• Aucun résultat trouvé

A Diffraction Method of Study of Thermal Quasiorder in a Finite Two-Dimensional Harmonic Lattice

N/A
N/A
Protected

Academic year: 2021

Partager "A Diffraction Method of Study of Thermal Quasiorder in a Finite Two-Dimensional Harmonic Lattice"

Copied!
11
0
0

Texte intégral

(1)

HAL Id: jpa-00247130

https://hal.archives-ouvertes.fr/jpa-00247130

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.

A Diffraction Method of Study of Thermal Quasiorder in a Finite Two-Dimensional Harmonic Lattice

P. Aranda, B. Croset

To cite this version:

P. Aranda, B. Croset. A Diffraction Method of Study of Thermal Quasiorder in a Finite Two- Dimensional Harmonic Lattice. Journal de Physique I, EDP Sciences, 1995, 5 (9), pp.1213-1222.

�10.1051/jp1:1995192�. �jpa-00247130�

(2)

J. Phys. I France 5

(1995)

1213-1222 SEPTEMBERI995, PAGE1213

Classification Physics Abstracts

61.10Dp 68.35Ja

A Diffraction Method of Study of Thermal Quasiorder in

a

Finite Two-Dimensional Harmonic Lattice

P. Aranda and B.

Croset(*)

Groupe de Physique des Solides, URA CNRS 17, Tour 23-13, 2 place Jussieu, 75251 Paris cedex 05, France

(Received

20 February 1995, revised 8 April 1995, accepted 17 May

1995)

Abstract. Due to the non-existence of long-range order, the diffraction peaks of 2D-solids

are considered to have a power-law shape

q)~~.

Taking into account the limite

size effects and calculating the powder average, we show that this power-law behaviour appears Orly for high qp

and then for very small intensities. It is therefore quite dificult and hazardous to characterise the quasiorder by usmg this asymptotic behaviour. Although the shape of the central part

of the peak cannot be used to characterise the quasiorder, we show that, for

a fairly good resolution, it is possible to deterrnine ~ using this central part. This deterJnination cari be dore irrespectively with the other details of the system by comparing trie peak width to its value at low temperature, i-e-, at low value of ~. By usmg two diffraction peaks, we propose the simple

relation:

~(QBI )/Qài

"

~(QB~)/Qà~

as a check of the two-diJnensional quasiorder.

Among

the extensive theoretical and

experimental

studies of order in two-dimensional

(2D)

sohds

[1-6]

and of two-dimensional

melting [7-14] significant

contributions were

brought by

diffraction

techniques. Nevertheless,

in contrast with three-dimensional

(3D) studies,

most of the 2D diffraction works are based on the

investigation

of trie first diffraction

peak only.

Whereas this restriction has important technical reasons

(low intensity,

presence of substrate

peaks

at

high Q),

it seems quite

surprising especially

in comparison with 3D usual results:

1) On the one

hand,

the thermal activation of

phonons

leads to a

non-perfect long-range quasiorder.

It results for an infinite

crystal

in the

replacement

of the

à(Q QB) singularity by

a

power-law singularity

(qp(~~~

= (Q

QBÎ~~~,

and

thereby

in a

broadening

of trie 2D diffraction

peaks [1,2]

for finite

crystals.

The linear

dependence

of ~ on the square of the

peak position QB predicts

that the structure factor of a 2D solid should

look,

at least

qualitatively,

like a 3D

liquid

one. To our

knowledge,

no attempt to check this

prediction

has been made.

ii)

On the other

hand,

the

liquid (10) peak

width has often been

interpreted

as the inverse of

a coherence

length,

L

[10,13];

such an assumption

implies equal

width for ail the

peaks,

which is in strong contradiction both with the usual

shape

of the structure factor of 3D

liquids

and

with the

fitting

of the

intensity

of the

liquid (10) peak by

a Ll~~ law.

(")

corresponding author

© Les Editions de Physique 1995

(3)

iii) Lastly,

Gunther et ai. [3],

Weling

et ai. [4] and Dutta et ai. [5] have calculated the influence of finite size eifects and their results show that the

power-law shape

cannot be

easily verified, being

valid

only

for

large qpL.

For these reasons, it seems

quite

useful to

study

at least two diisraction

peaks during

2D

melting

and to check trie linear

dependence

of ~ on

Qà.

We will examine how one can determine

~

independently

of trie precise nature of finite size eisects.

By calculating

the

iuteusity

diisracted

by

a

powder

in various cases, we shall show

firstly

that trie

asymptotic

behaviour of trie

peak profiles

cau be used neither as a check of trie thermal

quasiorder

nor as a measure of ~,

secoudly

that the thermal

quasiorder

leads to a

broadeniug

of the central part of the

peaks

which can be used in trie determination of ~

independently

of trie

precise intensity profiles.

For a 2D harmonic

crystal

with

periodic boundary conditions,

trie diifracted

intensity

can be written as

1(Qp)

=

~jexp(-iop (Iti Rj))

exp <

(Qp.u(Rz Rj))~ >l

~~

2

where

Qp

is trie

projection

of trie diffusion vector

Q

in trie

crystal plane,

~1(r) is trie relative thermal

displacement

of atoms distant from r.

To be able to

perform

an

analytical calculation,

we must assume that

I(Qp)

can be wntten as a sum of

peak profiles

centred around each vector

QB

of the

reciprocal

lattice:

Ijop)

=

LIQ~ jqp)

=

L L expj-iqp.jJt~ Rj))

exp

(-j

<

(QB.ujRz Rj))~ >)

QB QB 1,J

with qp =

Qp QB.

This

assumption

is

questionable

for broad diffraction fines but it allows the continuum ap-

proximation

for the two direct lattice sums so that

IQ~

is written

(for

the sake of

simplicity,

we will omit the index

QB

for

IQ~

):

I(qp)

=

~°~) / / exp(iqp r) f(r)c(r)dr,

L

where

f(r)

is the

generalised Debye-Waller

function,

c(r)

describes the cut-off in the direct space and accounts for the number of the

pair

of

crystalline

sites

separated by

r. Within a

Debye

model and

neglecting

the diiference between transverse and

longitudinal

sourd [6],

f(r)

is written:

kn

f(r)

= exp -~

L d~~~~

[1

Jo(~r)]

with ~

=

[kBTQ[(3~1+

À)]

/[4àr~1(2~1+

À)], ~1 and are the Lame coefficients.

In order to calculate

I(qp),

both

Weling

et ai. and Dutta et ai. substitute for

f(r)

its

asymptotic

behaviour for

large

r. For an infinite

crystal

this

"asymptotic" approximation gives f jr)

=

jr laD)~~

with aD the lattice

spacing. Together

with

c(r)

= 1, this

approximation

allows us to write

I(qp)

=

qp~+~,

the well-known

power-law shape.

For an infinite

crystal

the

asymptotic approximation

is

quite

valid for distances greater than second

neighbour.

For a limite

crystal,

limite size eifects have consequences on the one hard on the behaviour of

f(r)

and on the other hard in

c(r)

which is no

longer

constant.

Doing

the

"asymptotic"

approximation, Weling

et

ut.,

then Dutta et ai. have shown that the variation of

c(r)

dominates the behaviour of

I(qp)

and that

f(r)

may be

replaced by

its value for an infinite

crystal.

To

produce

more realistic fine

shapes

than

Weling

et

ut.,

Dutta et ut. made the choice

c(r)

=

exp(-r~ /L~),

a

simple

but

surprising

choice as we will see. This choice allows them to

(4)

N°9 A DIFFRACTION METHOD OF STUDY OF 2D QUASIORDER 1215

show that the

approximation

of

f(r) by

its value for an infinite

crystal

is reasonable. In other

words,

except near the

edges,

the atoms

really

present in the

crystal

of dimension L vibrate like the atoms of a cluster of dimension L immersed in an infinite

crystal.

This conclusion may

seem

surprising

when one thinks of the existence of surface modes in a 3D

crystal.

A correct calculation without

periodic

conditions of the vibration modes will exhibit

edge

modes. These

edge

modes are Dot due to the limite size but

only

to the existence of an

edge

and will appear for a semi-infinite

crystal.

At the very most,

they

have the

only

consequence to reduce the

apparent size of the

crystal.

Approximating f(r) by jr laD)~~, I(qp)

is written

Iiqp)

=

iLlaD)~-~rii n/2)i~/fi)-~ ifiii

~J/2;1;

-q(L~/4)

where ifi is the confluent

hypergeometric

function.

Several remarks must be

pointed

out:

For q = 0,

1(qp)

has a Gaussian

shape

because of the choice of

c(r).

Within the

logarithmic approximation,

the whole

dependence

on aD appears in the in-

tensity prefactor.

Within the

logarithmic approximation,

the

I(qp) profile

is a universal function of

qpL.

The relative

broadening

of this

profile compared

to its value at q = 0

(T

= 0

K)

is

independent

of

L,

as

long

as the

logarithmic approximation

is valid. This

paradoxical

result is

mainly

due to the

logarithmic divergence

ofu~

jr)

in a small

crystal,

the farthest

pairs

are better correlated than in a

big

one but

they

are in

larger

proportion

relatively

to ail other pairs.

The

asymptotic

form of the

degenerate hypergeometric

function allows us to find the

qp~+~

law established

by Imry

et ut. Nevertheless this behaviour is observed

only

at

large qpL

and with intensities near 1% of the maximum.

In his seminal paper

[15],

Warren chose a Gaussian approximation for

c(r),

a choice also made

by

Dutta et

ut.;

this is Dot the

only possibility leading

to

simple analytical

calculations.

However,

in

spite

of its

simplicity,

a Gaussian

shape

is Dot a

good approximation

either for

sin~Lz/sin~~,

or for its average on diiferent

L,

since it does Dot accourt for the

qp~ asymptotic

behaviour of these "raturai"

profiles.

This

profile

appears to be

satisfactory

in diffraction

techniques only

when the resolution is worse than the limite size eifect

broadening. Generally,

in the opposite case where raturai

profile

is

observed,

a Lorentzian

shape

is

preferred

as an approximation.

To discuss the consequences of the choice for

c(r),

we will calculate the

intensity

observed in

studying

diffraction

by

a

powder

of 2D

crystallites.

To obtain

J(Q),

the

intensity

diifracted

by

a

powder,

two

integrations

on

Q

orientations must be

performed:

«/2 «/2

JIQ)

"

1/Q / dll / dTIlQp)Qp

"

1/Q / dllBlop)

where ~1 is the

angle

of

Q

with the normal to the

crystal plane

and T the

directing angle

of

Qp.

The

integral

on ~1

gives

the

typical

saw-tooth

shape

of 2D

peaks

in

powder technique

[15] and must

usually

be

numerically performed.

The

integral

on T cari be

performed analytically by

doing

the

tangential approximation

which is usual in

crystallography

[16]: the

integral along

(5)

the circle of radius

Qp

is

replaced by

the

integral along

its

tangent.

This

approximation

is valid if

ôq

is small with respect to

Qp,

1-e-, if q is trot too close to 2. Then:

B(qp)

=

(uD)~ Î

ce

exp(iqpr)r~~c(r)dr

While

I(qp)

is a two-dimensional Fourier

transform, B(q~)

is a one-dimensional one. In

fact,

the

experiment

does not

give

direct access to

J(Q)

but to

K(Q),

the convolution of

J(Q)

with the

experimental

resolution

R(Q).

K(Q)

=

J(Q)

~p

R(Q)

=

i/Q

£~~ (qp)d~lj

~p

R(Q)

o

By reversing

the order of

integration

on q and

~1 and

by noting T(r)

the inverse Fourier transform of

R(Q), K(Q)

is written:

K(Q)

=

1/Q ~~~(B(qp)19

R(qp))dtl= 1/Q ~~~

M(qp)dtl

with

ce ce

M(qp)

=

(aD)~ cos(iqpr)r~~c(r)T(r)dr

=

(aD)~ cos(iq~r)r~~W(r)dr

Î Î

The

profile

of the diifracted intensity

depends

Dot

only

on q but also on

W(r)

which certains at once the limite size eifects and the limite resolution ores. In the

following

part, we will try to propose a method

allowing

the measure of q, as

independent

of

W(r)

details as

possible.

We cari

already

notice that the

analytic shape

of

M(qp)

is very similar to that of

I(qp).

Ail the remarks on the

universality

in

qpL

of the

I(q~) profile apply

to

M(qp),

within the

validity

limits of the

approximations.

Because a

least-squares

fit of

K(Q)

at low temperatures is the only way to obtain

W(r),

its

shape

for small r

depends mainly

on the far tait of

M(qp,

q =

0)

and therefore is in some way

arbitrary

from an

experimental point-of-view.

Nevertheless for

large

values of q the whole

peak profile depends

on this

shape

at small r, because of the

rapid

decrease of r~~. To circumvent this

difficulty,

we have

analytically

calculated the curves

M(qp,q)

for three choices of

W(r),

we have

numerically

determined the relative

broadening

à'

=

à(q) /à(0)

for each choice and

we have looked for a universal transformation

g(à')

almost

obeying

the relation

g(à')

m q for any

W(r).

a) W(r)

=

exp(-r IL),

the "Lorentzian case".

Such a choice

leads,

for q

= 0, to a Lorentzian

profile:

M(qp,

q =

0)

=

L/(q(L~

+

1).

As we have seen

before,

this

profile

is the raturai one, 1-e-, the

profile

obtained in the absence of thermal vibrations

(q

=

0)

for a

disperse powder

of 2D

crystallites

of

general shape. However,

the choice of an

exponential

for

W(r)

is also

suggested by

Nelson et ut. for the hexatic

phase I?i.

For q non zero, we have:

~~QP'~Î~ (qÎÎÎ

+

ÎÎÎ2-1/2

~°~

((~ 4)~~~~~~(QP~))

P

~

b) W(r)

=

exp(-r~ /L~),

the "Gaussian" case.

(6)

N°9 A DIFFRACTION METHOD OF STUDY OF 2D QUASIORDER 1217

Such a choice

leads,

for q

= 0, to a Gaussian

profile:

M(qp,

q =

0)

=

exp(-q(L~ /4).

As we have seen

before,

this

profile

is the instrumental one, 1-e-, the

profile

obtained in the absence of thermal vibrations when resolution is broader than the raturai

shape. Although

every

experimentalist

tries to avoid such a

situation,

we have studied this case after Dutta et ut.

For q non zero, we have:

M(qp, q)

=

r(1/2 q/2)L~~~

iFi

Il /2 q/2; 1/2; -q(L~ /4).

c) W(r)

=

il

+ r

IL) exp(-r/L),

the

"square-Lorentzian"

case.

Such a choice

leads,

for q

= 0, to a

square-Lorentzian profile:

M(qp,

q

=

0)

=

2L/(q(L~

+

1)~.

This

shape

is very similar to a Lorentzian except in the tait and cari be

regarded

as an intermedi- ary

shape

between Lorentzian and Gaussian. For

good

resolutions,

W(r)

is an autocorrelation

function: its discontinuities for r

= 0

depends

on the nature,

abrupt

or trot, of the coherence

disappearance.

This is

why,

to accourt for less

abrupt

situations than

crystallites

of finite

size with

Sharp edges,

we have chosen a function without discontinuities of the derivative for

r = 0 but

preserving

the

exponential

behaviour for

high

r. This is the case of the function

W(r)

=

il

+

r/L) exp(-r/L).

For q non-zero, we have:

~~~P'~~ q)ÎÎ

+

ÎÎÎ~ Î/2

~°~

((~ 4)~~~~~~(QP~))

+ ~~

iq(ÎIÎ i)/~ÎÎ

~ COS

i12 ~l)arctallioPL) i

For ail these cases, we cari see that:

The critical value of q, for which the

broadening diverges,

is q = 1 instead of 2 for

I(q~).

For these values of q, the

tangential approximation

may seem

questionable

but the

validity

of the

approximation depends

on the width of

I(qp)

and trot on the width of

B(qp).

Nevertheless for small

L,

the

tangential approximation

should be invalid.

For every value of q trot

equal

to zero, the

asymptotic

behaviour of

M(qp, q)

is

qj~~.

For q = 0, there is a

discontinuity

in the

asymptotic

behaviour

(as

an

example

the

asymptotic

behaviour of

M(qp,

q =

0)

is

qp~

in the Lorentzian

case).

This

discontinuity already

existed in the

profile

calculated

by

Dutta et ut.: their

profile

had a Gaussian

behaviour for q

= 0 and a power law

shape

for q non-zero. In practice, this result means the

following:

the

validity

domain of the

asymptotic

behaviour in

qpL

is

pushed

away to

infinity

for the low values of q.

In the central part of the

peak,

the thermal

quasiorder

leads to a

broadening

of the

peak.

While

hardly

visible on a

log-log plot,

this

broadening

is

easily

noticeable on a linear

plot

and is

strongly dependent

on q as cari be seen in

Figure

2. This

broadening

must be taken into accourt in order to measure the coherence

length

of a hexatic or

liquid phase

from the

(10) peak

width and invalids the direct interpretation of this

peak

width

as the inverse of coherence

length.

The increase of the

broadening

with q and then with

(7)

0.

z~

'( O.Oi

'

j

O.OOI

~'~~~Î o i oo i ooo

q~L

Fig. l. Comparison between M computed in the Lorentzian case and the asymptotic behaviour for ~ = ù-1 in

a log-log plot.

M(qp,~

= ù-1); ): asymptotic behaviour

(qpL)~°'~;

): best fit of

M(qp,~

= ù-1) between qpL = 10 and qpL = 10ù by a

(qpL)~"

law with

ce = 1.19

(such

a fit provides a wrong and unphysical value of

~).

QB

allows us to expect

liquid peaks

of

increasing

width with

QB

in accordance with 3D results.

No dear characteristic of the

peak shape

in its central part seems to be attributed to thermal

quasiorder.

A tentative way to characterise the 2D thermal

quasiorder

and to

measure q should be to use the

qj~~ asymptotic

behaviour of

M(qp, q)

since this behaviour

does not

depend

on

c(r).

But our

profile

calculations show that it seems

problematic

to characterise the thermal

quasiorder by

the

asymptotic

behaviour of the

profiles.

As

can be seen in

Figure

1, the

validity

domain of the

asymptotic

behaviour starts on and after

qpL

= 100.

Moreover,

in a

log-log diagram,

a

quasi-linear intermediary

behaviour

leading

to a wrong evaluation of q appears.

Lastly,

as can be seen in

Figure

2, the sum of

a Lorentzian and a fine of base fits well the

profiles

obtained

induding

thermal

disorder,

for low q.

(The

fine of base does not exceed 7% of the maximum intensity for q

=

0.2).

This

diiliculty

to

identify

without any

ambiguity

the nature of the disorder from the

shape

of a unique diffraction

peak

was also mentioned

by Heiney

et ut.

[loi

in their

study

of the

melting

of Xe

monolayer

adsorbed on ZYX. These authors obtained as

good

results

with Lorentzian

profiles

as with Dutta's

profiles.

The very low value of the intensity in the

asymptotic

domain

explains

the strong

dispersion

of their q measure.

This

diiliculty

suggests

measuring

q

by

use of the central part of the diffraction

peaks

and

characterising

thermal

quasiorder by

use of two diffraction

peaks

and

by

the check of the linear

dependence

of q on

Q[.

This method needs the observation of two

peaks

for which the relation:

1(QBI)/1(QB2)

"

QÎI/QÎ2

should be verified. It also needs to find a relation

g(à')

between the relative

broadening

and q.

To find g, the first step is to look for the

asymptotic

behaviour of the à'

divergence

for q

= 1.

It is easy to see, both in the "Lorentzian" case and in the

"square-Lorentzian"

case, that:

lim (1

q) In(à'~ +1))

= 2In 2

~-i

(8)

N°9 A DIFFRACTION METHOD OF STUDY OF 2D QUASIORDER 1219

0.8

1

0.6 ~~~'~

~l=0.2

(

0.4

0.2

°0 2 4 6 8 10 12 14

q~L

Fig. 2. Comparison between M computed in the Lorentzian case and the best fit with the sum of

a Lorentzian and a fine of base, for three dilferent values ~: ~

= o, ~ = ù-1 and ~

= 0.2. ):

M(qp, ~)

for ~

= 0, ~ = ù-1 and ~

= 0.2 in the Lorentzian case; ): best fit with the sum of a

Lorentzian and a fine of base.

0.8

é ~~ ,v"Î"

~

0.4

_.1°

o

_~ u

0.2 ~ .°'°

0.2

0A

0.6 0.8

~

Fig. 3. Companson between

g(~)

and ~ for three dilferent cases.

(-

. -); Lorentzian case;

(-V-)

square-Lorentzian case;

(-o-):

Gaussian case; ): g = ~ fine.

For q = 0, we have

by

construction à'

= 1 and (1

q) In(à'~

+ 1)

= In 2.

If the transformation g is chosen to be

g(à')

=

~~~ ~

+ l then the relation

In(à'2

+

1)

+ In 2

g(à')

= q is

obeyed

for q

= 0 and q

= in the last two cases mentioned.

Figure

3

displays

g(à')

as a function of q for the latter and the "Gaussian" case. The

linearity

is very

good

for the whole domain in the "Lorentzian" case while it is not too bad for other cases. It then appears that we achieved our purpose of

finding

a

quasi-universal g(à')

function and that this

measure is very

good

for

M(qp,

q =

0)

close to a Lorentzian.

The method that we

proposed

for measunng q suifers from two defects:

firstly

it requires

a measure of the

peak

width at

suiliciently

low q in order to define the relative

broadening;

secondly

it does Dot exhibit any

particular

feature in a

single profile allowing

the thermal

quasiorder

to be characterised. This second defect is common in

crystallography Ii?i,

and is

(9)

generally

overcome

by comparing

the width of two diiferent

peaks.

This

comparison

is very

simple

in our case since it should lead to:

g(/~~) l(QBI) QÎI

g(/~~)

SÎ(QB2)

QÎ2

To

identify

trie thermal

quasiorder,

we propose to use

g(à'),

the measure of q, for two

peaks

and to

verify

this relation.

In conclusion several

points

must be discussed:

i)

Dur whole

study

uses the universal

dependence

of the

profile M(qp)

on

qpL, irrespective

of the value of L. It may seem worthwhile both to

explain

this

universality

and to

study

the

approximations leading

to such a result.

To understand the

dependence

of the

profile M(q~)

on

L,

the best way is to come back to its Fourier transform representation:

M(q~)

=

/ ~ Cos(iq~r)(r/uD)~~W(r)dr

By using general properties

of Fourier

transforms, M(qp)

can be written as the convolution of two functions: one

being

the

intensity profile

for a

powder

of infinite

crystals presenting

quasiorder

observed with an infinite resolution

Mr~(q~)

and the other

including

ail the eifects of finite size and finite resolution

T(q~).

M(qp)

=

Moe(qp)

19

T(qp)

with

~ ~

Mr~(qp)

=

Î cas(iq~r) f(r)dr

=

Î cos(iq~r)(r/uD)~~dr T(Lqp)

=

/

ce

cas(iqpr)W(r)dr

o

Scaling

arguments

clearly

show that the power law

dependence

with

r/uD

of

f(r)

results

in a power law

dependence

on uDqp of

Mr~(qp).

On the other

hard, W(r)

and

T(qp)

are

characterised

by

a coherence

length

L

(or

a

peak

width

àqp

=

1IL)

and a

shape

SO that

T(qp)

may be written as

S(Lqp).

Because of the power law

shape

of

Mr~(qp)

and of the

properties

of the convolution

product,

it is dear that the

shape

of

M(qp)

in a

1IL

scale

depends only

on

S(Lqp)

whatever the ratio

LlaD.

Let us take two

examples

to illustrate this fact: on the

one

hand,

suppose the

crystal

size L~r to be quite

large,

the intensity

profile

will be the same in a qp

/àQ

scale as

long

as the resolution

àQ

is

large compared

with

1/L~r

and a resolution improvement will not make easier the

study

of the 2D

quasiorder;

on the other hand, suppose the resolution to be quite

good,

the

intensity profile

will be the same in a

L~rqp

scale as

long

as the

crystals

size is small

compared

with

1/àQ

and an increase of the

crystals

size will not make the

study

of thermal

quasiorder

easier. These assumptions have limitations and

they

are correct in the limit of

validity

of the

approximations

that we have made in our

analytical

calculation. These approximations are

mainly

three: the

asymptotic approximation

for the thermal motion of atoms, the

tangential approximation

m the calculation of the

powder

average and the

separability

of the diffraction

peaks.

Ail these

approximations

are wrong for very broad

peaks,

1-e-, for low L or very bad resolution. In practice, the asymptotic

approximation

is

good

for any

crystal

size smce

f(r)

is near its

asymptotic logarithmic

branch for

rlaD equal

to 2. This

large validity

domain of the asymptotic approximation is the same in 3D calculations for which

f(r)

reaches its asymptotic constant value for quite low values of

rlaD leading

to a validity of

(10)

N°9 A DIFFRACTION METHOD OF STUDY OF 2D QUASIORDER 1221

the well-known

Debye-Waller

formula for any

crystal

size. The

tangential approximation

may be renounced at the price of more detailed calculations. But to renounce the

separability

of the

peaks

is to renounce the concept of

quasiorder

itself.

ii) Though

the

g(q)

function that we

proposed

seems to be

quite satisfactory

as a measure

of q for any low temperature

profile (Lorentzian,

square

Lorentzian, Gaussian)

when examined in the whole domain of variation [0,

Ii,

the

discrepancy

to

linearity

rnay be

problematic

when

verifying

the

q(QBI )/n(QB~

"

Qà~ /Qà~

law in the "Gaussian" case: for

Qà~ /Qà~

" 3

(which

corresponds

to the first two

peaks

of a

hexagonal structure), g(q(QBI ))/g(n(QB~ ))

take values between 2.5 and

5.8,

its average value

being

3.9. Three ways may be used to overcome this default.

Firstly

a numerical determination of a

special

measure of q in the "Gaussian" case may be clone

using

the theoretical

profile

we have determined

ifi Il /2 q/2;1/2; -q(L~/4),

this method seems to be realistic

only

in the case of very bad resolution.

Indeed,

in the case of mean

resolution,

the use of a Gaussian

profile

at low q will

neglect

the

high

qp tait of the

intensity profile

due to finite size eifects and this

high

qp tait will have consequences on the

broadening

for

large

values of ~/.

Secondly,

when ever

possible,

trie ratio

g(~/(QBI ))/g(~1(QB~ ))

should be studied on a wide range of

~/ values and its

good,

even not

perfect,

fit with its theoretical

value, Qà~/Qà~>

may be considered as not fortuitous and therefore as a

good

test of the thermal

quasiorder

when

compared

with other disorders

(paracrystalline order,

pure finite size

eifects, etc.). Thirdly

because the q measure that we propose is quite

good

bath in the "Lorentzian"

case and in the

"square-Lorentzian"

case, 1-e-, when the finite size eifects dominate the

profile,

the determination of q and the check of the thermal

quasiorder

will be

highly

facilitated

by

a resolution

improvement.

This result is

quite physical

and

stimulating

for

experimentalists.

Nevertheless,

the

fairly good

agreement of

gin)

with q in the pure "Gaussian" case

dearly

shows that the measure of q does net need an excellent resolution and a very fine

analysis

of the

profile.

Acknowledgments

The authors are

grateful

to C.

Aslangul,

C. Marti and C. Simon for

helpful

discussions.

References

[1] Jancovici B., Phys. Re~. Lent. 19

(1967)

20.

[2] Imry Y. and Gunther L., Phys. Re~. B 3

(1971)

3939.

[3] Gunther L., Imry Y. and Lajzerowicz J., Phys. Re~. A 22

(1980)

1733.

[4] Weling F. and Grilfin A., Phys. Re~. Lent. 46

(1981)

353.

[Si Dutta P. and Sinha S-K-, Phys. Re~. Lent. 47

(1981)

50.

[6] Huse D.A., Phys. Re~. B 28 (1983 fil la. In this communication a calculation taking into account bath the transverse and longitudinal sound velocity is performed.

[7] Nelson D.R. and Halperin B-I-, Phys. Re~. B 19

(1979)

2457.

[8] Young A.P., Phys. Re~. 819

(1979)

1855.

[9] Strandburg K.J., Re~. Med. Phys. 60

(1988)

161.

[loi

Heiney P-A-, Stephens P-W-, Birgeneau R-J-, Horn P-M- and Moncton D.E., Phys. Re~. B 28

(1983)

6416.

[Il]

Specht E-D-, Sutton M., Birgeneau R-J-, Moncton D.E. and Horn P-M-, Phys. Re~. B 30

(1984)

1589.

(11)

[12] Specht E-D-, Birgeneau R-J-, D'Amico K-L-, Moncton D.E., Nagler S-E- and Nagler P-M-, J.

Phys. Lent. 46

(1985)

L561.

[13] Nielsen M., Ais.Nieisen J., Bohr J.,

MacTague

J-P-, Moncton D.E. and Stephens P-W-, Phys.

Re~. B 35

(1987)

1419.

[14] D'Amico K.L., Bohr J., Moncton D.E. and Gibbs D., Phys. Re~. B 41

(1990)

4368.

[15] Warren B-E-, Phys. Re~. 59

(1941)

693.

[16] Bertaut E-F-, Acta Gryst. 3

(1950)

14.

[17] see for exarnple the charactensation of paracrystalline order in Guinier A., Théorie et technique de la radiocnstallographie, 3e ed.

(Dunod

Ed., Paris,

1964).

Références

Documents relatifs

Largement répandu dans les établissements d’enseignement, le photocopillage menace l’avenir du livre, car il met en danger son équilibre économique et prive les auteurs

Our final implementation uses a red-black tree to store fil- tered transactions, item order is ascending according to their support, simultaneous traversal is used as a routing

➋ L'image initiale pour insérer une image nommée image1 dans la zone de travail : deux points A et B.. sont automatiquement créés aux coins inférieurs

Cubic hypersurfaces, lines in hypersurfaces, finite fields, zeta functions, Tate conjecture, Fano varieties.. The second author was partially supported by Proyecto FONDECYT

Zhou, ‘Analytical Solution for Transient Heat Conduction in Hollow Cylinders Containing Well- Stirred Fluid with Uniform Heat Sink’, International Journal of Heat and Mass

La raison : on trie bien les grandes bouteilles à la maison, mais on a moins le réflexe avec les plus petits formats que l’on emporte dans les transports, en balade, au sport…

Although a compact suffix trie has a bit more nodes than the corresponding suffix tree, all of its arcs are labeled by single symbols rather than factors (substrings).. Because of

To solve plagiarism problem, offsets specification of word tokens and also collecting noun words synsets and words stems are basic satellite data to be used in our proposed tree