• Aucun résultat trouvé

Copolymer Melts in Disordered Media

N/A
N/A
Protected

Academic year: 2021

Partager "Copolymer Melts in Disordered Media"

Copied!
22
0
0

Texte intégral

(1)

HAL Id: jpa-00247218

https://hal.archives-ouvertes.fr/jpa-00247218

Submitted on 1 Jan 1996

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.

Copolymer Melts in Disordered Media

S. Stepanow, A. Dobrynin, T. Vilgis, K. Binder

To cite this version:

S. Stepanow, A. Dobrynin, T. Vilgis, K. Binder. Copolymer Melts in Disordered Media. Journal de

Physique I, EDP Sciences, 1996, 6 (6), pp.837-857. �10.1051/jp1:1996245�. �jpa-00247218�

(2)

Copolymer Melts in Disordered Media

S.

Stepanow (~,*)

,

A-V-

Dobrynin (~,~),

T-A-

Vilgis (~)

and K. Binder

(~)

(~) Martin-Luther-Universitit

Halle-Wittenberg,

Fachbereich Physik, D-06099

Halle/Saale,

Germany

(~) Groupe de Physico-Chimie Th60rique, E-S-P-C-1-, 10 Rue Vauquelin,

75231 Paris Cedex 05, France

(~) Department of Chemistry, University of North Carolina, Chapel Hill, NC 27599, USA (~) Max-Planck-Institut ffir

Polymerforschung,

Postfach 3148, D-55021 Mainz, Germany (~) Instiiut fir Physik, Johannes-Gutenberg Universitit Mainz, Staudinger Weg 7,

D-55099 Mainz. Germany

(Received

3 July1995, revised 5 January 1996, accepted 5 March 1996)

PACS.05.20.-y Statistical mechanics

PACS.64.60.-I General studies of phase transitions PACS.61.41.+e Polymers, elastomers, and plastics

Abstract. We have considered a symmetric AB block copolymer melt in a gel matrix with

preferential adsorption of A monomers on the gel. Near the point of the microphase separation

transition such a system can be described by the random field Landau-Brazovskii model, where

randomness is built into the system during the polymerization of the gel matrix. By using the

technique of the 2-nd Legendre transform, the phase diagram of the system is calculated. We

found that preferential adsorption of the copolymer on the gel results in three effects:

a)

it

decreases the temperature of the first order phase transition between disordered and ordered

phase;

b)

there exists a region on the phase diagram at some small but finite value of the

adsorption energy in which the replica symmetric solution for two replica correlation functions

is unstable with respect to replica symmetry breaking; we interpret this state as a glassy state

and calculate

a spinodal line for this transition; c) we also consider the stability of the lamellar phase and suggest that the long range order is always destroyed by weak randomness.

1. Introduction

The

problem

of the

rigid gel

G immersed in a

symmetric

fluid mixture

A/B

with

preferential adsorption

of one component, say

A,

on the

gel

matrix has been considered

by

de Gennes (1j.

He found that near the

demixing

transition temperature T~ of the

A/B

mixture there exists

a range of parameters where the behavior of the system becomes

dependent

upon the sample

history

and

interpreted

that as a

glassy-state.

There is a very

simple physical explanation

of such behavior. With

decreasing

temperature

T, A-particles

will cover the

gel

matrix in order to decrease the number of

BIG

unfavourable contacts

forming

A-rich domains near the

gel

rods. The

boundary

of this

regime

can be found

using simple scaling arguments balancing

the

energetic gain

due to

A/G

contacts, ENg, with surface energy between A-rich and B-rich

(*) Author for correspondence (e-mail:

stepanow©physik.uni-halle.d400.de)

© Les

(ditions

de Physique 1996

(3)

regions, N(/(~,

where

(

=

T~~=((T T~)/T~)~~

with v m

2/3 being

the correlation

length

of the

A/B

mixture and

Ng

the number of monomers on the rods. This

immediately

results

in a condition for the

strength

of the

preferential adsorption

energy E >

NgT~/~

for which

A-rich domains will form near the

gel

rods and the characteristics of the system will

depend

on the

preparation

condition of the

gel

matrix. One should note that at the theoretical level this

problem provides

an

example

of the Random Field

Ising

Model

(RFIM) [2-7],

where

randomness is built into the system

during polymerization

of the

gel

matrix.

Let us now ask a

question:

"How will the behavior of the AB mixture in the

rigid gel

matrix

change

if we connect A and B monomers into

poly.mer

chains?" There are two different ways to

do it. We can create either chains

consisting only

of A and B monomers or chains which have

two types of monomers. In the first case the

problem

can be reduced to the case considered

by

de Gennes (I]

corresponding

to a

simple

renormalization of the

coupling

constant on the

length

N of the

polymer

chains. Another case when A and B

particles

form for

example

a

symmetric

diblock

copolymer chain,

AjBj can not be reduced to the case of the

A/B

mix- ture and is worth

special

consideration. It is important to note that

already

without the

gel, copolymer

systems behave

differently

from a mixture of Aj and Bj

polymers (8-12].

As the temperature

decreases,

the

copolymers

form one, two or three dimensional domain patterns

depending

on the

composition

of the

copolymer

chains, I.e. the fraction of A monomers on the

chain. The domain structure appears as a result of competition between

short-range

monomer-

monomer interactions that want to decrease the number of unfavorable contacts between A

and B monomers and

long-range

correlations due to chemical bonds between those parts of the

chains that tend to segregate into domains. The first stage of the domain pattern formation

can be described in terms of Landau-Brasovskii effective Hamiltonian

(13-15]

when the

ampli-

tude of the local

composition

fluctuations is small in

comparison

with its average value. This Hamiltonian describes

phase

transitions in the systems such as

weakly anisotropic

antiferro- magnets

(14],

the

isotropic-cholesteric

and nematic-smectic C transition in

liquid crystals

[16]

and pion condensation in neutron stars [17]. The

homogeneous

state of this Hamiltonian is unstable with respect to fluctuations of finite wave number qo that result in the formation of the domain structure with

period

L =

27r/qo

below the transition

point.

The

copolymer

melt in a

gel

matrix discussed in this paper is another

example

of such a system.

The paper is

organized

as follows. In Section 2 we discuss a model and demonstrate that the

problem

of the

copolymer

system in the

gel

can be reduced to random field Landau-Brasovskii model. After that

using

the 2-nd

Legendre

transformation

[18-21(

we calculate the free energy of the system. Section 3 presents the universal

phase diagram

of the

copolymers

in the

gel

for

replica symmetric

solution. In Section 4 we calculate the

spinodal

of the

replica symmetric

solution with respect to

replica

symmetry

breaking.

Section 5

explores

the

stability

of the lamellar

mesophase exposed

to random fields. In conclusion we discuss our results.

2.

Polymer

Formalism

We

begin

with the

assumption

that the

copolymer

melt can be described in terms of the individ- ual coordinates r~(s~)

(the

index I = I,...n~

,

counts the

copolymer chains,

sz

gives

the

position

of the segments

along

the

chain) by using

the Edwards Hamiltonian [22].

Replacing

these individual coordinates

r~(sz)

with the densities of

polymer

segments

(collective coordinates) according

to

pi

(r)

=

f /~~ dsb(r

rz

is))

and

p2(r)

=

f j~

dsb(r

r2

Is))

z=i °

z=i fN

(4)

where

f

is the fraction of

species

I, say, N the chain

length

of the total

copolymer,

and is the statistical segment

length

of the

polymer,

the

partition

function of n~

copolymer

chains

can be

represented

as

Z =

/ Dr(8)b(i uP(r)) exPl-HllP(r)I)) (1)

with

HllPlr)I)

=

( ~j /~ d810rz18)/08)~

12)

+

~ / pa(ri)V~p(ri r2)pp(r2)

+

/ p~(r)U~(r),

where

[

=

f

d~r and the sum convention over Greek indices is assumed. The delta-function in the r.h.s. of

(I)

takes into account the

incompressibility

condition. The matrix V~p

jr)

describes

interactions between the segments, U~

jr) represents

the interaction of

polymer

segments with the random environment. For

example,

if we will consider AB

copolymer

that is immersed in

gel

matrix, the random field U~

jr)

describes

adsorption

of monomers of ath type. In this case

the random field

U~(r)

is

U~

jr)

=

-~E~4lg(r) (3)

where E~ is

adsorption

energy of ath monomer in units of

kT,

~ is the excluded volume of the

gel-copolymer interaction,which

is assumed to be the same for all types of interactions,

4lg(r)

is the local

gel

concentration. The

gel

structure can be characterized

by

two first correlation functions

<

4lg(r)

>

=

§

=

Ng IRS (4)

and

<

4lg(r)4lg(r')

> =

G(r r') (5)

where

Ng

is the number of monomers between cross-links and llm is the mesh size distance which is

proportional

to

Ng

for a

rigid

network and

-~

N(/~

for a Gaussian one.

Taking

into account the relations

(4-5),

we can write two first moments of the random fields

U~(r)

as

follows

< U~

jr)

> =

~Enj (6)

<

Un(r)Up(r)

> =

~~EnEpG(r r')

=

~~EaEpN(Rjj~b(r r'). ii)

One should note that the last

equality

in the r-h-s- of

equation ii)

is valid as

long

as the characteristic

length

scale of the

polymer

melt L is

larger

than Rm.

Introducing

an

auxiliary

field

4l(r)

we can rewrite

(I)

as follows

z =

/ D4l(r) expj- j

j~

Am

jri )V~p jri

r2)Alp

jr2

18)

/ Ua(r)Ala(r))

<

b(( P)

>o

where

V~p(ri r2)

= ~

~ ~

b(ri r2)

is the matrix

representing

interactions

l + x 1

between monomers, x is the

Flory-Huggins

interaction parameter between the monomer

species

of the two blocks of the

polymer,

the brackets <>oin

(8)

represent the average over the

(5)

configuration

of the ideal

copolymer chains,

which is

given by

the first term in the

exponential

of

(I).

The average <

bill pi

>o can be written as

[8,9]

exPl-) / b4lal-qlsi[plqlb4lplq) Rut lb4lll,

so that the

partition

function takes the form

z =

/ D~(rj exp(- ~~ j-qjv~p jq

j4~p

jqj (9

/ b4~al~q)Silfll~)b~fllql / U*l~q)~* lql

ant

lb~))1

where f~ =

f d~q/(27r)~, b4ln(q)

is the Fourier transform of the function

b4ln(r),

the matrix

Sp(p(q)

is the inverse of the matrix with elements

being

the structure factors of a Gaussian

coiolymer

chain

(see

for

example

[8]

),

the functional Rnt

(ill)

is a series in powers of the order parameter ill. In the case of an

incompressible copolymer

melt

(~pm

-J

I),

which we will consider in this paper, b4l2 =

-b4li,

so that in this case we have a scalar order parameter

density

Ah

(r)

+ Ah

(r) fp,n. However~

since the

ordering

does not occur at wave vector k

= 0

in

reciprocal

space, this remark should be taken as a statement on the classification of the order parameter in the sense of

universality

classes in

phase

transitions. The effective Hamiltonian reads in this case

Hill)

=

j/All-q)Gi~lq)4llq)

+

~ / / / 4llqi)4llq~)4llq314ll-qi

-q~

q3)

+

/hl-q)4llql> (lo)

where uo is the fourth-order vertex function

computed

at

qRg

= qo,

h(r)

=

Ui(r) U2(r),

pm

is the average monomer

density being

for melt pm

-~ I

In,

N is the

degree

of

polymerization

of

copolymer chains,

and the Fourier transform of the inverse propagator Go

(ri r2)~~

is

given by

Gi~lq)

"

lNPm)~~121xs xlN

+

~llql ~0)~li Ii ii

where we have

approximate Go(q) by

its

Taylor

series up to the second order terms, and x~

is the value of

Flory-Huggins

parameter on the

spinodal computed by

Leibler

(xsN

= lo.495 at

f

=

1/2).

The constant e in equation

ill

is

given by

e ci

R(,

with

Rg being

the

gyration

radius of the

copolymer chain,

qo

It 1.94/Rg

at

f

=

1/2)

is the

peak

position of the

scattering

factor. Because of the unique scale of the

problem,

qo

°~

Rj~,

it is useful to introduce new dimensionless variables

q =

qR~, ~2jq)

=

~2jq) /jp~NRj),

>

=

uoN/jp~Rj),

~

=

2jx~ x)N j12)

In these new variables the effective Hamiltonian reads

Hill)

=

/lllql

qo)~ +

T)4llq)All-q) l13)

+

( fl / 4llq~)bl~j qi)

+

/ ~@/2 ~l-q)4llq)

~i

Q~

~i

q g

~

(6)

One can see from

equation (13)

that the order parameter

4l(q) couples linearly

to the ran- dom field

h(q). Thus,

the

copolymer

melt in a disordered medium is described in terms of the

Leibler's free energy

coupled linearly

to the random external field

h(q).

This

problem

is analo- gous to the random field

Ising (RFIM)

model [2]. The difference of the

copolymer

Hamiltonian in

comparison

to that of the RFIM consists of the

following.

While the propagator in the

RFIM achieves the maximum for zero momentum, the Leibler propagator,

Go (q)

(8], achieves his maximum at the some finite momentum qo.

The average over the random

potential h(q)

can be carried out

by using

the

replica

trick.

As a result the

multireplica

effective Hamiltonian is obtained as

Hnl14lll

=

/ f 4lalq)lbabGi~lq) ~)4lbl-q) l14)

+

( f ij / 4lalqz)blf qz),

a=iz=i Q~ ~=i

where n is number of

replicas,

and /h

=

Npm~2

(El

E2)~Nj IRS.

To calculate the free energy Fn we use a variational

principle

based on the second

Legendre

transformation

(18,19].

In accordance with these references

(see

also

Appendix A)

the

n-replica

free energy is obtained within the framework of the second

Legendre

transformation as

Fn = min

Wnl14lalq)1, lGablq)1) lis)

with

Wn "

-(Sp / lnjGabiq))

+

I f

/ibabGp~io) /h)Gabi-q)

~

a,b=1 ~

+

f /

<

4~ai~q)

>

i~abGi~io) ~)

<

4~biqi

>

+Snii< 4~ai~) >ii i~abi~)i)1 (16)

where the minimum is to be

sought

with respect to functions <

Ala(q)

> and the renormalized Green's function

Gab(q)

that are considered as

independent

variables at the fixed parameters I, T, /h. The

quantity Sn((< Ala(q) >), (Gab(q)))

is the so-called

generating

functional of all

2-irreducible

diagrams

that cannot be cut into two

independent

parts

by removing

any two lines between vertices I

(see Appendix

A for

details).

3.

Replica Symmetric

Solution

In this section we will consider

minimizing

the free energy

by taking

into account

only

the

one-loop

contribution to

Sn(...).

One can

easily

see

that,

in this case,

only

a

replica

symmetric solution exists. In the

replica

symmetric case, the effective propagator is

expected

to have the

following

form

Gab(q)

"

9(q)bab

+ (1

nab)flq)i (ii)

where

g(q)

and

f(q)

are

arbitrary

functions of the form

given

below. The

representation (17)

is

orthogonal

in the sense that the

diagonal

elements of the matrix Gab are g and the

off-diagonal

elements are

f.

The inverse of G is

given by

~~~~~~~ A~~~ ig i)

g

/+ ~i

~~~~

(7)

which in the limit n ~ o

simplifies

to

Gabio)~~

=

i) f

~~~

ibab f

~~~

il

bob).

i19)

In the ordered

phase,

there is a nonzero average value of the order parameter <

4l(q)

>, the Fourier transform of which has the form

< 4~aIQ) > "

Ai~io

Q0) +

~i~

+ Q0)).

12°)

Here we will consider

only

the ordered

phase

with the lamellar type of symmetry of the mi-

crophase

structure because it has the lowest free energy for the effective Hamiltonian

(14)

with- out the cubic term.

Substituting

the trial functions

(19-20)

into

expression

of the free energy we

can write the free energy as a functional of

A, f,

and g. In order to find the values A,

f,

g cor-

responding

to the extremum of Fn we have to take variational derivatives of Fn with respect to these functions. In the

one-loop approximation

for the functional

Sn((< Ala(q) >), (Gab(q)))

this results in the

following

extremum

equations A(T

+

'

/

G~aa~(q)

+ '~

= o,

(21)

2

~

2

~

ig

j~~

/~ = ° 122)

We note that the equations for g and

f

can be also obtained

by inserting (19)

into

(A.3)

and

comparing

the coefficients in front of (bob

I)

and bob.

Equation (21)

can also be obtained

directly

from equation

(A.20) using

equation

(20).

For the disordered phase

(A

=

o)

this system can be

simplified

to

19

f)~~

"

Gi~lq)

+

/

G(aa)lq)1

124)

f

=

/hig f)~. 125)

One can see

that,

in the

one-loop

approximation, the function

(g f)~~

has to have the form

((q(

qo)~ + r, because the

integral

in the r-h-s- of equation

(24)

results in renormalization of the effective temperature T.

Taking

into account the equations

(17,25)

we can write the

two-replica

correlation function G~abjIQ) as

G(ab)IQ) " 98(Q)bab +

iig~

IQ)

(26)

with

gB(q)

=

I/(((q(

qo)~ +

r).

It is convenient to introduce new reduced variables t =

T/(lq( /27r)~/~,

b

=

/h/(lq( /27r)~/~,

z =

r/ (lq( /27r)~/~ (27) Substituting

the function

given by (26)

into

(24)

we

get

~ ~ ~

47r@~~

~ 2r~' ~~~~

(8)

which,

on

substituting

with dimensionless

variables, gives

z = t +

)z~~/~(l

+

)). (29)

Repeating

the same calculations for the ordered

phase (A # o),

for which A~

= 2

/(1(g ii),

we

finally

obtain

~

~ ~ ~

4~$~~

~ ~' ~~~~

which

gives

in dimensionless variables

t = z +

~z~~/~(l

+

)). (31)

First we consider the case of weak field /h

IT

< I.

Analyzing equation (30)

one can see that the solution of this

equation

appears

only

below some critical temperature T~ that can be

considered as the

"spinodal"

of the ordered

phase.

The critical value of r at which

(30)

has a solution

corresponds

to the minimum of the r-h-s- of this

equation.

For a weak random field the

spinodal

r is obtained as

~

r~ =

l'~° )~/~, (32)

47r

which

gives

the

following

critical value for the effective temperature T~

Extrapolating

equation

(30)

to

large

fields

(/h IT

>

I)

we get the critical value of r~ as

~~

l~7r~'~~~~~~

The effect of

the

next-order terms on (34) is

discussed

at

the

end of

To

calculate the

phase

diagram of the

system under

onsideration, the energies

mogeneous

and

ordered phases have

to

be ompared. Substituting trial functions given

by uations (26) into thepression

for the

-replica free energy F~ andtaking

the limit

<

n-o

less riables)

and for the ordered one

The

phase

agram of

the

system

in

andom media is

ketched in Figure I in

the plane

16

= /h/((lq]/27r)2/~), t =

In

the limit of a weak

the irst-order phase ransition occurs

at

Ttr Cf -1.3(lq(/27r)~/~. At

the

strong random field

limit

the emperature

of the

phase transition

is

proportional to /h~/5 The latter behaves

(9)

lo

8

6

DIS GLASS

LAM

0 -1 -2 -3 -4

t

Fig. 1. The phase diagram of the copolymer melt in

a random field environment obtained within

the one-loop replica symmetric solution. The continuous curve is the coexistence curve between the ordered

(lamellar)

and disordered states. It is computed by setting equal equations

(35-36

). The dashed line is the stability line of the replica symmetric solution. It is derived by solving equations

(52,29).

Imry-Ma

arguments (23]. The statistical fluctuation of the random field

h(q)

will

destroy

the domain structure as

long

as the amplitude of the

4l(q)

due to this random field

h(q)

is

larger

than the

amplitude

of the dense wave A. In other

words,

as

long

as the energy of the domain

structure (T(~

IA

is lower than the fluctuation energy

/hq( / @~,

the

following

criteriuIn holds

true:

l~trl ~

(iql~h)~/~

137)

In the

approach

used above, we

neglected

the contribution of the second-order

diagrams

in powers of the vertex I in the

Dyson

equation

(24).

The contribution of this

diagram

is

1(2)

=

j>2 /48gr)r-3/2jji

+

z~/j2r))3 jz~/j2r))3), j38) Comparing

I12) with the contribution of the

one-loop diagram,

one can write

In the case of strong

fields,

/h

IT

> I, we can

neglect

the second order

diagram

as

long

as the

following

parameter is small

~lj

l))~~~ b/Z~

I1.

j40)

The first-order

phase

transition to the ordered

phase

occurs at

Zc "

~b)~/~

(41) Substituting

the latter into

equation (40)

we get an

inequality

for b where our approximation works

~~~

j~~)2/3

~

~ ~~~

34/581/5 ~qo ~ ~' ~~~~

(10)

In the weak random

field, /h/r

< I,

equation (40)

reduces to

loo

~_

(27r)2/3

1

~~~

~~~ ~2

Qo ~ ~'

(43)

4.

Stability

of the

Replica Symmetric

Solution

In this section, we consider the

stability

of the

replica symmetric

solution

Gab(Q)

"

/hf(q)

with respect to

replica

symmetry

breaking.

With this purpose let us write the renormalized

two-replica

correlation function GabIQ) in the

following

form

~abi~)

" 9BiQ)~ab +

~911Q)~i~b

+

boabio) i~~)

where e =

(I,..,I)

and

boab(q)

is a function that

equals

zero for a

= b.

Substituting

the function

Gab(q)

into the r-h-s- of the n -replica free energy

(A.21)

and

expanding SpIn(Gab)

in powers of the function

boab(q)

one finds

~Fn =

11 lib

~ele/boablq)bocalq)1

145)

( / / / boab(k)boab(P)flq)flq

k P)I

where in the r-h-s- of equation

(45)

the summation over all repeat indices is assumed. The derivative of

Sp In(Gab

with respect to

boab

IQ) is easy to obtain

using

the relation

bGjj /bGbc

"

-Gj/ Gj/

and

neglecting

terms

proportional

to n~. One should note that the last term in equa- tion

(45)

is the second-order term in powers of the vertex I. The form of the

perturbation boab(Q)

can be found

by analyzing

the

Dyson

equation for the

off-diagonal

part of the two-

replica

correlation function

including

the second-order

diagram

powers of the vertex I. This

analysis

shows that the fluctuations with (q( = qo

give

the main contribution to the renormal- ization of the bare characteristic of the system, so we can choose

boab(Q)

in the form

~oabio)

"

Qab9iiQ)1

1~6)

where

Qab

is a n x n matrix.

bitroducing

Parisi's function

q(x)

(24] defined in the interval (o,

I]

and connected to

Qab by

~ Q~ix)dx

=

it jn~

i~

j Qi~

v k

14i~

a,

the

quadratic

part of the free energy in power of

Qab

is obtained as

~F =

-(1 /~ /~llli -13)blx

Y)

+12)qlx)qlY)dxdYl, 148)

~~~~~

ii "

/9~(Q) ~~

~~~' ~~~~

q

~~

1~7r2'~~~~~~

~'

~~~~

12 " 2~l

/ g((q)

=

~

q(/hr~~/~ (51)

~

87r

(11)

The

replica symmetric

solution is unstable in the

region

of parameter values where the matrix

(Ii -13)b(x y)

+12 has a

negative eigenvalue

1-=11+12-13 <0

or in terms of reduced variables

1-

-~

(1+

~ c

~~ § o

(52)

where c

=

((27r)2/~/64)(1/qo)~/~

< l. Equation

(52)

determines the

spinodal

line of the

replica symmetric

solution. In the case of the weak random field

/h/r

< I we can

neglect

the second term in the r-h-s- of

(52)

and write

r~p =

(c/h~)~/~,

/h < c~/~

(53)

Extrapolating

equation

(52)

to

large /h,

/h

IT

> I or, /h > c~/5

we have

r~p =

(~c/h)~/~,

/h > c~/~

(54)

Substituting

the solutions r~p into

equation (28) giving

r as function of T for the disordered sthte, we get the

following

boundaries for the

replica

symmetric solution in the (T,

/h) plane

in the limit of small and

large fields, respectively

Tsp =

c2/9z~4/9 )c-1/9z~2/911

+

)iz~/c2/5)5/9),

/~ <

c2/5 155)

Tsp =

(c/h)~/~

~

/h~/~(2c/3)~~/~(l

+

2((2c/3)~/~llh)~/~),

/h > c~/~

(56)

4

However,

in order that this

region

with the RSB solution exists on the

phase diagram,

the effective temperature T as function of /h has to be

higher

than the effective temperature of the first-order

phase

transition Ttr. The

boundary

of the RS

solution,

which is

computed by

using equations

(52,29)

for the value c

= o.5, is also

plotted

in

Figure

I. It follows that RSB occurs

already

in the disordered

phase.

5. Disorder vs.

Ordering

in the Lamellar Phase

The previous section discussed the behavior of the block

copolymer

melt in the one

phase regime, I.e.,

at

high

temperatures before the micro

phase

separation transition. The

possibility

of the existence of the

glass phase

before the micro

phase

separation transition was shown. In the remainder of the paper we discuss the effect of random fields at temperatures below the micro

phase

separation transition. For

simplicity,

we consider

again only

symmetric diblock

copolymers.

In this case it is well known that the structure of the melt consists of the lamellar

phas~jy,

which can be viewed as

periodic

arrangements of

interphases

between thin films of A-rich a d B-rich domains. The

interesting

question that emerges is what is the effect of the random field on the

phase boundaries?,

I-e-, the interfaces between the domains. In the

following

we

d(tinguish

between the weak Segregation

lin~it

and the so called strong segregation limit

deep

below

th~,MST

transition temperature. In the latter case the

phase

boundaries are well formed and

the'density profiles

are

sharp,

whereas in the case of weak segregation the

density profiles

can be diodelled

by trigonometric functions,

at temperatures below but close to the MST transition temperature. The

following

consideration of the

stability

of the lamellar

phase

in the strong

segregation

limit is based on

Imry-Ma

arguments (23]. The

stability

of the lamellar

phase

in the weak

segjegation

limit is

cinsidered by mapping

the

copolymer

Hamiltonian onto the Hamiltonian of the random field XY model [25].

(12)

interfaces

o

B-rich -rich B-rich A-rich

P

--z h

A-rich

i

' flit@rfQC@S

' al bl

'

Fig. 2. The distortion of lamellas due to random fields.

5.I. COPOLYMER MELTS: STRONG SEGREGATION LIMIT. Consider block

copolymer

melts

with

f

=

1/2

in the strong

segregation

limit. The

ordering

consists of

parallel

interfaces

separating

thin films of A-rich and B-rich composition,

arranged

as a one-dimensional lattice of

spacing

D

(see Fig.2).

The interfacial free energy is

independent

of chain

length N,

while D scales as

N~/~.

We argue that this

long

range order is unstable

against arbitrary

weak random fields

(which locally prefer A,

or

B, respectively.

This

instability

occurs

against

weak

long

range distortions of the interfacial pattern relative to the ideal pattern, which acts as a local deviation bD of

the thickness of the lamella. These local deviations add up in a random-walk-like fashion over

a

large

distance z

perpendicular

to the local tangent

plane

to the interfaces. Hence, there is

no

long

range order over

large

distances.

Consider first the energy balance for a distortion of one interface

by

a small distance h

(compared

to

D)

over a radius p

parallel

to the interface

(in

d dimensional

space)

[26]. The volume of the shaded

region

in

Figure

2 is

V =

constp~-~h. (57)

The

gain

in random field excess energy scales like the square root of this volume

(Hrf

is the

strength

of the random field and is

given roughly by

the root of its variance, I.e., Hrf ci

4).

Erandom

field CC

~HrfW

CC

-HrfPi~~~~/~/l~/~ j58)

We estimate the energy cost for distortions of the interface from the

capillary

wave

description (assuming

a small

angle

8 for the local

bending

of the interface: 8 m

2h/p

<

I)

H~w =

a~l~~~~ / ~c(Vh)~df. (59)

2

Here df is the element of the

id I)-dimensional interface,

a the

underlying microscopic length

(size

of

polymer segment),

~c the interface stiffness

(which

is

equal

to the interfacial free energy

(13)

for block

copolymers).

For the case

considered,

Vh c~

8,

and hence

(we

henceforth measure

all

lengths

in units of

a)

(H~w)

c~ ~cp~~~8~ c~ ~cp~~~h~.

(60)

For

suppressing

such distortions of the

interface,

we would need

lHcw)

>

lErandomfieldl

~tP~~~/l~ >

HrfPi~~~~/~/l~/~ 161)

P~~~~~/~h~/~ >

Hrf/~t. 162)

We see that for d < 5 this

inequality

is violated for an

arbitrarily

small

strength

of the random

field,

for

large enough wavelengths

p at a finite distortion h. The thermal fluctuations would

only

induce a

divergence

of the local interfacial width for d < 3

(in

d

= 3 (h~)~~~~~~~ c~ In p, due to the

thermally

excited

capillary waves).

The random field creates such an

instability already

in d < 5 dimensions. In

particular,

for d

= 3 the condition that interfacial distortions must

satisfy

in order to occur is

p~~h~/~

<

Hrf/~c

or

V$8

< Hrf

IN. (63)

The balance between the random field energy and the

bending

energy

gives

a connection between the distortion h and the

length

scale p

along

the interface

/l Cf

(H

f/lQ)~/3

p(5-d)/3 j6~)

The lamellar

mesophase

will become instable, when the distortion h in

(64)

will become com-

parable

with the lamella

spacing

D.

Equation (64)

with h

= D

-~

N~/~ gives

the

length

scale at which distortion of the lamellar structure is relevant

id

=

3)

pm~x cf N~C

/Hrf. (65)

The

instability

of the lamella due to random fields should also exist in

mesophases

of cross- linked

polymers

such as

interpenetrating polymer

networks [27].

It must be

emphasized

that the considered distortions of the interfacial pattern constitute

only

one class of defects in the order stabilized

by

the

randomness,

but there may be other distortions that also are worth

considering.

One is the fluctuation in the direction of

q(x),

over distances

ix x'(

» D. This defect needs consideration of the

"bending rigidity'

of the pattern of

parallel

interfaces as a whole. Also it may be of interest to consider

"topological

defects" in the interface pattern, such as shown in

Figure

3, which

possibly

also could be stabilized

by

random field energy

gains.

These

problems

are left for future works, because even if it is found that such defects are also

stabilized,

this result would

only strengthen

and not

weaken our conclusion that lamellar

phases

are destabilized

by

random fields.

5.2. COPOLYMER MELT: WEAK SEGREGATION LIMIT. In the weak

segregation

limit for

which the fluctuations in the local

composition

are small in comparison with its average

value,

we can use the Hamiltonian

(13)

to describe the system below the

microphase separation

transition.

Keeping only gradient

and random field terms, the effective Hamiltonian is

H

ii~i IT)))

=

( / iiv~

+

Qii

~

ir))~ ~li/l~~ / hit)~

IT) 166)

For the case of a

periodic

structure in the z

direction,

the average value of the order parameter (4l

IT))

is

ill jr))

= 2A cos (qo

(z

+

u(r)))

,

(67)

(14)

liiil

al

fibl

A-r,ch

Fig. 3. Example of topological defects.

where the scalar function

u(r)

describes the deformation of the

layers

in the z-direction. Sub-

stituting equation (67)

into the r.h.s. of equation

(66),

after

averaging

over all

possible

con-

figurations

of the random field

h(r),

one can write the effective Hamiltonian for the function

~(T)

H

ilUa

IT)

I)

=

[ ) /

(ili~,yUo ir))~

+

4qi iV~Uo ir))~)

~ /hAaAp /

CDS iqo

iUoir) Upir)))

168)

a,p r

where the first term in the r.h.s. of

equation (68)

describes the deformation of the lamellar

layers

and the second one

couples

this deformation in different

replicas.

It is

interesting

to note that the cosine-like

coupling

term between fluctuations of the order parameter in two different

replicas

appears in disordered

physical

systems such as an array of flux-lines in type II

superconducting films,

in

magnetic

films [28], in

crystalline

surfaces with disordered substrates (29] and in the random field-XV model [25]. In all these systems the cosine term results in the

breaking

of

long-range

order and spontaneous

replica

symmetry

breaking (30, 31].

In the framework of the Gaussian variational

principle

(31, 32] the contribution to the free energy of the system due to fluctuations of the

displacement ua(r)

in different

replicas

is

Fvar "

~jSP /'~lGab(q))

+

lH11"a(q)1) H0)0

169)

where we have introduced

~0

~ / ~a/(~)~a(Q)~b(~Q) (I°)

a,b

and

Gj/(q)

is the two

replica

trial function whose form has to be defined

self-consistently

and the brackets

(...)~

denote the thermal

averaging

with

weight

exp

(-Ho ).

After thermal

(15)

averaging

the variational free energy reads

~f~

"

)sp / illjGab(q))

+

jsp / (Gi~ lq)bab G&/

IQ))

Gab(q)

~ lhAaAb

eXp

(- )

2

ab) Iii)

b#a

Where

Bab "

/ iGaaio)

+

Gbbio) Gabio) Gbaio)) 172)

and the inverse bare propagator

Gp~ (q)

is

(~ ((q(

+

q()~

+

4q(q])

The trial function GabIQ)

can be found from extremal equations

Gj/ (q)

=

Gp~ (q)

+ 2

~j /hq(AaAb

exp ~~Bab

,

(73)

a#b

~

~a/

IQ)

~21~Q~~a~b

eXP

(~j

~

~abj (14)

Analyzing

the last equation one can conclude that function

Gj/(q)

does not

depend

on the

momentum q for a

#

b.

So,

we can define

Gj/(q)

= -aab

la # b).

In the case of the one-step

replica

symmetry

breaking

for which the elements of the matrix aab are

assumid

to have two

different values ao and al

depending

on whether the two indices a and b

belong

to the same blocks of

length

m or not, one can rewrite the extremal equations

(see

(31] for

details).

al " Y exp

(~

ln

t)

,

(75)

ao = Y exp ~ lnt

(2~

In

Lqo

+ ~ ln

t)

,

76)

m

where L is the linear size of the system and we have introduced the

following

parameters

assuming

Aa to be the same in all the

replicas

and

equal

to A

~

16~A2'

~~~2q(~~~'

~'

~~~~~~' (77)

In the

thermodynamic

limit L ~ cc equation

(76) gives

ao

" o.

Taking

this fact into account

the trial function

Gab(q)

is

~~~~~~

p~(q~+

mar

~

Gj~(q) (G ~(q)

+

mail

~~~~

GabIQ) =

~ i

(

for

a, b E

diagonal

blocks m x m

(79) I

IQ)

(Gi

IQ) +

mail

Gab(Q)

= o, for a, b E

offdiagonal

blocks m x m

(80)

Substitution of the solution

(78-80)

for the trial function

Gab(Q)

into the

expression

for the variational free energy yields (31]

f~~r = lim ~

(F~~r(t) F~~r(o))

=

'f~

(l

)~t

+

Y'(I )t~)

,

(81)

"-° n

~ ~

m

(16)

where Y'

=

2Y/(A~q()

=

4/h/q(.

The

equilibrium

values of the parameters m and t can be found from the system of the

equations

~ ~t Y't~

= o,

(82)

(1 +

Y'(I m)t~~~

= o.

(83)

For ~ > l this system has

only

the trivial solution m

= I and t = o that

corresponds

to the

replica symmetric

solution with all

off-diagonal

elements of the matrix

Gab(Q) equal

to zero.

The nontrivial solution for ~ < l is

given by

m = ~,

(84)

1

t =

(Y'~)1-~ (85)

In other

words,

at ~ = l the system

undergoes

a

phase

transition at which the correlation function GabIQ)

changes

form. We can rewrite the condition ~

= l in terms of the parameters

of the system. In the mean field

approximation

A~ = 2 (T(

IA,

which

gives

the effective temperature of the

phase

transition (T( =

qol/327r. Comparing

this temperature vJith the temperature of the first order

phase

transition (T( re

(q(1/h)~/~one

can see that for /h >

q(/~l~/~

there is a first order

phase

transition between the disordered state

IA

=

o)

and the ordered

phase IA # o)

with one step

replica

symmetry

breaking

for the correlation function

I"al~)"b(~~))

To

complete

the

analysis

we now calculate the correlation function

(4l(r)4l(r'))

in the or-

dered

phase

below the

phase

transition

(~

<

l).

Due to fluctuations of the

displacement u(r)

the correlation function is:

14l(r )4l(r'))

c~ A~ cos

(qo(z z'))

exp

~-

~~

((u(r) ~t(r') )~))

,

(86)

2

with

(jujr) ujr'))2j

= 2

Iii

cos

[qjr r'))) jujq)uj-q)). j8i) Substituting

the

expression

for the correlator

(u(q)u(-q)),

which is

given by

the

diagonal

part of the two

replica

correlation

function,

one can write

, ~

2 (1

~)

In t + 2 In

z '~

qo, for z

z' (jl

> I

((u(z)-u(z)))

= m , , ,

qo 2~In

z

z qo, for z z

(j~

<

z~

z)

= o

(88)

and

((iL(Xl) MIX) ))~)

"

j

~~

j~~~

~ ~~

~~ ~~

~~~

~

~~

~ ~~ Ii

o ~ n xi x

~ qo, or xi x ~

~

~z-z'~

= o

(89)

Références

Documents relatifs

Holding meduc and f educ xed, by how much does sibs have to increase to reduce expected years of education by one year?. (A noninteger answer is acceptable

Tandis que le deuxième chapitre a eu pour objectif principal l’élaboration de la relation hauteur- débit ainsi que celle du coefficient de débit par une approche théorique en

The proof of the central limit theorem for the position of the tagged particle in the asymmetric exclusion process relies on the dual spaces and operators introduced in Section

We can see clearly, by the end of the Translator’s Note, that this is more a preface than a note, and one which allows for tension resolution (résolution d’une tension)

In the case when all the singular fibers of f are of Kodaira type I 1 , Gross-Wilson have shown in [GW] that sequences of Ricci-flat metrics on X whose class approaches c 1 (L)

there is a great need for more purposeful quantitative and qualitative research on effective classroom adolescent literacy practices-research that informs teachers about the kind

Chairperson of the first session of the General Assembly of the States Parties to the 2003 Convention for the Safeguarding of the Intangible Cultural Heritage,.. Your Excellency

Let (M, J) be a compact, connected almost complex manifold, endowed with the action of a connected compact Lie group G, and let L be a positive G-equivariant line bundle on M.. One