Granular Matter (2007) 9:407414

DOI 10.1007/s10035-007-0062-2
The ow rate of granular materials through an orice
C. Mankoc A. Janda R. Arvalo J. M. Pastor
I. Zuriguel A. Garcimartn D. Maza
Received: 12 February 2007 / Published online: 12 September 2007
Springer-Verlag 2007
Abstract The ow rate of grains through large orices
is known to be dependent on its diameter to a 5/2 power
law. This relationship has been checked for big outlet sizes,
whereas an empirical tting parameter is needed to repro-
duce the behavior for small openings. In this work, we pro-
vide experimental data and numerical simulations covering a
wide span of outlet sizes, both in three- and two-dimensions.
This allows us to showthat the laws that are usually employed
are satisfactory only if a small range of openings is consid-
ered. We propose a new law for the mass ow rate of grains
that correctly reproduces the data for all the orice sizes,
including the behaviors for very large and very small outlet
sizes.
Keywords Granular ow Silos Jamming
1 Introduction
The ow of granular materials through an orice has been
widely studied during decades due to the great interest for
industrial applications ranging from silos to hoppers [14].
Contrary to the uids, when a silo is discharged by gravity,
the ow rate does not depend on the height of the granular
layer. Indeed, when the thickness of the layer is greater than a
value close to 1.2 times the diameter of the silo, the pressure
at the bottom of the silo saturates due to the Janssen effect
and hence, the ow rate remains about constant. In a rst
order approximation it has also been shown that the ow rate
is independent on the diameter of the silo L if two conditions
C. Mankoc A. Janda R. Arvalo J. M. Pastor I. Zuriguel
A. Garcimartn (B) D. Maza
Departamento de Fsica y Matemtica Aplicada,
Facultad de Ciencias, Universidad de Navarra,
31080 Pamplona, Spain
e-mail: [email protected]
are fullled: L is greater than 2.5 times the diameter of the
outlet orice D
0
and also greater than D
0
+30 d
p
, where d
p
is the diameter of the particle [1].
The most widely accepted law that predicts the ow rate
of grains through an orice and its dependence on different
parameters was proposed by Beverloo et al. [5] and has the
form:
W = C
b

g(D
0
kd
p
)
5/2
(1)
where W is the average mass discharge rate through the ori-
ce, C and k are empirical discharge and shape coefcients
respectively,
b
is the apparent density, and g is the accelera-
tion of gravity. As C may depend on the friction coefcient,
sometimes this is explicitly stated by writing C(). Equa-
tion (1) is known as the Beverloo law and its validity has
been tested for mono-sized granular samples with d
p
larger
than 0.5 mm and D
0
big enough to avoid intermittencies in
the ow due to jamming. This means that the ow rate of
grains through orices has been found to follow the Bever-
loo law only for D
0
d
p
, well beyond the critical value
belowwhich the owcan be interrupted due to the formation
of arches or domes [6].
One of the most interesting issues concerning Eq. (1) is
the dependence of the ow rate with a 5/2 power of the
diameter of the orice. This relationship can be obtained by
dimensional analysis and can be physically explained if it
is assumed that the granular ow is driven by the behavior
of grains near the outlet. Following this line of reasoning, it
seems plausible to believe that, just above the outlet, there
is a free-fall zone limited by an arch. Above the arch the
grains are well packed and their velocities are negligible,
whereas below the arch the particles accelerate freely under
the inuence of the gravity. If the characteristic size of this
arch is somehow proportional to the radius of the orice, the
velocity of the grains through the outlet of the silo is the one
1 3
408 C. Mankoc et al.
corresponding to a particle falling without initial velocity
from a distance proportional to the radius of the outlet. A
primary consequence of this assumption is that the velocity
of the grains is proportional to D
1/2
0
and therefore the ow
rate proportional to D
5/2
0
. An equivalent equation to Eq. (1)
can be easily derived for the two-dimensional case and a ow
dependence on the diameter of the silo of D
3/2
0
is obtained.
The Beverloo law, and particularly the 5/2 power depen-
dence of the ow rate on the diameter of the orice, has been
found to be robust for different kind of particles, indepen-
dently of their packing fraction, density, surface properties
or shape [7, 8] for D
0
d
p
. It is also remarkable that this
equation can describe the ow rate of grains through orices
for different ow patterns developed inside the silo, i.e. mas-
sic ow, funnel ow or mixed ow. A modication has been
implemented for the ow of powders when the size of the
particle is smaller than 0.5 mm. In this case a term should
be included in Eq. (1) to reproduce the effect of the pressure
gradient generated by the air passing through the interstices
between the grains. Other modications have been developed
to predict the solid discharge rate for binary mixtures [2, 9]
and even the dense ow of air bubbles in a two-dimensional
silo [10].
Despite the above mentioned robustness of the Beverloo
law and the fact that has been successfully used by engi-
neers since 1961, the basic physical principles behind the
ow of grains through an orice remain elusive and the two
empirical coefcients (C and k) are required to be determined
experimentally for every single kind of grains and container
properties. The value of C, the so-called discharge coef-
cient, depends on the bulk density and it was found to be
in a range of 0.55 < C < 0.65 by Beverloo et al. [5]. The
shape coefcient k is generally agreed to be dependent on the
particle shape as well as the slope of the hopper. However,
the meaning and origin of the term kd
p
has provoked a
great controversy. The rst and most widely accepted inter-
pretation of this term was done by Brown and Richards [11]
who claim that the centers of the particles cannot approach
the edge of the orice within a distance of kd
p
/2. There-
fore the particle centers must pass through a effective orice
of diameter D
0
kd
p
. The value of k has been found to
be independent of the size of the particle [12] in a range of
1 < k < 2 depending on the particle and hopper properties.
Yet there are some exceptions, like the ow of sand, where
the value of k turns to be 2.9. However, Zhang and Rudolph
[13] claimthat the only plausible value for k is 1 and propose
an alternative expression where a new term c

is introduced.
The value of c

depends on D
0
and physically represents the
effect on the ow rate of shear friction between owing and
non-owing particles at the edges of the orice.
The experimental and numerical studies where the ow
rate has beenfoundtobe ina reasonablygoodagreement with
the Beverloo t are abundant in the literature. Yet it is difcult
to nd any work where the ow rate is measured covering, at
least, 2 decades of outlet sizes. Some of the works deal with
very big orices in real silos where the value of k has only a
weak inuence in the t, whereas others explore the region of
small orices but do not reach high values of D
0
[8]. This fact
may lead to a misinterpretation of the results as the t may
seem suitable just because the range of D
0
considered is not
large enough and the values of C and k are chosen arbitrarily.
In this work we will show that if the ow rate is measured
for a wide range of orice sizes, the Eq. (1) is not able to
describe the whole behavior as different values of C and k
must be used for big and small orices. In order to solve this
problema newexpression for the owrate is proposed where
the constant k is altogether eliminated. The new equation
is able to t the experimental results of the ow rate for
diameters of the orice ranging from 1.5 to 100 times the
diameter of the particle. Furthermore with this modication
it is also reproduced the behavior of the ow rate for very
small orices, when the ow is not continuous due to arch
formation. Its validity has been checked using different kind
of particles in three- and two-dimensional experiments as
well as in two-dimensional molecular dynamics simulations.
The manuscript is organized as follows. In the rst sec-
tion a description of the experimental setup and simulation
techniques is presented. The method used to measure the
ow of grains is discussed for both big and small orices.
In the next section the numerical results in two dimensions
and the experimental results in two- and three-dimensions
are reported. Different ts of Eq. (1) will be displayed to
stress that it is unable to t the ow rate for the whole range
of outlet sizes. Then, after measuring the deviations from the
Beverloo proposal, we introduce a modication consisting
on the elimination of the parameter k and the correction of
W by a multiplicative term. Finally we show that this cor-
rective factor may have its physical origin in a dependence
of either the velocity or the apparent density near the outlet
of the silo.
2 Materials and methods
In order to investigate the ow through a small orice under
the action of gravity, two scaled silos have been built. One
of them is cylindrical, and the other is two-dimensional (the
beads are contained between two glass panes so as to provide
direct optical access to the particles inside). Numerical simu-
lations using molecular dynamics have also been performed.
In the following subsections a description for the procedures
used in each case is provided.
2.1 Three-dimensional silo
The experimental setup for the three-dimensional silo was
described in detail in a former article [14]. It consists of an
1 3
The ow rate of granular materials through an orice 409
Table 1 Characteristics of the different beads that have been used in
the cylindrical silo
Material d
p
(mm) mass (mg) (g/cm
3
)
Glass 0.42 0.05 0.16 0.05 2.2 0,1
Glass 1.04 0.01 1.3 0.4 2.4 0.1
Glass 2.06 0.02 10.1 0.3 2.2 0.1
Glass 3.04 0.02 34.7 0.4 2.4 0.1
Lead 1.98 0.06 46.0 3.8 11,4 0.5
Lead 3.0 0.1 150 14 10.9 0.5
Delrin 3.00 0.02 18.9 0.3 1.34 0.05
All the beads are spherical
scaled, automated cylindrical container with an orice in the
base. The silo is made of stainless steel, but vessels made of
glass were also used whenever observation of the particles
was required. The size of the silo is such that its nite size
can be neglected. Moreover, the level of the granular matter
contained in it was never allowed to descend below a cer-
tain level (about twice the silo diameter) to ensure that the
pressure at the base was approximately constant due to the
Janssen effect, as explained above. The piece forming the at
bottom of the silo is changeable. This allows us to modify
the size of the outlet, which ranges from 2 to 50 mm. The
granular material falls through the orice and is collected
in a box placed on a scales. Eventually the particles can get
jammed. Let us call an avalanche the ensemble of particles
fallen between two jamming events. The size of the avalanche
is calculated by dividing the weight of the avalanche by the
weight of one bead. The time that the particles have been
owing is measured by means of a microphone that registers
the noise made by the falling grains. Thus, with our experi-
mental setup we can obtain both the size of avalanches and
its time span. After each avalanche, the system is unjammed
by directing a compressed air jet to the silo outlet; with this
device, the compaction fraction of the grains inside the silo
is not altered.
We usedspherical grains of different sizes andmade of dif-
ferent materials. A summary of some of their relevant prop-
erties is shown in Table 1.
Flow measurement was accomplished in two different
ways depending on the features of the silo discharge. It has
been shown that for big sizes of the outlet orice the outpour-
ing is almost continuous, while for a small orice jamming
events can occur. The border between the two regimes is
sharply dened at R
c
5, where R is the ratio between the
diameter of the outlet orice and the diameter of the beads
[14]. For R > R
c
, the ow can be measured just by divid-
ing the number of fallen beads by the time elapsed. As the
ow is continuous, this measurement can be done at any
moment with any number of beads (or elapsed time), and it
gives consistent results. However, when R < R
c
, the owwas
obtainedbydividingthe size of avalanches bytheir respective
duration. In order to illustrate the difference between both
cases, in Fig. 1 we show the measurements (number of fallen
beads as a function of time) for two outlet orices, one bigger
than R
c
and other smaller.
In both cases, a measurement of the mean owis obtained
from the slope of the straight line tting the data. The main
difference is that for R < R
c
uctuations are not completely
smoothed out as in the case R > R
c
. This is not due to a
lack of resolution in the measurements but to the fact that
at the involved time scales the ow uctuations can not be
neglected for short avalanches. For orice sizes where both
methods are feasible it has been checked that the measured
ow rate does not depend on the procedure.
2.2 Two-dimensional silo
A two-dimensional silo consisting of two sheets of glass was
also used. The silo is built so that the gap between the panes
is a little bit larger than the diameter of the beads. In order to
meet the stringent tolerances of these assembly, we only used
stainless steel beads with a diameter 1.00 0.01 mm, while
the separation between the two glass sheets was 1.10 mm.
As the separation between walls is just a little bigger than the
size of the particles, they only can arrange themselves in one
layer. The setup therefore allows us to record the movement
of individual particles using a high speed camera. The grains
owout of the silo through a slot in the base whose length can
be changed at will. The automation and measuring devices
are similar to those used in the three-dimensional silo. A
detailed description of this setup will be presented elsewhere.
The time span of the avalanches is registered with a pho-
todetector at the silo exit. We placed optical bers at both
ends of the slit forming the outlet orice; a light beam is
emitted from one of them and the other collects it and feeds
the photodetector, so a falling bead is detected when it blocks
the light beam. The time resolution of this assembly (better
than 1 ms) is smaller than the time that it takes for a parti-
cle to cross the light beam. The weight used also allows us
to measure the avalanche size with a resolution of just one
particle.
The width of the silo is larger than 200 particle diameters,
in order to prevent any inuence of the lateral walls. As in
the three-dimensional case, the measurement of the owwas
accomplished using two different methods, as described in
the previous subsection, depending on whether the outlet gets
jammed or not.
2.3 Numerical simulations
We have used soft particle molecular dynamics [15] simu-
lations of disks in two dimensions. In this method, we con-
sider that a collision has taken place when two disks slightly
1 3
410 C. Mankoc et al.
Fig. 1 Number of particles
fallen as a function of time for a
small (R = 3.5) and a big
(R = 5.07) outlet orice,
measured in the
three-dimensional silo
overlap (the distance between their centers is smaller than
the sumof their radii). Arepulsive force, in the normal direc-
tion of the collision, proportional to the overlap, prevents the
grains from traverse each other; besides, a term proportional
to the relative velocity of the particles accounts for the dissi-
pation of energy during the collision. A force perpendicular
to the collision direction implements the Coulomb law of
friction. Details of the algorithm can be found in [16]. The
velocity-Verlet scheme along with neighbor lists [17] were
used to integrate the equations of motion.
A simulation begins with 5,000 disks placed on a reg-
ular lattice that are given random velocities taken from a
gaussian distribution. The disks are allowed to fall under
gravity through a hopper. Below the hopper there is a at
bottomed silo in which the grains are deposited. This is the
preparation phase, which is aimed to break the correlations
that the initial regular arrangements of the grains may induce
in their dynamics. Once all the grains have settled in the at
silo and after most of the kinetic energy is dissipated, the
outlet at the bottom is opened, allowing the grains to fall.
The ow is measured by counting the number of beads that
go out of the silo for each simulation step.
The walls of both the silo and the hopper used are con-
structed with grains. The interaction graingrain is the same
than the interaction grainwall, but the latter keep their posi-
tions xed. Thus the walls are rough and give rise to dissipa-
tive collisions.
The at silo is 50 grains diameter wide and the level
reached by the grains is approximately twice that value when
the silo is lled. During a simulation, the level of the grains is
kept constant by reintroducing the exiting grains at the top of
the silo. These grains are placed as close to the free surface
as possible and with velocities similar to those of the exiting
grains, so as not to perturb the ow. These conditions allow
us to neglect any effect from the wall or lling method on
the ow rate.
3 Results
3.1 Validity of Beverloos law
Let us start with the presentation of the ow measurements
as a function of R in the three-dimensional silo. The exper-
imental data are shown on Fig. 2. As different materials are
included, we have taken the mass ow rate divided by the
mass of one bead W
b
= W/m
b
. By using W
b
, which is the
number of beads fallen per unit time, all the data can be plot-
ted together (as expected, W
b
shows only a slight dependency
on the material fromwhich the beads are made). Note that the
size of the outlet orice spans over almost two decades, from
R = 2 to 100. Besides, the results are a practical demonstra-
tion that the relevant parameter is indeed R = D
0
/d
p
. The
owrate of beads made of different materials can thus be rep-
resented in the same graph without any additional concern.
The best t obtained with Eq. (1) is W
b
= 50.5(R
1.16)
5/2
and is also represented in Fig. 2. Note that this t
fails for large R, where it underestimates the ow rate (see
the inset in Fig. 2). Indeed, the disagreement between the
measured and the predicted ow rate amounts to about one
million beads per second for R = 100 (a relative error of 10%
approximately). Note that the parameter k merely shifts the t
along the horizontal axis. Depending on the particular value
chosen, the t will be closer to the data in a different zone
of R. If the range of R is small, this may seem acceptable;
but if a large range of exit orices is considered, it becomes
clear that Beverloos law is inadequate.
1 3
The ow rate of granular materials through an orice 411
Fig. 2 Logarithmic plot of the ow rate of particles W
b
for a three-
dimensional silo (W
b
is the number of beads fallen per unit time). The
solid line is the best t of Eq. (1): W
b
= 50.5(R1.16)
5/2
. (Note that g
and are included in the numerical constant C

= 50.5). The material

and the diameter of the beads corresponding to the different symbols
are provided in the legend. The inset shows an expansion of the region
from R = 60 to 90, where it can be seen that k = 1.16 does not produce
a satisfactory match
Fig. 3 Logarithmic plot of the ow rate of particles W
b
for a three-
dimensional silo (same data than in Fig. 2). The solid line is a t of
Eq. (1) with k = 1: it gives W
b
= 64.59(R 1)
5/2
. Inset a zoom of the
same data and t for the range 60 < R < 90
Adifferent approach is to t Eq. (1) for large R with k = 1.
In this case at least the asymptotic behavior W R
5/2
for
R is recovered (see inset of Fig. 3). But then for
small orices this prediction and the experimental data do
not agree: the ow rate is overestimated (see Fig. 3).
The same holds for the two-dimensional case, where the
asymptotic behavior for large R is now W
b
R
3/2
(Fig. 4).
Note, however, that in this case we have explored a smaller
range of R. By varying the free parameters of Eq. (1), i.e.,
C and k, one can often get a seemingly acceptable result.
Nevertheless, close inspection of the t with k = 1 reveals
Fig. 4 Logarithmic plot of the ow rate of particles W
b
for a two-
dimensional silo. Circles represent experimental measurements, while
squares are the result of numerical simulations. The error bar for the
experimental data is of the same order as the symbol size. The solid line
is a t of Beverloos law with k = 1: W
b
= 102.2(R 1)
3/2
the same qualitative disagreement: Beverloos lawwithk = 1
overestimates the ow for small R. This can be more easily
seen if a large range of R is considered; but even for a smaller
range of outlet sizes, such as in in Fig. 4, the overestimation
is suggested for R 10. In the same gure the result of the
numerical simulation is displayed. The agreement with the
experimental results is excellent for all the explored range.
In summary, the experimental results both in two- and
three- dimensions as well as the numerical simulations in
two-dimensions clearly show a discrepancy with the values
predicted by Eq. (1). If a small range of large outlet orices is
considered, the free parameters of Beverloos equation allow
for a reasonably good t. However, if a large range of R is
included, the t does not yield acceptable results.
3.2 Modication to Beverloos statement
In order to study more closely the discrepancy between the
measured data in the three-dimensional case and the values
predicted by the Beverloo equation, we show in Fig. 5a
versus R, where = W
m
/W
p
is the ratio between the mea-
sured ow W
m
and the value W
p
predicted by Beverloos law
with k = 1, obtained from the t shown in Fig. 3.
It can be seen that saturates in a non linear fashion.
For large R indeed 1, i.e., W
p
= W
m
, as expected by
construction. In order to check the exponential dependence,
we plot 1 versus R 1 on a semilogarithmic scale
(Fig. 5b). A t is also provided showing the good general
agreement of this functional dependence:
= 1
1
2
e
b(R1)
(2)
From the t, one can obtain that b = 0.051.
1 3
412 C. Mankoc et al.
Fig. 5 a versus R for the three-dimensional silo. b Semilogarithmic
plot of 1versus R1 and the proposed t with Eq. (2), fromwhich
the value b = 0.051 has been obtained
We then propose another expression for the mass ow
rate in number of beads per unit time W
b
, which stems from
Eq. (1) W
p
after the inclusion of the correction factor :
W
b
= C

1
1
2
e
b(R1)

(R 1)
5/2
(3)
Recall that the constants C,
b
and

g from Eq. (1) have

been grouped into C

.
InFig. 6we showthe t of equation(3) toour experimental
data. As can be seen, the match is very good in all the range
of R, which spans for almost 2 decades. It should be stressed
that the proposed law ts the mass ow rate even for very
small orices i.e., R < 5), where the ow can be interrupted
by jamming events.
In two dimensions, the functional dependence is the same
as in three dimensions except for the exponent 5/2 which is
now 3/2 (coefcients C

and b are obviously different). In

Fig. 6 Fit of the experimental data for the 3D silo using the proposed
equation for the ow rate (Eq. (3)) with C

= 64.6 and b = 0.051. The

data are the same than in Fig. 3, taking only the ones corresponding to
glass beads
Fig. 7 Fit of the experimental data for the 2D silo (circles) and the
numerical simulations (squares) using the proposed equation for the
ow rate (Eq. (3)) with an exponent 3/2. The values of the parameters
are C

= 108 and b = 0.23

Fig. 7 we show the experimental data and the results for the
numerical simulations along with the proposed t.
The proposed functional dependence of the ow is more
satisfactory than tting the data by using the parameter k.
Let us recall that this parameter was introduced by Beverloo
in order to t the ow for small orices. The way that this
is usually done is the following. As W (D
0
kd
p
)
5/2
in
Eq. (1), thenplotting W
2/5
versus R shouldgive a straight line
andthe intercept withzerowouldprovide the value for k. This
is carried out, for example, in Ref. [18]. Our experimental
data for the owrate when the orice is small (Fig. 8) provide
a strong evidence that the intercept with zero is at, or very
near to, R = 1: we have observed that grains ow until near
that value. It should be noticed that if the ow rate is not
1 3
The ow rate of granular materials through an orice 413
Fig. 8 W
2/5
b
versus R for the three-dimensional silo near R = 1. The
solid line corresponds to a linear t for the data corresponding to R > 50
(the whole range of data is shown in the inset). The dashed line is the
t using the proposed ow rate equation (3)
measured for small R this method can be misleading, as it
depends on an unwarranted extrapolation. Indeed, the value
of k obtained depends on the range of R used to perform the
t. For instance, the solid line in Fig. 8 is a linear t obtained
using just the points for R > 50, and the value obtained in
this case is k 3. Any value between k = 1 and k 3
can be obtained with our data simply by considering just a
suitable range of R.
As explained above, the only plausible value for k is unity.
This is reasonable since W must vanish for D
0
d
p
, and so
one could just write W (D
0
d
p
), with k = 1. But then it
is impossible to t correctly the data using Eq. (1), as shown
before (see Fig. 3). Yet the proposed Eq. (3) neatly ts the
data for large values of R and it gives the correct behavior
for R 1.
Furthermore, our data show that the so-called empty
annulus effect is not evident at all. It is obvious that no
bead can pass through the orice in such a way that its center
is separated by less than d
p
/2 from the orice border. Thus,
the bead centers always pass through the orice in a region
which is effectively determined by a diameter D
0
d
p
. If
anything beyond this obvious logic is considered, such as an
empty annulus dened by D
0
kd
p
with k = 1, then it
should be checked that the ow vanishes for D
0
= kd
p
. We
have found no hint of this; instead we nd W
b
> 0 for values
of k often given in the literature.
In our proposal, embodied in Eq. (3), we have got rid
of the parameter k and we have instead introduced a correc-
tion term, namely, the parenthesis containing the exponential.
One of the more appealing possibilities is that this correction
term is related to the density of grains near the outlet. In both
the experiments and the simulations the compaction fraction
was the always the same at the beginning of the runs, and
Fig. 9 The compaction fraction as a function of R for a two-
dimensional silo, obtained from numerical simulations. The value of
the density depends on the size of the region where it is measured: cir-
cles and squares correspond to two square regions just over the outlet
orice. The tting function in both cases is

1
1
2
e
b(R1)

we have checked that it remains approximately unchanged

during the discharge. But it may be different near the outlet
orice. As this measurement is difcult to obtain experimen-
tally, we resort tonumerical simulations ina two-dimensional
silo. The value of the density depends on the size of the
region where numerical data are collected, and this should
be so because there exist spatial density variations near the
outlet. We show the results for the density in Fig. 9, where
we plot the local compaction fraction in a zone close to the
outlet orice as a function of R. The measurements for two
regions of different size can be t with a termjust like the one
proposed for the ow rate, i.e., (1
1
2
e
b(R1)
). The only
variation between the ts is a multiplicative factor. We can
then conclude that the correction factor that we introduced is
related to the density variation of the granular material near
the outlet orice.
Finally, we show the robustness of the proposed expres-
sion by displaying the residues of the best t using both
Eqs. (1) and (3). The t given by Beverloo is essentially
awed in the sense that it cannot reproduce faithfully the
ow for an extended range of R (so the residues are only
small for one particular value of R). As mentioned above,
it is remarkable that for R = 100 the difference between
the value given by Eq. (1) and the measured value amounts
to about one million beads per second. On the contrary, the
residues for Eq. (3) are consistently small.
4 Conclusions
In this paper we have looked into the behavior of the ow
of granular materials during the discharge of a silo. We have
1 3
414 C. Mankoc et al.
Fig. 10 The residues for the ts are shown as the difference between
the logarithms of the measuredowW
m
andthe value W providedbythe
best t with the two equations considered. Open circles the differences
for the t provided by Beverloo [Eq. (1), with the parameters as used in
Fig. 2]. Solid squares the differences for the t provided by Eq. (3) for
the the mass ow rate
checked that the ow W does not depend signicantly of
the material from which the beads are made, and that the
relevant control parameter is the ratio R between the diameter
of the orice and the diameter of the beads. We have collected
experimental data for the largest range of outlet sizes we have
been able to nd in the literature, which comprises about two
decades in R and six decades in W
b
.
Concerning the ow rate, the Beverloo expression pro-
vides the right asymptotic behavior W R
5/2
for large R.
However, the Eq. (1) is not valid for a sizeable range of R
including small sizes of the outlet orice, i.e. R 10. In this
case, the inclusion of the termkd
p
in the owequation, which
was an arbitrary remedy for tting the data, cannot account
for the experimental results. We propose instead a more
robust expression in which the ow depends on the prod-
uct of R
5/2
times an exponential corrective factor. We have
also shown that it ts neatly the data for a large range of R,
including very small orices, in two- and three-dimensions.
In the new proposal for the mass ow rate, we have got
rid of the factor k and introduced instead a correction term
which saturates exponentially with R, so that the asymptotic
behavior R
5/2
is recovered for large R. Apart from demon-
strating that the concept of empty annulus is wrong and
unable to explain the experimental results, we have provided
an explanation for the correction introduced in terms of the
local density variations near the outlet.
Acknowledgments This work has been partially supported by
Project FIS 2005-03881 (Spanish Government) and PIUNA (Univer-
sidad de Navarra). C.M., J.M.P. and R.A. thank Asociacin de Amigos
de la Universidad de Navarra for a fellowship. A.J. thanks Fundacin
Ramn Areces for a fellowship.
References
1. Nedderman, R.M., Tzn, U., Savage, S.B., Houlsby, G.T.: The
ow of granular materials I: discharge rates from hoppers.
J. Chem. Eng. Sci. 37, 15971609 (1982)
2. Tzn, U., Houlsby, G.T., Nedderman, R.M., Savage, S.B.: The
ow of granular materials II: velocity distributions in slow ow.
J. Chem. Eng. Sci. 37, 16911709 (1982)
3. Savage, S.B., Nedderman, R.M., Tzn, U., Houlsby, G.T.: The
ow of granular materials III: rapid shear ows. J. Chem. Eng.
Sci. 38, 189195 (1983)
4. Tighe, B.P., Sperl, M.: Pressure and motion of dry sand: translation
of Hagens paper from 1852. Granul. Matter 9, 141144 (2007)
5. Beverloo, W.A., Leniger, H.A., Van de Velde, J.: The ow
of granular material through orices. J. Chem. Eng. Sci. 15,
260296 (1961)
6. Zuriguel, I., Pugnaloni, L.A., Garcimartn, A., Maza, D.: Jam-
ming during the discharge of grains from a silo described as a
percolating transition. Phys. Rev. E (RC) 68, 30301 (2003)
7. Hirshfeld, D., Radzyner, Y., Rapaport, D.C.: Molecular dynam-
ics studies of granular ow through an aperture. Phys. Rev.
E 56, 44044415 (1997)
8. Hirshfeld, D., Rapaport, D.C.: Granular ow from a silo:
discrete-particle simulations in three dimensions. Eur. Phys. J.
E 4, 193 (2001)
9. Humby, S., Tzn, U., Yu, A.B.: Prediction of hopper discharge
rates of binary granular mixtures. J. Chem. Eng. Sci. 53, 483
494 (1998)
10. Bertho, Y., Becco, C., Vandewalle, N.: Dense bubble owin a silo:
an unusual ow of a dispersed medium. Phys. Rev. E 73 (2006)
11. Brown, R.L., Richards, J.C.: Principles of Powder Mechan-
ics. Pergamon Press, Oxford (1970)
12. Nedderman, R.M., Laohakul, C.: The thickness of the shear zone
of owing granular media. Powder Technol. 25, 91100 (1980)
13. Zhang, J.Y., Rudolph, V.: Effect of shear friction on solid ow
through an orice. Ind. Eng. Chem. Res. 30, 19771981 (1991)
14. Zuriguel, I., Garcimartn, A., Maza, D., Pugnaloni, L.A., Pastor,
J.M.: Jamming during the discharge of granular matter from a
silo. Phys. Rev. E 71, 051303 (2005)
15. Schfer, J., Dippel, S., Wolf, D.E.: Force schemes in simulations
of granular materials. J. Phys. I(Fr) 6, 5 (1996)
16. Arvalo, R., Garcimartn, A., Maza. D.: Anon-standard statistical
approach to the silo discharge. Eur. Phys. J. ST 143, 191197
(2007)
17. Rapaport, D.C.: The Art of Molecular Dynamics Simula-
tion. Cambridge University Press, London (2004)
18. Brown, R.L., Richards, J.C.: Prole of ow of granules through
apertures. Trans. Inst. Chem. Eng. 38, 243256 (1960)
1 3