THE PHYSICS OF GRANULAR

MATERIALS
The rich dynamics of these ubiquitous and important materials are
just beginning to be understood. Now there are suggestions that
processes taking place on astrophysical scales may mirror those
occurring in a pile of sand.

Heinrich M. Jaeger, Sidney R. Nagel and Robert P. Behringer

Who could ever calculate the path of a molecule'? How do material remains at rest even though gravitational forces
we know that the creations of worlds are not determined create microscopic stresses on its surface. If the pile is
by falling grains of sand? —V. Hugo, Les Miserables tilted several degrees above the angle of repose, grains
start to flow. (See the cover of this issue.) However, this
V ictor Hugo suggested the possibility that patterns
created by the movement of grains of sand are in no
small part responsible for the shape and feel of the natural
flow is clearly not that of an ordinary liquid because it
exists only in a boundary layer at the pile's surface, with
no movement at all in the bulk. We might view granular
world we live in. Certainly, granular materials, of which flow as that of a dense gas, because gases too are made
sand is but one example, are ubiquitous in our daily lives. of discrete particles with negligible cohesive forces be-
They play an important role in industries, such as mining, tween them. In contrast to ordinary gases, however,
agriculture and construction. They also are important in thermal energy, kT, is completely insignificant in granular
geological processes, such as landslides and erosion and, materials. The relevant energy scale for a grain of mass
on a larger scale, plate tectonics, which determine much m and diameter d is its potential energy mgd, where g is
of Earth's morphology. Practically everything we eat Earth's gravitational acceleration. For a typical sand
started out in a granular form and the clutter on our grain, this energy is at least 1012 times kT at room
desks is often so close to the angle of repose that a chance temperature, and ordinary thermodynamic arguments be-
perturbation can create an avalanche onto the floor. come useless. For example, many studies have shown
Hugo hinted at the extreme sensitivity of the macro- that vibrations or rotations of a granular material will
scopic world to the precise motion or packing of the induce particles of different sizes to separate into different
individual grains. We may nevertheless think that Hugo regions of the container.1"3 Because there are no attractive
overstepped the bounds of common sense in relating the forces between the particles, this separation would at first
creation of worlds to the movement of simple grains of appear to violate the principle that entropy must increase,
sand. This article is intended to show there is sufficient which normally favors mixing. However, in a granular
richness and complexity to granular motion that the pos- material the fact that kT = 0 implies that entropy consid-
sibility of Hugo's metaphor having literal meaning might erations can easily be outweighed by dynamical effects,
no longer appear so far-fetched: What happens in a pile which now become of paramount importance.
of sand on a tabletop is relevant to processes taking place Temperature allows a system to explore phase space.
on astrophysical scales. The fact that kT is negligible in a granular material
Granular materials are simple. They are large con- precludes such exploration. Unless perturbed externally,
glomerations of discrete macroscopic particles. If they are each metastable configuration of the material will last
noncohesive, the forces between them are strictly repulsive indefinitely, and no thermal averaging over nearby con-
so that the shape of the material is determined by external figurations will take place. Because each configuration
boundaries and gravity. If they are dry, any interstitial has its unique properties, it is difficult to achieve repro-
fluid, such as air, can often be neglected in determining ducibility of granular behavior, even on large scales and
many but not all of the flow and static properties of the certainly near the static limit where friction is important.
system. Yet despite this seeming simplicity, granular ma- Another role of temperature in ordinary gases or fluids is
terial behaves differently from the other familiar forms of to provide a microscopic velocity scale. Again, in granular
matter—solids, liquids or gases—and might therefore be materials this role is completely suppressed, and the only
considered an additional state of matter in its own right. velocity scale is the one imposed by any macroscopic flow
At the root of the unique status of granular materials itself. It is possible to formulate an effective "granular
are two characteristics: Ordinary temperature plays no temperature" in terms of velocity fluctuations around the
role, and interactions between grains are dissipative be- mean flow velocity.4 Yet, as we will see, such approaches
cause of the existence of static friction and the inelasticity do not always recover thermodynamics or hydrodynamics
of collisions. For example, a sandpile at rest with a slope because granular collisions are inelastic.
less than the angle of repose behaves like a solid: The The science of granular media has a long history.
Much engineering literature is devoted to understanding
how to deal with these materials. Nevertheless, the tech-
HEINRICH JAEGER and SIDNEY NAGEL are professors of nology for handling and controlling granular materials is
physics at the University of Chicago. ROBERT BEHRINGER is a not as well developed as that for handling conventional
professor of physics at Duke University in Durham, North fluids. Estimates suggest that we waste 40 percent of the
Carolina. capacity of many industrial plants because of problems

32 APRIL 1996 PHYSICS TODAY American Institute of Physics, S-0031-9228-9604-020-3

 FORCES WITHIN GRANULAR MEDIA.
a: Forces between pyrex spheres
surrounded by a water-glycerol mixture
with a matching refractive index can be
visualized when the three-dimensional
assembly is viewed between crossed
circular polarizers. Stress-induced '? r \ \ (
birefringence makes the beads visible

i
when a force exerted on a piston
covering the top surface of the
container, b: The forces/exerted by
spheres at the bottom of a cylindrical
container filled with spheres can be
measured by lining the bottom of the
container with carbon paper and
applying a force to the top surface of
•V*\S
spheres. The distribution of forces P(f)
decreases exponentially as the force
increases, as indicated by the excellent
fit to the curve P(f) = cexp{-f/fy.
(Adapted from ref. 9.) FIGURE 1

encountered in transporting these

materials from one part of the factory
floor to another.5 Over the last dec-
ade there has been a resurgence of
interest in this field within physics
t
(for overviews, see reference 6). The
science of these materials has clear
relevance to what is being done in
other areas of condensed matter phys-
ics, with the result that sandpiles
have become a fruitful metaphor for
describing many other, and often
more microscopic, dissipative dynami-
cal systems. One finds slow relaxations in vibrated sand-
piles that resemble those found in glasses, spin glasses
and flux lattices. Fluidlike behavior, similar to that ex-
hibited by conventional liquids, can be induced7 in these
materials. Some of the observed nonlinear dynamical
phenomena are relevant to breakdown phenomena in
semiconductors, stick—slip friction and earthquakes. A
recent, powerful use of sand as a metaphor has been the
idea of self-organized criticality, originally described in
terms of avalanches in a sandpile.8 This self-organization
paradigm was postulated to have applicability to a wide
realm of natural phenomena.
The following three sections explore the unique be-
havior of granular materials and contrast it with that of 4 6
ordinary solids, liquids and gases. FORCE /(newtons)

An unusual solid: sand at rest play an important part in many of the properties of the
Even in the resting state, granular materials exhibit a granular material, such as the transmission of sound.6
host of unusual behaviors. For example, when granular The distribution of forces within the pile is not clear
material is held in a tall cylindrical container, such as a from figure la alone. It can be found by simply placing
silo, the pressure head is not height-dependent as it would a piece of carbon paper in the bottom of the container and
be in a normal fluid; that is, the pressure at the base of the measuring the areas of the marks left by the forces, f, of
container does not increase indefinitely as the height of the the individual beads on the paper. The distribution of
material inside it increases. Instead, for a sufficiently tall forces is
column, the pressure reaches a maximum value that is
independent of the height. Because of contact forces between = cexp(-f/f0) (1)
grains and static friction with the sides of the container, the 9
where c and f0 are constants. (See figure lb.) The
container walls support the extra weight. fluctuations in the force are large and scale with depth
We can investigate the network of forces within the in the same way as does the mean force—rather than as
pile in greater detail. Figure la shows the stresses in a its square root as one might initially expect. Such behav-
three-dimensional arrangement of particles. The forces ior has been explained9 in terms of a simple model, in
appear to be very heterogeneous, forming chains along which masses placed on a lattice distribute their weights
which the stresses are particularly intense. These chains unevenly and randomly to the particles on the layer below

APRIL 1996 PHYSICS TODAY 33

them. This model can be solved exactly in a number of of an averaging process over length and time scales that
cases and, in agreement with experiment and simulation, are much larger than typical microscopic scales, and much
it yields an exponential distribution of large forces. Shear smaller than macroscopic scales—a separation of scales
experiments also attest to the extreme heterogeneity of that may not occur in granular flows.
granular media.6'11 Dense slow flows and rapid gaslike flows are useful
A fundamental issue in the physics of granular media idealizations for the development of models. Because real
concerns their packing. Depending on how the container granular systems rapidly dissipate energy, they often si-
is filled, a random assembly of spheres can be packed with multaneously exhibit both flow types in different spatial
a volume fraction anywhere from TJ = 0.55 to 17 = 0.64.6 domains, and one unresolved question is how to model
Through static friction, force chains can hold the assembly the transition between the two. When the density is low,
in a metastable configuration between these limits and kinetic theory models4 can be used to describe granular
prevent it from collapsing. How does the system pass materials. To maintain this state, however, energy must
between these states? Because kT is negligible, the only be continuously supplied, for instance by shaking.
way the density can change is as a result of external The other extreme is treated by models for quasistatic
disturbances of the container, for instance by vibrations. plastic deformation, based on Osborne Reynolds's principle
Recent studies of granular material settling under of dilatancy, and on the idea that deformations in com-
vibrations indicate that the relaxation in these systems pacted materials are typically irreversible.610 Dilatancy
is, in fact, logarithmically slow.12 Even after 100 000 occurs because the grains interlock under applied normal
vibration cycles, a tube filled with granular material may stress, and the material will begin to deform for shear
still undergo significant compaction before a steady state stresses only above a yield point that is determined by
is reached. A variety of models have been proposed to the applied normal stresses. Specific models consist of
account for this extremely slow settling.13 Perhaps the conservation laws in the form of partial differential equa-
most plausible explanation is based on the consideration tions augmented by constitutive models. Thus there is
of excluded volume effects.12 A simple corresponding pic- the standard continuity equation for mass conservation,
ture is that of a parking lot without assigned slots, and an energy equation and a momentum equation. The
with a high density of equal-sized, parked cars (or parti- momentum equation is perhaps the most revealing. It
cles). For the person wishing to park an extra vehicle (or relates the stress tensor TtJ and the strain rate tensor
insert an extra particle into the bead pack), the all-too-
familiar situation is that there exist large, but not quite (2)
large enough, voids between objects already in place. How dxt
many other cars (or particles) have to be moved just a bit where vt is the (th component of the velocity field. In one
for the additional one to fit in? If all densification occurs of the simplest versions of these models,10
by random "parking" and "unparking" events, it takes the
cooperative motion of many objects to open up new slots.
The resulting approach to the steady-state density is
?» = * jvr (3)
logarithmic in time. Here, I VI 2 = £ V | and k is a constant that is characteristic
for each material. A comparison of this equation with the
An unusual fluid: granular hydrodynamics Navier-Stokes equations reveals that the ordinary viscous
Qualitatively, granular materials can flow like fluids, and terms, proportional to the viscosity and the velocity gra-
there are a variety of theoretical models used to describe dients, have been replaced by terms independent of the
such flows. We refer to these models as granular hydro- shear rate. This feature is quite remarkable because it
dynamics (even though there is nothing wet here), in the implies that an overall increase in velocity leaves the
sense that they are continuum theories consisting of par- stress unchanged.
tial differential equations, analogous to the Navier-Stokes Models like that specified by equation 3 are used in
equations for Newtonian fluids. However, models for soil mechanics and in the design of materials-handling
granular flow do not have the stature of the Navier—Stokes devices, such as hoppers.6 However, experiments on flow
equations, because the Navier-Stokes equations arise out in thin hoppers14 have revealed dynamic behavior not

X-RAY IMAGES OF WAVES in a hopper offlowingsand show (two images at left) that waves form when the sand is composed of
rough grains—in this case, sieved construction sand; darker regions in the image correspond to lower density. When the grains are
sufficiently smooth—in this case, Ottawa sand—the waves are absent (two images at right). Typical grain sizes of the two
materials were identical, ranging from 0.6 to 0.7 millimeters. (Adapted from Baxter and Behringer in ref. 14.) FIGURE 2

34 APRIL 1996 PHYSICS TODAY

explained by such approaches. Fig-
ure 2 contrasts x-ray images of flow
out of a hopper for rough and smooth-
grained materials. These experi-
ments show density waves for the
rough materials, but not for the
smooth, nearly spherical materials.
In the case shown here for rough
grains, the waves propagate upward,
but the propagation direction reverses
if the hopper angle is made steep
enough. The role of grain shape in
these processes is critical and needs
to be better understood.
One exciting aspect of the present >
state of the physics of granular media
is the lively debate about the causes o
of some of the most prominent behav-
iors that these materials exhibit when
vibrated. Two such debates, dis-
cussed briefly here, concern the cause
of vibration-induced convection and
heaping and the cause of vibration- 20
10 15
induced size separation.
DEPTH (bead diameters)
Although convective flow in vi-
brated granular material was first ob- MAGNETIC RESONANCE IMAGES can reveal the behavior of
served by Michael Faraday 160 years particles hidden deep within a granular medium, in this case a
ago, the underlying mechanisms are cylinder filled with white poppy seeds, a: In this image, each
at best only partially understood. visible grain is an individual seed measuring roughly 1 millimeter
Both segregation and convection occur across. The coating of seeds glued to the inner surface of the
when the material is shaken in the container walls serves as a stationary reference and is visible
vertical direction. Typically the con- above the filling level, b: When the spatial modulation of the
tainer is shaken sinusoidally, as spin polarization is changed to the vertical or z direction, the
z=Acos(cot). When the maximum peaks of the modulation (bright bands) label narrow regions
acceleration of the container is within the granular material, c: This image was prepared in the
greater than the gravitational accel- same way as image b, except the container was given a single
eration—that is when Y=Aw2lg is vertical shake. The curvature of the horizontal stripes directly
larger than 1—the material rises gives the flow profile of the seeds (displacement per shake) and
above the floor of the container for can be seen by referring to the seeds glued to the container walls.
some part of each cycle, dilating in d: For a wide range of accelerations F, the vertical velocity, v2< of
the process, and allowing a macro- the central region of each band decreases rapidly as a function of
scopic convection roll that continuously
depth, z, below the top surface. In this plot, a straight line
transports grains. In cylindrical or rec-
tangular vessels, the flow is usually indicates an exponential depth-dependence of the velocity.
upward in the center and downward in (Adapted from Enrichs et al. in ref. 3.) FIGURE 3

APRIL 1996 PHYSICS TODAY 35

a thin stream along the sidewalls, leading to the formation the mixing of powdered drugs with a binder, when a
of a central heap, • • down which there is a steady ava- well-controlled and homogeneous mixture is essential.
lanche of grains. With different boundary conditions, such When granular materials are shaken, particles of
as sidewalls that slant outward, it is possible to reverse the different sizes tend to separate, with the largest particles
sense of the convection roll and induce downward flow in moving to the top, independent of their density.23 Several
the center.316 mechanisms have been associated with such size separa-
At least two mechanisms have been proposed to ex- tion, including sifting—in which small particles fall
plain these rolls for experiments in which the entire layer through the gaps between large particles if the gaps are
is shaken uniformly. One mechanism involves friction large enough—and local rearrangements—in which large
with the walls of the container. Several experiments and particles are wedged upward as smaller grains avalanche
simulations have shown that there is a ratchet effect that into voids beneath them during the dilation phase of each
produces a thin, rapidly moving boundary layer near the shake.2 In vertically shaken systems, however, experi-
walls and leads to circulating flow.1 Recent experiments ments3 have shown a direct link between convection and
have determined both the depth-dependence and the de- size separation: Large particles become entrained in the
tailed shape of the convection velocity profile using mag- upward convective flow, but, once on the top surface, they
netic resonance imaging to probe noninvasively the granu- remain stranded because they cannot follow the smaller
lar motion deep inside the material:117 (See figure 3.) grains in the thin layer of downward flow along the
The fastest flow occurs in the thin layer near the walls—a container walls.
situation different from what occurs for a conventional
fluid, for which the no-slip condition applies at the walls. An unusual gas: inelasticity, clustering, collapse
This raises a number of issues about what boundary condi- One critical difference between granular media and ordi-
tions are appropriate for convection and other granular flows. nary gases and fluids deserves particular attention: Be-
A second mechanism, first suggested by Faraday, for cause interactions between grains are inherently inelastic,
convection and heaping depends on the presence of inter- some energy is lost in each collision. As a result, novel
stitial gas. This effect is particularly apparent when features arise for the statistical mechanics of these sys-
friction with the container walls is reduced. Experiments tems. Any seemingly fluidlike behavior of a granular
to clarify the role of gas have yielded conflicting results. material is a purely dynamic phenomenon. For example,
(See the papers by Stefan Fauve et al. and Pierre Evesque the surface waves do not arise as a linear response to
et al. in reference 7.) One set of experiments indicated external energy input, but are a consequence of a highly
that the flow stopped when the surrounding gas pressure nonlinear, hysteretic transition out of the solidlike state.
was reduced, while another indicated that convection was Fluid behavior sets in only above a threshold excitation
virtually unchanged at pressures down to 4 Torr. Hyuk level, and inelastic collisions will bring the granular me-
K. Pak et al.15 have shed some light on this conflict through dium to rest almost instantly after the energy input is
experiments in which the pressure was held fixed with stopped. An individual grain, such as a single marble,
values between atmospheric pressure and vacuum. The dropped onto a glass plate, will bounce for quite a while.
convective heap persisted for pressures down to 10 Torr. Identical marbles, loosely filled into a sack, will stop dead
Further pressure decreases steadily diminished the height on the plate. This strikingly different collective behavior
of the heap. These results apply for grains of diameter arises from the exceedingly large number of rapid inelastic
up to about 1 millimeter, and the effect is more pronounced collisions among neighboring grains. In fact, we often use
for large amplitudes of oscillation. Developing a theory this energy absorption in such applications as packaging
that incorporates both the friction and gas effects remains fillers, recoilless hammers and Hacky Sacks.
a theoretical challenge. Because real granular materials are inelastic, energy
The free surface of a vibrated granular material can input from the boundaries, as in an ordinary heat bath,
exhibit several different wave phenomena, as well as more may not be sufficient to thermalize the system. If clus-
complex irregular (and possibly chaotic) states.15-7 The tering begins to occur inside the system, it may indicate
different waves can be either traveling (for material with a breakdown of Newtonian hydrodynamics because those
a steeply sloping heap) or standing, when heaping is weak aggregates will not be able to melt away. Recently, this
or nonexistent. Figure 4 shows examples of the subhar- clustering in the regime of finite inelasticity has become
monic standing waves; well-defined wave patterns and a focus of much interest.19 The conditions under which
their superpositions occur that are strikingly familiar from clustering should appear have been estimated: If a system
Faraday instabilities in ordinary liquids. In the first two of linear extent L is started in a uniform state with grains
parts of this figure, the waves are confined to a narrow occupying a volume fraction, TJ, the solutions provided by
rectangular container. Part c shows the striking patterns Newtonian hydrodynamics become linearly unstable to
that evolve when the container is a large open cylinder. cluster formation once the product r)L exceeds a constant
(See PHYSICS TODAY, October 1995, page 17.) that depends on the degree of inelasticity. If L is large
Another key feature of vibrated or flowing granular enough, the system always becomes linearly unstable
material is its unique mixing and size-separation ("unmix- toward cluster formation, no matter how small the inelas-
ing") behavior. Separation phenomena occur in long, ticity of each collision. In figure 5, we clearly see the
slowly rotating cylinders, where the cylinder axis is hori- tendency of inelastic collisions to produce particle cluster-
zontal; particles with different dynamical angles of repose ing. This snapshot is taken from a two-dimensional simu-
aggregate into sharply delineated regions.1 Conversely, lation by Isaac Goldhirsch and Gianluigi Zanetti.19 Here,
an important question concerns the rate of mixing in a system of hard disks was started with random initial
horizontally rotating cylinders or drums that are partially velocities, in the absence of gravity, and without external
filled and start from an inhomogeneous state. In these forcing.
situations, a steady succession of discrete avalanches A particularly exciting development has been the
cause mixing, and the degree of mixing as a function of recent recognition of a special type of clustering called
filling level can be calculated from geometrical argu- inelastic collapse. Sean McNamara and William Young19
ments.18 Both mixing and unmixing bear directly on such showed that inelasticity can lead to an infinite number of
technically important processes as the separation of "fines," collisions occurring in a finite time. In one dimension,
or dustlike particles (which may or may not be desirable), or such a collision sequence leaves the particles "stuck"

36 APRIL 1996 PHYSICS TODAY

 COMPLEX WAVE BEHAVIOR is easily observed in thin layers of
granular materials shaken vertically with frequency/and
1
•Mr" 'm •• -M, acceleration T = Aur/g. a: Side view of a subharmonic
standing wave in a quasi one-dimensional container filled with
glass beads. ( / = 20 hertz, T = 3.5.) (From Douady et at. in
ref. 7.) b: A different standing wave in an annular container.
The two snapshots are taken at successive oscillation periods
of the container and illustrate the subharmonic nature of the
response, which repeats after two drive cycles rather than one.
(From Pak and Behringer in ref. 7.) c: Two-dimensional
systems of granular material organize into a variety of
stationary subharmonic patterns that depend on/and T.
These photos show top views of large cylindrical vessels filled
with 165-micron-diameter brass spheres to a nominal depth of
about 8 spheres. Stripe (upper left) and square patterns (upper
right) form for / = 40 Hz, F «• 3. Highly curved interfaces
(lower right) between essentially flat, featureless plateaus
coexist 180° out of phase in different parts of the cell (/= 40
Hz, F = 5). Hexagonal patterns (lower left) occur when the
system is driven at two frequencies (16 and 32 Hz) and
F = 2.8.. (Photos courtesy of P. Umbanhowar.) FIGURE 4

Perhaps the most remarkable aspect of clustering,

which is also true in the regime of inelastic collapse, is
that it leads to long stringlike grain configurations, rather
than to a shapeless blob of particles. As noted by Gold-
hirsch and Zanetti, figure 5, in this sense, is qualitatively
similar to density maps of the visible universe. It is worth
speculating that the attractive gravitational potential
plays the role of a confining container to keep the density
high enough for clusters to form. Hence, on very large
scales, the structures created by repeated inelastic colli-
sions may also be responsible for the coagulation observed
in "gases" made up of stars or galaxies.
Perhaps the quotation from Victor Hugo at the begin-
ning of this article is not so far-fetched after all. The
motion of grains of sand may indeed be relevant to the
creation not just of worlds but also of galaxies and the
structure and formation of the astronomical landscape.

Scientific challenges ahead

This article has only touched on some of the distinctive
properties of granular materials. The preceding sections
have shown that, depending on how we prepare and excite
these materials, they act as highly unusual solids, liquids
or gases. Clearly, the physics of granular materials spans
a wide variety of phenomena, with applications ranging
from the mundane to the celestial. The array of experi-
mental techniques used to study these systems is also
quite broad and spans a considerable range of sophistica-
tion—from counting spots left by carbon paper to magnetic
resonance imaging and x-ray tomography. Despite their
apparent simplicity, these materials display an intriguing
range of complex, nonlinear behavior, and explanations of
this behavior often appear to challenge existing physics
wisdom. Many of the new ideas and techniques being
together in close contact with no relative motion. ls Re- developed within the specific context of granular materials
markably, inelastic collapse also persists in higher dimen- are applicable to a much wider range of systems of similar
sions, where it produces dense chainlike clusters. The metastable characteristics and in which the thermal energy,
precise relationship between inelastic collapse and the kT, is irrelevant—for example, in foams or in the tunneling
phenomenon of clustering, which is the initial signature regime for superconducting vortex arrangements.
of a breakdown of ordinary hydrodynamics, needs clarifi- The recognized deficiency of our understanding of
cation. One plausible scenario is that once the system granular matter offers tremendous opportunities for new
forms clusters, the occurrence of inelastic collapse requires insights from physics to make a strong, technologically
additionally that the energy loss per collision exceed a relevant impact. However, many scientific challenges
critical value. Another conceivable scenario envisions all must first be resolved. For example, for the case of
clusters as transients that eventually must terminate in packing, although we know that the packing history is
either inelastic collapse or the formation of shear bands. 19 relevant, we do not know how to include it in theories of

APRIL 1996 PHYSICS TODAY 37

 TYPICAL CONFIGURATION of 40 000 inelastically colliding
panicles exhibits clustering in two dimensions. Here the
coefficient of restitution is 0.6, the time corresponds to 500
collisions per particle and the average area fraction occupied
by particles is 0.05. Note the chainlike structures in the
photo. (From Goldhirsch and Zanetti in ref. 19.) FIGURE 5

Knight, V. Yu. Kuperman, S. R. Nagel, Science 267, 1632

(1995).
4. S. B. Savage, Adv. Appl. Mech. 24, 289 11984). P. K. Haff, J.
Fluid Mech. 134, 401 (1983). J. Rheol. 30, 931 (1986). 0. R.
Walton, R. L. Braun, J. Rheology 30, 949 (1986). J. T. Jenkins,
in Nonclassical Continuum Mechanics: Abstract Techniques
and Applications, R. J. Kops and A. A. Lacey, eds., Cambridge
U. P., New York (1987), p. 213. C. K. K. Lun, S. B. Savage, J.
Appl. Mech. 54, 47 (1987). C. S. Campbell, Ann. Rev. Fluid
Mech. 22,57(1990).
5. B. J. Ennis, J. Green, R. Davis, Chem. Eng. Progress 90, 32
(1994).
6. Physics of Granular Media, D. Bideau, J, Dodds, eds., Les
Houches Series, Nova Science, Commack, New York (1991).
Granular Media: An Interdisciplinary Approach, A. Mehta,
ed., Springer, New York (1991). H. Hayakawa, H. Nishimori,
S. Sasa, Y.-H. Taguchi, Jpn. J. Appl. Phys. 34, 397 (1995).
H. M. Jaeger, S. R. Nagel, Science 255, 1523 (1992). S. R.
Nagel, Rev. Mod. Phys. 64, 321 (1992). H. M. Jaeger, J. B.
Knight. C.-H. Liu, S. R. Nagel, Mater. Res. Soc. Bull. 19, 25
compaction or stress patterns within the medium. Like- (1994). R. P. Behringer, Nonlinear Science Today 3, 1 (1993).
wise, when attempting a hydrodynamic approach to granu- R. P. Behringer, Proc. Mater. Res. Soc. 367, 461 (1995).
lar flow, it is obvious that the ordinary hydrodynamic 7. S. Fauve. S. Douady, C. Laroche, J. Phys. (France) 50, 187
nonslip boundary assumptions are invalid, but we are still (1989). P. Evesque, J. Rajchenbach, Phys. Rev. Lett. 62, 44
at a loss as to how to treat the boundaries correctly. (1989). S. Douady, S. Fauve, C. Laroche, Europhysics Lett. 8,
It also remains to be seen to what extent Newtonian 621 11989). H. K. Pak, R. P. Behringer, Phys. Rev. Lett. 71,
hydrodynamics must be (or can be) modified to describe 1832 (1993). F. Melo, P. Umbanhowar, H. L. Swinney, Phys.
granular media correctly in the first place. Certainly the Rev. Lett. 72, 172 (1993). C. R. Wassgren, C. E. Brennen, M."L.
Hunt, preprint. F. Melo, P. B. Umbanhower, H. L. Swinney,
authors of this article debate this issue among themselves. Phys. Rev. Lett 75, 3838 (1995).
The debate extends to a related question—namely whether 8. P. Bak, C. Tang, K. Wiesenfeld, Phys. Rev. A 38, 364 (1988).
the idea of inelastic collapse is more than a beautiful 9. C.-H. Liu, S. R. Nagel, D. A. Schecter, S. N. Coppersmith, S.
theoretical concept or whether it has real experimental Majumdar, O. Narayan, T. A. Witten, Science 269, 513 (1995).
ramifications, and whether the difference between inelas- 10. R. Jackson, in The Theory of Dispersed Multiphase Flow. R.
tic collapse and the more general case of inelastic cluster- Meyer ed., Academic, San Diego, (19831. E. B. Pitman, D. G.
ing is experimentally observable. The authors hope that Schaeffer, Commun. Pure Appl. Math. 40, 421 (1987). D. G.
the recent surge of interest in the basic physics of granular Schaeffer, J. Diff. Eq. 66. 19 (1987). D. G. Schaeffer, M.
media will produce advances that can not only lead to Shearer, E. B. Pitman, SIAM J. Appl. Math. 50. 33 11990).
11. G. W. Baxter, R. Leone, R. P. Behnnger. Europhysics Lett. 21,
improved applications for technological processes, but also 569 (1993). B. Miller, C. O'Hern, R. P. Behringer, preprint.
deepen our understanding of the many related aspects of 12. J. B. Knight, C. G. Fandrich. C. N. Lau, H. M. Jaeger, S. R.
microscopic and macroscopic physics for which sand has Nagel, Phys. Rev. E 51, 3957 11995). H. M. Jaeger, E. Nowak,
been used as a metaphor. E. Ben-Nairn, J. B. Knight, S. R. Nagel. preprint.
13. G. C. Barker, A. Mehta, Phys. Rev. E 47. 184 (1993). D. C.
Because of space limitations, the list of references in this article Hong, S. Yu, J. K. Rudra, M.' Y. Choi, Y. W. Kim, Phys. Rev. E
only provides a glimpse of some of the relevant literature, and omits 50,4123(1994).
many important contributions. A more complete account can be 14. G. W. Baxter, R. P. Behringer, T. Faggert, G. A. Johnson, Phys.
found in the full version of this article, which is to appear in the Rev. Lett. 62, 2825 (1989). G. W. Baxter. R. P. Behringer, in
"Colloquia" section of Reviews of Modern Physics. We would like Tivo-Phase Flows and Waves, D. D. Joseph, D. G. Schaefer, eds.
to thank our collaborators for many exciting discussions on the Springer, New York, (1990), p. 1.
topics mentioned in this article. They include G. W. Baxter, E. 15. H.K. Pak, R. P. Behringer, Nature 371, 231 (1994). H. K. Pak,
Ben-Nairn, S. N. Coppersmith, E. van Doom, E. E. Ehrichs, L. P. E. van Doom, R. P. Behringer, Phys. Rev. Lett. 74, 4643 (1995).
Kadanoff, G. S. Karczmar, J. B. Knight, V. Yu. Kuperman, C.-H. 16. Y.-H. Taguchi, Phys. Rev. Lett. 69 1367 (1992). J. A. C. Gallas,
Liu, B. Miller. E. Nowak, C. O'Hern, H. K. Pali, D. A. Schecterand H. J. Herrmann, S. Sokolowski. Phys. Rev. Lett. 69,1371 (1992).
T. A. Witten. We are grateful to J. Cina and to W. Young and S. S. Luding, et ai, Phys. Rev. E 50, 1762 and 3100 (1994). T.
Esipov for a critical reading of the manuscript. Poeschel, H. J. Herrmann, Europhys. Lett. 29, 123 (1995). M.
Bourzutschky, J. Miller, Phys. Rev Lett, 74, 2216 (1995). H.
Hayakawa, S. Yue, D. C. Hong. Phys. Rev. Lett. 75, 238 (1995).
References 17. M. Nakagawa, S. A. Altobelli, A. Caprihan, E. Fukushima,
1. O. Zik, D. Levine, S. G.Lipson, S. Shtrikman, J. Stavans, Phys. E.-K. Jeong, Experiments in Fluids 16, 54 (1993).
Rev. Lett. 73, 644 (1994). K. M. Hill, J. Kakalios, Phys. Rev. 18. G. Metcalf, T. Shinbrot, M. M. McCarthy, J. M. Ottino, Nature
E 49, 3610 (1994), and references therein. 374,39(1995).
2. A. Rosato, K. J. Strandburg, F. Prinz, R. H. Swendsen, Phys. 19. O. R. Walton, in Particulate Tivo-Phase Flow. Part l.M.C. Roco,
Rev. Lett. 58,1038 (1987). R. Jullien. P. Meakin, A. Pavlovitch, ed., Butterworth-Heinemann, Boston (1992), p. 884. I. Gold-
Phys. Rev. Lett. 69, 640 (1992). J. Duran, J. Rajchenbach, E. hirsch, G. Zanetti, Phys. Rev. Lett. 70, 1619 (1993). S.
Clement, Phys. Rev. Lett. 70, 2431 (1993). McNamara, V. R. Young.'Phys. Rev. E 50, 28 (1994). Y. Du, H.
3. J. B. Knight, H. M. Jaeger, S. R. Nagel, Phys. Rev. Lett. 70,3728 Li, L. P. Kadanoff, Phys. Rev. Lett. 74, 1268 (1995). P. Constan-
(1993). E. E. Ehrichs, H. M. Jaeger, G. S. Karczmar, J. B. tin, E. Grossman, M. Mungan, Physica D 83, 409 (1995). •

38 APRIL 1996 PHYSICS TODAY