• Aucun résultat trouvé

Experimental evidence of self organized criticality in the stick-slip dynamics of two rough elastic surfaces

N/A
N/A
Protected

Academic year: 2021

Partager "Experimental evidence of self organized criticality in the stick-slip dynamics of two rough elastic surfaces"

Copied!
14
0
0

Texte intégral

(1)

HAL Id: jpa-00246899

https://hal.archives-ouvertes.fr/jpa-00246899

Submitted on 1 Jan 1994

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.

stick-slip dynamics of two rough elastic surfaces

S. Ciliberto, C. Laroche

To cite this version:

S. Ciliberto, C. Laroche. Experimental evidence of self organized criticality in the stick-slip dynam- ics of two rough elastic surfaces. Journal de Physique I, EDP Sciences, 1994, 4 (2), pp.223-235.

�10.1051/jp1:1994134�. �jpa-00246899�

(2)

Classification Physics Abstracts

05.45 9i.30 05.40

Experimental evidence of self organized criticality in the

stick-slip dynamics of two rough elastic surfaces

S. Ciliberto and C. Laroche

Laboratoire de Physique

(*),

Ecole Normale Supérieure de Lyon, 46 Allée d'Italie, 69364 Lyon, Cedex 07, France

(Received

18 August 1993, revised 1 October 1993, accepted 4 November

1993)

Résumé. Nous décrivons une expérience sur la dynamique du frottement solide entre deux surfaces élastiques. La dynamique révèle un comportement critique auto-organisé sur un grand intervalle des paramètres de contrôle qui sont la vitesse de glissement et la pression. L'effet de la taille du système a été étudié. Nous faisons aussi une comparaison avec la dynamique produite

par un système à une seule dimension et par rapport au cas les deux surfaces sont rigides.

Abstract. An expenment on the stick slip dynamics of two elastic

rough

surfaces is de- scnbed. This dynamics produces a self organized cntical state on a wide range of control parameters, namely loading speed and pressure. The dependence on the system size has also been studied. A comparison with the stick-slip dynamics produced by a one d1nlensional system and by two rigid surfaces is made.

1 Introduction.

Mechanical

models,

based on the stick

slip dynamics

of blocks connected

by springs,

and

sliding

on a surface with a

given

friction law, bave been

recently widely

studied to

give

more

insight

into trie mechanisms of

earthquakes

and friction

iii.

Several of these models [2-6] are

directly

derivable from trie

original

one of

Burridge

and

Knopoif

[7], however tue

stick-slip dynamics

bas been also

investigated

using cellular automata

[8-12].

Ail of these models are characterized

by

a

loading period

where tue system accumulates energy and a

slip period,

" tue

earthquake",

where part of this energy is released. An important issue m this

study

is to understand under what conditions these models present self

organized

critical

(S.O.C.)

states [8,

13].

In a

geophysical

context S-O-C- could

help

to rationalize severa1observations and a detailed discussion on this

point

can be found in references

[16,

13]. For what concerns tue above mentioned models one

(*) URA 1325.

(3)

wants to know whether it is

possible

to

reproduce

tue

Gutenberg-Richter

Iaw

(G.R.)

[14] for the number of occurrence N of

earthquakes

with a

given

moment

E, specifically NIE)

cc

E~~~~

We recall that for real

earthquakes

trie average value of B is around 1

iii.

Except

for a few numerical simulations where the G-R law is satisfied tilt tue maximum energy [4, 5,

iii,

in

general NIE)

presents an anomalous

peak

for

large

E [2], that is to say trie

scaling

is self-similar

only

for small

earthquakes.

There are many reasons [2, 10]

why

this may

happen

but here we want to focus

only

on the

experimental

verification of G-R- and on trie

comparison

with tue numerical results.

Indeed,

in spite of trie

large

amount of numerical data

on trie above mentioned

models,

to trie best of our

knowledge,

there are

only

a few

laboratory

experiments [12, là,

16],

where trie stick

slip dynamics

bas been tested from trie point of view of its statistical

properties.

Trie great diiliculties of these experiments are that trie surfaces, which are

shding

one over

another, change

their properties as a function of time because of trie

erosion due to trie continuous friction. Therefore statistics should be constructed on very short

time

intervals,

otherwise

they

coula be influenced

by

trie detenoration of trie surfaces. This effect is

clearly pointed

out in reference [15] where is described a very

interesting experiment

which allows one tu measure not

only averaged

properties but aise the localisation m space and time of trie

stick-slip dynamics.

This experiment, as many other numerical simulations [2, 9,

loi,

shows very

dearly

for a

large

range of trie control parameters,

namely

pressure and

sliding speed,

a power law distribution for trie

slip amplitudes

with an enhanced

probability peak

at

high energies.

Trie other experimental paper [12] shows a power law on trie

integrated probability

distribution function for

only

a value of trie central parameter, thus we do net know

if this coeurs in a wide range of central parameters. Furthermore tue

frequency

spectrum of the

friction force versus time presents trie triv1al power law

f~~

In other words we do net know if that system is in a state of self

organized criticality

[8]. In this paper we describe an expenment

on

stick-slip dynamics

in which ail the above mentioned

problems

are solved and the system presents a self

organized

critical state in a wide range of contrai parameters. Tu obtain this result much care has been taken in order tu construct the

sliding

surfaces which are described

m section 2

together

with tue experimental apparatus. Trie rest of trie paper is

orgamzed

as

follows. In section 3 we report trie statistical properties of trie

stick-slip dynamics.

in section 4 we compare systems with different properties,

namely elasticity

and

dimensionahty. Finally

trie results are discussed in section 5.

2. The

experimental

set up.

The very

simple experimental

apparatus is shown in

figure

1. Two expenments with different

geometries

bave been

performed,

but no relevant quantitative difference bas been noticed except

for trie maximum value of trie

scaling

exportent.

2.1 THE ANNULAR EXPERIMENT. in trie first une,

figures

la, 16, le, trie

sliding

surfaces

bave an annular

shape

of width h

= 1 cm and mean diameter D. We used three values of D in order tu

verify

if trie results are

independent

of trie size of trie surfaces. Trie two

sliding

surfaces

are mounted on two aluminium disks which are 2 cm thick in order to avoid deformations.

One of trie two disks is

kept

fixed and the torque necessary to compensate trie friction force Ff between trie two

sliding

surfaces is measured

by

a torque transducer.

Specifica1Iy

we measure Ff x

D/2.

Trie transducer

signal smtably amplified

and filtered is couverted

by

a 16 bits

AID

converter. Trie minimum detectable variation of trie friction force Ff is about 3 x lo~2 N. Trie other aluminium disk was driven at constant

speed

Q

= 2 x

V/D by

a

reduction-gear

and a de electric motor, here V is trie

loading speed.

A micrometric device a1Iows us to

change

trie

(4)

a)

~~

~~

ÎÎfl 5''

'C°n ~U°b~~

F-D

2 D

,

i'

',

i

j 1

C)

1

steel sJ~eres

~

d

)

§j Wood table

flfl Silicon Rubber

Fn v

steel s ~ere ~

Fj

Fig. l. Expenmental apparatus.

a)

Schematic drawmg of the disk expenment, trie torque F x D necessary to keep the disk fixed is measured by a torque transducer while the other disk is maintained

at constant angular speed V x D. b) Disk surface. c) Expanded cross section of a disk showing how

the steel spheres are glued inside the groove with the silicon rubber.

d)

Cross section of the linear geometry expenment; the sbde is moved at

a constant speed V and the friction dependent force Ff is nleasured by a force transducer. The wood table is 180

cm long.

pressure with which the surfaces of trie two aluminium disks are

pushed against

une another and trie axia1force is aise measured. Trie two annular

shding surfaces, figures 16,

le, mounted

on the two aluminium disks were

prepared

in the

following

way. First of ail an annular groove of

depth

1.5 mm and width 1 cm is made on trie disk surface

(see Fig. l).

Three grooves with different mean diameters D bave been

made, specifically

9.3 cm,cm and 29.3 cm. We put inside trie grooves steel

spheres

of diameter 2 mm m such a way that trie distance between

(5)

sphere centers was about two diameters.

Finally

trie grooves were filled with silicon rubber.

In this way an artificia1surface with controlled

roughness

and

elasticity

is constructed. Trie maximum

roughness

is

exactly

o-à mm

equal

tu trie

sphere

diameter minus trie

depth

of trie groove. Furthermore each

single sphere

is rather free tu move around its

equilibrium position

because of trie

elasticity

of trie rubber. We will call this surface soft surface tu be

distinguished

from an other type of surface which is very hard and whose

roughness

dues trot

change

toc much because of trie friction. This surface is constructed

by dring

a mixture of

glass

microspheres

(10

~m of

diameter)

and

"stycast" glue which,

once it is dried, becomes a very

rigid

surface with a well defined

roughness.

We will call this surface

rigid

surface.

Incidentally

we remind that

surfaces,

very similar tu trie soft une, bave been used tu make mathematical models of friction laws between two

rougir

surfaces

il?,

18].

Unfortunately

in these two references tue

Ioca1interactions are different from those of this experiment, thus a direct

comparison

is net

possible.

However

experiments

of this kind associated with numerical models

il?,

18] can be useful in order tu

give

more

insight

into trie

study

of friction laws.

2. 2 THE LINEAR EXPERIMENT. Trie second

experimental

set-up, shown in

figure id,

used

instead a linear geometry. A slide is moved at constant

speed

V on a

rougir

surface constructed with steel

spheres emerging

of o-à mm from a wood

plate

180 cm

long

and 20 cm wide. Trie force necessary to move trie slide at constant

speed

is measured

by

a force transducer. The normal force

acting

on the surface of the slide can be

changed by modifying

trie

weight

of trie slide. Trie surface of trie shde was constructed in trie same way as trie soft surface described

above. Two slides are used in order to

study

trie

dependence

on trie system size, one 9.5 x18 cm~

and trie other 9.5 x 10 cm~. Trie

depth

of trie groove where steel

spheres

are inserted is 1.5 mm

as for trie annular surfaces.

We

point

ont that there is a very

important

diiference between trie disk experiment and trie linear one. Indeed in trie disk experiment it is trie distance between trie two disks that

is

kept

fixed and trie normal force

changes

because of trie

dynamics.

In contrast, in trie slide experiment, it is trie normal force that remains constant.

However,

in

spite

of this

important diiference,

trie results are similar m trie two cases, thus we will net in

general

make a distinction

between them. Trie reason for this

similarity

between trie two

geometries

is

probably

due to trie very

large compliance

of trie matrix and could be net true m a very

rigid

case.

It is aise important to

explain

how trie

spheres actua1Iy

move.

By Iooking

at

figure

1 we see that trie

spheres

are

touching

trie bottom of trie

plate. Thus,

when a

sphere

is

pushed against

an

asperity

of trie opposite

surface,

it Carnot pass over this asperity but it can

only

move

transversely

to trie direction of trie

loading velocity

m order to go around trie obstacle.

This bas been verified

by making

in trie hnear experiment a transparent slide and

looking

from

above,

with a

videocamera,

trie motion of several

spheres.

This is useful aise to estimate the maximum

displacement

of a

sphere

that is about 0.2 mm.

2,3 PHYSICAL PROPERTIES OF THE SURFACES AND TYPICAL TIME SCALES. Trie stick-

slip dynamics produced by

trie interaction of trie diiferent surfaces put in contact bave been studied in trie case of trie disks. Three cases bave been considered: trie

ngid-ngid, rigid-soft

and soft-soft interactions. Trie

rigid-rigid

interaction is net very interesting from a statistical point view and it wiII be discussed in section 4.2.

Trie

rigid-soft

and soft-soft interactions, which are trie most important,

give

similar results and

they

will be

analysed

in fuII detail in section 3. Their

dynamics

lias been studied as a

function of trie

applied

pressure and of tue

loading speed

V. Trie values of trie

apphed

param-

eters must be

compared

with those obtained

using

trie

physical properties

of trie experiment.

Tue mass M~ of tue

spheres

is 4 x 10~~

kg

and trie total number of

spheres

is m between 200

(6)

and 300. The total mass

MT

involved is about 1

kg.

With trie total mass we indicate trie

sum of

sphere

masses

plus

either that of tue disk or that of trie slide

depending

on tue

experi-

mental geometry. Tue

Young

modulus of tue E of tue rubber where

spheres

are embedded is 5 x 10~

N/m~. Therefore, using

as a

typical Iength

scale tue

sphere

diameter d

= 2 x

10~~

m,

trie elastic constant k

= E x d is 10~

N/m.

Two

typical

time Saales can be

defined,

one based

on trie acceleration of Ma and trie other on MT: TT =

(MT /k)~/~

and Ta =

(Ma /k)~/~

We find that Ta ce 2 x 10~4

s whereas TT m 3 x 10~~s. As it wiII be discussed in section 3, this value of TT is very close to trie

sliding

time for trie smallest events measured on tue friction

force,

meaning that tue motion is dominated

by

tue total inertia. Tue

darnping

time of tue

oscillation of M~ and MT are of trie same order as Ta and TT

respectively.

This means that tue system is over

darnped

and oscillations are not allowed.

Another important time scale is

Tv

"

d/V

which

corresponds

to tue time needed to move

tue

plate

of a characteristic

Iength;

we wiII see that this time

corresponds

to the

Ioading

time between two very

large

events.

Z.

o a

a.z6 o

@.

~

o

a

o a

@. o

o on

@.

~.5 1. I.E 2.a 2.5

v (mm/Gec)

Fig. 2. Friction coefficient for soft-soft surfaces as a function of the loading speed. The different symbols correspond to measurements made with different normal forces.

We bave measured au average friction coefficient ~

= <

Ff(t)

>

/Fn

of our artificial surfaces

as a function of V

by

measunng tue mean value of tue friction force

Ff(t)

divided

by

tue mean normal force Fn

(here

< > stands for time

average).

We find that ~ increases

slightly

as a

function of V. Tue same property bas also been

reported

in reference [15]. Tue measurements

of ~ versus V are shown in

figure

2 for tue soft-soft case. It is important to

point

out that this

is trot a

precise

measurement of ~ because a measurement of this type is

strongly

influenced

by

tue stick

slip dynamics

whose

amplitude

is

velocity dependent.

A more accurate evaluation of tue friction coefficient would require tue suppression of tue

dynarnics;

this is not easy in tue

case of trie soft surfaces. Thus the behaviour of ~ as a function of V shown in

figure

2 is not

quantitative

but it is useful

just

to give an idea of the behaviour of the ratio between the mean friction force and the

applied

normal force.

(7)

15@a

~m

z i Jzz

~

~

~L~

~j 1@Z

ù

'

/~~

/-.

~

~~

/

>/ ' ~.

T

Î,~~~~~/

/~

~

/?~

/

/~

/

i

~ 7zz i

ij

àze ~----<--,

2C ~'z fié Ba "@3 2~ ù@ 't-C'

i se

c

-~~~~ ~

c

z 3@z

j /

~ /~i

/ /

$ Z@(

~~/ j/

~

~

/l~__ ,/

~

/~ /~ /

~ ~°~

/ /

J

£ /

1' ,/

~

~ôzl

~,

/ /~

~

/

/

~,'

l/

>,/

~~~à

l~ le m 4Q

~1 f(ÎÎ~ iÎ

se c

iz@@ -j

~

C Z

ii ji

aZ@~

~ )~ j j

j1)

j

/

/j

/ ' ,

o Il

j)

' ~

( ' ' '

Ezz

,

i ' [ iii

~

j

?~j, lj~l ii

'

~m ~ Ii ,j< jj

jj

jj

j'

~

~jji

'

f , j, '

7 1'

'

~@@ Î

~ ' Î

l

@@~»>

»>,» »»<>~"'>'>

5 je i~ 2@ 25 2@ 25 ~C

1Sec)

Fig. 3. Friction force versus t1nle for diierent loading speeds. a) V = 0.04

mal/s;

b) V

= o-1

nlm/sj

c) V =1.25

mal/s.

(8)

3. The statistical

properties

of stick

slip.

We now

analyse

tue statistical

properties

of tue stick

slip dynamics

in our system. To do this

we first record

Fit)

and an

example

for tue soft-soft case is shown in

figure

3. We

dearly distinguish loading cycles

where the two surfaces are stuck and

Fit)

increases up to time t;

where it reaches a relative maximum. At that

point

tue

slip period begins

and Iasts until time te where

Fit)

reaches a relative minimum and then it

begins

to mcrease

again.

We define the

amplitude

of tue

slip

event as

M(ti)

=

Fit; F(te).

We can estimate tue Iateral

displacements

of

spheres corresponding

to tue minimum Mm;n

and maximum

Mmax

value of

Mit).

We find

Mm;n

= 3 x10~~ N and

Mmax

= 3 N

respectively.

Assuming

that tue minimum events are

produced by only

one

sphere

we find that ~lmjn

=

Mm;n /k

= 3 x10~~ mm which is a very reasonable value because it is about

1/7

of the maximum

displacement

we have measured

(see

Sect.

2.2).

In contrast ~lmax =

Mm~x/k

= 3 mm which is not reasonable because

spheres

cannot be moved on such a

long

distance. Therefore one deduces that trie

largest

events are

produced by

an ensemble of at least 100

spheres

which in average are

displaced

of about ~min. Notice that ~lmin

/TT

identifies a characteristic

sliding velocity

( m

mm/s.

Indeed we will see in trie next section that ail

interesting

statistical

features

disappear

when V >

(,

that

is,

when

loading speed

becomes

comparable

to

sliding speed.

3. 1 PROBABILITY DISTRIBUTION. We measure tlie

probability

distribution function

P(M)

of M

finding

a transition to a power Iaw distribution for sma1I velocities V. At Ieast 10~ events bave been recorded to construct

P(M).

This is shown m

figure

4a for tue soft-soft surface.

We see that at

hîgh

velocities

P(M)

presents

just

a

large plateau

with an

exponential

cut-off whereas at small velocities a power Iaw appears with an exponent i which is

increasing

for

decreasing

V and reaches trie value of about 2 for V - 0. This is seen in

figure

4b where trie exponents i of tue power Iaws are

reported

as a function of V for two different types of surface

interactions

(rigid-soft

and

soft-soft).

We found that tue cut-off of

P(M),

1-e-tue

amplitude

of tue

largest

events,

depends

on the norma1force. This is shown in

figure

5a where trie

P(M)

obtained

by using

diiferent normal forces are

reported.

We

clearly

see that trie exponent of power law does not

depend

on trie

normal force but trie

amplitude

of trie

largest

events increases with

increasing Fn.

Furthermore

we bave checked that trie cut-off of

P(M) depends

on the system size. This is done

by using

either different D in trie disk

experiment

or trie small and

big

slides in trie Iinear

experiment.

In

figure

5b trie

P(M)

obtained from trie

large

slide is

compared

with that obtained from the small one. The

Ioading speed

was V

= o-à

mm/s

and the normal pressure p =

Fn/S

was trie same in trie two cases, here S is trie area of trie

sliding

surface. It is important to

stress that tue normal pressure, and not tue normal

force,

remains constant

by changing

tue

system size. Indeed tue relevant parameter is tue normal force

acting

on each

sphere

which is about

Fn /S,

because tue number of

spheres

per unit area is

always

tue saine. This is also tue

case in numerical simulations [2-7] where trie normal force

acting

on each block is tue control parameter. In

figure

5b we see that tue maximum

amplitude

of M increases with tue system size.

3. 2 POWER SPECTRA. TO stress trie fact that trie system is

really

in a state of S.O.C. when trie

Ioading speeds

are slow we show m

figure

6 trie power spectra of

Fit)

for two different V.

We

clearly

see a

large

diiference m tue two spectra. At

high

V trie spectrum is broadened with

a

Sharp

cut off. In contrast at slow V trie spectrum

decays

in

f~~

for at least une order of

magnitude.

This means that trie derivative of

Fit) decays

in

f~~

Indeed

by looking

at tue

JOURNAL DE PHYS>QUE T 4, N' 2 FEBRUAR~ >994 9

(9)

~

~.~

/

~

~.e

~

~

-i 1-Il ~ O

r5 ~

~ i ~ Q

' ~

0-

*

©

~

~ r

co

j

o U b

~ #

'

o z-

>,m~-,<-,> ~~m~~Tr ,~, i

Z

d-B 1-Z >.5 2.~ P-b 3.~

~ ~'m/~éCJ

~ i i

Lo g

ie(

h)

Fig. 4.

a)

Event size distribution

P(M)

for different loading speeds. Starting from the bottonl curve the corresponding V are, Ù-1 n1n1/s, 0.2

nlm/s,

Ù-S n1n1/s, 1-S

nlm/s

and 2.2

nlm/s.

The vertical scale

is arbitrary because the curves have been shifted in order to clearly show their shapes on the saule

graph, b) Exponent of the power laws of

P(M)

as a function of the loading speed for different normal forces and different experimental set-ups.

lu)

and

IA)

correspond to the soft-soft interaction in the disks and linear expennlent respectively. (*) correspond to trie soft-ngid case on the disk expennlent.

i

~

b

~ W5

~

i 1

~ o

j c5

o 1 o

~-3.

o

i

1 ~

8

~, i ~

La g~~ ~i) La

g~~ ~i)

Fig.

5.

a)

Companson of

P(M)

obtained at the same loading speed

IV

= Ù-Sn1n1/s) but for different nornlal forces using the large slide; Fn

= SA N

(continuous fine);

Fn

" 10.8 N

(long

dashed fine;

Fn = 20.4 N

(short

dashed fine). The straight fine is a fit M~~'~. b) Conlpanson of P(M) obtained

for the saule loading speed IV = Ù-S

nlm/s)

but for two different system sizes m the linear geonletry;

(continuous fine)

small shde,

(circles)

large shde. The normal pressure Fn/S was the saule

in the two

nleasurements.

(10)

f-3

h

-9

m O.l la

Fig. 6. Power spectra of

F(t)

for two different loading speeds V,

a)

V = Ù-1

nlm/s

and

b)

V

=

1.4

mal/s.

suape

of trie

signa1in figure

3a we understand tuat tue

Iargest

contribution to trie derivative of

F(t)

is

coming

from trie fast

decreasing

parts

(trie slip phase)

whose

amplitude

M bas a power law distribution as we bave seen in

figure

4. Notice however that a series of discontinuities in trie

signal

it is not

enough

to get an exponent -3 in trie power spectra but it is important

that also trie time of occurrence of events bas a non

trivia1distribution,

otherwise tue triv1al

exponent -2 is obtained. Tuerefore we bave also measured tue distibution

P(r)

of tue time

r between two successive events of any

amplitude.

Trie results are

reported

in

figure

7 for a sma1I and a

high velocity.

We see

tuat,

at V = 0.1

mm/s, FIT)

presents a clear power Iaw

r~~

with

fl

m 1. In contrast at V = 1A

mm/s FIT)

is

simply

broadened with no

specific

ieature.

Thus one concludes that at sma1I V not

only P(M)

bas a power Iaw

dependence

but also tue power spectra and

P(r).

3.3 COMPARISON oF POWER LAW EXPONENTS. Trie

relationship

between M and trie

actualenergy

released

during

tue

slip period

is not clear not

only

in our

experiment

but also in numerical models. M could be

proportional

eituer to tue number n of

spueres

involved in

tue event or to E. Tue

simplest equation

is M

= k ~ln witu ~1tue

slip

distance and k tue elastic constant. However tue same M could

correspond

eituer to a small number of

spueres

wuicu

undergo

a

large displacement

or to many

spueres

and a small

displacement.

Tuus tuis

relationship

is not obvious because it may

depend

on tue

specific

system and on tue most

probable configuration

alter an

eartuquake.

Indeed in reference [2] tue

jump

in friction force is related witu tue

"eartuquake"

moment, wuereas tuis seems not to be tue case in tue

experiment

of reference [15] wuere tue measure of tue total force is Iess sensitive tuan tue local

analysis.

Furthermore in the numerical model of reference [4] it has been found [19] tuat M cc

nf

witu

(

+~

O(i).

In tuis model tue

eartuquake

moment E cc n° with I.à < a < 2

depending

on

control parameters.

In

general

we con write tuat E cc M~ witu 1 < c < 2 tuus B =

(i- ii /c. As1

- 2 for V - 0

we see tuat o-s < B < wuicu is in tue range of wuat is observed in

realeartuquake

statistics.

A more precise evaluation of B could be done

only by measuring

tue local

displacement

of tue steel

spueres during

tue events to

exactly

correlate tue friction force to tue moment.

(11)

~

~

~

z

, cj

z

n~~

n

m ~

_ m

~ mm

- g

~ ""

~m

iÎ ?

qjfl~Zm~

~ J

,~q%

, m

i ,-

~'~

o 1 2 3

lcgl~)

Fig. 7. Probability distribution function of the time 7 between two successive events of any ampli- tude. V =11.1

mal/s (black

squares) and V = IA

mm/s (white

squares).

4. Other systems.

4.1 THE ONE DIMENSIONAL CHAIN. We bave checked under what conditions Dur system

could

reproduce

trie

peak

of

P(M)

at

high

M observed in many numerical simulations [2].

There are indeed two ways to enhance trie

P(M)

for

large

M. One is to increase trie normal force and trie other to force trie system to be as close as

possible

to a one dimensional one. In the first case the

explanatîon

is

quite simple.

Indeed a

large

normal force makes the system

rigid

because trie

spheres

are

pushed against

the disk surface and are net free to move. Therefore the system is very similar to the interactions of

only

two

rigid

blocks

(see

Sect.

4b).

In the

second case, m order to reduce the system one dimensional we put

just

one row of

spheres

embedded with silicon m a very thin groove with a width of 3 mm. In such a way the

spheres

cari move

mainly

m the direction of the

loading speed

V. We studied the stick

slip

of this

single

row of

spheres sliding

on either the hard surface described above or another

single

row.

We found a

peak

for

large

M in trie

P(M)

and trie result is

reported

in

figure

8. The existence

of a power law for small M

depends

on V as in the system described in section 3. This means

that trie

peak

in trie

P(M)

at

large

M is

related,

at least in our

experiment,

to trie dimension of the system and to the fact that the

spheres

have not much freedom to move

transversely

to the direction of the

Ioading speed.

Indeed

by making

a

Iarger

groove the

peak

at

large

M

disappears.

A similar

explanation

is also discussed in in reference [2] for the block mortels and

m reference

iii]

for automata. However there are some other one dimensional chains [4] wuose

(12)

-1

~ i

~

~~

*

Ô~*,

r5

,

fi -2 (~L

~ l/~ § h,

~

' 'l'~ ;,/"~' ù- '~

~/~

~_ ,P ,/" & § ( é à

q~ j

R C0 * à

o Q ' ~

~

o ~ $

[

Î

~ ~

~ ~ ~

j ~ ô

~ * ~

~

~ ~ n

i

, à

©~ ô ~

~5.

i 5

LUg10( ti)

~~'

i i

Fig.

9

Lo g~~ ~

Fig.

8

Fig. 8.

P(M)

for a one-d1nlensional chain of elastically coupled steel spheres. The continuous fine corresponds to nleasurements obtained wuen the one-d1nlensional chain is sliding on the hard surface

described in the text at V

= 0.04

mm/s.

The straight fine has a slope 1.64. lnstead the dashed fine corresponds to measurements obtained when the one dimensional chain is sliding on another one-

dimensional chain at V

= 0.7

mm/s.

Fig. 9.

P(M)

m the case of the rigid-rigid surfaces. (*) and (o) correspond to nleasurements made

with the saule normal force Fn =10 N. The loading speeds were V

= 0.05

mm/s

and V

= 0A

mm/s,

respectively. Instead

IA)

correspond to measurements with V

= 0A

mm/s

but at a normal force

Fn = 20 N.

dynarnics

does not

produce

the

peak

in tue

P(M).

These cuains are Ioaded in a diiferent way,

suggesting

tuat also tue

loading

metuod may

play

a very

important

role.

4.2 THE RIGID-RIGID SURFACES.

Finally

we bave studied tue

dynamics produced by

tue

interaction of two

rigid surfaces,

described in section 2. In

doing

this

experiment

we took

care to bave tue same masses involved and

rouguly

tue same

elasticity

of tue

experimental

apparatus except

tuat,

tue surface

being rigid,

trie

elasticity

was

just

on tue disk axis mstead of

being

distributed ail over tue

interacting

surfaces. Time TT defined in section 2.2 is about

tue same. Tuis kind of

experimental

set-up presents no interest from tue point of view of G-R

because tue system is reduce to tue interaction of two blocks like in tue Turcotte model

il,

3]. Thus

only

a cuaotic behaviour bas been obtained At small

V,

these chaotic states are

perturbed by

trie surface randomness which is trot

enough

to construct a selfsimilar

scaling

law and

only

a broad distribution of

P(M)

is observed. Tuis is suown in

figure

9 wuere tue

P(M)

obtained for diiferent normal pressures and

loading speeds

are

compared.

We see tuat at

uigu

V a

periodic

state witu a well defined

amplitude

is recovered. Wuereas at small V tue

P(M)

are broadened because of tue existence of cuaotic states

perturbed by

a random noise.

We mention uere tuis expenment on tue

rigid-rigid

interaction

just

to state tuat randomness

atone,

eituer

produced by

tue

experimental

noise or

by

tue surface

rouguness,

cannot construct

(13)

power Iaws in

P(M).

Thus one condudes that trie results descnbed in trie previous sections are not

produced by

tue fact that trie steel

spuere

array is not

perfectly regular.

Trie

important

fact is that each

protuberance

of trie

surface, although interacting

with the other Que, is free to move in two directions.

5. Discussion and conclusions.

In this paper we bave

analysed

trie

dynamics

of

stick~slip produced by

trie friction of two elastic surfaces which bave an artificial and well defined

roughness

made of steel

spueres.

When one of these surfaces

(called

soft-surface m trie

text)

is moved on a similar one, trie

spheres produce

Iocal interactions which are

elastically coupled

among them. This kind of interaction is very

interesting

because it can be considered as trie

experimental

realisation of theoretical models used to

study

friction forces

il?,

18]. A direct

comparison

of our results with tuose of reference

[17,

18] is not

actua1Iy possible

because the local interactions in our

experiments

and in the

models are not the saine. However this expenment can be useful to check the

validity

of models

of friction where the local interactions are made more similar to the one described in this paper.

The stick-slip

dynamics

of the soft surfaces described in section 2 has been studied as

a function of the control parameters,

namely loading speed,

normal

force,

system size and geometry. We have found that the

dyuamics

is a

complex

one.

Specifically,

wheu the

loading speed

tends to zero tue

dynamics

presents, for a wide range of trie control parameters, a self

organized

critical state. Indeed tue event size distributions has a power law

dependence

witu

a cut-oi which is a function of the system size. The

frequency

power spectra of the friction force

decay algebraically

with a non-trivial exportent

showing

that the time evolution of trie force is

just

trie

integral

of a

il f

noise.

We bave also Iooked under what conditions trie

P(M)

presents a

peak

for

large

M. We

bave shown in section 4 that this

peak

can be

easily

observed if our system is reduced to a one-dimensional chain in which trie steel

spheres

are forced to move

mainly

m one direction.

This is in agreement with several numerical simulations [2, 9,

loi.

Finally

we have also shown that randomness atone is not

enough

to construct power laws but it is

really

necessary that the system is

composed by

many diiferent parts which may interact

by

means of an elastic medium.

The relevance of this kind of

experiments

in the

study

of

earthquake

statistics is

doubtful,

because friction is just an aspect of

earthquakes il,

13]. In contrast these experiments and the associated mortels can be more useful for

understanding

friction laws. In any case we have noticed in section 3 that the value of B obtained

by

the exponent i of the power Iaw m

P(M)

is o-à < B < as for

eartuquakes.

A more precise

relationsuip

between B and i can be found

by

measunng the Iocal

displacement

of the steel

spheres.

This will also be very

important

to

study

the mecuanisms of friction Iaws.

As a

conclusion,

tue most

important

result of tuis paper is tuat a very

complex dynamics

with many features of S-O-C- can be

produced by

tue

stick-slip dynamics provided

that a two dimensional system is

composed by

many diiferent

interacting

parts and that the

loading penod

is mucu smaller than tue cuaractenstic times of tue system defined in terms of elastic

constants. There were a few

numencalexamples

[4, 5,

iii

wuere tuis bas been shown but no

dear

experimental

evidence tuat tuis could be obtaiued ou a

laboratory

scale.

(14)

Acknowledgements.

We

acknowledge

S.

Fauve,

F. Lund, G.

Paladin,

M. Susa de Vieira and J. P. Vilotte for useful discussions. Tuis work bas been

part1ally supported by

tue

"Region Rhône-Alpes".

References

[ii

For a review see for example:

a)

Scholz G. H., The Mechanics of earthquakes and faulting

(Cambridge

University Press,

1990);

b)

Turcotte D. L., Fractal and Chaos in Geology and Geophysics

(Cambridge

University Press,

1992).

(2] Carlson J-M-, Langer J. S., Phys. Rev. A 40 6470

(1989);

Phys. Rev. A 44 884

(1991).

[3] Huang J., Turcotte D. Geophys. Res. Lent. 17 223

(1990);

Nature 348

(1990)

234.

(4] de Sousa Vieira M. Phys. Rev. A 46

(1992)

6288

(Si Knopofl L., Landoni J. A., Abinate M. S., Phys. Rev A 46

(1992)

7445.

[6] Schmitbulll J., Vilotte J. P., Roux S., Europhysics Lett. 21

(1993)

375.

[7] Burridge R., KnopoflL., Bull. SeismoJ. Soc. Am. 57

(1967)

341.

[8] Bak P., Tang C., Wiesenfeld K., Phys. Rev. Lent. 59

(1987)

381; Phys. Rev. A 38

(1988)

364.

[9] Nakanislù H., Phys. Rev. A 41

(1990)

7086.

[10] Crisanti A., Jensen M.H., Vulpiani A,, Paladin G. Phys. Rev. A 46

(1992)

7363.

(Il]

OLami Z., Feder H-J-S-, Christensen K., Phys. Rev. Let. 68

(1992)

1244.

[12] Feder H.J.S., Feder J,, Phys. Rev. Lent. 66

(1991)

2669.

(13] Sornette A., Sornette D., Europhys. Letters 9

(1989)

197;

Sornette D., Virieux J., Nature 357

(1992)

403; Sornette D., J. Phys. Hance 2

(1992)

2089.

[14] Gutenberg B., Richter C. F., BUJJ. SàsmoJ. Soc. Am. 46

(1956)

las.

ils]

Valette D. P., Gollub J. P., Phys. Rev. A 47

(1993)

820.

[16] Johansen J., Dimon P.,

Ellegaard

C., Larsen J. S., Rugh H. H., submitted to Phys. Rev. E.

[17] Lommtz-Adler J., J. Geophys. Res. 96

(1991)

6121.

[18] Pisarenko D., Mora P., to be published in Pure and Applied Geophysics, Special issue on local friction.

[19] de Sousa Vieira M., pnvate communication.

Références

Documents relatifs

By considering Brochard–de Gennes slip model and molecular chain dynamics at the wall, we propose an expression for the critical shear stress at the onset of stick–slip as a

We report that the value of the effective peeling angle has a strong effect on the triggering and amplitude of the stick-slip instability, even tough the mean fracture veloc- ity

Le défaut de &#34;stick-slip&#34; (ou défaut oscillant) est bien connu pour les polymères linéaires : sur un rhéomètre capillaire à débit imposé, on observe qu’à partir

For these reasons, both projects were carried out by means of high energy resolution X-ray spectroscopy, using two crystal spec- trometers, a DuMond Laue type bent crystal

Spontaneous case From the position signal of the peeling point in the roller reference frame, we compute the instantaneous velocity and plot it as a function of time (black dots

In fact, in the case of rough wetting surfaces (endowed with a no-slip condition), one can even show the following: if one puts the artificial boundary in a way that the flow rates

— Date UT is given for a mid-transit time, NLC is a number of light curves used, epoch is a transit number from the initial ephemeris, Tmid is mid-transit time in BJD based on TDB,