• Aucun résultat trouvé

Stability of two-dimensional packings of disks built with ballistic deposition

N/A
N/A
Protected

Academic year: 2021

Partager "Stability of two-dimensional packings of disks built with ballistic deposition"

Copied!
16
0
0

Texte intégral

(1)

HAL Id: jpa-00246410

https://hal.archives-ouvertes.fr/jpa-00246410

Submitted on 1 Jan 1991

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.

Stability of two-dimensional packings of disks built with ballistic deposition

Rémi Jullien, Paul Meakin

To cite this version:

Rémi Jullien, Paul Meakin. Stability of two-dimensional packings of disks built with ballistic de- position. Journal de Physique I, EDP Sciences, 1991, 1 (9), pp.1263-1277. �10.1051/jp1:1991201�.

�jpa-00246410�

(2)

Classification

Physics

Abstracts

61.42 46.30 68.70

Stability of two-dimensional packings of disks built with ballistic deposition

Rkmi Jullien

(I)

and Paul Meakin

f)

(~) Bit. 510,

Physique

des Solides, Universitd Paris-Sud, Centre

d'orsay,

91405

Orsay,

France

f)

Central Research and

Development Department,

E-I- du Pont de Nemours and

Company,

Wilrnington,

DE 19880-0356, U-S-A-

(Received

6 March 1991,

accepted

in

final

form 23

April1991)

Abstract. Two-dimensional

packings

of identical hard disks of unit diameter are

generated using

a ballistic model with

complete restructuring

and

starting

from a first row made with disks

horizontally regularly spaced by

a =

2 sin #, with #

ranging

from " to

I.

The

stability

with

6 3

respect to random vertical

displacements

for the disks in the first row was studied. A transition from a

pseudo-regular

array without defects to a random structure with defects is observed when

the amount of disorder 8

(amplitude

of the vertical

displacements)

passes a critical value

8~ and a

phase diagram

is measured in the

(#,

3

plane

which can be understood in terms of

simple geometrical

arguments. The random structure found for 3 ~ 3

~

does not

depend

too much

on #i and

strongly

ressembles to the one obtained when

starting

from a horizontal basal line the

density

of defects decreases as a power law with

height

and the

large-height

defect-free structure is characterized

by

a broad distribution of bond

angles.

1. Inboducfion.

Random

packings

of hard disks

(in

two

dimensions)

or

spheres (in

three

dimensions)

have been the

subject

of considerable interest for many years. The

major application

has been the

development

of a better

understanding

of the structure of

simple liquids [I]

and

glasses [2, 3].

The two-dimensional case is

particularly interesting

since the structure of the random

packing

is

highly

unstable and

strongly depends

on the

boundary

conditions

[4].

In some cases, the structure of defects may exhibit some fractal structure

[5].

Here we

study

a very

simple

two- dimensional

packing

which was first introduced

by

Visscher and Bolsterli

[6]

and reinvented

by

us in the context of random

deposition

processes

[7].

In this model hard disks are released

one after

another, along randomly positioned

vertical

trajectories.

Each disk follows the

path

of steepest descent on the

packing

until it reaches a final

position,

stable

against gravity,

before the next disk is released. lvhen the disks were

deposited

onto a horizontal basal

line,

the

density

of the

resulting

disordered

packing

was found to be

equal

to 0.818±

0.001

[7].

In a more recent work

[8],

we have

analysed

the

density

of defects p~, a defect

being

defined as a disk

having

a number of contacts different from four. This

density

was found to decrease with

height

h as a power

law, p~(h)

~

h ~, with an

exponent

a

numerically

close to

(3)

2. Moreover the bond

angle distribution,

in the defect-free random

packing

obtained in the

asymptotic

limit h

- tx~, was found to reach a characteristic stable broad

shape.

In this paper we extend the

preceeding study by considering deposits

on a first row of disk with a controlled amount of disorder. This allow us to

investigate

a transition between a

regular

array of disks and a disordered

phase.

2.

Description

of the model.

Explorating

small scale simulations in which hard disks were

deposited

onto a row of

equally spaced

disks with the same

height

indicated that ordered

deposits

were

generated

for center to

center disk

spacing

a in the range

do

~ a ~

/ do

where

do

is the disk diameter. However

as

the

spacing

a

approaches do

or

/ do (corresponding

to a

triangular lattice)

the hard disk

packing

becomes very sensitive to

perturbations

and in the limits a=

do

and a=

/ do

even the

perturbations resulting

from the round off errors in the double

precision

arithmetics used in our models was sufficient to induce a transition from a structure with

long

range translational order to a disordered structure that in the

asymptotic (h

- tx~

)

has

long

range orientational order but

only

short range translational order. To

explore

this transition from « ordered » to disordered structures in a more

systematic

manner a model in which the substrate disorder could be

continuously

tuned was

developed.

In this model the first row is made of L disks of unit diameter whose centers are

positioned

at horizontal coordinates

x~ =

(I

a with I =

1, 2,

,

L 2

and vertical coordinates :

Zj =

3fi

where

f~

are

independent

random variables

uniformly

distributed between 0 and I. In the

following

we will often make use of an

angle

parameter #i related to the horizontal

spacing

a

through

:

a =

2 sin $~

(1)

$~ is the

angle

from the vertical of the bond directions of the

regular

array obtained for 3

=

0,

whose

density

is :

g~ fir

(2)

~ 4 sin

(2 $r) fi~

Here bonds are defined as lines

joining

the centers of

contacting

disks in the

deposit.

We will

vary $r between the two limits 30° and

60(

which

correspond,

for 3

= 0 to the

regular

« horizontal »

(a

=

I

)

and vertical »

(a

=

/) triangular lattices, respectively.

The

density

is maximum and

equal

to p =

"

m 0.907 in these two limits while it is minimum and

equal

2

/

to

(

m 0.785 for the square lattice obtained for #i

=

45°

(a

=

/).

Then the

packing

is built within a

strip

of width La

(with periodic boundary

conditions at the

edges

of the

strip) by using

a

procedure

in which disks of unit diameter are released one at

a time. To

deposit

a disk we first choose a random number between 0 and La to define a random vertical

trajectory.

Then we release the disk from above the

top

of the

deposit along

(4)

this

trajectory.

When the

moving

disk contacts a

previously deposited

disk it starts to roll down on it until it either reaches an

equatorial position

where the two disks have the same vertical

coordinate,

or it finds a second contact with another disk of the

deposit.

If it reaches

an

equatorial position,

it looses its

previous

contact, falls

vertically

until it

again

contacts a

particle

in the

deposit,

etc... As soon as the

moving

disk has found two contacts, we

analyse

the

stability

of its

position

under

gravity by checking

if the horizontal coordinate of its center lies between those of the

contacting

disks. If the

position

is

stable,

the

moving

disk stops and

another disk is

deposited.

If not, the

moving

disk looses its contact with the upper

contacting

disk to roll on the lower one and the motion is continued as

long

as the

moving

disk does not reach a stable

position.

We

always

wait until the n-th disk has reached a stable

position

before

releasing

the

(n

+ I

)-th

disk.

Moreover,

when the stable

position

is

reached,

the disk stays

permanently

in this

position (we neglect many-particles

effects such as

avalanches).

In

practice,

in our

largest simulations,

we have taken the

quantity

La as an

integer,

La

=

16

384,

and we have varied the

integer

L in order to most

closely approximate

the selected

$~ values. For each $r

value,

we have

deposited

108 disks and we have calculated various

quantities

such as the coordinates of the

disks,

their number of contacts, the

orientations of the

contacting

bonds

compared

to the vertical and the

density

of the

packing

in the

large height regime.

3.

Qualitative

results.

In

figure la, b,

c and

d,

we show small

regions

the first rows for

typical packings

obtained with

$~ =

35(

and

increasing

values of 3

(3

=

0, 0.1,

0.2 and 0.5

respectively).

The bonds between

contacting particles

are shown on these

figures

and

only

the defect

particles,

I-e- those

having

a number of contacts different from

4,

are indicated

by

circles. For 3

=

0,

the

expected

stable

periodic

array is found, For 3

=

0.

I,

this array is

only slightly

distorted and all the disks inside

the bulk conserve four contacts i the structure is «

regular

» and does not exhibit any defect.

For 3

= 0.2 and 3

=

0.5 a

strongly

disordered structure is observed with many defects near the bottom. The

density

of defects decreases

rapidly

with

increasing height.

A similar transition between a

pseudo-regular

array without defects for low-a values and a disordered

packing

with defects for

large-

3 values is observed for all values of $~. In the

large-

3 disordered

phase

the

particular

structure of the defect and the

height dependence

of their

density

is more

clearly

seen in

larger

scale

pictures

such as that shown in

figure 2a,

b and c for 3

= 0.5 and

#

=

35(

45° and

55( respectively.

The defect

pattern

consists of streaks whose

mean orientation with the vertical is $r. The decrease in the defect

density

with

increasing height

is

quite apparent

in these

figures.

4.

Quantitative study

of tile Wansition.

A convenient way to define the location of the transition is to

plot

the total number of defects

as a function of 3 for a

given #

value and to determine the value

3~

at which the first defect appears. In

practice

we measured the number of defects

N~

found within a distance of 8 disk diameters from the substrate. Because the defect

density

decreases

quite rapidly

with

increasing height

this

quantity

is

approximately proportional

to the total number of defects

and can be measured much more

quickly.

The numerical estimates of

3~,

for

#

=

32.5( 35(

37.

5( 45(

55° are shown

by

dots in the «

phase diagram

» of

figure

3. Moreover we have tried to fit the behavior of

N~

near the transition

by

a power law of the

type

:

N~

= C

(3 at

)Y

(3)

The estimate of the exponent y appears to be

roughly equal

to

3,

for

# ranging

from 30° to

(5)

35°, 0.0 35°,0,1

« ~ < >

30 DIAMETERS 30 DIAMETERS

a)

b)

35°1 0.2 35°, 0.5

. 30 DIAMETERS ~ « 3o DIAMETERS >

c)

dj

Fig.

I.

-Typical examples

of

packings

obtained with $r =

35[ Cases a, b, c and d

correspond

to 3

= 0.0, 0.1, 0.2, 0.5,

respectively.

(6)

v 35°

'.

v 45°

0.5 ~l 0.5

'~

« - « ,

l 024 DIAMETERS 1024 DIAMETERS

a) b)

v 55°

= o.5

'~", ../

fl

* I024DtAMETERS ~

C)

Fig. 2.- Large scale view of the

packings

in which

only

defects are shown

by

black dots. These

packings

are built with 8

= 0.5 and parts a, b and c

correspond

to #i

= 35°, 45° and 55(

respectively.

(7)

DA

03

/

o-z

o-i

O-O

30 35 40 45 50 55 60

lt (deg.I

Fig.

3.-Phase

diagram

in the #i-3~ plane. The dots

correspond

to numerical results and the

continuous curve to

analytical

results as

explained

in the text. The error bars on 8~

0ess

than

0.001)

are

always smaller than the dot thickness.

about 43° and

roughly equal

to 2 for

# ranging

from 43° to 601

Examples

of fits are

given

in

figures

4a and b where

log N~

has been

plot

as a function of

log (3 3~)

for # = 35° and

# = 55°

respectively.

All these numerical results can be understood

using simple geometrical

arguments. The idea is that a defect is created as soon as the relative

heights

between

neighboring

disks in the first row are

sufficiently high

to create an

equilateral triangle

of bonds when other disks are

deposited

on the

top

of them. The critical

configurations

are of

quite

different nature when

#

is close to 30° or when it is close to 60[ In the first case, the

equilateral triangle

is formed as

soon as two

neighboring

disks of the second row come into contact. Since the random vertical

displacements

are

uniformly

distributed between 0 and

3,

the smallest value of 3 which can

creates such event

correspond

to the

symmetrical

situation

depicted

in

figure

5a which

involves three

neighboring

disks

(I,

I +

I,

I +

2)

of the first row with z~

= z~~~ = 0 and

z~~i = 3.

Taking

the

origin

as indicated on the

figure,

the coordinates of

point

A are 2 sin

#

and 3 and the first coordinate of

B,

x~ can be then calculated to be :

x~ = sin

#

3

i~ ~

(4)

~~~sin ~

Then the critical condition becomes :

xB #

(5)

In the other case, the first

equilateral triangle

occurs when a

deposited particles

touches

(8)

y ~35°

60= 0.106

#~

£ /

/ O-102

SLOPE

m 3.0

~

~) ~

l/&-&oj

~ ~

V=~~

6o= 0,136

f

~

l35 /

/ Slope=2.0

~

~ h

9 8 7 6 5 4

b) In(b -6c)

Fig.

4.- Plot of

log N~

as a function of log (8

8~).

Parts a and b

correspond

to #i = 35° and

# = 55[

respectively.

simultaneously

a disk of the first row and a disk of the second row, as shown in

figure

5b. Now the critical

configuration

involves

only

two

neighboring

disks

(I,

I +

I). Using

the notations of the

figure,

the critical condition is :

x~ =

2 sin $r

~

(6)

2

The two conditions lead to a second

degree equation

in

al,

which

can be solved to

give

:

3j

=

2(cos~ #

A~

~/(cos~ # A~)~

4 A~

sin~ # ) (7a)

with:

A

= sin

#

in the first case

(7b)

2 and :

A

=

~

sin

#

in the second case

(7c)

2

This

equation

determines the

phase diagram

shown

by

the continuous curve in

figure

5. In the intermediate

region,

the lower of the two

possible

values for

3~

obtained from

equation (7)

(9)

ti

A

6 6

o

a)

c

ti

A 6

o 2 sin y

b)

Fig.

5.- The critical

configuration corresponding

to the appearence of a defect. Parts a and b

correspond

to #i values close to 30° and

60( respectively,

and are characteristic of the two branches of the

phase diagram.

must be selected. Thus there is an

angular point

which is located near 43t The

expansions

of the

analytical

formulae at first order near 30° and 60°

give

3~

= $r

)

=

0.0175($r

30°

) (8a)

3~

=

I #

=

0.0302

(60 #

°

(8b)

The fact that

3~

tends to zero when

#

tends to either 30° or 60° means that the

triangular

array,

which is the most compact

arrangement

of

disks,

is unstable as soon as an infinitesimal amount of disorder is included. This was

already

noticed in our

previous

work [7] where we

observed

that,

when

starting

from a

regular

line of tangent

disks,

the round off errors of the

computer

were sufficient to induce the random structure.

The value of the

exponent

y can be also understood from the above

geometrical arguments.

In the first case, when 3 is

slightly larger

than

3~,

the condition to obtain defects writes :

zi+zi~~-2z~~i >23~. (9)

(10)

Since z~, z~~ i and z~~~ are distributed

independently

with a flat

probability

distribution between 0 and

3,

the

probability

to have a

configuration satisfying

the above condition is

proportional

to

(3 3~)~.

Since the number of defects is

proportional

to this

probability,

y =

3. In the second case, the condition for

creating

a defect involves

only

two

adjacent disks,

so that defects are formed if :

z~ -z~~i

>3~

leading

to a

probability proportional

to

(3 3~)~

and y

=

2.

S.

Density

of defects.

Some results for the variation with

height

of the

density

of

defects, p~(h),

in the disordered

phase,

are

given

in

figures 6a,

b and c, for different values of

#: #

=

35(

45° and 55° and different values of

3,

3

=

0.5,

1-o and

I-o, respectively.

The results are consistent with a

power law behavior of the

type

:

p~(h)

~h~"

with a

=1.68,

1.65 and 1.80 for

#

=

35(

45° and

55( respectively.

The same kind of behavior is found for all

wand

3 values within the disordered

phase

and the estimated a

exponent

is

always lying

in the range 1.6

~ a ~ 2.0. The

exponent

a

might

very well

depend

on the model parameters

but,

due to the numerical errors, we have not

strong

numerical evidence to support this conclusion. It is worth

noticing

that the value

a = 2

corresponds

to the one found when

starting

from a basal line

[8].

6. Distribution of bond orientations.

We have calculated the distribution

Ni (0

of the

angles

0

giving

the orientation of the bonds from the vertical

direction, Ni(0)

do is the number of bonds whose orientation

angle

lies

between 0 and 0 + do. In

figures 7a, b,

c and d we report these distributions for

#

=

35° and 3

=

0.1, 0.25,

0.5 and

I, respectively.

In

figures 8a, b,

c and d we report the

corresponding

curves obtained for

#

=

55t In each

figure,

four different

distributions,

obtained with a number n of

deposited particules lying

in the

respective

ranges n = 0.5 x 10? 1.0 x

lo?,

n = 1.5 x

lo?

2.0

x

lo?,

n = 3.5 x

lo?

4.0

x

lo?,

n =

7.5 x

10?

8.0

x

10?

have been

superimposed.

While for the 3 value below the

threshold,

we obtain a

relatively

narrow

peak

centered at 0

=

#,

with a

symmetric shape

and a finite

width,

for 3 values above the

threshold,

we

get

a broad

distribution,

which is more and more extended from 30° to 60°

when 3 increases. The small tails below 30° and above

60(

are

only

observed for

relatively

small

heights

and are due to the presence of defects at the bottom of the

packing. They disappear

for

larger heights

where the distribution reaches a stable

asymptotic

limit

characteristic of the disordered structure. It appears

that,

in the

large

3

(strong disorder) limit,

the

shape

of the distribution is almost

independent

on

wand

ressembles to the one found in the case where we start from a horizontal basal line

[8].

The

peak

at 30° is however a little bit less

pronounced

here.

Also,

for a

given sample,

the fine structure seen in the curve does not vary with

height

and is thus

typical

of the nature of the disorder

put

in the first row

(it changes

when one

changes

the seed of the random

generator).

This memory effect comes

from the fact

that,

when there is no

defects,

successive bonds stay

strictly parallel

from rows to

rows. This fine structure is

averaged by increasing

L so that we expect to get a continuous

curve in the infinite-L limit. Thus it appears that the disordered structure in the

large-h

limit is

of

quite special

nature since it exhibits strong

long

range orientational order and no

long

range

(11)

,-Slope=-1.68

C

y=35$

6=O.5

2 3 4 5 6 7 8 9

In(h) j)

,~slope=-1.65

C

y=45$6=1.o

2 3 4 5 6 7 8 9

In(h)

b)

2

~

slope « -1.80

C

y =

55$ 6 =1.o

2 3 4 5 6 8

In(h) c)

Fig.

6.-

Density

of defects

pd(h)

as a function of the

height

h

(log-log plot).

Parts a, b and c

correspond

to

(#i

=

35[ 8

=

0.5), (45(

1.0) and

(55(

1.0)

respectively.

(12)

1800000

y 35$ b 0.1

1400000

6~ 1000000

£ 800000

200000

~

30 35 40 45 50 55 60

al 6

y=35$ 6= 0.25

30 35 40 45 50 55 60

b)

o

y 35$ 6 0.5

30 35 40 45 50 55 60

C) 0

y=35i 6=1.0

30 35 40 45 50 55 60

d)

0

Fig.

7. Distributions

Ni (8 (not normalized)

for the bond orientation

angles

8 from the vertical for

#

=

35[ Parts a, b, c and d

correspond

to 8

= 0.1, 0.25, 0.5 and 1.0,

respectively.

JOURNAL DE PHYSIQUE I T I,M 9, SEPTEMBRE 1991 5o

(13)

v"55t 6=0.1

30 35 40 45 50 55 60

a) o

V "55$ b 0.25

~f

~400000

200000

30 35 40 45 50 55 60

b)

8

v =55°, 6 =o.50

30 35 40 45 50 55 60

~) o

v*55i &=1.0

30 35 40 45 50 55 60

d) o

Fig.

8. Distributions

Ni (9 ) (not normalized)

for the bond orientation

angles

9 from the vertical for

#

=

55°. Parts a, b, c and d

correspond

to 8

= 0.1, 0.25, 0.5 and 1.0,

respectively.

(14)

positional

order.

Moreover,

since the

density

of defects decreases

algebraically,

the

long

range orientational order takes

place quite quckly

when h increases and thus we have

quite

good

confidence that we have here

enough

informations on the

asymptotic limit,

in

contradiction with the conclusions of

Delyon

and

Levy [9]

on the same kind of model.

7.

Packing density.

We have also determined the distribution

N~(# )

for the

angle #

between the two bonds that a

given

disk makes with the two

contacting

disks

placed

below. This

angle corresponds

to the top

angle

of the rhombus made with four

neighboring

disks when there is no defect. The

knowledge

of this distribution allows a

precise

calculation of the

density

of the

packing through

the formula :

~

N2(# d#

~ ~

N~(#)

sin

# d#

~~ 31.25$32,5$35°,37.5°,40$42,5$45°

o.86

c~

0.2 0.4 0.6 0,8 1-o 1.2 1.4

a)

6

0.88 4~a 47_50, sol 52.5$57.5°,58.75°

o.86

~

o.84

0.82

o.8o

~ i 2 1.4

0.7§

~ ~ ~ o.4 °.6 ~'~

~ ~

Fig.

9. Bulk

density

p calculated in the

large-h asymptotic regime

plotted as a function of 8. In a) are shown the curves

corresponding

to #

= 31.25(

32.5( 35(

37.5( 40( 42.5° and 45[ each curve can be

recognized by

the fact that p

(8

=

0)

decreases

monotically

when # increases from 30° to 45°. In

b)

are

shown the curves

corresponding

to #

=

45( 47.5( 50( 52.5( 55(

57.5° and

58.75(

each curve can be

recognized by

the fact that p

(8

=

0)

increases

monotically

when # increases from 45° to Ml

(15)

Alternatively

the

packing density

p can be obtained

directly

from the

particle

coordinates and this

approach

was used to obtain the data

presented

here. In the

large

disorder

(large-3)

limit

we obtain p

= 0.806 ±

0.001, independent

of $r, a value

definitively

smaller than the value p =

0.818 ± 0.001 found for the disordered

packing

obtained when

starting

from a basal line.

This is a

small,

but

significative,

difference which

might

reflect the fact

that, although apparently

universal

(I.e. independent

on $r and 3 for

large 3)

the

present

disordered model

might

not be in the same

universality

class than the model which starts from a basal line. The transition can also be seen on the p

(3

curves which are

reported

in

figure

9a and b. These

curves exhibit a

sigmoidal shape

near the transition but the

singularity

at

30

is

certainly

weaker than in the case of

N~.

It is worth

noticing

that while for $r

lying

in the range 30~37.5°

or in the range 55~60° the order-to-disorder transition

corresponds

to a

sharp

decrease of the

density,

for 37. 5° ~

#

~

55(

it

corresponds

to an increase of the

density,

this means

that,

near

45(

the disordered

phase

is denser than the ordered

phase

~p45

=

~

=

0.78539...).

4 8. A model to describe the structure of defects.

In order to understand the characteristic

stripped

structure and

algebraic

decrease of the

density

of defects in the disordered

phase

we have built a

simplified

model which contains most of the features of the defect formation. The model starts from a first row made of L sites located at the

positions

x~

(I

=

1, 2,

...,

L).

Two semi-infinite

straight lines,

oriented to the left and to the

right,

with the same orientation compare to the vertical are

propagated

from each site. The intersections of these lines define a

regular

array. Periodic

boundary

conditions

are considered. Bond oRentations of

fit (I

and

0~(I )

are associated with the fines

emanating

to the left and

right

directions

respectively

from the I-th site. For the first row, we set

0t(I)

and

0~(I)

to be

randomly

and

uniformy

distributed between 0° and 901 Then we

investigate

all the line intersections

sequentially,

row

by

row,

starting

from the second row.

At a

given intersection,

which

corresponds

to a

crossing

between a left line I and a

right

line

j,

we calculate the

quantity #

=

fit (I )

+

0~Q

which represents the

angle

between bonds at this

site in the model. If

#

lies between 60° and

120(

there is no

«interaction»,

I-e-

0t(I )

and

0~(I ) keep

their

original values,

the intersection is not a defect

(this corresponds

in the

original

disk

packing

model to the fact

that,

when there is no

defect,

the top

angle

of the rhombus made with four

neighboring particles

lies between 60° and

120].

If

#

is smaller than 60° or

larger

than

120(

the site is considered as a defect and new values of

01(I)

and

0~@

are

assigned

to the

intersecting

lines. In the

original

model these new values

depend

not

only

on the

previous

values of

01(I)

and

0~Q)

but also on the bond orientations at the

neighboring

columns which

certainly depends

on the

original

3 and

#

value. We assume that the main feature is that the new distribution of the bond

angle Starting

from a defect is no

longer uniformly

distributed between 0° and 90° but is concentrated into a narrower range so

that in the

asymptotic (large height)

limit all

angles

will lie between 30° and 60t In the

simplified

model the bond

angles of (I)

and

0~Q) following

an interactive intersection at a

defect are selected at random from a uniform distribution between q and 90° q

(where

q is

a parameter of the

model).

In the actual disk

packing

process lines are created or

destroyed

at a defect and this is not taken into account in the

simplified

model.

With this

simplified

model we have found

that,

as soon as q is

larger

than a critical value

q~ of order

7(

the

density

of defect decreases

algebraically

with the

height

with a

characteristic exponent a that varies

continuously

with q.

a tends to zero when q tends to q~. The

algebraic

law is tested with

considerably

more

precision

than in the

original

model.

An

example

of the structure of defects obtained with the

simplified

model is shown in

figure

10 for L

=

000 and q

= 12° for which the a

exponent

is close to 2.

Although

there are less correlations between

defects,

the

stripped

structure of

figure

lo look like those shown in

(16)

y

~.

t ,

_/.~,

i~'

:

?

? ;.'

1,

.

, $'' ( _ '

~ ' .

~ l

I

-

'/

_ / ,' ' P

,

( $ , '

'

' ~ '

.

' '

o '_

' ~

L' . f

' ' -

Fig.

10. Structure of defects of the

simplified

model for L

= 1000 and

~D = 12[

figure

2. The results of this model indicate that the exponent a

might

vary

continuously

with

parameters 3 and

#

in the whole disordered

phase.

However the transition found at

q~ is of

completely

different nature than the one found at

3~

in the

original

model since the

geometrical ingredients

which

explain

this transition are not taken into account in the

simplified

model.

9. Conclusion.

We have obtained evidence for an order-disordered

phase

transition in a two-dimensional

packing

of disks when

including

disorder at a

boundary.

We have been able to recover the

phase diagram by simple geometrical arguments.

Since we

always

obtain a defect-free

structure in the

large-height regime,

we are convinced that the disorder

coming

from the

random choice for the successive vertical

trajectories

is irrelevant in the mechanism of this transition and thus our results

might

be very

general.

We intend to

apply

the same

procedure

to a variant of the Bennet model

[3]

in the

strip

geometry. In this model disks are added

sequentially always

at the lowest

position insuring

two contacts with

previously deposited particles

: the

procedure

is now

completely

deterministic so that we can be confident that

disorder comes

only

from the first row.

References

[Ii

BERNAL J. D., Proc. Roy. Sac. London A280 (1964) 299 ; FINNEY J. L., Proc. Roy. Sac. London A319

(1970)

479.

[2] CARGILL S., J.

Appt. Phys.

41

(1970)

2248.

[3] BENNETT C. H., J.

Appt. Phys.

43

(1972)

2727.

[4] BIDEAU D., GERVOIS A., OGER L. and TROADEC J. P., J. Phys. France 47 (1986) 1697.

[5] ONODA G. Y. and TONER J.,

Phys.

Rev. Len. 57

(1986)

1340.

[6] VISSCHER W. H. and BOLSTERLI H., Nature 239

(1972)

504.

[7] MEAKIN P. and JULLIEN R., J.

Phys.

France 48

(1987)

1651.

[8] MEAKIN P. and JULLIEN R.,

Europhys.

Lett. 14

(1991)

667.

[9] DELYON F. and LEVY Y. E., J.

Phys.

A 23

(1990)

4471.

Références

Documents relatifs

Consider an infinite sequence of equal mass m indexed by n in Z (each mass representing an atom)... Conclude that E(t) ≤ Ce −γt E(0) for any solution y(x, t) of the damped

The Principe in this function is, first it will transform the equation in to the Laplace-domain and second it solves the equation as a linear equations, third it transforms

In this paper, we propose a nonlocal model for linear steady Stokes equation with no-slip boundary condition.. The main idea is to use volume constraint to enforce the no-slip

Abstract—This article presents a detailed description of an algorithm for the automatic detection of the mid-sagittal plane in three-dimensional (3–D) brain images.. The algorithm

23 In a comprehensive meta~analysis of various correlates of absenteeism, Farrell &amp; Stamm (1988) categorized 72 studies with respect to the type of dependent measure (time- lost

the major prey items (1\tlallolll.\' villo.ws and shrimp) encountered were fairly similar in terms of both percent occurrence and percent biomass between all locations

(ed.), Flysch sedimentology in Nonh America, Geological Association of Canada, Special Paper 7, p. Collision along an irregular margin: a regional plate tectonic

To evaluate the strength of bow plating, three limit states (three-hinge collapse, permanent set and membrane collapse) were selected. Statistical distributions for