.

www.annualreviews.org • SPH and Its Diverse Applications 341


FL44CH14-Monaghan ARI 1 December 2011 8:2

SUMMARY POINTS
1. Problems involving disrupted free surfaces, whether produced by wave breaking or rigid-
body impact, can be simulated easily using SPH, and the results are satisfactory. They
can be improved by including surface tension and more particles.
2. SPH can be used with little change for problems involving several fluids, whether they
be liquid, gas, or dust fluids.
3. SPH has advantages for the simulation of elastic fracture because the particles carry the
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

damage with them, and a transition from the continuum to fragments is seamless.
4. SPH gives accurate results for thermal and matter diffusion and chemical precipitation.
5. Applications of SPH to physiological problems involving soft tissue give satisfactory
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

results. It may be possible in the future to use SPH as a research tool in surgery.
6. Because of its flexibility, SPH is a major tool in the special effects and graphics industry,
encompassing video games, movies, and TV advertising. Graphics in which a liquid moves
naturally, then morphs into something bizarre, have almost certainly been created with
SPH.
7. SPH codes are easy to program, even with many fluids and arbitrary boundaries.
8. Although not discussed above, the application of SPH to astrophysical problems is
widespread because of its flexibility and accuracy.

FUTURE ISSUES
1. Most SPH calculations generate short-length-scale noise, which reduces the accuracy.
This can be reduced by remeshing (Chaniotis et al. 2002), although remeshing is non-
trivial when there is more than one phase. The new turbulence models may have a
considerable effect in reducing the noise with little cost.
2. Boundaries are currently treated in different ways, and all careful studies of their effec-
tiveness are in two dimensions with a single fluid. It would be useful to extend the work
to three dimensions in systems with more than one fluid.
3. In many SPH simulations, turbulence is generated. A suitable turbulence model, possibly
along the lines of that described above, may prove effective.
4. It would be highly desirable to devise algorithms that would allow the user to split the
particles to give greater resolution and combine them to increase efficiency. This would
enable thermal boundary layers to be simulated more accurately.
5. SPH algorithms are much slower than finite-difference calculations, although they are
easier to program. No algorithms that would substantially increase the speed have been
developed in the past few years, although the use of parallel clusters has allowed simula-
tions with up to 109 particles.

DISCLOSURE STATEMENT
The author is not aware of any biases that might be perceived as affecting the objectivity of this
review.

342 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

LITERATURE CITED
Adami S, Hu XY, Adams NA. 2010. A new surface tension formulation for multi-phase SPH using a repro-
ducing divergence approximation. J. Comput. Phys. 229:5011–21
Antoci C, Gallati M, Sibilla S. 2007. Numerical simulation of fluid-structure interaction by SPH. Comput.
Struct. 85:879–90
Barriere-Fouchet L, Gonzalez, Murray JR, Humble RJ, Maddison ST. 2005. Dust distribution in protoplan-
etary disks. Astron. Astrophys. 443:185–94
Becker M, Teschner M. 2007. Weakly compressible SPH for free surface flows. In Proc. ACM
SIGGRAPH/Eurographics Symp. Comput. Animat., ed. D Metaxas, J Popovic, pp. 209–17. San Diego:
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

ACM SIGGRAPH
Becker M, Tessendorf H, Teschner M. 2009. Direct forcing for Langrangian rigid-fluid coupling. IEEE Trans.
Vis. Comput. Graph. 15:493–502
Benz W, Asphaug E. 1994. Impact simulations with fracture. I. Method and tests. Icarus 107:98–116
Benz W, Asphaug E. 1995. Simulations of brittle solids using smooth particle hydrodynamics. Comput. Phys.
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

Commun. 87:253–65
Birman V, Martin JE, Meiburg E. 2005. The non-Boussinesq lock-exchange problem. Part 2. High-resolution
simulations. J. Fluid Mech. 357:125–44
Bonet J, Kulasegaram S. 2001. Remarks on the tension instability of Eulerian and Lagrangian corrected
smoothed particle hydrodynamics. Int. J. Numer. Methods Eng. 52:1203–20
Bonet J, Lok T-S. 1999. Variational and momentum preserving aspects of smoothed particle hydrodynamics
formulations. Comput. Methods Appl. Mech. Eng. 180:97–115
Brackbill JU, Kothe DB, Zemach CA. 1992. Continuum method for modeling surface tension. J. Comput.
Phys. 11:335–54
Brookshaw L. 1985. A method of calculating radiative heat diffusion in particle simulation. Proc. Astron. Soc.
Aust. 6:207–10
Capone T. 2009. SPH numerical modelling of impulse water waves generated by landslides. PhD thesis. Univ.
Rome, La Sapienza
Capone T, Kajtar JB, Monaghan JJ. 2008. SPH molecules: a model of granular material. Presented at SPHERIC
Workshop, 3rd, Lausanne, Switz.
Chaniotis AK, Poulikakos, Koumoutsakos P. 2002. Remeshed smoothed particle hydrodynamics of viscous
and heat conducting flows. J. Comput. Phys. 182:67–90
Cleary PW, Ha J, Alguine V, Nguyen T. 2002. Flow modelling in casting processes. Appl. Math. Model.
26:171–90
Cleary PW, Monaghan JJ. 1999. Conduction modeling using smoothed particle hydrodynamics. J. Comput.
Phys. 148:227–64
Cleary PW, Prakash M, Ha J. 2006. Novel applications of smoothed particle hydrodynamics (SPH) in metal
forming. J. Mater. Proc. Technol. 177:41–48
Colagrossi A. 2004. A meshless Lagrangian method for free-surface and interface flows with fragmentation. PhD
thesis. Univ. Rome, La Sapienza
Colagrossi A, Landrini M. 2003. Numerical simulation of interfacial flows by smoothed particle hydrodynam-
ics. J. Comput. Phys. 191:448–75
Colicchio G, Colagrossi A, Greco M, Landrini M. 2002. Free surface flow after a dam break. Ship Technol.
Res. 49:95–104
Cross M. 1965. Rheology of non-Newtonian fluids: a new flow equation for pseudoplastic systems. J. Colloid
Sci. 20:417–37
Dalrymple RA, Rogers BD. 2006. Numerical modeling of water waves with the SPH method. Coast. Eng.
53:141–47
Das R, Cleary PW. 2010. Effect of rock shape on brittle fracture using smoothed particle hydrodynamics.
Theor. Appl. Fract. Mech. 53:47–60
Delorme L, Coa A, Souto-Inglesias A, Zamora-Rodriguez R, Bolta-Vera E. 2009. A set of canonical problems
in sloshing. Part I. Pressure fields in a forced roll: comparison between experiments and SPH. Ocean Eng.
36:168–78

www.annualreviews.org • SPH and Its Diverse Applications 343


FL44CH14-Monaghan ARI 1 December 2011 8:2

Eldredge JD. 2008. Dynamically coupled fluid-body interations in vorticity-based numerical simulations.
J. Comput. Phys. 277:9170–94
Feldman J, Bonet J. 2007. Dynamic refinement and boundary contact forces in SPH with applications in fluid
flow. Int. J. Numer. Methods Eng. 72:295–324
Ferrari A, Dumbser M, Toro EF, Armanini A. 2009. A new 3D parallel SPH scheme for free surface flows.
Comput. Fluids 38:1203–17
Gingold RA, Monaghan JJ. 1977. Smoothed particle hydrodynamics. Mon. Not. R. Astron. Soc. 181:375–89
Gingold RA, Monaghan JJ. 1982. Kernel estimates as a basis for general particle methods in hydrodynamics.
J. Comput. Phys. 46:429–53
Graham JP, Holm DD, Miinni PD, Pouquet A. 2007. Highly turbulent solutions of the Lagrangian-averaged
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

Navier-Stokes α model and their large-eddy-simulation potential. Phys. Rev. E 76:056310


Gray J, Monaghan JJ. 2004. Numerical modelling of stress fields and fracture around magma chambers.
Volcanol. Geophys. Res. 135:259–83
Gray J, Monaghan JJ, Swift RP. 2001. SPH elastic dynamics. Comput. Methods Appl. Mech. Eng. 190:6641–62
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

Grenier N, Antuono M, Colagrossi A, Le Touze D, Alessandrini B. 2009. An Hamiltonian interface SPH


formulation for multi-fluid and free surface flows. J. Comput. Phys. 228:8380–93
Gröbelbauer HP, Fanneløp TK, Britter RE. 1993. The propagation of intrusion fronts of high density ratios.
J. Fluid Mech. 250:669–87
Harlow F, Amsden A. 1975. Numerical calculation of multiphase flow. J. Comput. Phys. 17:19–52
Hayward ATG. 1967. Compressibility equations for liquids: a comparative study. Br. J. Appl. Phys. 18:965–77
Hieber SE, Koumoutsakos P. 2008. A Lagrangian particle method for the simulation of linear and nonlinear
elastic models of soft tissue. J. Comput. Phys. 227:9195–215
Holm DD. 1999. Fluctuation effects on 3D Lagrangian mean and Eulerian mean fluid motion. Physica 133:215–
69
Hosseini S, Manzari M, Hannani S. 2007. A fully explicit three-step SPH algorithm for simulation of non-
Newtonian fluid flow. Int. J. Heat Fluid Flow 17:715–35
Hu XY, Adams NA. 2006. An incompressible multi-phase SPH method for macroscopic and mesoscopic flows.
J. Comput. Phys. 213:844–61
Hu XY, Adams NA. 2007. An incompressible multi-phase SPH method. J. Comput. Phys. 227:264–78
Hunter JP. 1992. Surface tension in smoothed particle hydrodynamics. Honours thesis. Math. Dep., Monash Univ.
Ihmsen M, Akinci N, Gissler M, Teschner M. 2010. Boundary handling and adaptive time-stepping for
PCISPH. Proc. VRIPHYS, ed. J Bender, K Erleben, M Teschner, pp. 79–88. Copenhagen: VRIPHYS
Ihmsen M, Bader J, Akinci G, Teschner M. 2011. Animation of air bubbles with SPH. Presented at GRAPP
2011, Algarve, Port.
Issa R. 2004. Numerical assessment of the SPH gridless method for incompressible flows and its extension to turbulent
flow. PhD thesis. Univ. Manchester Inst. Sci. Technol.
Jutzi M, Benz W, Michel P. 2008. Numerical simulations of impacts involving porous bodies I. Implementing
subresolution porosity in a 3D SPH hydrocode. Icarus 198:242–55
Jutzi M, Michel P, Hiraoka K, Nakamura AM, Benz W. 2009. Numerical simulations of impacts involving
porous bodies II. Comparison with laboratory experiments. Icarus 201:802–13
Kanso E, Marsden JE, Rowley CW, Melli-Huber J-B. 2005. Locomotion of articulated bodies in a perfect
fluid. J. Nonlinear Sci. 15:255–89
Kajtar JB, Monaghan JJ. 2008. SPH simulations of swimming linked bodies. J. Comput. Phys. 227:8568–87
Kajtar JB, Monaghan JJ. 2010. On the dynamics of swimming bodies. Eur. J. Mech. B Fluids 29:377–86
Landau LD, Lifshitz EM. 1993. Course of Theoretical Physics, Vol. 6: Fluid Mechanics. Oxford: Butterworth-
Heinemann
Landrini M, Colarossi A, Greco M, Tuline MP. 2007. Gridless simulation of splashing processes and near-
shore bore propagation. J. Fluid Mech. 591:183–213
Libersky LD, Petschek AG. 1990. Smooth particle hydrodynamics with strength of materials. In Advances in
Free-Lagrange Method, ed. C Trease, pp. 248–57. Lect. Notes Phys. 395. Berlin: Springer-Verlag
MacDonald JR. 1966. Some simple isothermal equations of state. Rev. Mod. Phys. 38:669–79
Martin JC, Noyce WJ. 1952. Part IV. Experimental study of the collapse of liquid columns on a rigid horizontal
plane. Philos. Trans. R. Soc. Lond. A 244:312–24

344 Monaghan
FL44CH14-Monaghan ARI 1 December 2011 8:2

Monaghan JJ. 1989. On the problem of penetration in particle methods. J. Comput. Phys. 82:1–15
Monaghan JJ. 1992. Smoothed particle hydrodynamics. Annu. Rev. Astron. Astrophys. 30:543–73
Monaghan JJ. 1994. Simulating free surface flows with SPH. J. Comput. Phys. 110:399–406
Monaghan JJ. 1995. An SPH formulation of surface tension. Appl. Math. Rep. Preprints, Monash Univ.
Monaghan JJ. 1997. SPH and Riemann solvers. J. Comput. Phys. 136:298–307
Monaghan JJ. 2000. SPH without a tensile instability. J. Comput. Phys. 159:290–311
Monaghan JJ. 2002. SPH compressible turbulence. Mon. Not. R. Astron. Soc. 335:843–52
Monaghan JJ. 2005. Smoothed particle hydrodynamics. Rep. Prog. Phys. 68:1703–59
Monaghan JJ. 2011. A turbulence model for smoothed particle hydrodynamics. Eur. J. Mech. B Fluids 30:360–
70
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

Monaghan JJ, Cas RF, Kos A, Hallworth M. 1999. Gravity currents descending a ramp in a stratified tank.
J. Fluid Mech. 379:39–69
Monaghan JJ, Huppert HE, Worster G. 2005. Solidification using smoothed particle hydrodynamics.
J. Comput. Phys. 206:684–705
Monaghan JJ, Kajtar JB. 2009. SPH particle boundary forces for arbitrary boundaries. Comput. Phys. Commun.
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

180:1811–20
Monaghan JJ, Kocharyan A. 1995. SPH simulation of multiphase flow. Comput. Phys. Commun. 87:225–35
Monaghan JJ, Kos AM, Issa N. 2003. Fluid motion generated by impact. J. Waterw. Ports Coast. Ocean Eng.
129:250–59
Morris JP. 2000. Simulating surface tension with smoothed particle hydrodynamics. Int. J. Numer. Methods
Fluids 33:333–53
Parshikov AN, Medin SA. 2002. Smoothed particle hydrodynamics using an interpolation contact algorithm.
J. Comput. Phys. 180:358–82
Peskin CS. 1977. Numerical analysis of blood flow in the heart. J. Comput. Phys. 25:220–52
Poschel T, Buchholtz V. 1993. Static friction phenomena in granular materials: Coulomb law versus particle
geometry. Phys. Rev. Lett. 71:3963–66
Rafiee A, Thiagarajan K, Monaghan JJ. 2010. Numerical study of density ratio effects and compressibility of gas phase
in sloshing. Presented at Aust. Fluid Mech. Conf., 17th, Auckland
Rayleigh L. 1964. On the theory of surface forces. In Collected Papers, Vol. 3, Art. 176, pp. 397–425. New York:
Dover
Robinson M, Cleary P, Monaghan JJ. 2008. Analysis of mixing in a Twin Cam mixer using smoothed particle
hydrodynamics. AIChE J. 54:1987–98
Shao S, Lo EYM. 2003. Incompressible SPH method for simulating Newtonian and non-Newtonian flows
with a free surface. Adv. Water Resour. 26:787–800
Sirovich L. 1967. Initial and boundary value problems in dissipative gas dynamics. Phys. Fluids 10:24–34
Spillmann J, Tescher M. 2009. Cosserat nets. IEEE Trans. Vis. Comput. Graph. 15:325–37
Springel V. 2010. Smoothed particle hydrodynamics in astrophysics. Annu. Rev. Astron. Astrophys. 48:391–430
Tartakosky A, Meakin P. 2005a. Modeling of surface tension and contact angles with smoothed particle
hydrodynamics. Phys. Rev. E 72:026301
Tartakosky A, Meakin P. 2005b. A smoothed particle hydrodynamics model for miscible flow in three-
dimensional fractures and the two-dimensional Rayleigh-Taylor instability. J. Comput. Phys. 207:610–24
Tartakovsky A, Meakin P, Scheibe TD, Eichler-West RM. 2007. Simulations of reactive transport and pre-
cipitation with smoothed particle hydrodynamics. J. Comput. Phys. 222:654–72
Ulrich C, Rung T. 2010. SPH modelling of water/soil suspension flows. Presented at SPHERIC Workshop, 5th,
Manchester, UK
van Heijst GJF, Clercx HJH, Molenaar D. 2006. The effects of solid boundaries on confined two-dimensional
turbulence. J. Fluid Mech. 554:411–31
Veen DJ. 2010. A smoothed particle hydrodynamics study of ship bow slamming in ocean waves. PhD thesis. Curtin
Univ. Technol. West. Aust.
Vidal Y, Bonet J, Huerta A. 2007. Stabilized updated Lagrangian corrected SPH for explicit dynamic problems.
Int. J. Numer. Methods Eng. 69:2687–710
Violeau D, Issa R. 2007. Numerical modelling of complex turbulent free surface flows with the SPH method:
an overview. Int. J. Numer. Methods 53:277–304

www.annualreviews.org • SPH and Its Diverse Applications 345


FL44CH14-Monaghan ARI 1 December 2011 8:2

von Neumann J. 1963 (1944). Proposal and analysis of a new numerical treatment for hydrodynamic shock
problems. In Von Neumann Collected Works, ed. A Taub, pp. 361–79. Oxford: Pergamon
Wells MG, Wettlaufer JS. 2005. Two dimensional density currents in a confined basin. Geophys. Astrophys.
Fluid Dyn. 99:199–218
Wendland H. 1995. Piecewise polynomial, positive definite and compactly supported radial functions of
minimal degree. Adv. Comput. Math. 4:389–96
Whitehouse SC, Bate MR. 2004. Smoothed particle hydrodynamics with radiative transfer in the flux limited
diffusion approximation. Mon. Not. R. Astron. Soc. 353:1078–94
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

346 Monaghan
FL44-FrontMatter ARI 1 December 2011 22:45

Annual Review of
Fluid Mechanics

Contents Volume 44, 2012


Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

Aeroacoustics of Musical Instruments


Benoit Fabre, Joël Gilbert, Avraham Hirschberg, and Xavier Pelorson p p p p p p p p p p p p p p p p p p p p 1
Cascades in Wall-Bounded Turbulence
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

Javier Jiménez p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p27


Large-Eddy-Simulation Tools for Multiphase Flows
Rodney O. Fox p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p47
Hydrodynamic Techniques to Enhance Membrane Filtration
Michel Y. Jaffrin p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p77
Wake-Induced Oscillatory Paths of Bodies Freely Rising
or Falling in Fluids
Patricia Ern, Frédéric Risso, David Fabre, and Jacques Magnaudet p p p p p p p p p p p p p p p p p p p p p p97
Flow and Transport in Regions with Aquatic Vegetation
Heidi M. Nepf p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 123
Electrorheological Fluids: Mechanisms, Dynamics,
and Microfluidics Applications
Ping Sheng and Weijia Wen p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 143
The Gyrokinetic Description of Microturbulence in Magnetized Plasmas
John A. Krommes p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 175
The Significance of Simple Invariant Solutions in Turbulent Flows
Genta Kawahara, Markus Uhlmann, and Lennaert van Veen p p p p p p p p p p p p p p p p p p p p p p p p p p 203
Modern Challenges Facing Turbomachinery Aeroacoustics
Nigel Peake and Anthony B. Parry p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 227
Liquid Rope Coiling
Neil M. Ribe, Mehdi Habibi, and Daniel Bonn p p p p p p p p p p p p p p p p p p p 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
Dynamics of the Tear Film
Richard J. Braun p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 267
Physics and Computation of Aero-Optics
Meng Wang, Ali Mani, and Stanislav Gordeyev p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 299

v
FL44-FrontMatter ARI 1 December 2011 22:45

Smoothed Particle Hydrodynamics and Its Diverse Applications


J.J. Monaghan p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 323
Fluid Mechanics of the Eye
Jennifer H. Siggers and C. Ross Ethier p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 347
Fluid Mechanics of Planktonic Microorganisms
Jeffrey S. Guasto, Roberto Rusconi, and Roman Stocker p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 373
Nanoscale Electrokinetics and Microvortices: How Microhydrodynamics
Access provided by Korea Advanced Institute of Science and Technology (KAIST) on 02/25/20. For personal use only.

Affects Nanofluidic Ion Flux


Hsueh-Chia Chang, Gilad Yossifon, and Evgeny A. Demekhin p p p p p p p p p p p p p p p p p p p p p p p p p 401
Two-Dimensional Turbulence
Guido Boffetta and Robert E. Ecke p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 427
Annu. Rev. Fluid Mech. 2012.44:323-346. Downloaded from www.annualreviews.org

“Vegetable Dynamicks”: The Role of Water in Plant Movements


Jacques Dumais and Yoël Forterre p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 453
The Wind in the Willows: Flows in Forest Canopies in Complex Terrain
Stephen E. Belcher, Ian N. Harman, and John J. Finnigan p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 479
Multidisciplinary Optimization with Applications
to Sonic-Boom Minimization
Juan J. Alonso and Michael R. Colonno p p p p p p p p p p p p p p p p p p p p p p p p p p p 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
Direct Numerical Simulation on the Receptivity, Instability,
and Transition of Hypersonic Boundary Layers
Xiaolin Zhong and Xiaowen Wang p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 527
Air-Entrainment Mechanisms in Plunging Jets and Breaking Waves
Kenneth T. Kiger and James H. Duncan p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 563

Indexes

Cumulative Index of Contributing Authors, Volumes 1–44 p p p p p p p p p p p p p p p p p p p p p p p p p p p p 597


Cumulative Index of Chapter Titles, Volumes 1–44 p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 606

Errata

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

vi Contents

You might also like