• Aucun résultat trouvé

Geometrical Analysis of Avalanches in a 2D Drum

N/A
N/A
Protected

Academic year: 2021

Partager "Geometrical Analysis of Avalanches in a 2D Drum"

Copied!
17
0
0

Texte intégral

(1)

HAL Id: jpa-00247084

https://hal.archives-ouvertes.fr/jpa-00247084

Submitted on 1 Jan 1995

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Geometrical Analysis of Avalanches in a 2D Drum

Florence Cantelaube, Yann Limon-Duparcmeur, Daniel Bideau, G.H. Ristow

To cite this version:

Florence Cantelaube, Yann Limon-Duparcmeur, Daniel Bideau, G.H. Ristow. Geometrical Analy- sis of Avalanches in a 2D Drum. Journal de Physique I, EDP Sciences, 1995, 5 (5), pp.581-596.

�10.1051/jp1:1995107�. �jpa-00247084�

(2)

Classification Physics Abstracts

46.10 05.40 64.60C

Geometrical Analysis of Avalanches in

a

2D Drum

Florence Cantelaube

(~),

Yann

Limon-Duparcmeur (1),

Daniel Bideau

(1)

and

G-H Ristow (~)

(~) Groupe Matière Condensée et Matériaux

(*),

Université de Rennes I,

Campus

de Beaulieu,

35042 Rennes Cedex, France

(~) Fachbereich Physik, Philipps Universitàt Marburg, Renthof 6, 35032 Marburg, Germany

(Received

28 July 1994, revised 15 December 1994, accepted 2

February1995)

fibstract.

We present an experimental and numerical analysis of intermittent flow at the surface of a 2-dimensional granular medium. Two types of devices are used: a 2-dimensional drum and a long rectangular box. Two types of grains are used

(disks

and

pentagons),

and

comparison is made between the dilferent cases. The experimental data are compared with the results obtained from molecular dynamics simulations, usmg roughly trie same number of

partiales. Dynamical conditions of stability are determined from the angle of repose 6r and trie angle of maximum stability 6m. Intermittent flow at the "surface" appears ta be of two types, as

in 3-dimension: small avalanches whose time fluctuations are characterised by

a power spectrum obeying a power law, and large events with well-defined average size and duration time.

l. Introduction

Granular media are not

only

one of the most often used mater1als in industrial processes but are also

massively

present in our environment. Tueir flow

properties

have been [1] and

are still now tue

subject

of many studies m

puysics,

mecuanics and soit

mechanics,

civil and chemical

engineering.

Some

exciting puenomena

are

investigated,

sucu as the formation of

convection

cells,

spontaneous

heap

formation, surface

fluidization, avalanches, density

waves,

or size

segregation [2,3].

For a review, see for

example

the paper

by Jaeger

and

Nagel

[4].

Avalanches have

recently

received a great attention, as

being

a

paradigm

for Self

Organized Criticality (SOC

[SI. A

sandpile

would

by

itself

adjust

its

slope

to a critical

value, supposed

to be the

"angle

of

repose"

@r, defined as the

angle

between the horizontal and the free surface of the

sandpile.

This self

adjustment

is

suggested

to be

analog

to

SOC,

and then characterized

by long-range spatial

and

temporal correlations,

with a

typical 1If

power spectrum of the

fluctuations. Some cellular automaton models have confirmed this

assumption [6-9]. Except

in small-size systems,

experimental

results

[10-13]

are m contradiction with the

prediction

of SOC model:

peak

distributions of avalanche sizes on intervals are

observed,

instead

of1If

(*) URA CNRS804

@

Les Editions de Physique 1995

(3)

582 JOURNAL DE PHYSIQUE I N°5

noise. Tue essent1al reason of tuis

discrepancy

comes from tue fact tuat tue

stability

of tue

sandpile

cannot be described

by

a

unique angle

of repose @r, but

by

two

angles (and

a

hysteresis

is

observed):

the maximum

angle

of

stability

@m measured

just

when an avalanche starts and the

angle

of repose measured

just

after the avalanche. The average dilference between these two

angles

@m @r is of the order of 2° for sand in 3D.

Although

SOC is an

exciting problem,

numerical and

experimentalstudies

of avalanches are interesting because

they

are

essent1ally

linked to two not well-understood

problems concerning granular

materials: friction and

dilatancy.

The

dilatancy

was first introduced

by Reynolds

[14] and it is of great

importance

in the deformation of dense

granular

media: due to the

geometricalinterlocking

of

grains,

a shear-

ing

deformation cannot be

imposed

to a non-cohesive

granular

mater1al without

dilating

it.

Starting

from this

analysis,

a model for

sandpile

avalanches has been

proposed,

which takes

dilatancy

and friction

explicitly

into account

[11,15,16].

A macroscopic friction coefficient was introduced

by

Coulomb [17] as

equalto

~t

=

tan@r.

This mean value of friction does trot

explain

the two values of obtained

above,

and their relation to the geometry of the system, the

shape

of the

grain,

and the

microscopic

friction is not clear.

Because of its

cyclic

character, a

rotating

drum allows for

experiments

with

long-time

ob- servations in processes characterized

by

short-time scales

(e.g., avalanches),

and it has served

as a

good experimental

tool

recently [18,19].

This geometry is also

currently

used in industrial systems, such as

cylindrical

ovens.

In this paper we

explore experimentally

and

numerically

the rote of the friction coefficient and the

partiale shape

in

equilibrium

of the

heap,

and tue detail of the statistics of avalanches.

We restrict ourselves to two

spatial

dimensions in order to be able to observe the geomet- rical modifications in the system. Some

experiments

have

previously

been

performed using

pentagons,

showing clearly

the main importance of face to face contacts,

compared

to tuat of tue usual

punctual

contacts of

spueres

or disks [20]. In

fact,

in tue case of pentagons, tue number of

possible

stable

configurations

is

larger,

and some

large

dilferences are

expected

in tue definition of tue surface

equilibrium

of a

pile,

from botu the statics and

dynarnics.

Wuile it is

perfect

for

geometrical properties,

tuis

2-dimensionalexperimental

and numerical system poses

problems

m tue context of SOC: the 2D version of tue model

proposed by Bak, Tang

and Wiesenfeld

(BTW)

[Si for avalanches is not critical at

ail,

unlike some otuer cellular

automaton models [21]. A numencal simulation

by

molecular

dynamics

of a 2D

sandpile

does not

give

clear results

concerning

tuis cntical cuaracter [22].

Tuis paper is divided into turee parts: in tue first part, we present our

experimental

set-up and numerical

techniques.

Tue second part is devoted to tue

dynamicalequilibrium

conditions of a

pile,

and tue tuird part gives our

experimental

and numerical results on statistics of

avalanches.

2.

Description

of Our

System

2.1. EXPERIMENTAL SET-UP. Tue watts of the drum are two

parallel glass

disks with a

diameter of 60 cm. The width between these two disks is

(1+ e)

mm

(e

«

1)

and we have used

particles (disks

or

pentagons)

which are 1 mm thick. Thus the

overlap

of

partiales

is

impossible,

and our system is

actually

2-dimensional. The ensemble is driver in rotation

by

a

motor and a v-belt. Avalanches are studied

by filling

half of the drum with 1400 pentagons

with a side

length

of 6 mm, or with 1400 disks with a diameter of 8 mm. The

packing

fraction

(surface

of the

grain

over total

surface)

of the

pile

is found to be

equal

to o.77 + o.03, a value

(4)

OE~«-UloewoE~-UloeoE~-UloeoE~

-oeOE«Ulw~moe-oeoEUlwmoeooeoE ---oeoeoe number of pentagons

Fig. 1. Distribution of the mass of avalanches observed in the drum for pentagons.

comparable

to that obtained on dense

packings

of pentagons [20]. For

disks,

this

packing

fraction is 0.82 + o.02.

As stated

above,

the surface of the

packing

within the drum is characterized

by

two

angles:

tue

angle

of repose @r and tue maximum

angle

of

stability

@m. Tue dilference

A@(=

@m @r)

between tue two is

generally

of a few

degrees.

For small systems, tuere is a critical

lengtu

1,

defined as the

length

from which one

grain

is seen

by

an

angle

A@. If the size of the system is smaller than 1,

hysteresis

cannot be

observed,

and the nature of avalanches is

changed:

for

example,

some power law for the avalanche size distribution is observed [12]. In our case, this

length

1is 3 cm, which is small

compared

to tue diarneter of tue drum.

Tue rotation speed (uJ) is about o.25 rpm, wuicu is

sufficiently

slow to ensure tuat tue statistics of avalanches is

speed-independent.

A video film of tue drum in rotation is

analyzed by

an

image processing

program

(visilog 4.1.3)

on a work station. From

tuat,

we obtain the statistics of tue

weigut,

duration and

separation

time of

avalanches,

and we are able to measure the

angle

at tue

beginning

and at tue end of an avalanche. Tue results obtained on pentagons

by

tuis metuod are

compared

to tue results tuat we have

previously

obtained on

disks,

and

by

numerical simulations.

For eacu avalanche event, we record grey scale

images rigut

before and

rigut

after tue avalanche and

digitize

tuem using tue

image processing

prograrn. Tuese

pictures

are bina- rized witu a thresuold in order to represent pentagons

by

values of1 and the

background by

o.

On the

picture

taken after an

avalanche,

we

impose

a

negative

rotation

by #

= uJtd, where id is the duration of tuis

avalanche,

to correct tue drum rotation in our calculation of the mass of the avalanche. If one tilts tue voles m the pictures

by

a

geometrical

operation

(dilation)

and

tuen subtracts

tuem,

the

resulting

picture shows tue surface of tue

avalanche,

wuicu we will colt

weigut.

In order to separate

large

events from small events, we have carried out two types of

analysis

of these statistics. A

simple

observation of our

experiments,

or of avalanches at tue surface of an mclined box filled

by sand, clearly

shows that these two dilferent types of avalanches

coexist:

they

will be described later.

Figure gives

the mass distribution of avalanches. It is important to note tuat in tuis

figure,

tue

probability

to observe very

big

avalanches is cut off due to finite size elfects in the drum which prevent the system from

reaching

as

higu

a value

as m the static case. For the

analysis

of

large

events, we suppress events with a surface smaller

(5)

584 JOURNAL DE PHYSIQUE I N°5

o

@ C

©

2ce

OE

, o

~ o

O

~ ~ ~

@ ~

£l

E io0

o ~ o

° cxpcnmcnul

~ C

~ .

~cantcd

o

0

0 5 10 15 20 25 30

deltatheta

Fig. 2. Experimental

(0)

and tueoretical

(-)

relationships between the angle dilference and the avalanche mass, expressed in number of grains.

tuan 25 pentagons.

In order to

study

tue

spatial

distribution of avalanches more

precisely,

we bave built an

experimental

device very similar to tuat used in Drake's

experiment

[23]. It is a 2-dimensional

box, consisting

of two

rectangular glass planes,

200 cm

long

and 30 cm

uigh.

Tue widtu

between tue two

planes

is about 1.2 mm. Tue inclination

angle

of tuis box witu respect to tue horizontal cari be controlled

by

a system of

pulleys.

As in the

drum,

we tilt tuis cell witu styrene disks witu a diameter of 8 mm, or witu pentagons whose

sidelengtu

is 6 mm. It is necessary to use 2400

particles

to obtain

size-independent results,

witu a

pile

about ter

particles uigu.

Actually

tuere are never more tuan rive

layers

of

grains

whicu fait when an avalanche settles.

Thus tue surface on whicu the

partiales

flow is

really

constituted of otuer mobile

particles

like

m the drum.

2.2. EXPERIMENTAL VALIDATION.

Using

pentagons allows us to have an

amorphous

system

without

crystalline

zones, which con induce some non trivial correlations in collective rotations.

Crystalline

structures

clearly modify

the conditions of the static and

dynamic equilibrium.

We have also studied

amorphous

systems of

equal

disks in the

drum,

and in the 2D

box, by gluing

a

rough layer

of disks of dilferent sizes at the externat

watt,

or

by mixing particles

of dilferent sizes.

Figure

2 shows the avalanche mass versus A@, the dilference of the

angle

before and alter

an avalanche. The

points

have been obtained

by

the image

processing

program where we took tue rotation of the drum

dunng

tue avalanche into account. The

straight

fine is the theoretical

curve calculated via the relation

~

~

Î

~~

where L is the

length

of the

pile,

C the

packing fraction,

and

Sp

the surface of one pentagon.

It is obtained

directly by writing

the surface

occupied by

pentagons in an arch of radius

L/2

and of

angle

@.

The

experimental points

are the result of

averaging

the masses obtained in each interval of one

degree. Except

for small events

(which

are,

by

nature, dilferent irom the

large

events descnbed

by

this

formula),

a reasonable agreement is obtained

(the

fluctuations observed are

(6)

essent1ally experimental),

and we conclude that the

experimental weight

is

roughly proportional

to the

angle

dilference of the

large

avalanches.

2.3. NUMERICAL MODEL. In

1979,

Cundall and Strack

proposed

a model to describe the forces

acting

at the

microscopic

level on

grains

in an

assembly

of

spheres

which

they

called the distinct element method [24]. Since then this method has been extended and

slightly

modified to

describe,

among other

things,

the mechanical

sorting

of

grains

[25], the shear-induced

phase

boundaries

[26],

the

origin

of convection cells

[27-29],

and the outflow from a

hopper [30-32].

For the sake of

simplicity,

we will consider

spherical particles,

but if one includes a static friction mechanism

(thereby neglecting

the rotation of the individual

sphere),

we can well

reproduce

the

experimental

results obtained with pentagons, which we will discuss below.

Whenever two

partiales

are doser thon the sum of their

radii,

1-e-, when

they

have been

deformed, they

interact via normal and shear forces. The force on

partiale1,

caused

by partiale j

is

given by:

F~j

=

Fnii

+

Fsé (1)

Fn

=

-kn(r~

+ rj

(ri rj)ii)" ~inme~(f~ fj)11 (2)

Fs

=

-sign(ôs) min(ks jôsj,/J jfnj) (3)

where

m~mj

~~~

mi + mj

and

ôs =

Î

(é~

éj )é

dl

Here ri denotes tue

position

vector of tue i~~

partiale,

ù is tue unit vector

pointing

from the

partiale

to

j

and § is a unit vector

perpendicular

to ii

rotating

clockwise. Tue constant kn stands for tue

"Young modulus",

ks for tue

spring

constant in the suear direction to mimic static friction

[32],

and ôs for tue total shear

displacement during

the contact of the two

partiales.

me~ is the effective mass. Tue shear force Fs bas two

regimes

where tue transition is

given by

the friction coefficient ~t due to the Coulomb

Fs

critenon. For a

= 1, we are left with Hooke's law and for a =

3/2,

we

investigate

the Hertzian contact force for

slightly

deformable disks.

Our units are chosen in such a way that the average

partiale

diameter is unity with a mass

of

unity.

The

pseudo "Young

modulus"

kn

is

equal

to 10~

N/m~/~,

and we used a time step of Ai = 2 x 10~~ s which assured numerical

stability.

The

relationship

between our

damping

parameter ~in and the

experimentally

more

commonly

used restitution coefficient e con be derived

exactly

for a = 1 from equations

(1)

and

(2)

and is of tue form

In(e)

'/n C~

~

For a = 1.5, one

gets

a very weak

velocity dependence

but the above

equation

can still be used

over a wide range. Tuis has been verified for our model

by

one of us [30].

To avoid tue ordered

triangular lattice,

we use a distribution of

partiale

sizes wuere tue size of tue

biggest particle

is four times tue size of tue smallest

partiale.

The drum is ualf filled with

partiales

and contains

rouguly

1450 of tuem. We bave tested our simulation for dilferent rotation

speeds

and have found the three

experimentally

well-known

regimes

of distinct

avalanches,

continuous flow and

centrifugated regime

[18]. The rotation

speed

has been chosen

as uJ = 0.25 rpm in order to be in the distinct avalanche

regime

which is that

experimentally

used,

as discussed in the

preceding

section.

(7)

586 JOURNAL DE PHYSIQUE I N°5

0.2 0.2

P(8M)

(a)

P(8R)

(b)

0.ifi 0.ifi

0.12 0.12

0.08 0.08

0.04 0.04

0 0

30 40 60 60 30 40 60 60

BM

8R

Fig. 3. Probability distribution of the maximum angle of stability 6m

(a)

and of the angle of repose 6r

(b)

of an avalanche. These distributions are obtained experimentally with pentagons.

As it has been

argued

in Section 2.2, the mass of an avalanche is

proportional

to the

angle

dilference A@

= @m-@r. Since the latter is measured more

easily

in our

simulations,

we will now describe the

procedure.

At each iteration, the

rotating

drum is divided into 100 vertical

strips

and the position of the

highest partiale

in each strip is searched. From these 100

partiales,

we obtain the

angle

of the surface

by

a

least-squares

fit. The

angle

is

averaged

over 1000 iterations and we check our automated

algorithm by comparing

the results with a visual manual fit for

dilferent times.

3.

Dynamic Equilibrium

of trie Pile and

Geometry

We have seen that in the intermittent

regime (avalanches),

a

heap

of sand is cuaracterized

by

two

angles:

tue maximum

angle

of

stability

@m and tue

angle

of repose @r. Tuese

angles

are

non-trivial functions of tue coefficient of friction between

grains,

tue geometry of tue system

at tue individual scale

(shape

of tue grains, etc. and at tue

global

scale.

3.1. EXPERIMENTAL ANALYSIS. Each

experimental angle

is determined as trie result of

an

averaging

process

(least-squares metuod)

over tue drum

diameter,

in order to smooth the

roughness

of the surface.

Experimentally,

it is clear that there is a more or less broad dis- tribution of the measured

angle

of repose @r and of the maximum

angle

of

stability

@m. The distributions for pentagons obtained

by measuring

260 avalanches are

given

m

Figure

3: the

mean value of

@r is 37° and that of @m is 45°.

Experimentally

too, for

amorphous (1.e.,

with disk size

distribution) packings

of

disks,

we have obtained @r

= 24° and @m

= 33°. Two reasons

con

explam

this dilference:

first,

because of face-to-face contacts, each pentagon is more stable thon the

corresponding

disk. But there are also strong correlations in the relative rotations of pentagons [33], and tue

beginning

of an avalanche and its motion are more difficult in this

case: a

pile

of pentagons is more stable tuan a

pile

of disks.

As suown in

Figure

4, tue

experimental angle

dilference A@, wuicu measures tue range of avalanche

sizes,

is very wide: it ranges from

nearly

zero to

nearly

20

degrees,

witu a mean value of 9°. One notes tuat tue

angles

between 2 and 14

degrees

are almost

equally probable.

(8)

o.io

P(A8)

0.08

o-où

0.04

0.02

0

0 10 20

A8

Fig. 4. Probability distribution

P(A6)

of the dilference A6 between the starting and stopping angle

of an avalanche, obtained experimentally with pentagons.

Tuis

large

range is due to tue fact

tuat,

for pentagons, tue

possibility

to find an

equilibrium (individual

or

collective)

is greater

compared

to tuat for disks

3.2. NUMERICAL APPROACH. It is difficult to make a

quantitative expenmental analysis

of tue influence of tue friction coefficient between

grains

on the

angles

@r and @m. In return,

using

our numerical

simulations,

since the introduced

spring

constant ks is ratuer

artific1al,

we may obtain a reasonable value of tue

angles defining

tue

equilibrium

of the

heap

from

some test runs. We fill the drum

halfway

with

partiales

of dilferent diameters

(which

is the

way of

introducing

disorder in the numencal

simulations)

and start to rotate it. After some thousands of

iterations,

we stop the drum rotation and watt for the system to relax: we control the relaxation

by monitoring

the kinetic energy. We measure the obtained

angle

of repose @r for dilferent values of k~ and of tue Coulomb turesuold ~t. Tue results are

presented

in

Figure

5, in whicu the continuous litre

corresponds

to the Coulomb relation @r

=

tan~~

~t [32].

A linear

relationship

is obtained for @r as a function of ks for small values of p. For

larger

values of ~t, a saturation elfect is observed. In order to obtain

equilibrium

conditions in nu- merica1 simulations and in

experiments

on pentagons, we have

adjusted

our simulations

by choosing

~t

= o-G for a value of ks

= 1000 N.m~~

Doing this,

we obtain à numerica1value of

@m near the

expenmental

one.

Figure

6 shows the distribution of

angles

obtained

numerically

for

non-rotating

disks

averaged

over 112 avalanches: we obtain @m

= 45° and @r

= 40°.

Therefore one has to

keep

in mind that, in spite of an apparent

quantitative

agreement, there are strong

geometrical

dilferences between tue two systems

(disks

and

pentagons),

as

discussed above: tue numerical value of @r is dilferent from tue

experimental

one. Here too tue

experimental packings

of pentagons

comparatively

olfer a

larger

number of

equilibrium

states than in

numericalsystems

of

disks,

even with a strong friction: tuis

explains

tuat tue distributions of

angles

are

larger

for

experimental

pentagons

(see Fig. 3)

than for numerical disks. These distributions are trot

symmetrical,

and we believe that this is due for a

large

part

to limite size elfects

(see below).

4. Avalanches

4.1. DEFINITIONS. An avalanche is here defined as the motion of at least one

particle

down tue inclined surface.

Starting

from tuis

definition,

we must

distinguisu

between two

(9)
(10)
(11)

590 JOURNAL DE PHYSIQUE I N°5

(a) (b)

20000 ioooo

p J~

ioooo ioooo

0 0

0 20 40 fi0 80 100 0 20 40 60

t Îs) t (s)

a) b)

Fig. 8.

(a)

Fluctuations m pixel of two successive pictures obtained experimentally in the 2D drum filled with pentagons. Measurements are taken everyof rotation;

(b)

same result obtained

numerically. Measurements are taken every o.15° of rotation.

Tuis initiates a collective

back-avalanche,

wuicu is a

global

loss of

equilibrium involving

ail

partiales

of tue

pile leading

to a total reconstruction of tue

surface,

wuicu is

finally

inclined from tue horizontal

by

@r.

Small events

only

involve one or few

grains

m metastable

equilibrium

on the surface of tue

pile.

Tuis loss of

equilibrium

is

individual,

and is of the type observed if one

places

one

grain

on a

rougu

inclined surface built

by gluing glass spueres

on it [36,

37].

Wuen it loses its

equilibrium,

tue

grain

falls

down,

and

during

its

motion,

it bas some collisions witu otuer

grains,

and stops or may create a collective motion.

Orly

a weak and

generally

local modification in tue surface

morpuology

is observed

along

a

lengtu

1,

being

tue free

patu

of tuis event,

generally

smaller tuan tue

lengtu

of tue free "surface". Tue occurrence of tuese events

depends

on tue

morpuology

of the

pile surface,

on the inclination

angle

and on its fluctuations.

In the

following

part, we will descnbe tue statistics of tuese two dilferent events.

4.2. SMALL EVENTS

4.2.1.

Experimental A~ialysis.

Bretz et ai. [34] bave suown tuat tue surface of a

sandpile

is

subject

to small fluctuations wuen one

slowly

increases the

angle

of repose. Tuese small events

occur between

large

avalanches: this is what we observe in both

experiments

in the 2D drum

and in a 2D inclined chute.

They

evolve less than 30

particules (disks

or

pentagons).

Figure

8a illustrates the relative importance of each type of events, and how

they

are time- distributed: it

gives

the mass fluctuation at tue surface of tue 2D

pile

of

pentagons

m tue drum.

To obtain this

(small)

mass

variation,

we

analyse

the pictures, but

only

between identified

large

events. We

plot

the dilference in

pixels

between two successive

pictures

taken with an interval of time

equal

to 0.67 s

imposed by

the resolution of the

image processing

program and

exactly

the sonne

procedure

as descnbed in Section 2.1 is

applied

to measure the fluctuation of

weight.

The area obtained

by

tue subtraction of the two pictures results from the

reorganization

of the

(12)

Î t~

= z

~ n

m

~ n

fl o

u

20 40 60 80 100 120 140 160 180 200 20 40 60 80 100 120 140 160 180 200

a)

startIng position (cm)

b)

~~~~~'~~ P°~'~'°~ l~~l

Fig. 9. Probability for

a disk to be the seed of a small event at distance L from the top of the pile,

in the inchned chute:

(a)

pentagons;

(b)

disks.

heap

surface

during

tue rotation of1°

(rotàtion angle

of tue drum

during

tue interval of time of 0.67

s).

We

apply

tue same

procedure

at eacu rotation

by

one

degree

until the next

large

event. The statistics obtained are the sum of these

analyses.

Even if the

experimental

spectrum obtained from the square of the

amplitude

of these mass fluctuations could be fitted

by

a

straight

litre with a

slope

of the order

of1.5,

the statistics

are not sufficient to be affirmative

jbout

any power law behaviour. The numerica1simulations

can

explore

the fluctuations of the

pile

surface

during

an interval of time reduced to 0.1 s.

Tuus we suall use tuese numerical simulations to describe a better

analysis

of tuese small mass fluctuations in tue

following

section.

According

to results obtained on tue inclined

plane [36, 37],

it is reasonable to expect a

uomogeneous

distribution of tue

starting positions along

tue line for tuese small events: one

con believe tuat at tue surface of a

packing,

tue

probability

to find

partiales

in metastable

equilibrium

is tue same

everywuere

on tue surface.

We

plotted

on

uistograms presented

in

Figures

9a and 9b tue

probability

for a small event

to occur at a distance L from the top for pentagons

(a)

and disks

(b),

obtained in the indined

chute. This

probability

is not constant

along

the

surface,

neither for

disks,

nor for

pentagons.

On the contrary, we observe that the

probability

of the

starting position

decreases

nearly linearly

with L. This evolution is smoother for disks thon for pentagons. Nevertheless small events are intercalated between

large

events, wuich lead to a

complete reorganization

of the tue surface. After the reconstruction of the

pile,

the successive small events will

destroy

the metastable states below its

starting

position up to its capture

by

the bulk: the surface is

progressively

washed. Tuis

explains

trie results suown

Figure

9.

4.2.2. Numencal

Analysis.

In tue case of numerical studies of small events, we map every o.1 s our

grain configuration

to a soc x 500

binary grid

wuere a 1 in tue cell indicated that

more tuan 501~ of the cell area is covered

by

a

partiale.

Two successive grids are subtracted from another and the dilference is measured in cells. This dilference is our measurement for the small surface fluctuations.

A

typical

time trace of our numerical system is shown in

Figure

10. In part

(a)

we show the

angle

of the surface

averaged

over looo iterations

(which

has been obtained

by dividing

the

drum in loo vertical

regions, looking

for the coordinate of the

highest partiale

and

making

a

least-squares fit).

In part

(b)

we show the total kinetic energy of the

partides.

One sees how eacu

peak

in the energy

corresponds

to a drastic

change

in tue surface

angle corresponding

to a

iouxNAL DB pHYsioçB i T.s,w s,MAy 199s n

(13)

592 JOURNAL DE PHYSIQUE I N°5

2000

e 1',1

Il

,ji

'

jçoo

~~

.,~

~(

~Î. Ji'

~

~~ 1,~

,

' J 1200

v

j

Si t 'j

) ' soc

~~ 400

, ,

~

6o go 100 120

j~ ,~ fi0 80 100 120

1trie (si tinte (si

a) b)

Fig. 10.

(a)

Surface angle fluctuation obtained numerically;

(b)

corresponding total kinetic energy fluctuations.

cuaracteristic avalanche. Also notable are tue small fluctuations

(small events)

in botu

grapus (anotuer

source for fluctuations are tue statistical errors of our measurement

technique).

From these numerical results, we con draw

Figure

8b. The noise

(at

a

frequency roughly equal

to

Hz)

observed in this

figure,

then

concerning

numerical

simulations,

is

essentially

due to the numerical method. Between

large

events, we observe a succession of very small events

(one

has to

keep

in mind that the

large

event distribution is cut off

by

the finite size of

our

system).

As before we

plotted

the power spectrum of the square of the

amplitude

of the mass fluc- tuations. The data

only ranged

over one order of

magnitude

with an error of

roughly

o-à- The numerical

results,

shown in

Figure

ii are fitted

by

a

straight

line with a

slope of1.8,

not

alfected

by

the "numerica1noise" which appears at

frequency

15 Hz.

4.2.3. Conclusion. These numerical and

experimental results,

which

give

tue power laws for tuese small events, are

qualitatively

similar to tuose obtained in 3 dimensions

[12, 34]. However,

because of the small size of our systems, no definitive conclusion cari be drawn.

4.3. LARGE EVENTS.

Large

events involve trie whole

pile.

As it is a

global

process, which is trot modified

by

small events, the

probability P(L)

to observe an avalanche

starting

at a

distance L from the top of the

pile

is

expected

to be constant.

P(L),

determmed in our inclined

chute,

is

given

in

Figure

12a for pentagons and in

Figure

12b for disks.

Except

for small or

large

values of

L, P(L)

is

roughly

constant in both cases. Near the

bottom,

the slide of a

row of

partiales

is blocked

by

the wall and near the top, the pressure on a

grain

is screened

near the wall: this

explains why P(L)

is smaller at the two ends of the distribution. These two screening

lengths

are a little

larger

for pentagons than for

disks,

as the

stability

of these

grains

is

larger. They

are of about 40

particle

diameters.

Accordingly,

finite size effects are

expected

to be

significant

in our experiments in a drum: for disks in a

drum,

the

histogram

is

comparable

to that of

Figure

12a.

Figure

13 shows that

P(Ù)

is not a constant of the starting position for pentagons in the drum: tue first

partiale

wuicu flows down tue

pile

is located

mostly

at tue top of tue surface.

Thjs dilference in

P(L')

between disks and pentagons con be correlated to tue

geometrical

(14)

le-05

~p

ie-06

ie-llî

le-os

Ie-09

le-10

o-1 10 10']

f[H2]

Fig. Il. Power spectrum relative to surface fluctuations obtained numerically.

£~ ~

é ~

W £

n @

o n

ÉO

20 40 60 80 100 120 140160 180 200 20 40 60 80 100120 140 160180 200

startIng position (cm) startIng position (cm)

a) b)

Fig. 12.

P(L),

probability for an avalanche to start at a distance L from the bottom of the pile m

tue box, uprsus L:

(a)

pentagons;

(b)

disks.

configuration

of tue top of tue

pile just

before tue avalanche

(see Fig. 14):

for

disks,

tue surface of tue

pile

in tue drum is

regular (Fig. 14a),

wuereas for pentagons, tue

suape

is very dilferent near trie top

(Fig. 14b). Figure

14a shows tue

configuration

of tue top of tue surface for a mixture of small and

large disks;

but tuere is no

change

in tuis

pile geometry

wuen built

only

witu

large partiales.

Because tue

slope

is steeper in tuis part, tue top of tue pentagon

pile

becomes very instable wuen becomes of tue order of @m. Therefore tue avalanches start

preferentially

at tue top.

Figure

15a shows the distribution of tue duration time of an avalanche and

Figure 15b,

tue

separation

time between two running avalanches. The

uistograrn

of the duration looks

quite

suarp, and tue mean duration time is 2.7 seconds. Tue

separation

distribution is broader thon tue duration one and gives a mean separation time of 8-9 seconds.

(15)

594 JOURNAL DE PHYSIQUE I N°5

ce + W W O ce + W w O ce + W w

ce ce ce ce ce

Starting length (cm)

Fig. 13. Probability for an avalanche to start at a distance L' from the top of trie pile in the drum, for pentagons.

i

b) a)

Fig. 14. Photographs

showmg

the geometry of the surface of the pile m trie drum:

(a)

disks;

(b)

pentagons.

5. Conclusion

While made m

relatively

small systems, our

study

in 2 dimensions has confirmed the bimodal character of avalanches at the surface of a

pile:

small events, wuose size ditribution is described

by

a power

law,

and cuaracteristic

(large)

events wuose size and time duration are

well-averaged.

Tuese results are

qualitatively

tue same wuetuer one

uses

disks,

or

pentagons

in expenments.

Tue numerical simulations are

essentially

used as a

complementary

tool to

study

in more detail tue influence of some

puysical

parameters wuich cannot be reached

experimentally (friction,

energy, etc. ).

(16)

0.4

~~~~ l~) Plt) jb)

0.3 0.12

0.2 0.08

o-1 0.04

0 0

0 2 4 fi 0 4 8 12

t [s] t [s]

Fig. 15.

(a)

Distribution of avalanche duration;

(b)

distribution of time separation between two successive avalanches, obtained experimentally with pentagons.

Beyond

tuese results, we bave outlined tue strong influence of tue

partiale shape

and friction coefficient on tue

angle

of repose or of maximum

stability. Finally,

because of tue extreme

sensitivity

of tue avalanche to tue local and

global equilibrium,

we bave suown that

geometrical

limite size elfects are important in these systems.

Acknowledgments

This work has benefited from useful discussions with J-P-

Troadec,

A. Hansen and S. Roux. It has been

partly supported by

the GdR CNRS

"Physique

des Milieux

Hétérogènes Complexes".

References

[1] Faraday M., Phil. Trans. Roy. Soc. London 52

(1831)

299.

[2] Bideau D. and Hansen A., Disorder and Granular Media

(North

Holland,

1993).

[3] Mehta A., Granular Matter

(Spnnger,

1994).

[4] Jaeger H-M- and Nagel S-R-, Science 255

(1992)

1523.

[Si Bak P., Tang C. and Wiesenfeld K., Phys. Reu. Lent. 59

(1987)

381.

[6] Tang C. and Bak P., Phys. Reu. Lent. 60

(1988)

2347.

[7] Dhar D., Phys. Reu. Lent. 62

(1989)

1813.

[8] Gutfram R., Private Communication.

[9] Puhl H., Physica A 197

(1993)

14.

[10] Jaeger H-M-, Liu Ch. and Nagel S-R-, Phys. Reu. Lent. 62

(1989)

40.

[Il]

Evesque P., Phys. Reu. A 43

(1991)

270.

[12] Held G-A-, Solina D.A. Il, Keane D.T., Haag W-J-, Horn P-M- and Grinstein G., Phys. Reu. Lent.

65

(1990)

l120.

[13] Rosendahl J., Vekic M. and Kelley J., Phys. Reu. E 47

(1993)

1401.

(17)

596 JOURNAL DE PHYSIQUE I N°5

[14] Reynolds O., Phd. Mag. 20 (1885) 469.

[15] Evesque P., Fargeix D., Habib P., Luong M.P. and Porion P., Phys. Reu. E 47

(1993)

2326.

[16] Porion P., Thèse, Université des Sciences et Technologie de Lille

(1994).

[17] de Coulomb C.A., "Mémoires de Mathématiques et de Physique présentés à l'Académie Royale des Sciences" 7

(1773)

343.

[18] Rajchenbach J., Phys. Reu. Lent. 65

(1990)

2221.

[19] Caponeri M., Douady S., Fauve S. and Laroche C., prepnnt.

[20] Ammi M., Bideau D. and Troadec J-P-, J. Phys. D 20

(1987)

424.

[21] Kadanolf L.P., Nagel S-R-, Wu L., Zhou S., Phys. Reu. A 39

(1989]

6524.

[22] Lee J., J. Phys. I France 3

(1993)

2017.

[23] Drake T.G., J. Geophys. Res. 95

(1990)

8681.

[24] Cundall P. and Strack O.D.L., Geotechniq. 29

(1979)

47.

[25] Half P.'K. and Werner B-T-, Powder Technol. 48

(1986)

239.

[26] Thompson P-A- and Grest G-S-, Phys. Reu. Lent. 67

(1991)

1751.

[27] Ristow G-H-, J. Phys. I France 2

(1992)

649.

[28] Taguchi Y.H., Phys. Reu. Lent. 69

(1992)

1367.

[29] Gallas J-A-C-, Herrmann H-J- and Sokolowski S., Phys. Reu. Lent. 69

(1992)

1371.

[30] Ristow G-H-, Int. J. Mod. Phys. C 3

(1992)1281.

[31] Ristow G-H- and Herrmann H-J-, Phys. Reu. E, in press.

[32] Lee J. and Herrmann H.J., J. Phys. A 26

(1993)

373.

[33] Limon-Duparcmeur Y., pnvate communication.

[34] Bretz M., Cunningham J-B-, Kurczynski P-L- and Non F., Phys. Reu. Lent. 69

(1992)

2431.

[35] Carlson J-M-, J. Geophys. Res. 96

(1991)

4255.

[36] Riguidel F-X-, Jullien R., Ristow G-H-, Hansen A. and Bideau D., J. Phys. I France 4

(1994)

261.

[37] Riguidel F-X-, Hansen A. and Bideau D., Europhys. Lent., submitted.

Références

Documents relatifs

Deux d’entre nous sommes autorisés à rester avec l’équipe médicale aux côtés de notre chef d’intervention : une de mes collègues qui en train de devenir elle-même

On avait aussi vu une ribam- belle de jeunes journalistes en profiter pour se mettre en avant, soulignant le fait que Mikael Blomkvist n’était plus dans le coup, qu’il

• Gretel doit cuisiner pour engraisser son frère et nourrir la sorcière.. • Gretel tue

» C’est pour ça, quand sont sorties les nouvelles courbes de l’Insee, sans verser dans le complotisme, je me suis rappelé cette discussion.. Enfin, pendant notre

Par exemple, lorsque l’on entre ces deux mots dans SUDOC, la base de données universitaire française rassemblant plus de 13 millions de notices biblio- graphiques, aucun résultat

CHEVAL RENARD CRABE CANARD. CHAT CHAMEAU

RENARD NICHE CANAPE NUAGE.

MUR ECUREUIL JUMELLES AUTRUCHE. RUE TORTUE