Sie sind auf Seite 1von 22

Powder Technology 162 (2006) 208 – 229

www.elsevier.com/locate/powtec

Granular material flows – An overview


Charles S. Campbell ⁎
Aerospace and Mechanical Engineering, University of Southern California, Los Angeles, CA 90089-1453, USA
Received 6 June 2005; received in revised form 19 August 2005

Abstract

The paper attempts to give a critical overview of the field of granular flow with attention both to the history and the underlying physics that
govern the field. It starts with a discussion of the basic transport mechanisms in a granular flow. It continues with a discussion of contact
mechanics – the way that individual particles see each other mechanically. It then discusses the historical limiting regimes of granular flow, the
Quasistatic and the Rapid-Flow regimes. Finally, it concludes with a review of the Elastic picture of granular flow, which both unifies the
Quasistatic and Rapid regimes and fills in the intervening space. It shows that the rheological behavior of granular systems changes with system
scale constraints, and, in particular, that the materials behave differently under controlled-stress and controlled-concentration conditions. The
Elastic model defines an entire flowmap of granular flow and thus allows one to place boundaries on where the Quasistatic and Rapid-Flow
models (sometimes called kinetic theory models) are something of a red herring and cannot be applied to common granular flows.
© 2005 Elsevier B.V. All rights reserved.

Keywords: Transport mechanism; Quasistatic regime; Rapid-Flow regime; Granular flow

1. Introduction air. However, it will be assumed herein that the particles are
large and heavy in the sense that they are immune to effects of
Under the correct conditions, a granular solid can flow like a the interstitial fluid. For the most part we will also ignore
fluid. This was probably first recorded be Lucretius (ca. 98– cohesion between particles; cohesion arises from surface forces
55 B.C.), who wrote “One can scoop up poppy seeds with a or related phenomena such as liquid bridges, both of which act
ladle as easily as if they were water and, when dipping the ladle, on the surface area and thus can generally be neglected for large
the seeds flow in a continuous stream,” (quotation taken from particles with small surface area to volume ratios. Note that
Jacques [1]). As long as there has been mining and agriculture, these requirements collectively define what is meant by “large”
man has attempted to exploit the flowability of granular solids although those criteria cannot yet be quantitatively defined by a
to ease handling and storage problems. In particular, the ability set of dimensionless parameters.
of gravity to drive a granular flow, as noted by Lucretius, greatly This paper grew out of a long lecture given to the Ohio State
simplifies and provides a cost-free mechanism of transport. As a summer course on Powder Technology. It is an attempt to put
result, the most common granular handling devices, chutes and the state of knowledge of granular flows into perspective. It is
hoppers, are gravity-driven flows. Yet the design of granular not intended to be a review article, in the sense that I am not
systems is still something of a black art, in part because even the trying to mention every paper written on the subject, but instead
most basic flow mechanisms of granular materials are not well attempt to hit the highpoints and give a critical and balanced
understood. In fact, science has not identified the set of material view to the whole subject.
properties that control the flow behavior.
For the purposes of this article, a granular solid is taken to be 2. Internal force transmission
a collection of discrete solid particles. In general the spaces
between the particles are filled with an interstitial fluid, usually The unique features of granular material arise from the
manner in which force is internally transmitted. In continuum
⁎ Corresponding author. mechanics this is represented by a stress tensor τ, each
E-mail address: campbell@usc.edu. component of which τij represents the force in the i-direction
0032-5910/$ - see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2005.12.008
C.S. Campbell / Powder Technology 162 (2006) 208–229 209

3. Contact forces

In a dense granular flow, forces are largely generated by


interparticle contacts. The contact forces are how the particles
“see” one another mechanically. Imagine the two spherical
particles shown in Fig. 2. As long as linear elasticity applies,
the normal force on the contact will proportional to EAε, E is
the Young's modulus, A is the contact area and ε is the local
Fig. 1. The two mechanisms of internal momentum transport. (a) Contact
strain. The strain, ε = δ/ L where δ is the depth of the contact
transport: Here momentum crosses the imaginary surface (the dashed line) as the
result of the contact force Fc which can be thought of as transporting momentum deformation (the distance the contact has been compressed)
between the centers of the particles along the vector l. (b) Streaming transport: and L is an appropriate length scale. For the contact between
Here the momentum of the particles is carried across the imaginary surface due two spheres shown in Fig. 2, note that the contact area A is
to the random motion of the particles in a manner analogous to the transport of the zero on the unloaded contact to the left and increases
momentum in the kinetic theory of dilute gases.
when the particles are pressed together as on the right. The
fact that both the area A and the strain ε, change
on a surface with outward pointing normal unit vector in the j simultaneously as the particle is deformed, leads to non-
direction. There are two internal modes of stress transmission. linearity of the contact response.
The first or Contact Stress, τc is due to force transmission across In 1882, Hertz [3] derived an elastic solution for the contact
interparticle contacts. Thus a force Fc applied at a contact can be between bodies. The solution is not exact as often thought, but
thought of as being transmitted in the direction of the vector contains an implicit assumption that the square root of the
l that connects the centers of mass of the two particles involved. contact area is small compared to both the local radius of
(The length of l is the distance between the particle centers.) curvature and the overall dimensions of the body. The normal
When averaged 〈 〉 over time and volume, this yields the force exerted on a contact between two bodies of local curvature
Contact stress tensor: R is:
t c ¼ hF c li ð1Þ 4 E
fn ¼ R1=2 d3=2 ð3Þ
3 1  m2
(If the forces are transmitted collisionally, this is sometimes
called the Collisional Stress tensor). This is shown schemati- (from Johnson [4]), where E is the Young's modulus, and υ is
cally in Fig. 1a. Note that Fc need not point in the direction of l; Poisson's ratio. This corresponds to a normal stiffness:
when coupled with the fact that the averaging volume must be
larger than a particle (and not allowed to shrink to infinitesimal dfn E
k¼ ¼ 2R1=2 d1=2 ð4Þ
size as in standard continuum mechanics), this means that the dd 1  m2
contact stress tensor need not be symmetric. Any symmetry is
balanced by gradients in a couple-stress tensor that governs the Substituting for δ from (2), one can write the stiffness in terms
transmission of torques internal to the material (see [2]). of fn.
If the particles are moving, there will be some degree of  2=3
E
internal momentum transport due to the motion of an individual k ¼ 61=3 R1=3 fn1=3 ð5Þ
1  m2
particle as it moves relative to the bulk material, carrying its
momentum with it as illustrated in Fig. 1b. If u′ represents the
velocity of that relative motion, then one defines a Streaming
Stress Tensor τs analogous to the Reynolds stress tensor in
turbulent flow.
t s ¼ qp vhu Vu Vi ð2Þ
Here ρp is the density of the solid material and ν is the “solid-
fraction” or solid concentration, the fraction of a unit volume
occupied by the particles (so that ρpν is the bulk density of the
solid phase). The streaming stresses will only be significant at
small concentrations when contacts are infrequent and in cases
where the random particle velocity u′ is large.
Common granular flows, such as hoppers, chutes and
landslides are densely packed with solid concentrations well
above 50% by volume. It is possible to obtain flows at a small
concentration, but they are limited to high-speed laboratory
shear cells, computer simulations and perhaps the rings of Fig. 2. A schematic of the contact between two spheres of radius R generated by
Saturn. As such, the contact stresses will dominate and the the application of a force F. Here A is the area of the contact and δ is the
streaming stresses can usually be neglected. deformation. As shown the deformation is greatly exaggerated.
210 C.S. Campbell / Powder Technology 162 (2006) 208–229

Now Bathurst and Rothenburg [5] derived the bulk elastic fn1/2 or is linearly proportional to the deformation δ. That is
modulus of a random granular material from the contact behavior reminiscent of the interaction between the point of a
stiffness and showed that: conical contact and a surface. (Conical contacts are non-
Hertzian because they have zero radius of curvature at the
k
Ebulk ~f ðnÞ ð6Þ point.) Goddard [6] pointed out that the observed behavior can
R be explained if the particles initially interact across near-conical
where n is the coordination number. (The coordination number asperities on the surface, thus accounting for the conical
is the number of contacts between a particle and its neighbors; it behavior at low pressures. As the pressure increases, the
appears in the bulk modulus since the larger the number of asperities are compressed until the spherical surfaces of the
contacts on a particle, the larger the number of contacts particles come into direct contact eliciting a Hertzian response.
available to resist an applied force and, consequently, the stiffer Goddard also presents a model that encompasses both limits and
the material.) Note first that the bulk modulus depends on the the transitional region between. There have been attempts to
stiffness, not directly on the modulus E of the material that explain this behavior in terms of an increase in the coordination
makes up the particles because it is through the stiffness than the number n with pressure (e.g. Makse et al. [8]) and such an
particle see one another elastically. While the stiffness is linearly increase has been observed (see for example, Potapov and
dependent on E, it also depends on R, the local radius of Campbell [9]), but the Duffy and Mindlin data used materials
curvature and thus depends on the geometry of the contact. carefully assembled in an FCC packing so that the coordination
The bulk modulus in Eq. (6) makes it possible to use the number was fixed. It is a bit surprising to find a significant effect
soundspeed in a static granular material as p a way to probe ffi the
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi of asperities on high tolerance stainless steel balls, but if
contact stiffness. The soundspeed
pffiffiffi varies as E bulk =q bulk and is examined on a small enough scale, any surface will exhibit
thus proportional to k . Fig. 3 shows the soundspeed as a some asperities. One can only expect more severe behavior
function of the confining pressure p, applied to uniform-sized from common granular materials and indeed soundspeed
stainless steel spheres arranged in a regular face-centered cubic measurements in sands by Richart and coworkers [10,11] (See
packing (which, being nearly spherical, smooth and placed in an also [6]) show a pressure dependence more characteristic of
ordered arrangement, is about as nice a granular material as is conical contacts.
available). Data are taken from Goddard [6], who in turn took All of the above results indicate a purely elastic contact.
the data from Duffy and Mindlin [7]. Different behavior can be expected if the material yields
The pressure p, which is applied to the bulk assembly, must plastically under the application of the contact force. Analyses
be balanced by the forces on individual particle contacts. From of Hertzian contacts with plastic yielding performed by Walton
Eq. (5) one expects the stiffness k to vary as the cube root of the [12] and Thornton [13] show that, as the load on a contact is
normal force on the contact fn1/3, and thus to the cube root of the increased, beyond an initial period of elastic behavior, the
pressure, p. Thus the soundspeed should go as p1/6. As Fig. 3 is normal force fn, varies nearly linearly with δ, indicating a nearly
a log–log plot, this means that the sound speed as a function of constant normal stiffness k. However, the unloading follows a
pressure should have a slope of 1/6, represented by the solid different curve, again nearly linear, but with a steeper slope,
lines. Fig. 3 shows that this is true only at large confining indicating a larger but still constant k. (This last means that the
pressures. Surprisingly at low pressures the soundspeed varies force drops to zero before the particle centers are separated by
as p1/4, and only assumes the p1/6 predicted for Hertzian the sum of their radii. Physically, this occurs because the plastic
contacts for large pressures. Working backwards through the deformation leaves a flat indentation in the surface of the
above calculation, this would imply that the stiffness varies as particles so that they lose contact early.) This bi-linear behavior

Fig. 3. The sound speed as function of hydrostatic confining pressure in an FCC packing of 1/3 in. diameter steel balls with low tolerance (○, ±50 × 10− 6 in.) and high
tolerance (▵, ±10 × 10− 6 in.). The solid lines have a slope of 1/6, indicative of a Hertzian contact while the dashed lines have a slope of 1/4, indicative of a conical
contact. From Goddard [6] based on the data of Duffy and Mindlin [7].
C.S. Campbell / Powder Technology 162 (2006) 208–229 211

lead to the “latched-spring” contact model used in many The Mohr–Coulomb failure criterion is usually expressed in
computer simulations starting with Walton and Braun [14]. the form:
In addition, the result of many successive contacts on a
sV c þ r tan/ ð7Þ
particle surface may work harden or in other ways evolve the
character of the surface. For example, in one of the earliest Here c represents the cohesion of the material (for this paper, the
(1918) studies of impact behavior, Raman [15] observed that particles are assumed to be large and dry so that c is assumed to
repeatable results could only be achieved if the surfaces of the be zero.), σ is the normal stress, τ is the shear stress, and ϕ is the
particles were polished between experiments – presumably “friction angle”. When τ = c m σ tan ϕ, the material yields and
removing any plastic damage and work hardening of the begins to flow. The two constants c and ϕ are assumed to be
surface. Thus, each contact can change the local surface material properties that are measured in standard shear cell tests.
properties so that the properties on the surface may change with Coulomb yield, could be used to construct a plastic yield
both position and time. criterion, and only the adoption of a flow rule was required to
All of the above discussion involves only the normal force employ the methods of metal plasticity to granular flow. As a
on the contact. As a result of interparticle friction, each contact result, it is not necessary to consider the behavior on the level of
will also experience a tangential force which will evolve even individual particles (as in Eqs. (1) and (2)), but the material can
more dramatically with its history. Mullier et al. [16] showed be treated as a continuous plastic solid. The general principles
that, when a contact is loaded tangentially until it reaches “gross and governing equations are laid out in Sokolovski [19].
sliding” (when the surfaces of the two particles slip relative to The problem is further simplified by the idea of a “Critical
one another), the effect of gross sliding is to shear off asperities State”, the observation that a shearing granular material will
from the particle's surface, thus changing its frictional approach a “critical” concentration, νc, i.e. the fraction of a unit
properties. Mullier et al. found that the behavior before gross volume filled with solid material, whose value depends only on
sliding could be well described by the complex theory of the applied load and is again assumed to be a material property.
Mindlin and Deresiewicz [17], but not the behavior following (The critical state concept is probably due to Casagrande [20]
gross sliding. Presumably, this occurs because the removal of and is explored in detail in Schofield and Wroth [21].) This is
the asperities during gross sliding changes the surface friction shown schematically in Fig. 4. Fig. 4a shows a representative
coefficient (Mindlin and Deresiewicz assume the friction plot of the approach of a soil towards the critical concentration.
coefficient remains unchanged throughout the process). A material that is “under-consolidated” (i.e. starting with a
concentration below the critical concentration νc) will increase
4. Quasistatic or slow-flow theories its concentration as it sheared until it reaches the critical value.
Conversely, an “over-consolidated” material will decrease its
Granular flow modeling began with the 1773 paper by concentration as it sheared until it reaches the critical
Coulomb [18] who first described the yielding of granular concentration. As slow granular flows usually involve large
materials as a frictional process. He was not interested in flow, shear strains, it is reasonable to assume that the material is
per se, but in the prediction of soil failures for Civil Engineering shearing at the critical concentration. Fig. 4b shows the
applications. As such, the onset of failure in the soil, usually dependence of the critical concentration on the applied stress
meant that the structure collapsed; compared to such a σ. Note that the critical concentration is nearly constant over a
catastrophe, the subsequent motion of the soil, that part that wide range of σ, and only increases at large σ. The increase in νc
interests those of us working on granular flows, was of little at large σ can be attributed to the compressibility of the
interest. particles. At large applied stress, the solid particles are

Fig. 4. A schematic illustration of the critical stress concept. (a) Demonstrates how the overall concentration approaches the critical concentration νc, at large total
shear, γt. An over-consolidated material starts above the critical concentration while an under-consolidated material starts below the critical concentration, yet both
collapse to νc at large strains. (b) The variation of the critical concentration with applied stress σ. Note that the critical concentration is nearly constant at small loadings
and increases only for very large loadings when the applied stress is large stress is large enough to compress the particles.
212 C.S. Campbell / Powder Technology 162 (2006) 208–229

second great work in granular flow – which showed that beyond


a certain height the weight of a bed within a bin is supported by
friction on the sidewalls. Thus, the pressure on the bottom of the
bin is independent of bed depth. As that pressure controls the
flowrate through the orifice, the flowrate is depth-independent.)
But the techniques suffered from mathematical problems of
applying boundary conditions and the flowrate predictions
could have been better. Jackson [22] examines this in some
detail.
A likely source of the problems, is the assumption that ϕ is a
constant material property. Fig. 5 shows measurements of tanϕ
in two dimensional hopper flow simulation by Potapov and
Campbell [27]. In it, tanϕ can be seen to vary by more than a
factor of 3, violating the fundamental assumptions of quasistatic
flow theory. This variation of tanϕ can explain the discrepancies
between the theory and experiment. However, it is not
understood why tanϕ changes, as simple shear simulations on
similarly constituted materials indicate that tanϕ is a constant at
small shear rates (e.g. [28]).

Fig. 5. A contour diagram of the apparent friction coefficient, tanϕ, from a two- 4.1. The “frictional” nature of granular materials
dimensional simulation of a hopper with a 60° angle and a polydisperse granular
material, from Potapov and Campbell [27]. The annotations max and min
indicate the areas where tanϕ takes its maximum and minimum values,
Eq. (8) indicates that tanϕ is the ratio of shear to normal
quantitative values of which are written at the bottom of the plot. Note that tanϕ forces in the material and thus can be understood as an apparant
is far a constant, but changes by a factor of more than three. friction coefficient. Recently, it has become popular to refer to
quasistatic flows as “frictional”. However, this is misleading as
compressed together due to solid deformation at the contact the internal behavior of the material is not what one would
points and are squeezed into the interparticle pore space. But for classically call “frictional”.
a wide range of smaller loadings, the critical concentration is
independent of the applied stress. In many soil mechanics
applications, the applied stress can be large, (for example
beneath a large building). But in most granular flows, the
applied stresses are relatively small and as the total strains are
large, it is reasonable to assume that the flow is incompressible
and fixed at the critical concentration, νc.
These plasticity-derived techniques have been used widely in
soil mechanics to predict the failures of soils below foundations
and structures such as retaining walls and earthen dams. There
were also problems that became apparent. For example, the first
versions of these theories predict the material would continu-
ously expand with shearing and never approach a critical state
(e.g. [22]).
When extended to study granular flows, this technique has
had partial success in predicting the flow from hoppers (e.g.
Jenike and Shield [23], Davidson and Nedderman [24], and
Brennen and Pearce [25]). As that material flows within the
hopper, it is assumed that the material is always yielding so that:
s ¼ r tan/ ð8Þ
everywhere within the hopper. Furthermore, as the material
experiences large shear strains, it is always assumed to be at the
critical concentration, υc, and it is treated as incompressible.
There were many successes of these theories. In particular they
showed that the flowrate from a hopper was independent of the
depth of material, a characteristic that makes sand hourglasses Fig. 6. A photoelastic image of the force chains generated in the two-
an easily built method of timekeeping. (This is a direct dimensional shear cell of Howell et al. [29,30]. Here, the inner cylinder is
reflection of the 1895 analysis of Janssen [26] – perhaps the rotating counter-clockwise to force the particles together into chains.
C.S. Campbell / Powder Technology 162 (2006) 208–229 213

To see this requires understanding how the particles interact the stress ratio τxy/τyy, which in a non-cohesive material is the
inside a granular material. Fig. 6 shows a photoelastic picture of apparent friction coefficient or tanϕ, is related to the ratio Fx/Fy,
the interparticle forces inside a two-dimensional shear cell which depends only on the geometry of the force chain. As
[29,30]. For the photoelastic technique, the brightness of the chains form in the direction best suited to resist the applied
light surrounding a contact point is proportional to the force on forces, and as they collapse before they have rotated to any
the contact. This allows the force distribution within the significant degree, their geometry is roughly fixed and
material to be visualized. Notice that the forces are not evenly controlled by the applied force. Thus it is not surprising that
distributed throughout the material, but are concentrated in quasistatic flows demonstrate a friction-like response. However,
“Force Chains” (e.g. [31–33]). These are quasi-liner structures this response is not the result of frictional sliding as shown in
that support the bulk of the internal stress within the material. Fig. 7a, but a result of the internal structure of force chains
(Note that many of the interparticle contacts are unloaded, or shown in Fig. 7b.
nearly unloaded.) In a shearing material, these force chains are Interestingly, Rapid Granular Flow theories, which will be
dynamic structures. When the material shears, particles are described in the next section, also predict that the bulk friction
pushed together to form the chains. After it is formed, the chain coefficient τxy/τyy is a constant, (e.g. Lun et al. [34]), although
will be rotated slightly by the shear motion, but will quickly experiments (e.g. [35]) and computer simulations (e.g. [36])
become unstable and collapse. show it to be a weak function of the solid fraction ν. (The ν
While observing failure within a granular material, it was dependence is attributed to internal microstructure development
noticed that the failure occurs along narrow planes within the [37,38]). As the concentration increases, the particles arrange
material. This lead to a picture of two blocks of particles sliding themselves in a regular order that allows the material to shear at
over one another as shown in Fig. 7a. That picture evokes the a large concentration. That structure restricts the orientations of
standard model of sliding friction and thus accounts for the the available contacts between particles, the vector l in the
Mohr–Coulomb behavior (7). However, the slip planes are not contact stress tensor in Eq. (1) and thus affects the relative
true planes of infinitesimal thickness, instead are zones on the magnitude of the stress tensor components.) This occurs
order of ten particles across called “shear bands”. Inside the although there are no long duration solid–solid contacts in
shear band, the particles exist within force chains such as those Rapid Flows and thus, like Quasistatic flows, no true frictional
seen in Fig. 6. However, these still behave globally in a behavior. Thus Quasistatic and Rapid Flows are equally,
“frictional” manner, in the sense that the shear and normal frictional flows.
stresses are related. Consider the idealized force chain in a
simple shear flow, shown in Fig. 7b. Note that the x- and y- 5. Collisional or rapid granular flows
direction forces, Fx and Fy, are related by the force F acting
along the chain. Eq. (1) gives, Bagnold [39] was the first to try and model a granular
  material from the point of view of individual particles. He
sxy Fx ly imagined particles of radius d and density ρp at a solid
¼   ð9Þ
syy Fy ly concentration ν, in a shear flow with shear rate γ. (From here on,
it is assumed that γ is the magnitude of the gradient of x-
where lx and ly are the x- and y-direction components of the direction velocity u that points in the y-direction within a simple
vector l connecting the centers of the contacting particles. Thus shear flow. In common notation, γ = du/dy.) The stress tensor
must then vary as:
sij ¼ f ðm;qp ;d;gÞ ð10Þ
As ν is already dimensionless, the Buckingham Pi theorem
allows only one other dimensionless parameter:
sij
¼ fij ðmÞ ð11Þ
qp d 2 g2

or

sij ¼ fij ðvÞqp d 2 g2 : ð12Þ

Fig. 7. The source of the “frictional” nature of granular materials: (a) the
This is Bagnold's famous result indicating that the stresses
standard view that the material deforms as large blocks that interact frictionally should vary as the square of the shear rate γ. Bagnold justified
and slip occurs along slip planes within the material, thus interacting in a true the result with a simple model in which the first γ controlled the
frictional manner, (b) the forces are generated by the compression of Force degree of momentum exchange between particles and the
Chains. As by Eq. (1), the contact stresses are generated as averages of these second power of γ reflected the collision rate or the number of
forces, one can see that the ratio of shear to normal stress will be related to the
ratio of Fx to Fy, which are related through F and the angle of the chain. As the
momentum exchanges per unit time. The stresses then reflect
chain angles do not vary dramatically, this results in an apparent frictional the internal momentum transport due to interparticle collisions.
behavior. This is a valid interpretation of Eq. (12), but the equation itself is
214 C.S. Campbell / Powder Technology 162 (2006) 208–229

a result entirely of dimensional analysis and is completely


independent of any underlying model. In other words, any
model that uses the same dimensional quantities in Eq. (10), will
yield a result in the form of Eq. (12). Bagnold justified his
results with experiments performed on a suspension of particles
in a shear cell, but a recent re-examination of that data, Hunt et
al. [40], throws doubt on the interpretation of the experimental
results.

5.1. The granular temperature

Bagnold's picture of the individual granules moving in a


shear flow invariably brings comparison with the motion of
molecules in the kinetic theory of gases. Furthermore,
interparticle collisions will induce random velocities that are
reminiscent of the thermal motion of molecules. The magnitude
Fig. 8. The two modes of granular temperature production: (a) collisional mode;
of these fluctuating velocities is called the “granular tempera-
(b) streaming mode.
ture”.
1   1  2   2   2  
T ¼ j u V2 j ¼ uV þ vV þ wV ð13Þ
3 3 ature. While a product of the mean flow field, this appears as a
Notice that this is related to the trace of the streaming stress random velocity, because on the average, for every particle that
tensor described in Eq. (2). moves downwards in the velocity gradient – producing a
positive “random” velocity – another moves upwards –
1
T¼ Traceðt s Þ ð14Þ producing a negative “random” velocity. Although the gener-
3qv ated velocities are random in sign, they are not random in
As the streaming stresses arise from unsteady velocities, it direction. This mechanism only produces “random” velocity
should not be surprisingly that there is this close connection components in the direction perpendicular to the velocity
between τs and the granular temperature. The concept of a gradient, which is usually the direction of the mean flow
granular temperature was first introduced by Ogawa [41]. velocity (here, the x-direction). Thus the granular temperatures
If one takes the granular temperature as the exact analogue of are large in the mean flow direction. This results in normal-
the molecular temperature in the kinetic body theory, then one stress differences, with the largest normal-stresses in the flow
can use the formalisms of kinetic theory, for example the direction; at low concentrations, the flow direction temperature,
Chapman–Enskog method (e.g. [42]), to derive a set of can be several times larger than in the other directions.
governing equations for granular flow. From these ideas came Naturally, the streaming mechanism dominates at a small
the field of “Rapid Granular Flows”. As in the kinetic theory, it concentrations while the collisional mechanism dominates at
assumes that the granules interact by instantaneous collisions large concentrations.
and that all transport rates are governed by the granular As described above, granular temperature is only produced
temperature. As will be shown, both of these are problematic by velocity gradients. There are other mechanisms of granular
assumptions. But central to the whole issue is the granular temperature production. For example, shock waves, such as
temperature, which will be examined here in some detail. those produced in vibrated granular flows can generate granular
Just like the stresses in Eqs. (1) and (2), granular temperature temperature [9,43]. In multiphase flows, interactions between
is produced by a collisional mechanism (schematically shown particles and the interstitial fluid can also generate granular
in Fig. 8a) and a streaming mechanism (Fig. 8b). The collisional temperature. However, this paper will only consider shear-
mechanism is the most intuitively obvious. Because of the induced temperatures.
geometry of the impact, any collision between particles will act Granular temperature is one of the most confusing concepts
to randomize the impact velocity, thus converting the mean in granular flow. Typically, when granular temperature is
motion of the flow into granular temperature. The streaming measured, it is assumed to be a measure of the unsteady
mechanism (Fig. 8b) results from the motion of a particle components of velocity. But granular temperature reflects the
relative to a velocity gradient. velocity fluctuations on top of the mean velocity, and if the
Consider the streaming mechanism illustrated schematically mean velocity is itself unsteady, then one must be careful to
in Fig. 8b. Imagine a particle that starts in a high velocity region separate the unsteady mean velocity from the granular
of the flow and imagine also that its random motion has a temperature. To clarify these issues, one must understand the
component moving parallel to the velocity gradient, towards the role of granular temperature in kinetic theory modeling. There,
lower velocity regions of flow. In the time before its next the temperature drives the transport rate in the two principal
collision, the difference the between the mean flow velocity at modes of internal transport. Granular temperature provides the
its original and current positions appears as granular temper- relative velocity that drives particles together to collide,
C.S. Campbell / Powder Technology 162 (2006) 208–229 215

Unlike molecular systems, the interactions between macro-


scopic granules are inelastic. Hence the energy reflected in the
granular temperature is continually being dissipated away to
heat and must be re-supplied from the mean flow energy. As
both temperature generation mechanisms described in Fig.
8 derive from the shear motion, the granular temperature is
created by shear work. Thus rapid granular flows follow an
energy flow pattern shown in Fig. 9. Driving forces, e.g. gravity
or the motion of walls and other boundaries are converted into
the kinetic energy of the mean flow, which is converted to
granular temperature through shear work, and from there to
sensible heat through inelastic collisions or another dissipation
mechanism such as air drag. However, most rapid flow theories
only assume collisional dissipation. And, as collisions are
assumed to be instantaneous, the dissipation is represented as a
coefficient of restitution, ε, the ratio of the approach to recoil
velocities in the center of mass. Since granular temperature is
produced by shear work one might expect a relationship
between the shear rate γ and the magnitude of the granular
temperature. As given in Eq. (13), the granular temperature T
has units of (velocity)2. Thus a convenient dimensionless
scaling of the granular temperature and shear rate is the
parameter

dg
S¼ ð15Þ
T 1=2
Fig. 9. The energy flow pattern in a rapid granular flow (after Campbell [53]). first proposed by Savage and Jeffrey [44]. Measurements of S via
the computer simulations of Campbell [36] and comparison with
the predictions of Lun et al. [34] are shown in Fig. 10. The
general trend of the simulation data is predicted by the Lun et al.
resulting in the collisional transport mode (the contact stresses
in Eq. (1)). Also, granular temperature causes the diffusive
particle mixing that results in the streaming mode of transport
(Eq. (2)). Both arise because the granular temperature
generates relative motion between particles. But if the mean
flow is unsteady (for example, particles resting on a surface
that is gently vibrated with accelerations well below 1 − g), all
the particles will move together with the same unsteady
motions. But that unsteadiness does not drive relative motion
between particles and is thus not a granular temperature. A
slightly more confusing case is fluid turbulence, in which
particles may become entrained in turbulent eddies. As all the
particles within a given eddy move with the same mean, if an
unsteady mean, velocity, the turbulence does not generate
interparticle collisions and, in that sense, is not a granular
temperature. On the other hand, the eddies do induce
streaming transport of the particle momentum and, in that
sense, the turbulence acts like a granular temperature. In other
words, the unsteady motion in an eddy is only partially a
granular temperature as it generates streaming stresses but no
collisional stresses. Also, when two eddies collide the
difference in velocity between the eddies will induce collisions
between particles in different eddies, and will generate
collisional stresses. Thus, it is almost impossible to quantifi-
Fig. 10. The parameter S = dγ/T1/2 relating the magnitude of the granular
ably the portion of turbulence that fills the role of a granular temperature and shear rate as a function of the solid concentration, v. The lines
temperature. In general, it is difficult, if not impossible, to are the predictions of Lun et al. [34]. While the symbols are from the computer
determine the granular temperature in unsteady systems. simulations: (from Campbell [36], used with permission).
216 C.S. Campbell / Powder Technology 162 (2006) 208–229

theory, although the simulations are frictional while the theory


ignores friction. (Particle surface friction also dissipates energy;
this allows a simulation to produce steady state temperatures at
an otherwise elastic ε = 1.) There are a few interesting features
here. The first is that over most of the range of solid fractions ν, S
is of order 1. This means that the relative particle velocities
induced by the mean shear dγ, is of the same order as that
produced by the granular temperature T1/2. Also note that the
value of S generally decreases as the coefficient of restitution
increases, reflecting the fact that that the larger ε, the smaller the
dissipation rate and thus the larger the granular temperature
produced by a given shear rate. Perhaps the most striking feature
is that S → 0 as ν → 0, indicating that at small solids
concentration, the granular temperature becomes infinite. This
can be understood by first noting that, at small concentrations,
the stresses are largely generated by the streaming mechanism
(Eq. (2)) so that in a simple shear flow with flow in the x-
direction and the velocity gradient in the y-direction, the granular
temperature is generated by the shear work, τxyγ = ρρν b u′ν′ N γ ∼ ν
and is directly proportional to the solid fraction ν. The dissipation
of the granular temperature occurs through collision. The Fig. 11. The ratio of the x- and y-direction temperatures, Tx/Ty . Note the
probability of a collision is proportional to finding two particles anisotropy at small ν (after [36]).
in contact, and at small concentrations, where the position of one
particle has little effect on the position of any other particle, the
probability of finding a particle at any location is proportional to ν; 5.2. Rapid granular stresses
thus the probability of finding two particles in contact is
proportional to ν2, so that the collision rate (and the dissipation Rapid granular flows provide a forum in which it is possible
rate) are proportional to ν2. Consequently, the ratio of granular to observe the interplay between the collisional stresses (Eq. (1))
temperature production to dissipation ∼ν/ν2 ∼ 1/ν and becomes and streaming stresses (Eq. (2)). (This is largely because the
infinite as ν → 0. Thus, the granular temperature becomes infinite flow will be in the rapid regime whenever the concentration is
as ν → 0. Note that this only occurs because collisions are assumed small enough that streaming stresses become significant.) Fig.
to be the only dissipation mechanism. If there are other ways of 12 shows the dimensionless shear stresses, scaled as in Eq. (11),
dissipating energy (air-drag for example), this limit would not generated in a simple shear flow. Note that the simulation data
appear. In fact, there are no documented observations of this limit asymptotes to ∞ both as ν → 0 and as ν approaches a “random
other than in computer simulations. close pack” (the concentration of a randomly assembled volume
It should also be noted that, unlike the thermodynamic of spheres) at ν ≈ 0.64. The latter limit occurs because in a
temperature, the granular temperature is not isotropic. This random close pack, the particles are in intimate contact and the
should be obvious because the streaming mechanism of collision rate is approaching ∞; thus this asymptote occurs as a
temperature generation (Fig. 8b) only generates temperature singularity in the contact stress tensor (Eq. (1)). It is also a
in the direction perpendicular to the velocity gradient. Fig. 11 byproduct of the rigid particle assumptions that lie at the heart of
shows the ratio of Tx and Ty the x- and y-direction temperatures, rapid flow ideas; because the particles cannot deform, it would
from Campbell [36]. Note that at small ν, where the streaming require an infinite stress to shear a material near the random
mode of temperature generation is largest, Tx/Ty is significantly close pack. But the singularity would disappear for real particles
larger that unity reflecting the streaming generation of x- with finite elastic moduli, as by deforming their shape, the
direction temperature. But at large ν, where the collisional mode particles can be forced to shear at any concentration. In addition,
dominates, Tx/Ty is near unity indicating the more isotropic the stresses demonstrate a singular behavior as ν → 0. This is a
temperatures produced collisionally. Note also that the reflection of the fact that the granular temperature becomes
coefficient of restitution ε has a strong effect, with the largest infinite in that limit (as seen in Fig. 10). As the granular
anisotropies appearing at the smallest ε. The larger dissipation temperature is closely related to the streaming stresses, obvious
at the smaller ε reduces the collisional generation of temperature from Eq. (14), the stresses demonstrate the same asymptotic
and thus enhances the relative importance of the streaming behavior as the granular temperature.
mechanism, leading to larger temperature anisotropies. Note
that while higher order kinetic theories can produce normal 5.3. Rapid granular flow models
stress differences [46,47], they are strictly only valid in the limit
ν → 0; furthermore, they predict that the differences are The formalisms of gas kinetic theory can be used to derive a
functions only of the inelasticity and are independent of the set of equations for granular flow if the particles are assumed to
concentration. be rigid. In turn, the rigid particle assumption implies that all
C.S. Campbell / Powder Technology 162 (2006) 208–229 217

As the viscosity η has units of (mass)/(length-time), one can


write the viscosity as:

g ¼ qdfg ðm;eÞT 1=2 ð20Þ


Following the same ideas, the conductivity of the granular
temperature is:

a ¼ qdfa ðm;eÞT 1=2 ð21Þ


and the dissipation is:
qfC ðm;eÞ 3=2
C¼ T ð22Þ
d
Where fp, fη, fα and fΓ are unknown functions of the
dimensionless, concentration ν, and coefficient of restitution,
ε, that must be determined from experiment or analysis.
There are several problems that should be immediately
apparent with this formulation. The most obvious is that the
range of applicability of rapid flow theory is limited. Note that
from Eq. (17), the solid phase stresses are viscous in nature.
Thus systems such as hoppers cannot be modeled by rapid flow
theory. A key feature of hoppers is the frictional support of
material by the vertical walls of the bin (the effect explained by
Fig. 12. The dimensionless shear stress in a rapid flow from computer Janssen [26]). A viscous material produces no forces unless the
simulation. The open symbols are from the rough particle simulations [36]. The material is in motion and can thus provide no such support. As
solid symbols are from smooth particle simulations [45]. The lines are
predictions of rapid flow theory [35] (after [36]).
Janssen's frictional support appears in static materials, it cannot
be modeled the type of viscous material predicted by rapid-flow
contacts occur instantaneously. Thus, there is vanishing theory. There have been attempts to add a frictional response to
probability of multiple simultaneous contacts and only binary rapid flow models, most notably Johnson et al. [49,50]. But
contacts need be considered. From there, kinetic theory these are ad hoc models based on self-contradictory assump-
formalisms yield a set of Navier–Stokes like equations. This tions as they are formed by simply adding a rapid-flow model,
should not be surprising as the same formalism applied to gases based on instantaneous collisions, to a frictional model, based
also yield the Navier–Stokes equations. While there are minor on long duration contacts.
variations between theories, the basic equations are: Also, the gas kinetic theory on which rapid-flow models is
based, assumes that the molecular collisions are elastic in the
Conservation of mass: sense that they do not dissipate energy. In converting the theory
to granular materials, the inelasticity of granular impacts is
Dqm largely accounted for by the granular energy Eq. (18). But one
þ qmjdu ¼ 0 ð16Þ
Dt must compute integrals over a velocity distribution functions in
order to compute the constitutive relationships, (19–22) and the
Conservation of momentum: methods for computing the distribution function restrict the
Du system to “nearly elastic particles”, roughly ε=0.9 and above.
qm ¼ jpðq;m;T ;eÞ þ jdðgðq;m;T ;eÞjuÞ ð17Þ This severely limits the materials that may be modeled with
Dt
these methods.There is also a more subtle problem. Notice that
Conservation of granular energy (granular temperature): all of the constitutive laws in (19–22) obtain their rates through
the granular temperature T. This implicitly assumes that the
DT magnitude of the thermal velocities (T1/2) is much larger than
qm ¼ jdðaðq;m;T ;eÞjT Þ þ s:ju  Cðq;m;T ;eÞ: ð18Þ the relative velocities induced by the shear (dγ). In terms of Fig.
Dt
10 this means that S ≪ 1, which is only observed at extremely
In this last equation, α is the conductivity, τ : ∇u is the small solid concentrations. However, Fig. 10 shows that over
temperature production by shear work, and Γ is the dissipation most of the range of solid concentrations, S ≈ 1, indicating that
by inelastic collisions. Haff [48] gave a wonderful heuristic dγ ∼ T1/2. Thus, the mean shear and the temperature are equally
analysis that through dimensional analysis gives a set of scaling important in driving the relative motion between particles, the
laws for the various constitutive coefficients. As the pressure collision rate, and thus the transport rates in a rapid granular
has units of (mass)/(length-time2): flow. As the kinetic theories depend only on the temperature to
govern transport, they most likely to either underpredict the
p ¼ qfp ðm;eÞT ð19Þ transport rate or overpredict the granular temperature.
218 C.S. Campbell / Powder Technology 162 (2006) 208–229

Furthermore, as the relative velocities induced by the shear rate orientation into the problem, which similarly complicates the
are only in the direction perpendicular to the velocity gradient, notion of molecular chaos.
this introduces anisotropies in the angular distribution of Rapid-Flow theory has been used to generate models for
collisions about a particle (as measured by Campbell and binary mixtures, pioneered by Jenkins and Mancini [54].
Brennen [37]). Interestingly, the collisional anisotropy was Kinetic theory based models of segregation have also been
included in the earliest, albeit incomplete, rapid flow models, developed, but the results are quite mixed (see the review of
Savage and Jeffrey [44] and Jenkins and Savage [51], which Ottino and Khakhar [55]). Being probabilistic models, they
only considered contact stresses. To include streaming stresses, agree well with probabilistic Monte Carlo simulations, but the
required modifying the velocity distribution function, which agreement breaks down if compared against more realistic
proved intractable if the collisional anisotropy was included. As deterministic simulations. Some degree of agreement could be
the theories predict that S ≈ 1 (Fig. 10), they are not self- had only if the granular temperature is used as a fitting
consistent in that their predictions conflict with their implicit parameter. Khakhar et al. [56] argue that this is due to a
assumptions. Goldhirsch [52] cites Sela and Goldhirsch's [47] breakdown in the underlying kinetic theory, in that the frictional
comparison with normal stress difference data as evidence that dissipation is so large. But if kinetic theory assumptions
this effect is unimportant; but this argument is not applicable as breakdown in determining the granular temperature, it is
the comparison is done at ν = 0, the only point where the Sela difficult to argue that the same assumptions work well in
and Goldhirsch calculation is valid. There, Fig. 10 shows S ≈ 0 predicting segregation within the same flow. Besides, if the
(T = ∞) so, of course γd ≪ T1/2, and, while there may be theory cannot handle frictional particles, then it can handle no
collisional anisotropy due to the anisotropic granular temper- realistic materials.
ature, there will be no shear-induced collisional anisotropy Much of the recent effort in the areas of rapid granular flow
under the conditions of the Sela and Goldhirsch analysis. or kinetic theory have been directed towards issues that are of
Finally, at the heart of all kinetic theories is the largely academic interest, either because of unrealistic assump-
assumption of Boltzmann's “Stosszahlansatz” or molecular tions or because they are only of interest at small particle
chaos, that there are no correlations in the velocities or concentrations that are never found outside the laboratory.
positions of colliding particles. This is troubling because These include items such as the “cooling” of homogeneously
common granular flows occur at such large concentrations thermalized granular “gas” (first introduced by Haff [48]) which
that any given particle will interact many times with its can of course never be found in reality because there is no way
neighbors and it is likely their velocities will be strongly to create a homogeneously thermalized granular gas. Some of
correlated. In addition, the aforementioned microstructures the higher order kinetic theories (e.g. Sela and Goldhirsch [47])
[37,38] correlate the relative positions of particles. Thus true are valid only in the ν → 0 limit and thus inapplicable to any
molecular chaos is unlikely in real granular systems although realistic granular flow. The development of “inelastic micro-
it is difficult to estimate the degree of error introduced by this structure”, a clustering instability, first observed by Hopkins and
assumption. Lounge [57] has received much attention. At low concentra-
In 1990, I wrote a review article on the field of rapid granular tions, particles are observed to not be homogeneously
flows [53]. The article ended with a list of “Pressing Concerns” distributed but to form higher concentration clusters surrounded
designed to push the field towards more realistic systems and it by regions that are near voids. But this has little effect for the
is worth a paragraph to comment on the progress of the last 15 large concentrations of common granular flows, simply because
years. The concerns were: Material properties, Microstructure, the particles are already so tightly packed, there is no room for
Non-spherical particles, Non-uniform particle size and segre- the clusters and voids to grow. (It is somewhat inappropriate to
gation, Interstitial fluid effects and Solid/Fluid behavior of even refer to these disordered clusters as “microstructures”, as
granular systems. However, it should have been obvious, even that name implies an ordered “structure” of particles; as a result,
in 1990, that the first 3 topics would be almost intractable, either these are often confused, e.g. Goldhirsch, [52], with the ordered
because they complicated the collision integrals from which the high concentration microstructures that strongly affect the
constitutive properties are derived or because they violate the relative magnitude of the stress tensor components [37,38].)
assumptions of molecular chaos. For example, even simple In fact, much of the progress has been negative in the sense
properties such as a stick–slip surface friction make a that we are learning that more and more granular systems
discontinuity in the collision integrals; as a result, friction is cannot be described by rapid-flow theory. For example, many
only approximately incorporated in Rapid-Flow theories models have been developed that use rapid granular flow ideas
through a tangential coefficient of restitution. Also, friction to model the solid phase stresses in multiphase systems (one of
dissipates energy and as discussed above, if the energy the “Pressing Concerns” from Campbell [53]) mostly in gas-
dissipation is large enough, it may be possible to accurately fluidized systems (e.g., [58–60]). But direct measurements in
assess the velocity distribution function. Like the collisional fluidized beds [61] show that they cannot be modeled by kinetic
anisotropy, the development of internal microstructure affects theory [62,63]. The transition from solid-like to fluid-like
the contact angle between particles and it is difficult to include behavior, such as is seen at the boundaries of funnel flows in
in the kinetic theories, partially because of the complications to hoppers (the last Pressing Concern) has been shown not to be
the collision integrals and partially because it violates the the phase change suggested by Campbell [53], but instead
Stosszahlansatz. Non-round particle shapes bring the particle occurs outside the realm of rapid granular flows as a quasistatic-
C.S. Campbell / Powder Technology 162 (2006) 208–229 219

like failure [64,65]. And even vibrated beds, at least those of of the force is independent of the shear rate. But the chain
commercially significant depth, cannot be described by kinetic rotates at a rate ∼ γ and eventually become unstable and self-
theory [66]. destructs. Thus, the lifetime of the chain is proportional to 1/γ.
Consequently, the product of (formation − rate) × (lifetime) for a
6. Elastic granular flows chain is γ-independent and because the force generated is also
γ-independent, The stresses generated are quasistatic. This is
Recently, Campbell [28,67,68] has been able to unify the the Elastic–Quasistatic regime. It is the same as the old
various granular flow theories and, in particular, fill in the gap Quasistatic regime; the word Elastic is added to indicate that it
between the quasistatic and rapid flow regimes, and draw is a subregime of the global Elastic regime.
complete flowmaps for shearing granular materials. The But at high shear rates, the elastic forces in the chain must
missing link was to include the elastic properties of the particles absorb the inertia of the particles that are gathered in the chain,
into the models – in effect to put the solid back into granular requiring an extra force required to accelerate the particles in the
solids. The principle quantity is the interparticle stiffness k, as it chain so that it rotates at a rate proportional to the shear rate.
governs how particles “see” one another mechanically and, as Thus, even though the particles are locked in force chains, the
mentioned previously, determines the bulk elastic properties of forces generated must reflect the particle inertia. The force F
the granular material. It requires the exercise of little generated in the chain must have the form F = a + bγ where a is
imagination to see the importance of the stiffness to the the baseline elastic force and bγ is the inertial augmentation.
rheology of dense granular flows. At the large concentrations of Still the (formation − rate) × (lifetime) of the chain is γ-
common granular flows, particles are locked into force chains, independent so that the resultant stresses τ ∼ F ∼ a + bγ increase
such as those seen in the shear cell in Fig. 6. Assuming that the linearly with the shear rate. (This is shown in Campbell [67].)
walls the cell are rigid, then the degree of compression of the Naturally, there will be some inertial effect throughout the
force chains and thus the deformation of each contact, is Elastic regime, but for a wide range of flows, bγ ≪ a and the
determined only by the need for the material to shear at a given flows appear quasistatic. However, when bγ becomes of the
concentration. Thus, if each particle in Fig. 6 was removed and same the same order as a, i.e. the inertial forces become of the
replaced with one with, say, twice the stiffness, the forces on same order as the elastic forces, the flow transitions into the
each contact and thus the contact stresses would double. As the Elastic–Inertial regime in which the forces are linearly
streaming stresses are insignificant at such large concentration, proportional to the shear rate γ.
this means that the stresses are proportional to the contact The ratio of elastic to inertial effects is govern by a
stiffness. dimensionless parameter:
Now imagine a high-concentration granular material, with
the forces distributed in force chains, undergoing shear at k
k⁎ ¼ ð23Þ
constant volume. The shear will force particles together to form qd 3 g2
the chain, cause the force chain to rotate until it becomes
unstable and collapses. As the chain rotates, it will want to dilate Note that k / (ρd3γ2) = (τ / ρd2γ2) / (τd / k) and is thus the ratio of
the bulk material, but is prevented from doing so by the constant Bagnold's inertial to the elastic stress scalings. A similar
volume constraint; instead, the rotation compresses the chain parameter was first proposed by Babic et al. [69]. Campbell [28]
and generates an elastic response. If Δ is the overall gives several interpretations for this parameter, but the most
deformation of the chain, then the deformation of each contact useful is that k⁎ represents (d / δi)2 where δi is degree of
is δ = Δ/N = Δd/L where N = L/d is the number of contacts in a deformation expected from the impact by a particle moving at
chain of length L composed of particles with characteristic size the shear velocity, dγ, making k⁎ a measure of inertially induced
d. Thus the force F = kδ = kΔd/L. The generated stress τ ∼ F/ deformation, much as (τd/k) is a measure of elastic deformation.
d2 ∼ k/d. Hence an appropriate dimensionless scaling for the As such the value of k⁎ reflects the relative effects of elastic to
stress in the regime is τd/k. Note that following the above inertial forces, i.e. in principle at large k⁎ elastic forces dominate
argument, τd/k ∼ F/kd = δ/d; that is, τd/k can be interpreted as the and at small k⁎, inertial forces dominate.
particle deformation δ represented as a fraction of the particle Fig. 13 shows the dimensionless normal stress, scaled
diameter, d. elastically as τyyd/k plotted against the stiffness parameter. The
Campbell [28] divided the entire granular flow field into two plot is marked to show the division into the Elastic–Quasistatic
broad regimes the Elastic and the Inertial. The Elastic regimes and Elastic–Inertial regimes, which are differentiated solely by
encompass all flows in which force is transmitted principally the fact that the stresses are independent (Elastic–Quasistatic)
through the deformation of force chains for which the natural or dependent (Elastic–Inertial) on the shear rate γ, and thus on
stress scaling is τd/k. Continuing from the above arguments, the parameter k / (ρd3γ2). As expected, the flow transitions from
follow a force chain through its life cycle. The chain will form Elastic–Quasistatic to Elastic–Inertial as the shear rate
when particles are driven together by the shear rate γ and thus increases, (k/(ρd3γ2) decreases). Each point was taken for
the rate of chain production is proportional to γ. The chain then three particle diameters and as many as three stiffnesses and
rotates and is compressed. The degree of compression and thus thus each point represents up to 9 overlapping points illustrating
the force generated in the chain is determined by the necessity the robustness of the scaling. These data are for a single solid
of meeting the constant volume constraint; hence the magnitude fraction ν = 0.6, which lies below a random close-pack
220 C.S. Campbell / Powder Technology 162 (2006) 208–229

Fig. 13. The dimensionless normal stress τyyd/k, as a function of the stiffness parameter k/(ρpd3γ2), showing the separation into the Elastic–Quasistatic and Elastic–
Inertial regimes. This data was taken in shear flow simulations with constant stiffness at a concentration v = 0.6 and particle surface friction μ = 0.5 (after [28]).

concentration; hence, these stresses are a byproduct of the shear μ = 0.5. At μ = 0.1, force chains are weak and the flow
flow. In the absence of shear, the particles need not to be in transitions from Elastic to Inertial behavior. The Inertial
contact and no stress would be generated. Note that the regime encompasses flows where force chains cannot form
dissipation rate reflected through the coefficient of restitution ε, and the momentum is transported largely by particle inertia as in
is only of importance for the smallest k / (ρd3γ2) (or the largest Rapid-Flow theory (although as shall be shown, inertial
shear rates, γ), making these flows very different from Rapid behavior does not necessarily mean a Rapid Flow). In Inertial
Flows in which the granular temperature is critically dependent flows, the stresses are independent of the stiffness k and the
on the dissipation rate. (This should be apparent from the strong natural scaling is the Bagnold scaling τ / ρd2γ2.
effect of the coefficient of restitution in Fig. 12.) But notice that the μ = 0.1 data in Fig. 14, is only Inertial at
Fig. 14 shows the effect of the particle surface friction large k/(ρd3γ2) and transitions back to Elastic–Inertial behavior
coefficient μ on the dimensionless normal stress, τyyd / k. While as γ increases (as k/(ρd3γ2) is reduced). This leads to the ironic
μ has only a weak effect on the shear to normal stress ratio τxy / conclusion that Inertial flows occur at only small shear rates. It
τ yy (the bulk friction coefficient), it has a major effect on the also implies that increasing the shear rate can compel the
normal stresses [28]. The μ = 0.1 data in Fig. 14, appears to go formation of force chains at lower concentrations than one might
zero at large k / (ρd3γ2), but actually the values are about two normally expect. This can be understood from an alternate
orders of magnitude smaller than the corresponding points for interpretation of k / (ρd3γ2), in particular that k / (ρd3γ2) ∼ 1

Fig. 14. The effect of the particle surface friction coefficient μ on the dimensionless normal stress τyyd/k, as a function of the stiffness parameter k / (ρd3γ2) for a fixed
concentration, ν = 0.6. Surprisingly, the surface friction has a strong effect on the normal stress because it affects the strength of force chains. In fact, for μ = 0.1, the flow
transitions from Elastic to Inertial behavior at large k / (ρd3γ2) (after [28]).
C.S. Campbell / Powder Technology 162 (2006) 208–229 221

(γTbc)2 where Tbc is the binary contact time, the duration of a The stresses for Elastic Flows, dominated by force chains,
contact between two freely colliding particles. (Physically, this naturally scale as τd / k. Thus, if an Inertial Flow, for which the
last can be thought of as the ratio of 1 / γ, the time scale relevant scaled stress τ / ρd2γ2 is independent of k/(ρd3γ2), is plotted on
to how quickly particles are drawn together by the shear flow, to an elastically scaled log–log plot, τyyd/k vs. k/(ρd3γ2), the
Tbc, a scale characteristic of how quickly the elastic contact inertial flows should plot with a slope of − 1. At the same time if
forces push the particles apart). In other words, at small k / an Quasistatic–Elastic flow, τd/k = const., were plotted on an
(ρd3γ2), the shear rate is driving particles together at rates inertially scaled plot τyy/ρd2γ2 vs. k/(ρd3γ2), it would have a
comparable to those at which the elastic forces are driving them slope of 1. Fig. 15 shows plots of the normal stress τyy, for
apart, so that at large shear rate, force chains may form at smaller various solids concentrations with both scalings. It is easy to
concentrations than in relatively quiescent systems. distinguish the Elastic and Inertial regimes in each figure.

Fig. 15. Normal stress data for a variety of solid concentrations, ν. (a) Elastic scaling, τyyd / k: here Elastic–Quasistatic flows plot as flat lines and Inertial flows plot with
a slope of − 1. (b) Inertial Scaling, τyy/ρd2γ2: Here Inertial flows plot as horizontal lines and Elastic–Quasistatic flows have a slope of 1. In all cases, the surface friction
μ = 0.5 (after [28]).
222 C.S. Campbell / Powder Technology 162 (2006) 208–229

But notice in Fig. 15 that the change from elastic to inertial many particles are interacting simultaneously. This means that
behavior occurs quite abruptly with as little as 1% change in the Inertial regime is divided into two sub-regimes, Inertial–
concentration. At large k/(ρd3γ2), the flow is elastic at ν = 0.59, non-Collisional and Inertial–Collisional (the Rapid-flow
but inertial at ν = 0.58. And that change is accompanied by huge regime).
changes in stress. At k/(ρd3γ2) = 107, the stress changes by more Thus granular flows can be divided into two global regimes,
than two orders of magnitude from Elastic flow at ν = 0.59, to Elastic and Inertial. The Elastic regime is dominated by force
the Inertial flows at ν = 0.58. This reflects the fact that the elastic chains and is divided into the Elastic–Quasistatic regime and
force chains can support much larger forces than can particle the Elastic–Inertial regime depending on whether there is a
inertia. Notice also that for each concentration shown, the flow noticeable dependence of the stresses on the shear rate. Both
behaves elastically for small k/(ρd3γ2) even those that behave Elastic subregimes have the same physical underpinnings and it
inertially at large k/(ρd3γ2), demonstrating once again that high is difficult to draw a precise division between the two. The
shear rates can force a transition from Inertial to Elastic– Inertial regime, which is free of force chains and had has
Inertial behavior. stresses that scale with the square of the shear rate, can be
For the constant stiffness models used here, the binary divided into the Inertial–non-Collisional regime and the
contact time Tbc is a fixed constant. In other words, for any Inertial–Collisional (or Rapid-Flow) regime depending on
collision between two particles, the contact time must be exactly whether the dominant particle interaction is binary collision.
Tbc. The only way that a contact can endure for a longer time is Campbell [70] examines the confinement of a contact by
if, during the contact period, one or more additional particles infinitely chains of surrounding particles (a non-collisional
make contact with one of the colliding particles and prevents case) and found that as long as the chain is unloaded, the
them from moving apart. On the other hand, if particles are confinement only the extended the contact time by a few
locked in force chains, each particle is interacting simulta- percent. Thus, one can conclude that tc/Tbc must be almost
neously with, at least, its near neighbors in the chain and the exactly one for the flow to be truly collisional. With this
contact will endure for the lifetime of the chain: as force chains information one can draw the entire flowmap encompassing all
are created, rotated and destroyed by the shear, the inverse shear the regimes of granular flow, such as that shown in Fig. 17.
rate 1 / γ, is a characteristic time of the chain lifetime. Thus, There are several interesting features of this figure. Note that
another property that can be used to probe the flow is the it is generally assumed that, at small shear rates, a flow will
average contact time, tc. If the ratio, tc/Tbc = 1, the particles are behave quasistatically, and that by increasing the shear rate, one
interacting collisionally, i.e., it is a rapid flow. On the other will eventually end up in the Rapid-Flow regime. But his flow
hand, if tc/Tbc ∼ 1/γ(∼[k / (ρd3γ2)]1/2), the flow is dominated by map shows that picture to be false. If one starts in the Elastic–
force chains, i.e. it is an Elastic flow. Fig. 16 shows tc / Tbc for Quasistatic regime and increases the shear rate (decreasing k /
the same data plotted in Fig. 15. It is easy to see the Elastic (ρd3γ2)) at a fixed ν, one eventually reaches Elastic–Inertial
behavior as the points line up with slope 1/2. The points behavior. The Rapid-Flow, or that matter any of the Inertial
corresponding to inertial behavior deviate from the slopping flow regimes, will never be reached except by reducing the
lines. But note that for all the data shown, tc/Tbc N 1. Thus even concentration. This makes sense because at such large
the Inertial flows are not Collisional – they are not Rapid concentrations, there will always be force chains, and increasing
Flows. Instead, they are dominated by conditions in which the shear rate will not make them go away. Thus at constant

Fig. 16. The ratio of the average contact time to the binary contact time, tc / Tbc as a function of the stiffness parameter, k / (ρd3γ2). The data is from the same set shown
in Fig. 15 (after [28]).
C.S. Campbell / Powder Technology 162 (2006) 208–229 223

Fig. 17. A flowmap showing the division into the four sub-regimes as a function of the concentration ν, and the stiffness parameter, k/(ρd3γ2). The data are for μ = 0.5
(after [28]).

volume, one cannot leave the Elastic regimes by changing the pressure on the granular contacts and the pressure in the water.
shear rate, one can only cause transitions between Elastic– In a drained test, the water is allowed to leak out, so that all the
Quasistatic and Elastic–Inertial behaviors. forces are supported across the particle contacts. And therein
Also note that Rapid-Flows (the Inertial–Collisional regime) lies a possible difference between controlled-stress and
ironically appear at low-shear rates, so that, in effect, Rapid- controlled-volume systems. In controlled-stress systems, the
Flows are the least rapid of the flow regimes. Starting with a material must, through the action of elastic and/or inertial
Rapid Flow, and increasing the shear rate at constant ν, one forces, support the applied load under all conditions; at small
progresses through Inertial–non-Collisional flow and finally to shear rates, even a small applied stress forces particles into
Elastic–Inertial flow. This transition reflects the phenomena contact, and form force-chains. Under controlled volume, it is
discussed earlier that very large shear rates can force particles possible for the stress to fall by orders of magnitude with a small
together at a rate comparable to the time it takes the elastic concentration change, as seen in the transition from Elastic to
forces to break them apart, making it possible to form force Inertial behavior at large k/(ρd3γ2), because there is nothing to
chains at remarkably small concentrations if the shear rate is force the particles into contact.
large enough. Fig. 18 shows the dependence of the solid concentration ν,
But most granular flows do not occur at fixed concentration. on the dimensionless applied stress τ0d/k, and the stiffness
Common granular flows such as chutes and hoppers, as well as parameter k/(ρd3γ2). Remember that τ0d/k represents the
landslides, all have a free surface. Thus the stress level is deformation of a particle as a fraction of its diameter. Thus,
controlled by an overburden of material. But the concentration the large concentrations seen at large τ0d/k are simply a
is not fixed because the materials are free to expand toward the reflection of the compressibility of the particles; i.e. at τ0d/
free surface as needed to balance the applied stress. The k = 0.1, the particles are pressed until their centers are 10%
concentration may vary slightly, even immeasurably so, but it closer together. Note that for a range of the stiffness
still may have an important effect on the rheological behavior. parameter, the concentration is independent of k/(ρd3γ2).
Thus it is possible that the behavior of stress-controlled systems This is Casagrande's critical state behavior illustrated in Fig.
is different from those for which the concentration is controlled. 2. Then one by one, starting with the smallest k/(ρd3γ2) (the
There are indicators of this from the long history of soil largest shear rates), the lines diverge from the critical state
mechanics. There it is common to perform “drained” and line. The line marked “Low-Stress Critical state Concentra-
“undrained” test on water-saturated soil samples – which tion” is particularly interesting. It shows that there is a range
demonstrate remarkably different behaviors even when per- of τ0d/k and k/(ρd3 γ2 ) for which the concentration is
formed on the same soil samples. In a drained test the water is independent of either parameter. This is partially because
allowed to leak out as the material is sheared, while in undrained the stresses are too small to cause a noticeable change in the
test, the water is confined within the sample. The significance is concentration; for example one would expect an insignificant
that the water is a nearly incompressible fluid and if the pore- change in concentration between τ0d/k = 10− 3 (0.1% defor-
space is filled with water (i.e. the soil is saturated), then the mation) and τ0d/k = 10− 6 (0.0001% deformation). And at the
water keeps the volume of the sample from changing. In such same time, the shear rate is too small (k/(ρd3γ2) is too large)
cases, an applied load can be supported by a combination of the for the inertia of the particles to support the applied load.
224 C.S. Campbell / Powder Technology 162 (2006) 208–229

Fig. 18. The relationship between the solid concentration ν, applied stress, τ0d/k, and k⁎ = k/(ρd3γ2) in a stress-controlled simulation with μ = 0.5 (from Campbell [68],
reprinted with permission).

One interesting correspondence between controlled stress ior. (Another indicator can be found from changes in the ratio of
and controlled volume flows is that the Low-Stress Critical shear to normal stresses.) And as in Fig. 16, contact time data is
State concentration in controlled stress flows corresponds to the used to determine the transition to Inertial behavior. (See [68]
bounding concentration between the Elastic–Quasistatic– for details.)
Elastic and Inertial regimes in controlled-volume flows. This This allows flowmaps, like that in Fig. 19, to be drawn for
makes sense, if one thinks of the transition as the minimum controlled-stress flows. (Note that vertical ordinate in Fig. 19 is
concentration at which force chains can form without help from the dimensionless applied stress, τ0d/k, instead of the solid
the particle inertia. Thus in controlled volume flows, it fraction, ν used in Fig. 17.) The behavior is very different from
represents the transition between flows with and without force the controlled volume case in Fig. 17. Most noticeable is that
chains, that is between Elastic and Inertial Flows. In controlled- collisional flows are found at small k/(ρd3γ2) instead of at the
stress flows, it represents the minimum concentration that can large k/(ρd3γ2) in Fig. 17. This means that increasing the shear-
support the applied load without inertial help. But this will turn rate causes the expected transition from Elastic–Quasistatic →
out to be one of the few correspondences between controlled- Elastic–Inertial → Inertial flows, nearly the opposite of what is
volume and stress flows. And indeed the difference makes itself seen at constant volume. Here one sees this behavior because
apparent even here. As it is possible to create elastic structures the increased shear rate causes changes, often small changes, in
with a wide range of concentrations (Fig. 17 shows elastic the concentration. One might also notice that there are no
behavior as low as ν = 0.45), the first appearance of inertial Inertial–non-Collisional flows delineated in Fig. 19. The data
effects in controlled-stress flows – the first deviation from the [68] indicate that the transition from Elastic–Inertial to
low-stress critical state line – does not correspond to the Inertial–Collisional behavior occurs very rapidly in controlled
disappearance of force chains, only the first sign of inertial stress flows. The Inertial–non-Collisional regime is very
support. Thus, it is an Elastic–Quasistatic/Elastic–Inertial narrow and can only be observed in a few isolated cases.
transition instead of the Elastic–Quasistatic/Inertia–non-Col- In addition, Elastic–Inertial behavior appears over a much
lisional transition seen in controlled volume tests. Hence, wider range of k/(ρd3γ2) at constant stress, than at constant
although the transitions occur at the same concentrations in volume. This is also easy to understand. Remember that
controlled-stress and volume, they are physically distinct Elastic–Inertial flows appear when the inertial forces become
transitions. comparable to the elastic forces. At constant volume, the
In controlled-stress flows, the magnitude of all the stresses elastic forces are generally so large that it takes very large
are more or less fixed by the applied stress τ0. It is not possible shear rates to generate comparable inertial forces. But under
to examine, say the effect of strain-rate γ on the stresses, controlled stress, the magnitude of the elastic force at zero
because the magnitude of the stress is fixed by τ0 and changing shear rate, is the applied load and can be as large or small as
γ will only change the concentration. Thus indirect means are desired. Hence at small applied loads, it is easy for the inertial
used to determine the flow-regime by making analogies to the forces to become of the same order as the elastic forces and
observations made in controlled-volume flows. For example, thus possible to have Elastic–Inertial behavior at relatively
the deviation from critical state behavior in Fig. 18 is taken as small shear rates. For the fixed concentration data in Campbell
one indicator of Elastic–Quasistatic to Elastic–Inertial behav- [28] at ν = 0.6, Elastic–Inertial behavior was observed for k/
C.S. Campbell / Powder Technology 162 (2006) 208–229 225

Fig. 19. A flowmap of controlled-stress granular flows. The particle coefficient of restitution, ε = 0.7 and particle surface friction, μ = 0.5 (from Campbell [68], reprinted
with permission).

(ρd3γ2) b 104 which, for 1 mm sand (which Campbell [68] the maps indicate why rapid granular flows are seldom
estimates to correspond to about k = 1.5 × 104 N m− 1), required observed. Fig. 19 indicates that, for τ0d/k = 10− 5 (which
about 775 s− 1 of shear – an unrealistically large shear rate corresponds to an overburden of a 10-particle-deep layer), the
indicating that Elastic–Inertial behavior will never be observed flow becomes collisional at k/(ρd3γ2) = 106. Continuing with
for sand under controlled-volume conditions. (However, 2-mm the 1 mm sand numbers, this corresponds to a very large shear
plastic beads should assume elastic inertial behavior at a more rate of about 100 s− 1. This is similar to the minimum shear
accessible, γ = 10 s− 1.) But that is not so for the stress- rates observed for larger glassbeads in 10 particle deep shear
controlled studies shown here. For example, Fig. 19 indicates cell experiments [71]. Now, Campbell [68] points out that the
that for μ = 0.5 and τ0d/k = 10− 5 (ten particles deep), Elastic– transition shear rate should scale with particle diameter as
Inertial behavior can first be observed k/(ρd3γ2)∼5 × 107 or γd = 0.1 m s− 1. While there are no direct experiments to
about γ = 14 s− 1 in 1 mm sand-like materials. The reason for compare these numbers to, Wang and Campbell [71] found
this is obvious from the underlying physics. Elastic–Inertial from measurements on three sizes of glassbeads, that their
behavior occurs when the inertial forces are of the same order minimum achievable shear rates were somewhat higher at
as the elastic forces. In fixed-concentration flows [28], the γd = 0.34 m s− 1 (requiring a minimum of about 340 s− 1 of
magnitude of the elastic forces were dictated by the shear for 1 mm particles). Deeper depths would require even
requirement that the particles undergo shear at the large larger shear rates and while such shear rates can be obtained in
concentrations where force chains form. Reducing the laboratory shear cells and computer simulations, they are not
concentration to the point that the force chains disappeared found in practice. One should note that these limiting shear
eliminates the elastic forces and the stresses drop by orders of rates are relatively insensitive to the particle stiffness. Going to
magnitude; this is possible because the concentration was a softer material would decrease k and with it k/(ρd3γ2),
independent of the stresses generated. In constant-stress flows, pushing the system towards an inertial flow, but at the same
as long as the shear rate is small enough, force chains can time would proportionally increase τ0d/k, pushing the system
always form even for very small applied stresses, simply away from inertial behavior, so that, from Fig. 19, a change in
because if the inertia effects are small, force chains are the only k is unlikely to cause a change in flow regime. While high
method available to the balance the applied load. And for small speed laboratory shear cells can achieve the hundreds of
applied stresses, the corresponding elastic forces are small and inverse seconds of shear rate required to achieve Rapid Flow,
it only takes a small shear rate to generate inertial forces that shear rates in the hundreds of inverse-seconds, will not be
are comparable to the elastic forces, pushing the flow into the found in any common granular flows, such as chutes and
Elastic–Inertial regime. As a result, Elastic–Inertial behavior hoppers, all of which are gravity driven. One might argue that
is much more accessible in controlled stress flows. Further- under reduced gravity conditions, say on the Moon or Mars,
more, under controlled stress, when progressing from Elastic– smaller shear rates are required. While this is certainly true,
Quasistatic to Inertial behavior, the flow must always go reducing gravity also reduces the driving force for the bulk
through an intermediate Elastic–Inertial regime. motion and reduces the available shear rates proportionally.
One advantage of having complete flowmaps, is that one Hence, reducing gravity does not make rapid flows any more
can set bounds on that various flows regimes. For example, accessible.
226 C.S. Campbell / Powder Technology 162 (2006) 208–229

7. Conclusions But the largest problem with the rapid flow theory is that huge
granular temperatures are required to support any reasonable
This paper is an attempt to summarize what is currently overburden of material. And as the granular temperature
known about granular material flows. It began with a derives from the shear rate, huge shear rates are required to
discussion of basic transport mechanisms. It then goes on to produce enough pressure to support even small overburdens of
discuss contact laws and points out that real contact behavior material. Estimates of the required granular temperature
is much more complex than Hertz and Hertz–Mindlin models. derived from the Elastic models, and by the limited
This was followed by a discussion of the two limiting granular experimental evidence that is available, indicate that rapid
flow models, the Slow-Flow or Quasistatic model, and the flows are unobtainable outside of high-speed laboratory shear
Rapid-Flow or Collisional model. It finally describes the cells and gravity-free computer simulations.
Elastic model, which, by including particle elasticity, allows Surprisingly, despite its 27-year history, as of this writing,
the entire flowmap for granular flow to be drawn, including there have been no other attempts to validate the fundamental
the Quasistatic and Rapid-Flow regions and the area in assumptions of rapid granular flow theory – a long silence that,
between. in itself, speaks volumes. Monte Carlo simulations provide and
The Quasistatic models were derived from metal plasticity thus assume rapid flow behavior; they cannot be used to probe
theory using a Mohr–Coulomb frictional yield condition. As its limits. Neither can rigid particle (or event-driven) computer
applied to granular flows, these generally assume that the simulations, such as those used to derive Figs. 10–12, as they
friction angle ϕ (tanϕ is the apparent friction coefficient of likewise, presuppose a collisional flow. As it stands, Rapid Flow
the material) is a constant material property, and that the is a theory in search of an application as steadily, more and more
material is in the critical-state and thus behaves incompres- systems are shown not to be rapid flows. In addition to the fact
sibly with the density fixed at the critical-state concentration. that common granular flows cannot generate large enough shear
While there are mathematical problems in using these models, rates to enter the rapid flow regime, there have also been direct
particularly in the application of boundary conditions, they measurements of the particle forces in bubbling fluidized beds,
have been used to model hoppers and are able to predict the indicate that they cannot be modeled by kinetic theory ideas
qualitative features of hopper flow and the predicted flow [61–63]. Even vibrated beds, at least those of commercial
rates are in the same ballpark as experiments. The deviations depth, several hundred to thousands of particle deep, cannot be
may be attributed to the aforementioned problems in thermalized and are not rapid flows [66]. After 27 years, it is not
boundary conditions, etc., but simulations of hopper flow outrageous to ask for some verification of the Rapid-Flow
suggest that the error lies in the assumption of a constant theories. Thus, it probably best to abandon rapid-flow studies, at
friction angle ϕ. It is not yet understood why ϕ changes least until a practical application for the results of the theory can
within the hopper. be found.
The Rapid-Flow models that have been so popular in the By including the particle stiffness – which governs how
last few decades, are also derived using models from another particles see each other mechanically and thus governs the
branch of science, in this case the kinetic theory of gases. elastic properties of the bulk material – it is possible to draw
They assume that the granules behave like molecules and, to out the entire flowmap that connects the Quasistatic and
apply kinetic theory ideas, assume that the particles interact by Rapid-Flow regimes, at least for simple granular materials.
instantaneous collision. Superimposed on the velocity field is Flows are divided into two global regimes, the Elastic and the
another field quantity, the granular temperature representing Inertial. Elastic flows are dominated by force chains and the
the random, thermal-like, kinetic energy of the particles. stresses scale directly with the interparticle stiffness as τd/k.
Because the particle collisions dissipate energy, the granular which is interpreted as the elastic contact deformation as a
temperature must be continually re-supplied from the energy fraction of the particle diameter. This regime is further
of the mean flow field through the mechanism of shear work. subdivided into the Elastic–Quasistatic (the old Quasistatic
The granular temperature is assumed to control all the regime) and the Elastic–Inertial regimes, depending on
transport rates within the granular material; in particular, the whether the shear rate significantly affects the magnitude of
pressure and all the internal normal stresses are proportional to the stresses. The relative magnitude of inertial and elastic
the granular temperature. However, there is a problem with effects are governed by a parameter, k⁎ = k/(ρd3γ2), interpreted
this assumption; as the granular temperature is produced as the inverse-square elastic deformation on a contact due
through shear work, it has roughly the same magnitude as the solely to the particle inertia, again as a fraction of the particle
square of γd, the shear rate times the particle diameter. Thus, diameter. Inertial Flows are free of force chains and internally
the transport rates are equally controlled by the shear rate as transmit force by interparticle inertia and the stresses scale
by the granular temperature. As such, Rapid-Flow theory is with the Bagnold scaling, τ/ρd2γ2. They can also be divided
not self-consistent with its basic assumptions. Also, just as into two flow regimes, the Inertial–non-Collisional for which
kinetic theory derives the equations of fluid motion, rapid flow many particles interact simultaneously, and the Inertial–
theory predicts a viscous-like response from a granular Collisional regime (the old Rapid-Flow regime), where
material, meaning that rapid-flows cannot demonstrate any particles interact by binary collisions.
of the solid-like properties (such as Janssen's frictional Somewhat surprisingly the rheological behavior appears to
support) commonly associated with bulk granular material. change with system-scale constraints, in particular whether the
C.S. Campbell / Powder Technology 162 (2006) 208–229 227

concentration is determined by controlling the volume of the 8.1. Rigid sphere models are inapplicable to dense systems
system or by controlling the applied stress. This has a direct
analog in drained and undrained tests on water-saturated soils. It Dense systems interact by force chains and transmit force
is generally assumed that at small shear rates, the behavior is along the chain by elastically deforming the interparticle
quasistatic and, by increasing the shear rate, will eventually contacts. The elasticity of the particles is intrinsically important
become rapid. However, that behavior was only observed under to the internal processes of the system. Such systems cannot be
controlled-stress conditions. For controlled-volume flows, there modelled as rigid spheres and any such model would be missing
is no path between quasistatic and rapid flows, except by essential physics. Also, contrary to intuition, granular materials
decreasing the density. By simply varying the shear rate, one are, in the bulk, quite soft. This is evidenced in their sound
can only move between Elastic–Quasistatic and Elastic– speed which is of the order of 100 m s− 1, roughly 50 times
Inertial flows or between Elastic–Inertial and Inertial flows slower than the soundspeed in their constituent solid material,
depending on the specified concentration. Surprisingly, Rapid and indicates that the bulk granular material has an apparent
Flows (Inertial–Collisional flows) appear only at small shear elastic modulus, more than three orders of magnitude smaller
rates under controlled volume conditions; increasing the shear than its constituent solid.
rate forces particles into force chains and causes a transition to
Elastic–Inertial behavior. In controlled-stress flows, the volume 8.2. Particle surface friction is essential to modelling dense
can change to balance the applied stress. Thus starting with a systems
quasistatic flow and increasing the shear rate, one can end up
with a rapid flow, because the transition brings about a slight Fig. 14 shows that the simple act of removing surface
increase in volume. friction can cause a transition between an Elastic and Inertial
At this point, one can only speculate whether the elastic flow regime. Friction is vital to the strength of force chains
effects can explain the variation in tanϕ seen in the hopper flow and force chains are essential for the Elastic flow regimes.
plots shown Fig. 5. Campbell [53] shows that the apparent Thus, while it is convenient to remove surface friction in
friction coefficient τxy/τyy (the equivalent of tanϕ in simple many analyses or computer simulations, it is removing
shear), decreases with k/(ρd3 γ2), throughout the Elastic– essential physics from the problem and will lead to erroneous
Inertial regime and becomes constant in the Elastic–Quasistatic behavior. At the same time, the importance of surface friction
regime. An examination of the data of Babic et al. [69], may be dwarfed by the importance of particle shape. Angular
indicates that effect is even more severe for two-dimensional particles have a greater tendency to lock together and will
disc flows such as those used in the hopper simulations in Fig. 5. thus both more easily form force chains and form stronger
Thus, if the hopper flow were in the Elastic–Inertial regime, this force chains. The effect of particle shape has yet to be
indicates that tanϕ should increase with shear rate. This is explored in detail. However, results on frictional spheres have
consistent with Fig. 5 as the largest tanϕ are found in the hopper some value. One can at least buy frictional spherical particles;
throat where the shear rates are the largest, and the smallest tanϕ frictionless particles, spherical or otherwise do not appear in
are in the bin section where the shear rate is small. But the nature.
sensitivity to system-scale constraints may provide another
possible explanation. A hopper flow is far from a simple shear 8.3. Dense flows are not “frictional”
flow and it may not be possible to directly apply simple shear
results to hoppers. This was dealt with in detail in the text, but is worth
It is interesting that the elastic properties of the particles do repeating here. While it is true that slow granular flows (and for
not appear in either the quasistatic or rapid flow theories. But that matter rapid flows) appear frictional because in the bulk, the
remember that both quasistatic and rapid-flow modeling were ratio of shear to normal stress is a constant, it is not a result of
based on the formalisms derived from other branches of science, solid–solid friction. Instead the frictional behavior of dense
of metal plasticity and kinetic theory respectively, and neither granular materials is a reflection of the evolution of the force
formalism allows the introduction of elastic properties. For chain structure.
quasistatic theory, all that is required is a yield surface and a
flow rule. In rapid flow theory, particle interactions are assumed 8.4. All important granular flows are dense
to be instantaneous collisions and thus assume a infinitely rigid
solid with no allowance for finite elasticity. Thus, it would be This is not so much an observation of this paper, as much as
most difficult, if not impossible to adapt these theories to an observation of real life. Hoppers, reasonably deep chute
include elastic effects. flows and even commercial scale vibrated boxes, all operate in a
near packed state.
8. Some lessons
8.5. Rapid Flows are seldom, if ever encountered outside the
As noted in the Introduction, this paper was derived from a laboratory
lecture given to the summer Powder Technology class at Ohio
State. Hence, it is appropriate to highlight some important This has also been dealt with in great detail above, but
lessons learned in the course of the paper. deserved repeating here if only because the scarcity of rapid
228 C.S. Campbell / Powder Technology 162 (2006) 208–229

granular flows is the physical reason behind the last lesson, [18] C.A. Coulomb, Essai sur un application de règles de maximis et minimis à
that all important flows are dense. Simply put, the shear rates quelques problèmes de staique relatifs à l'architecture, Mém. Math. Phys.
Acad. R. Sci., Paris 7 (1773) 343–382.
obtained in common flows cannot generate enough granular [19] V.V. Sokolovski, Statics of Granular Media, Pergammon Press, 1965.
temperature to support the material at small concentrations, [20] A. Casagrande, Compaction Tests and Critical Density Investigations in
leaving all flows in or near the Elastic regimes. Cohesionless Materials for Franklin Falls dam, U.S. Engineer Corps,
Boston District, 1938.
[21] A. Schofield, P. Wroth, Critical State Soil Mechanics, McGraw-Hill, New
8.6. System-scale constraints can affect the rheological
York, 1968, 310 pp.
behavior [22] R. Jackson, Some mathematical and physical aspects of continuum models
for the motion of granular materials, in: R.E. Meyer (Ed.), Theory of
This is reflected in the different rheological behaviors Dispersed Multiphase Flow, Academic Press, New York, 1983,
observed for controlled-volume and controlled stress flows as pp. 291–337.
well in the difference between drained and undrained tests in [23] A.W. Jenike, R.T. Shield, On the plastic flow of Coulomb solids beyond
failure, J. Appl. Mech. 26 (1959) 599–602.
soil mechanics. There may be many other system constraints of [24] J.K. Davidson, R.M. Nedderman, The hour-glass theory of hopper flow,
importance that are yet to be found. Trans. Inst. Chem. Eng. 51 (1973) 29–35.
[25] C. Brennen, J.C. Pearce, Granular material flow in two-dimensional
Acknowledgements hoppers, J. Appl. Mech. 45 (1978) 43–50.
[26] H.A. Janssen, Versuche über getreidedruck in silozellen, Z. Ver. Dtsch. Ing.
39 (1895) 1045–1049.
The author would like to thank Professor L. S. Fan for [27] A.V. Potapov, C.S. Campbell, Computer simulation of hopper flows, Phys.
inviting me to give the lecture that inspired this article and for Fluids, A Fluid Dyn. 8 (1996) 2884–2894.
his hospitality while I was at Ohio State. [28] C.S. Campbell, Granular shear flows at the elastic limit, J. Fluid Mech. 465
(2002) 261–291.
[29] D.W. Howell, R.P. Behringer, C.T. Veje, Fluctuations in granular media,
References Chaos 9 (1999) 559–572.
[30] D.W. Howell, R.P. Behringer, C.T. Veje, Stress fluctuations in a 2D
[1] D. Jacques, Sands, Powders, and Grains: An Introduction to the Physics of granular Couette experiment: a continuous transition, Phys. Rev. Lett. 26
Granular Materials, Springer, New York, 2000. (1999) 5241–5244.
[2] C.S. Campbell, Boundary interactions for two-dimensional granular flows: [31] A. Drescher, G. De Josselin de Jong, Photoelastic verification of a
Part I. Flat boundaries, asymmetric stresses and couple stresses, J. Fluid mechanical model for the flow of a granular material, J. Mech. Phys. Solids
Mech. 247 (1993) 111–136. 20 (1972) 337–351.
[3] H. Hertz, Über die Berürung fester elasticher Körper, J. Reine Angew. [32] P.A. Cundall, O.D.L. Strack, A discrete numerical model for granular
Math. 92 (1882) 156–171. assemblies, Geotechnique 29 (1979) 47–65.
[4] K.L. Johnson, Contact Mechanics, Cambridge Universtiy Press, 1985 [33] D.M. Mueth, H.M. Jaeger, S.R. Nagel, Force distribution in a granular
452 pp. medium, Phys. Rev., E 57 (1998) 3164–3169.
[5] R.J. Bathurst, L. Rothenburg, Micromechanical aspects of isotropic [34] C.K.K. Lun, S.B. Savage, D.J. Jeffrey, N. Chepurniy, Kinetic theories for
granular assemblies with linear contact interactionss, J. Appl. Mech. 55 granular flow: inelastic particles in Coutte flow and slightly inelastic
(1988) 17–23. particles in a general flowfield, J. Fluid Mech. 140 (1984) 223–256.
[6] J.D. Goddard, Nonlinear elasticity and pressure-dependent wave speeds in [35] S.B. Savage, M. Sayed, Stresses developed by dry cohesionless granular
granular media, Proc. R. Soc. 430 (1990) 105–131. materials in an annular shear cell, J. Fluid Mech. 142 (1984) 391–430.
[7] J. Duffy, R. Mindlin, Stress–strain relations and vibrations of a granular [36] C.S. Campbell, The stress tensor for simple shear flows of a granular
medium, J. Appl. Mech. 24 (1957) 585–593. material, J. Fluid Mech. 203 (1989) 449–473.
[8] H.A. Makse, N. Gland, D.L. Johnson, L.M. Schwartz, Why effective [37] C.S. Campbell, C.E. Brennen, Computer simulation of granular shear
medium theory fails in granular materials, Phys. Rev. Lett. 83 (1999) flows, J. Fluid Mech. 151 (1985) 167–188.
5070–5073. [38] C.S. Campbell, The effect of microstucture development on the collisional
[9] A.V. Potapov, C.S. Campbell, Propagation of elastic waves in deep stress tensor in a granular flow, Acta Mech. 63 (1986) 61–72.
vertically shaken particle beds, Phys. Rev. Lett. 77 (1996) 4760–4763. [39] R.A. Bagnold, Experiments on a gravity-free dispersion of large solid
[10] B.O. Hardin, F.E. Richard, Elastic wave velocities in granular soils, J. Soil particles in a Newtonian fluid under shear, Proc. R. Soc. Lond., A 225
Mech. Found. 94 (1963) 27–42. (1954) 49–63.
[11] F.E. Richart, R.D. Woods, J.P. Hall, Vibration of Soils and Foundations, [40] M.L. Hunt, R. Zenit, C.S. Campbell, C.E. Brennen, Revisiting the 1954
Prentice Hall, New York, 1970. suspension experiments of R.A. Bagnold, J. Fluid Mech. 452 (2002) 1–24.
[12] O.R. Walton, Numerical simulation of inelastic, frictional particle–particle [41] S. Ogawa, Multitemperature theory of granular materials, Proc. US–Japan
interactions, in: M.C. Roco (Ed.), Particulate Two-Phase Flow, Butter- Seminar of Continuum–Mechanical and Statistical Approaches in the
worth-Heinemann, Bouston, 1993. Mechanics of Granular Materials, 208, Gakujutsu Bunken Fukukai, Tokyo,
[13] C. Thornton, Coefficient restitution for collinear collisions of elastic 1978.
perfectly plastic spheres, J. Appl. Mech. 64 (1997) 383–386. [42] S. Chapman, T.G. Cowling, The Mathematical Theory of Non-Uniform
[14] O.R. Walton, R.L. Braun, Viscosity and temperature calculations for Gases, third editionCambridge University Press, 1970.
shearing assemblies of inelastic, frictional disks, J. Rheol. 30 (1986) [43] A. Goldshtein, M. Shapiro, C. Gutfinger, Mechanics of collisional motion
949–980. of granular materials: Part III. Self-similar shock-wave motion, J. Fluid
[15] C.V. Raman, The photographic study of impact at minimal velocities, Mech. 316 (1996) 29–51.
Phys. Rev. 12 (1918) 442–447. [44] S.B. Savage, D.J. Jeffrey, The stress tensor in a granular flow at high shear
[16] M. Mullier, U. Tuzun, O.R. Walton, A single-particle friction cell for rates, J. Fluid Mech. 110 (1981) 255–272.
measuring contact frictional-properties of granular materials, Powder [45] O.R. Walton, R.L. Braun, Stress calculations for assemblies of inelastic
Technol. 65 (1991) 61–73. spheres in uniform shear, Acta Mech. 63 (1986) 73–86.
[17] R.D. Mindlin, H. Deresiewicz, Elastic spheres in contact under varying [46] I. Goldhirsch, N. Sela, Origin of normal stress differences in rapid granular
oblique forces, J. Appl. Mech. 20 (1953) 327–344. flows, Phys. Rev., E 54 (1996) 4458–4461.
C.S. Campbell / Powder Technology 162 (2006) 208–229 229

[47] N. Sela, I. Goldhirsch, Hydrodynamic equations for rapid flows of smooth [61] K. Rahman, C.S. Campbell, Particle pressures generated around bubbles in
inelastic spheres, to Burnett order, J. Fluid Mech. 361 (1998) 41–74. gas-fluidized beds, J. Fluid Mech. 455 (2002) 103–127.
[48] P.K. Haff, Grain flow as a fluid-mechanical phenomenon, J. Fluid Mech. [62] C.S. Campbell, Granular flows and gas fluidization, in: U. Arena, R.
134 (1983) 401–430. Chirone, M. Miccio, P. Salatino (Eds.), Fluidization, vol. XI, ECI NY,
[49] P.C. Johnson, R. Jackson, Frictional–collisional constitutive relations for 2004, pp. 21–36.
granular materials with applications to plane shearing, J. Fluid Mech. 176 [63] C.S. Campbell, Granular flows and gas fluidization, Int. J. Chem. React.
(1987) 67–93. Eng. 2 (2004) 1 http://www.bepress.com/ijcre/vol2/P1.
[50] P.C. Johnson, P. Nott, R. Jackson, Frictional–collisional equations of [64] Y. Zhang, C.S. Campbell, The interface between fluid-like and solid-like
motion for particulate flows and their applications to plane shearing, J. behavior in two-dimensional granular flows, J. Fluid Mech. 237 (1992)
Fluid Mech. 210 (1990) 501–535. 541–568.
[51] J.T. Jenkins, S.B. Savage, A theory for the rapid flow of identical, smooth, [65] C.S. Campbell, The transition from fluid-like to solid-like behavior in
nearly elastic particles, J. Fluid Mech. 130 (1983) 187–202. granular flows, in: C. Thornton (Ed.), Powders and Grains 93: 2nd
[52] I. Goldhirsch, Rapid granular flows, Annu. Rev. Fluid Mech. 35 (2003) International Conference on Micromechanics of Granular U.K. Media,
267–293. Birmingham, July 12th–16th, 1993, pp. 289–294.
[53] C.S. Campbell, Rapid granular flows, Annu. Rev. Fluid Mech. 22 (1990) [66] S. Klongboonjit, The Effect of Particle Elasticity on the Convection in
57–92. Deep Vertically Shaken Particle Beds, PhD thesis, University of Southern
[54] J.T. Jenkins, Mancini, Kinetic theory for binary mixtures of smooth, nearly California, Los Angeles, 2005.
elastic, spheres, Phys. Fluids, A Fluid Dyn. 1 (1989) 2050–2057. [67] C.S. Campbell, Elastic granular flows, Int. J. Chem. React. Eng. 2 (2004) 2
[55] J.M. Ottino, D.V. Khakhar, Mixing and segregation of granular materials, http://www.bepress.com/ijcre/vol2/P2.
Annu. Rev. Fluid Mech. 32 (2000) 55–91. [68] C.S. Campbell, Stress controlled elastic granular shear flows, J. Fluid
[56] D.V. Khakhar, J.J. McCarthy, J.M. Ottino, Mixing and segregation of Mech. 539 (2005) 273–297.
granular materials in chute flows, Chaos 9 (1999) 594–610. [69] M. Babic, H.H. Shen, H.T. Shen, The stress tensor in granular shear flows
[57] M.A. Hopkins, M.Y. Lounge, Inelastic microstructure in rapid granular of uniform, deformable disks at high solids concentrations, J. Fluid Mech.
flows of smooth disks, Phys. Fluids, A Fluid Dyn. 3 (1991) 47–57. 219 (1990) 81–118.
[58] J.L. Sinclair, R. Jackson, The effect of particle–particle interactions on the [70] C.S. Campbell, A problem related to the stability of force chains, Granul.
flow of gas and particles in a vertical pipe, AIChE J. 35 (1989) 1473–1486. Matter 5 (2003) 129–134.
[59] J. Ding, D. Gidaspow, A bubbling fluidization model using kinetic theory [71] D.G. Wang, C.S. Campbell, Reynolds' analogy for a shearing granular
of granular flow, AIChE J. 36 (1990) 523–538. material, J. Fluid Mech. 244 (1992) 527–546.
[60] D.L. Koch, A.S. Sanagni, Particle pressure and marginal stability limits for
a homogeneous monodisperse gas-fluidized bed: kinetic theory and
numerical simulations, J. Fluid Mech. 400 (1999) 229–263.

Das könnte Ihnen auch gefallen