Segregation of large particles in dense granular flows: A granular Saffman effect?
K. van der Vaart,1, 2 M. P. van Schrojenstein Lantman,2 T. Weinhart,2 S. Luding,2 C. Ancey,1 and A.R. Thornton2
Environmental Hydraulics Laboratory, École Polytechnique Fédérale de Lausanne, Écublens, 1015 Lausanne, Switzerland
2
Multi-Scale Mechanics, ET and MESA+, University of Twente,
P.O. Box 217, 7500AE Enschede, The Netherlands
(Dated: May 22, 2017)
We report that the lift force on a single large particle segregating in a monodisperse dense granular
flow is correlated with a downstream velocity lag. This correlation suggests a viscous-inertial origin
for the lift force, similar to the Saffman lift force in (micro) fluids. This insight is relevant for modelling of particle-size segregation and our approach opens up a new avenue for the numerical study
of granular flows. For dense granular flows our findings suggest an interplay between polydispersity
and viscous-inertial forces, striking new parallels with fluids and suspensions.
Size-polydispersity is intrinsic to non-equilibrium systems like granular materials. It gives them the ability to
size-segregate when agitated, a process which spatially
separates different sized grains [1–3], but is different from
phase separation in classical fluid systems. Particle-size
segregation in dense granular flows [4, 5] has been intensively studied [e.g. 6–22], but a fundamental question
remains unanswered: why do large particles segregate?
It is generally understood that in dense granular flows
both small and large particles are pushed away from high
shear regions [7, 8] or pulled by gravity [9, 10]. The reason for the eventual separation of the two species is that
small particles move more effectively. They can carry
proportionally more of the kinetic energy [12–15] and are
also more likely to move into the gaps between larger
particles; a process referred to as kinetic sieving [9, 10].
However, the concept of kinetic sieving breaks down when
the large-particle concentration is very low.
Current models for size-segregation perform well when
the small and large-particle concentrations are nearly
equal [8, 23–27]. When accounting for the effect of sizesegregation asymmetry [28, 29], models have been extended to more unequal concentrations, but they remain
inaccurate in the limit of low large-particle concentrations. Extending models to this limit is important since
the presence of a few large particles strongly effects the
bulk flow [19]. Moreover, current models are either completely or partly phenomenological. Thus, to advance
modelling of this non-equilibrium system, we must understand the microscopic origin of segregation to derive
the free-variables from their microscopic quantities.
Recent constitutive models for dense granular flows include a particle-size dependence [30, 31]. However, they
only account for an average size. If we are to implement
size-distributions in these constitutive models a better
understanding of micro-scale effects between large and
small particles also seems crucial.
In contrast, particle migration in suspensions, in the
limit of low concentrations, is generally well understood [e.g. 32, 33]. Arguably this progress has been aided
by the fact that the fluid forces acting on a particle can
be calculated, which can not be said for granular media.
Thus, we are inspired to treat the particles that surround
an intruder as a continuum and attempt to understand
the forces acting on this particle based on the measured
continuum fields.
In this letter we study single large intruders in monodisperse dense granular flows down inclines, in which
large particles are well-known to segregate to the freesurface. In order to allow for long-time averaging of
forces, we attach an intruder particle to a spring perpendicular to the plane, as shown in Fig. 1. This prevents
the intruder from rising continuously.
Methods.—We use MercuryDPM, based on discrete
particle methods (MercuryDPM.org; [34, 35]), and investigate three-dimensional (3D) flows of mixtures of
spherical dry frictional particles flowing down an incline of θ = 22°. All simulation parameters are nondimensionalized such that the particle density is ρp = 6/π
and the gravitational acceleration is g = 1, with vertical
component gz = cos θ. The simulations are conducted in
a box with dimensions (x, y, z) = (30, 8.9, 40), with periodic walls in the x and y directions. The particles that
zi
h ⇡ 32
arXiv:1705.06803v1 [cond-mat.soft] 18 May 2017
1
Fsp
z
y
30
zI
x
✓ Fg
FIG. 1. Schematic of the simulations: 3D mono-disperse granular flow down an incline, with angle θ = 22◦ . Only base and
surface particles are shown. The flow contains three intruder
particles that are held with springs around three different zpositions zI (intruder positions in the schematic are to scale),
but move freely in the x-y plane.
2
make up the bulk of the flow have a diameter db = 1.
The intruder diameter is varied between di = 0.5 and
3.2, where the subscript i is not an index but refers to
“intruder”. The rough base consists of particles of diameter 0.85 and the flow height is h = 32 ± 0.5.
We place three identical intruders in the flow at vertical
positions zI = 5, 15 and 23 (see Fig. 1). Each intruder
is attached to a spring [36], which applies a vertical force
Fsp = −k(zi − zI ) proportional to the vertical distance
between the intruder position zi and its corresponding
zI . Here k = 20 is the spring stiffness. We also simulate
k = ∞ by fixing the intruder at zi = zI . Our findings
are independent of k, so unless stated otherwise all data
reported are for k = 20. We do not discuss the data for
zI = 5 because the intruder experiences boundary effects,
likely due to layering near the bed, as reported in [37].
A linear spring-dashpot model [37, 38] with linear elastic and linear dissipative contributions is used for the
normal forces between particles. The restitution coefficient for collisions rc = 0.1 and the contact duration
tc = 0.005. This results in a different stiffness depending on the particle size. We verified that our findings
are not the result of this difference in stiffness nor the
dependence on rc and tc . The friction coefficient for contacts between bulk particles µbb and between bulk and
intruder particles µbi equals 0.5, unless otherwise stated.
Applying coarse-graining (CG) [37, 39, 40], after a
steady state has been reached, we obtain time-averaged
3D continuum fields for ν the local solids fraction,
and σ the stress tensor, which satisfy the conservation
laws. The CG-width w is 1.0 [39]. We approximate
the bulk solids fraction at the position of the intruder
ν(xi , yi , zi ) = νi = Vi /Ṽi using the ratio of the particle volume Vi = 34 π(di /2)3 and the Voronoi volume
Ṽi , which we obtain through 3D weighted Voronoi tessellation (math.lbl.gov/voro++;[41]). All error-bars correspond to a 95% confidence interval.
Results.—The time-averaged spring force Fsp (downwards) exerted on an intruder is a measure for the
magnitude of a segregation lift force (upwards). Figure 2(a) shows that Fsp grows when the intruder size
ratio S = di /db increases. Guillard et al. [36] recently
introduced this method and applied it to large intruders
in 2D flows. They showed that the total upward force
Ftot = −Fsp + Fg , with Fg = ρp gz Vi the gravity force,
has two components: one scaling with a shear gradient
and the other with a pressure gradient. These scaling
laws predict the direction of segregation of large particles
in different flow configurations depending on whether a
shear or pressure gradient has the strongest contribution.
Unfortunately, they do not explain the origin of the lift
force. We take a different approach and decompose the
total upward force as Ftot = FL + Fb , where FL is a lift
force, which differs from Fsp , and Fb is a buoyancy force,
which results from the hydrostatic pressure and is not
2
3
0
2
10-3
1
-2
0
-4
-1
-6
-2
-8
0
1
2
3
0
1
2
3
FIG. 2. (a) The force Fsp experienced by an intruder as a
function of S. (b) The intruder lag λx as a function of S and
normalized by the average bulk velocity at zi . Dashed line is
a fit of λx = Sa − a, with a = 2.7 × 10−3 .
always equal to Fg .
Our first crucial finding is that the downstream velocity vxi of a large intruder experiences a lag λx =
vxi − vx (zi ) with respect to the average downstream velocity vx of the medium. Figure 2(b) shows that a large
intruder, S > 1, lags (λx < 0), while for a same sized
intruder, S = 1, there is no lag, within the fluctuations. The lag appears to scale as λx = Sa − a (with
a a constant), which suggests that it tends to a limiting
value. Interestingly, but outside the scope of this letter,
for S < 1, when the intruder is expected to sink, λx flips
sign and becomes a velocity raise (increase).
Pressure.—We look for the origin of the lag in the pressure field P = Tr(σ)/3. Figure 3(a) shows the crosssection P (x, 0, z) for different size ratios. For S ≤ 1
the pressure is hydrostatic, i.e., P ≈ PH = νρp gz (h − z),
which is expected for this type of flow [42]. Here ν is constant through-out the flow. For S > 1, P changes and
a high pressure region develops at the bottom-front side
of the intruder. Pressure fluctuations of lower magnitude
also appear around the intruder. This demonstrates that
the presence of a large particle modifies the local pressure around it. Although it is known that pulling an
object through a granular medium affects the local pressure [43, 44], here the intruder is not actively pulled but
can freely flow in the x-y plane.
In order to isolate the non-hydrostatic effects in the
pressure we visualize PL = P − PH . Figure 3(b) shows
that for S ≤ 1 PL is zero, within the fluctuations, while
PL increases for S > 1 and is characterized by positive
regions (over-pressure) in the lower right and upper left
quadrants, and negative regions in the lower left and upper right quadrants. It seems reasonable now to correlate the lift and the lag of a large intruder to the nonhydrostatic pressure.
Buoyancy & Lift Force.—Different definitions for granular buoyancy forces exist [e.g 36, 45]. In order to deter-
3
FIG. 3. (a) Cross-sections P (x, 0, z) around the intruder, centered at the origin, at zI = 15. The blue circle (diameter di )
corresponds to the intruder. The edge of the white circle (diameter di + db ) corresponds to the position of the first layer of
bulk particles. (b) Cross-sections PL (x, 0, z), where PL = P − PH , around the intruder at zI = 15.
mine the lift force FL on the intruder we derive a new
definition of the buoyancy force Fb resulting from PH .
Taking inspiration from [45] and using our approximation ν(xi , yi , zi ) = νi at the intruder position, we integrate PH over the surface Ãi of Ṽi . With the divergence
theorem we find:
Z
Z
Fb =
PH n · ez dÃi = νρp gz
dṼi = νρp gz Ṽi (1)
and tends to a finite value above S = 2. The plot of
(Fb + FL )/Fg shows that there is an optimal size ratio for
segregation, in agreement with experimental findings [6]
and predictions [46].
Saffman Lift.—Next we investigate whether the lag
and the lift force are correlated. Such a correlation is
known to exist in fluids: The Saffman lift force is found to
scale with the velocity lag of a particle in a fluid [47, 48]:
Here n is the normal outward vector to Ãi and ez is the
upward unit vector. Substituting Ṽi = Vi /νi we obtain:
FSaffman = 1.615 ρ(η|γ̇|)0.5 λx d2i sgn(γ̇)
V˜i
Ãi
Fb =
ν
ρp gz Vi
νi
(2)
Figure 4(a) shows that the measured νi strongly depends on S and is in fact bigger than ν for S > 1.
Therefore the important consequence of the ratio ν/νi
in Fb is that for S > 1 the buoyancy force will be less
than the gravity force Fg on the intruder, as shown in
Fig. 4(b) where Fb /Fg < 1. When S = 1, ν equals νi ,
and the buoyancy force balances gravity. In the limit of
S → ∞, νi → 1 and thus Fb corresponds to the buoyancy force in a fluid with density ρ = νρp . Our treatment
results in a buoyancy force that differs from the classical Archimedean buoyancy definition Fb = νρp gz Vi in a
granular fluid, which has two problems: it is independent of S, and more importantly, predicts that Fb < Fg
if S = 1. If this were true, there must be an additional
force to balance Fg , but we measure no such force.
We can now determine the lift force FL similar to the
way we obtained PL . We subtract Fb from the total
upward force on the intruder: FL = Ftot − Fb . Where
we have obtained Ftot = Fnz + Ftz from the measured
vertical normal and tangential contact forces, Fnz and
Ftz , respectively. In Fig. 4(b) we see that FL /Fg is approximately zero for S = 1, increases rapidly for S > 1
(3)
where η is the fluid viscosity, γ̇ = ∂z vx (zi ) the shear-rate
and ρ the fluid density. Saffman [47] considered the limit:
ρλx d
2η
ργ̇d2
4η
0.5
1
(4)
where the first term is the Reynolds number for the lag
Reλx and the second term is the shear-rate Reynolds
number Reγ̇ . This relation physically corresponds to the
viscous and inertial forces in the fluid being of similar
magnitude. The square root in Eqs. (3) and (4) follows
from this assumption. Whether Eq. (4) is valid for dense
granular flows in general remains to be seen, nonetheless
it is valid for our system; we find Reλx = O(10−3 ) and
Reγ̇ = O(10−1 ), using ρ = νρp , with η and γ̇ measured
from CG-fields far away from the intruder particle.
In order to test if a Saffman-like relation exists between
FL and λx we take the functional form for λx , fitted in
Fig. 2(b), and multiply it with d2i sgn(γ̇):
FL = bλx d2i sgn(γ̇) = b(
a
− a)d2i sgn(γ̇)
S
(5)
where a and b are likely functions of η and γ̇ as in Eq. (3).
We fit Eq. (5) to FL in Fig. 4(b), using the value for
4
1
2
0.9
1.5
0.8
1
0.7
0.5
Ṽi
0.6
1
0.5
0
0.5
0.4
1.5
-0.5
0
1
2
3
4
-1
0
1
2
3
0
FIG. 4. (a) Local intruder solids fraction νi versus S. Different (almost collapsing) symbols correspond to intruders with
µbi = 0.5, µbi = 0, zI = 15, zI = 23, k = 20 and k = ∞. Solid line is a fit of (ν − 1)S c + 1, with c = −1.2. The schematic
depicts the Voronoi volume Ṽi (dotted octagon) of the intruder (dashed circle). (b) The forces Fb , FL and FL + Fb , normalized
by Fg , for zI = 23. The solid red line is a fit of Eq. (5) with a = 2.7 × 10−3 and b = 3.7 × 103 . (c) The forces Fb , FL , Fnz and
Ftz , normalized by Fg , for S = 2.4, µbi = 0 and 0.5, at zI = 15.
a obtained in Fig. 2(b), and find that it captures the
trend in FL . This demonstrates that the lift force and
the lag are linked through a Saffman-like relation, thus
suggesting a viscous-inertial origin for the segregation of
large particles at low large-particle concentrations.
We think that a buoyancy force not compensating for
the weight of a large intruder is a thought-provoking insight. To further support this concept we present one
additional result. If the intruder-bulk friction µbi = 0 we
find that FL is reduced, as shown in Fig. 4(c), and critically, that a large intruder sinks instead of rises. The
measured total upward force Ftot = Fnz + Ftz on the intruder is lower than Fg , which demonstrates that Fb is
less than Fg . Interestingly, since FL does not completely
disappear it should have both a geometric and frictional
component. We also confirmed that PL is reduced but
does not disappear. The sinking of a frictionless particle
agrees with a recent study on the segregation of monodisperse mixtures of high and low friction particles, where
it was found that the low friction particles sink [49].
Discussion.—We report that a single large particle in
a dense granular flow gives rise to a non-hydrostatic pressure field around it. This coincides with a velocity lag λx
and a lift force FL , coupled through Eq. (5), causing the
particle to rise against gravity. These findings suggest
that the ‘squeezing’ in the mechanism of squeeze expulsion [10]—which has been used to qualitatively explain
the segregation of large particles in dense granular flows
since 1988—is a granular equivalent of the Saffman effect;
a viscous-inertial lift force [47, 48].
While FL scales with λx and the particle surface area,
the buoyancy force Fb on a large particle is reduced by a
factor ν/νi (Eq. (2)). We predict that as the intruder size
increases, λx will reach a constant value and because the
buoyancy force ultimately tends to the fluid buoyancy
with density ρ = νρp , gravity will outgrow the upward
force and the particle will sink. This could provide a
physical explanation for the observations of sinking of
very large particles [50, 51], as well as an optimal size
ratio for segregation [6, 46] and the unexplained trend
for Ftot (S) in Fig. 6 by Guillard et al. [36].
A Saffman-like lift force suggests that viscous-inertial
effects play a role in polydisperse dense granular flows.
This certainly warrants further investigation because of
the relevance for our understanding of these flows. The
fact that our system falls within the limits for the fluid
Saffman effect (Eq. (4)) is reassuring. However FSaffman
is also characterized by a square root dependence on the
shear-rate γ̇ (Eq. (3)). It remains to be verified if this
holds for dense granular flows to irrefutably establish a
granular Saffman effect as the cause for segregation of
single large particles. Thus, determining the full dependency of the lift force on η, γ̇, and µbi is as crucial as
investigating the change in FL when the large-particle
concentration increases.
The observed lag must originate from a horizontal drag
force, but drag forces on a free-flowing object in a granular medium have received little attention so far [45]. We
are investigating whether a Stokesian drag, found by Tripathi and Khakhar [45] for a heavy sinking mono-disperse
intruder, also holds for segregating large intruders. Obtaining the drag on free-flowing particles is an important
endeavour for the rheology of granular flows in general,
but foremost because drag is a cornerstone of models for
particle-size segregation in dense granular flows.
We have not discussed small size ratios S < 1, because
our continuum assumption of the medium breaks down
and the model predicts a downward force. This seems unphysical since a small particle falls through gaps between
the larger bulk particles. Nonetheless, the measured velocity raise (instead of a lag) warrants investigation.
The approach of treating the granular medium sur-
5
rounding a particle as a continuum and obtaining the
force laws can potentially be extended to every particle
in a system, leading to new micro-inspired macroscopic
models.
KV and MS contributed equally to this study. The
authors would like to acknowledge Chris G. Johnson
for suggesting the Saffman effect and François Guillard for helpful email correspondence and commenting
on the manuscript. The authors acknowledge support
from the Swiss National Science Foundation grant No.
200021 149441 / 1 and the Dutch Technology Foundation STW grant STW-Vidi Project 13472.
[1] J. C. Williams, Powder Technol. 15, 245 (1976).
[2] A. Rosato, K. J. Strandburg, F. Prinz, and R. H. Swendsen, Phys. Rev. Lett. 58, 1038 (1987).
[3] J. M. Ottino and D. V. Khakhar, Annu. Rev. Fluid.
Mech. 32, 55 (2000).
[4] GDR-MiDi, Eur. Phys. J. E 14, 341 (2004).
[5] P. Jop, Y. Forterre, and O. Pouliquen, Nature 441, 727
(2006).
[6] L. A. Golick and K. E. Daniels, Phys. Rev. E 80, 042301
(2009).
[7] Y. Fan and K. M. Hill, Phys. Rev. E 81, 041303 (2010).
[8] Y. Fan and K. M. Hill, New J. Phys. 13, 095009 (2011).
[9] J. A. Drahun and J. Bridgwater, Powder Technol. 36, 39
(1983).
[10] S. B. Savage and C. K. K. Lun, J. Fluid Mech. 189, 311
(1988).
[11] C. R. K. Windows-Yule, B. J. Scheper, A. J. van der
Horn, N. Hainsworth, J. Saunders, D. J. Parker, and
A. R. Thornton, New J. Phys. 18, 023013 (2016).
[12] K. M. Hill and D. S. Tan, J. Fluid Mech. 756, 54 (2014).
[13] T. Weinhart, S. Luding, A. R. Thornton, A. Yu, K. Dong,
R. Yang, and S. Luding, in AIP Conf. Proc., Vol. 1542
(AIP, 2013) pp. 1202–1205.
[14] L. Staron and J. C. Phillips, Phys. Rev. E 92, 022210
(2015).
[15] D. R. Tunuguntla, T. Weinhart, and A. R. Thornton,
Comp. Part. Mech. 1, 1 (2016).
[16] S. Wiederseiner, N. Andreini, G. Épely-Chauvin,
G. Moser, M. Monnereau, J. M. N. T. Gray, and C. Ancey, Phys. Fluids 23, 013301 (2011).
[17] L. Staron and J. C. Phillips, Phys. Fluids 26, 033302
(2014).
[18] M. Harrington, J. H. Weijs, and W. Losert, Phys. Rev.
Lett. 111, 078001 (2013).
[19] L. Staron and J. C. Phillips, Comp. Part. Mech. 3, 367
(2016).
[20] A. N. Edwards and N. M. Vriend, Phys. Rev. Fluids 1,
064201 (2016).
[21] Y. Fan, K. V. Jacob, B. Freireich, and R. M. Lueptow,
Powder Technol. (2017).
[22] L. Jing, C. Y. Kwok, and Y. F. Leung, Phys. Rev. Lett.
118, 118001 (2017).
[23] J. M. N. T. Gray and A. R. Thornton, Proc. R. Soc. A.
461, 1447 (2005).
[24] A. R. Thornton, J. M. N. T. Gray, and A. J. Hogg, J.
Fluid Mech. 550, 1 (2006).
[25] B. Marks, P. Rognon, and I. Einav, J. Fluid Mech. 690,
499 (2012).
[26] D. R. Tunuguntla, O. Bokhove, and A. R. Thornton, J.
Fluid Mech. 749, 99 (2014).
[27] Y. Fan, C. P. Schlick, P. B. Umbanhowar, J. M. Ottino,
and R. M. Lueptow, J. Fluid Mech. 741, 252 (2014).
[28] K. van der Vaart, P. Gajjar, G. Epely-Chauvin, N. Andreini, J. M. N. T. Gray, and C. Ancey, Phys. Rev. Lett.
114, 238001 (2015).
[29] P. Gajjar and J. M. N. T. Gray, J. Fluid Mech. 757, 297
(2014).
[30] K. Kamrin and G. Koval, Phys. Rev. Lett. 108, 178301
(2012).
[31] D. L. Henann and K. Kamrin, Proc. R. Soc. A. 110, 6730
(2013).
[32] R. J. Phillips, R. C. Armstrong, R. A. Brown, A. L. Graham, and J. R. Abbott, Phys. Fluids 4, 30 (1992).
[33] F. Boyer, O. Pouliquen, and É. Guazzelli, J. Fluid Mech.
686, 5 (2011).
[34] A. R. Thornton, D. Krijgsman, A. Voortwis, V. Ogarko, S. Luding, R. Fransen, S. Gonzalez, O. Bokhove,
O. Imole, and T. Weinhart, in DEM6 - Int. Conf. DEMs
(Colorado School of Mines, 2013).
[35] T. Weinhart, D. R. Tunuguntla, M. P. van SchrojensteinLantman, A. J. van der Horn, I. F. C. Denissen, C. R.
Windows-Yule, A. C. de Jong, and A. R. Thornton, in
Proc. 7th Int. Conf. DEM (Springer, 2017) pp. 1353–
1360.
[36] F. Guillard, Y. Forterre, and O. Pouliquen, J. Fluid
Mech. 807 (2016).
[37] T. Weinhart, R. Hartkamp, A. R. Thornton, and S. Luding, Phys. Fluids 25, 070605 (2013).
[38] P. A. Cundall and O. D. L. Strack, Geotechnique 29, 47
(1979).
[39] D. R. Tunuguntla, A. R. Thornton, and T. Weinhart,
Comp. Part. Mech. 3, 349 (2016).
[40] I. Goldhirsch, Granular Matter 12, 239 (2010).
[41] C. Rycroft, Lawrence Berkeley National Laboratory
(2009).
[42] T. Weinhart, A. R. Thornton, S. Luding,
and
O. Bokhove, Granular Matter 14, 531 (2012).
[43] F. Guillard, Y. Forterre, and O. Pouliquen, Phys. Fluids
26, 043301 (2014).
[44] F. Guillard, Y. Forterre, and O. Pouliquen, Phys. Rev.
E 91, 022201 (2015).
[45] A. Tripathi and D. V. Khakhar, Phys. Rev. Lett. 107,
108001 (2011).
[46] J. M. N. T. Gray and C. Ancey, J. Fluid Mech. 678, 535
(2011).
[47] P. G. T. Saffman, J. Fluid Mech. 22, 385 (1965).
[48] H. A. Stone, J. Fluid Mech. 409, 165 (2000).
[49] K. A. Gillemot, E. Somfai, and T. Börzsönyi, Soft Matter
13, 415 (2017).
[50] N. Thomas, Phys. Rev. E 62, 961 (2000).
[51] G. Félix and N. Thomas, Phys. Rev. E 70, 051307 (2004).