• Aucun résultat trouvé

The random phase approximation for polymer melts with quenched degrees of freedom

N/A
N/A
Protected

Academic year: 2021

Partager "The random phase approximation for polymer melts with quenched degrees of freedom"

Copied!
19
0
0

Texte intégral

(1)

HAL Id: jpa-00246520

https://hal.archives-ouvertes.fr/jpa-00246520

Submitted on 1 Jan 1992

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.

The random phase approximation for polymer melts with quenched degrees of freedom

M. Brereton, T. Vilgis

To cite this version:

M. Brereton, T. Vilgis. The random phase approximation for polymer melts with quenched degrees of freedom. Journal de Physique I, EDP Sciences, 1992, 2 (5), pp.581-598. �10.1051/jp1:1992168�.

�jpa-00246520�

(2)

Classification

Physics

Abstracts

05.40 61.25H

The random phase approximation for polymer melts with

quenched degrees of freedom

M. G. Brereton

(I)

and T. A.

Vilgis (2)

(1) WC in

Polymer

Science and

Technology,

The

University

of Leeds, Leeds LS2 9JT, G-B- (2) Max Planck Institut fur

Polymerforschung,

Postfach 3148, 6500 Mainz,

Germany

(Received I December I99I,

accepted

5 February 1992)

Abstract. There are many concentrated

polymer

systems of interest (e.g. crosslinked rubber networks) where the chains retain their

configurational

freedom but the translational

degree

of

freedom is absent. To

investigate

the effect of the loss of this

degree

of freedom on the

density

fluctuations we have considered a

simple

case of a melt of polymer chains where one end of each chain is anchored to a fixed

point

in space. The random

phase approximation (WA)

used for melts cannot be

immediately applied

to this

problem,

since the loss of translational freedom

couples

together density fluctuations

belonging

to different wavevectors. The quenched chain end

variables must first be accounted for

using

a «

replica

» calculation, then the WA can be used for the annealed

configurational

degrees of freedom. An

explicit analytic

result has been obtained,

which shows that

despite

the loss of the translational

degree

of freedom the

density

fluctuations in the

quenched

system remain similar to those in the true melt out to

spatial

scales of the order of N I/~R, where R is the size of a

single

chain coil (~ N ~°) and N is the

degree

of

polymerisation.

We have also considered the concentration fluctuations a binary blend of chains with anchored chain ends. The situation is shown to be very similar to a diblock

polymer

melt in so far as when the

Flory

interaction parameter exceeds a

given

value

microphase

separation occurs.

Agaul

the loss of the translational freedom of the chains makes little difference to the

stability

criterium.

I. Introduction.

The random

phase approximation (WA)

was introduced into

polymer physics by

de Gennes

[I]

and has been

responsible

for a whole series of

important

theoretical and

experimental

advances in the

understanding

of concentrated

polymer systems [2-9].

The

approximation

deals

directly

with the statistical mechanics of

density

and concentration fluctuations in tennis of a mean field

concept.

An essential

part

of the WA is the

assumption

that there are many

overlaping polymer

chains and that the chains must be free to be

anywhere

in space and to

adopt

any

configuration

consistent with the chain nature of the molecule.

Correspondingly

its

application

is

largely

restricted to annealed systems such as melts or concentrated solutions.

However there are many concentrated

polymer

systems of interest where not all of these conditions are fullfilled. For

example

in a crosslinked rubber the chains

overlap

each other and have

nearly complete configurational

freedom but are constrained

by

crosslinks to be in

localised

regions

of space. In this case the crosslinks have

effectively quenched

the

(3)

translational

degrees

of freedom associated with the centre of mass of the chains. Of

particular

interest are the cases of networks formed from

polymer blends,

co-block

polymers

or

inter-penetrating

networks

[10,

II

].

Other

examples

of concentrated

polymer systems

with

quenched

variables which

negate

the direct

application

of the RPA are : flexible side chain

polymers

attached to immobile groups such as colloidal

particles, liquid crystalline

stems or surfaces

[12, 13].

In all cases the location of the

polymer

chain is localised in space and has to be considered as a

quenched

variable. Recent attempts

[14, 15]

have been made to

apply

the

RPA,

as used for

melts, directly

to

quenched (network)

systems without theoretical

justification.

In this paper we wish to address the statistical mechanical

problems posed by polymer

melts when the translational freedom of the molecules is

quenched.

To

accomplish

this we will

consider the

simple

case of a melt of

polymer

chains where one end of each chain is fixed in space,

apart

from this

single

restriction the chains will have

complete configurational

freedom. For statistical

systems

with

quenched variables,

the Partition Sum is

ideally

calculated

by averaging

over the annealed variables for a fixed set of values of the

quenched

variables. Then some

physical quantity

of

interest,

such as the free energy, is found and

averaged

over the distribution of the

quenched

variables. In

practice

it is more convenient to average the

physical quantity

over the

quenched

variables first

by

means of the «

replica

trick ». This

technique

is now well established since the

early

work of Edwards on

spin glasses [16, 17]

and networks

[18].

It has

recently

been used

by Panyukov [19]

to

investigate

the effect of

topological

defects on

density

fluctuations in

polymer

networks.

By using replicas

the

quenched

average is done before the annealed average and a variation of the conventional RPA is then used to

perform

the annealed averages. For the

problems

considered in this

paper the final

replica

space calculations can be done

explicitly

and without further

approximations.

Before

presenting

these calculations we will consider a very

simple quenched

system where the

density

fluctuations can be found

directly

without

using

the

replica

trick.

This system consists of

point impurities

fixed in space and surrounded

by

the

polymer

melt.

The

impurities

are assumed to interact with the monomers

through

an

arbitrary

interaction

potential.

The form of this result, obtained without

using replicas,

is

important

for the later work and

provides support

for the later results where use of the

replica

method is essential.

2.

Polymer density

fluctuations in the presence of

quenched impurities.

The

system

of fixed

impurities interacting

with a

polymer

melt is shown

schematically

in

figure

I

together

with the notation to be used.

An

impurity

fixed at the

position R~

interacts with the s segment of a

polymer

chain

a located at

R)

with an interaction

strength U(R~ R[ ).

The

polymer

chains

comprising

the melt interact with each other

through

a

potential

V

(R) R().

A

generalisation

to chains of

different

species

is

quite straightforward

and will not be

pursued

here. The

polymer-impurity potential U(R)

is

expressed

in terms of Fourier components as

U(R)= n~l£U~expiq.R (2,I)

q

where n is the volume of the system.

The total

polymer-impurity

interaction energy can be

expressed

as

z U(R~ R) )

= n l

z U~

exp

iq. (R~ R)

a. n wn

=

z U~

p

~ exp

iq R~ (2.2)

qn

(4)

yawn «cMh

«

'°'"'~~

($~~nti

at

h

ga vj

i~

R~

S ~~

~~

l' /~

jJ(t~-11')

~

'

~m ant

@ ~

R,'

f~t$~~~$~n ~ ~

Fig,

I.

Impurities

located at fixed

points

R~

interactblg

with a

polymer

melt

through

an interaction

potential

U. The polymer molecules are

interacting

with each other

through

a

potential

V.

where

p_~

is the Fourier component of the

polymer density

fluctuations

given by

p_ ~ =

n l

z

exp

iq. R) (2.3)

In tennis of the collective

density

variables

(p~)

the interactions between the chains is

given by

z

v

~ p

~ p_

~

(2.4)

q

The

partition

sum Z

(R~)

and hence the free energy In Z

(R~), depends explicitly

on the

position

of the

impurities (R~)

and is

given by

Z

(R~)

=

exp

~

z V~ p~

~ +

pq

U

~

£

~~~ ~~

~"

+ c,c.

(2.5)

kT

q «

Jl

o

The

averaging (.. )~

is over the

unperturbed

chain

configurations (R)),

which are contained in the collective

density

variables

(p~).

The RPA for

polymer

melts consists in

directly averaging

over these variables

by treating

them as Gaussian random variables. In

general

a Gaussian distribution function is deternlined

by

the second moment correlation

function,

which in this case is

given by (p~ p~)

~, where

(p~ p~)~

=

z (exp (q. R[

+ k.

R$))

~

«flss'

=

£ (exp

I

(q

R ~ + k.

Rfl ))

~

(exp

I

(q r)

+ k

r$ ))

~

(2.6)

aflss'

Here

R)

=

R~ +

r)

and the

position

vectors

(R")

of the chain ends are

independent

of the

configurational

vectors

(r~)

(5)

In a melt the system is

translationally invariant,

where each chain can be

anywhere,

hence the first ternl on the

right

hand side of

(2.6) gives

(expi(q.R~ +k.Rfl))~= 8(k+q)/n. (2.7)

This result

(2.7)

is the crucial

point

of

departure

when we come to consider chains where one

end is fixed.

Only

when the average over the

quenched

variables

(denoted by

a

bar)

is

perforated

does the

system

show translational

invariance,

I-e-

exPi(q.R" +k.Rfl)= n-iexpi(q.~Ra -Rfl))&(k+q). (2.8)

In the

present

case of an annealed melt the distribution of

density

fluctuations P

(p~)

is

given by

the Gaussian distribution :

P

(p ~)

exp

l~ z (~~ (2.9)

q q

where

S(

is the structure factor for

non-interacting chains,

I-e- V

= 0.

S(

=

£ (exp iq (R[ R) ) (2,10)

~

ass'

~

Using

the distribution function

(2.9)

and

(2.5)

for the

partition function,

the free energy functional F

(p~

;

R~)

for the

system

in the presence of the

impurities

can be identified as

F

(p~

;

R~)

m ~

£ i

p~ ~~

+

~~

+

p~ U_~ £

~~~ ~~

~"

+

.c,j

(2,I1)

2 s

~

n

The

density fluctuations,

which

depend

on the

positions R~

of the

impurities,

are then calculated in the usual way as

fl ldp~[p~[~exp

F

(p~ R~) llPqlRni l~)

=

~

(2.'2)

fl dp~

exp F

(p~

;

R~)

~

The

integrals

are standard Gaussian

integrals

and

give

for the full structure factor

S~(R~) =Sf+ (SfU~(~n~l£expiq. (R~-R~) (2.13)

where

~o

Sf

= ~

~

(2, 14)

+

'~q Sq

is the structure factor for the pure melt. The average over the

position

of the

quenched

variables

(R~)

is

straightforward

to do and

gives S~ (R~

=

Sf

+

Sf U~

~ C

(q) (2, 15)

(6)

where

C(q)

is the structure factor of the

impurities C(q)

=

n

£

exp

iq. (R~ R~ ) (2.16)

To discuss the result

(2,15)

we use a random distribution of

impurities

so that

C(q)

= the

impurity

concentration

= q~ and an

approximation [4]

for

Sf

which is valid for Gaussian chains of N segments and step

length

b. When N is

large

:

~

£~

12 c

(2,17)

~~ b~q~ f~+

where

I

is the Edwards

screening length f~

=

b~/(12 cV) (2,18)

and c is the concentration of

segments.

For

simplicity

we take the

impurity polymer

interaction

U~

to be the same as the

polymer polymer

interaction

V~

and consider the

particular

case when

i~

=

b~/12

then the result

(2,15)

for the

density

fluctuations in the

polymer melt/quenched impurity

system can be

written as

~~ i~ni

q2 ~l

+ i +

q2 ~'

+ i

~ (2. '9)

The

impurities

contribution

according

to their number but

acquire

a

scattering

structure

(form factor)

due to the interactions with the

polymer

chains. This is reflected in the

spatial

variation of the

density fluctuations, S~(R

;

R~),

which is

given by

the fourier transform of

(2,19)

as

S~

(R

R

~)

=

f I

exp

(- R/f )

+

I

exp

(- R/f )j (2.20)

f

R

f

The

polymer density

fluctuations associated with the

impurities

are different from those in the pure melt.

In the next section we will consider the case of a melt of

polymer

chains where one end of each chain is anchored in space, I-e- the

position

variables R~ for each chain end are now the

quenched

variables. We will show that the fixed ends behave

similarly

to the frozen

impurities

we have

just considered,

however there are also

significant

differences. The situation of the anchored chains is

sufficiently

more

complicated

that a

replica

calculation is necessary.

3. A melt of anchored chains

In the

simple

case of

non-interacting

chains there is

already

a

major

difference in the

density

fluctuations of a

single

anchored chain

compared

to one which can be

anywhere

in space.

Before

dealing

with the effect of interactions and the method of

handling quenched

chain end variables it will be useful to consider this case first.

3, I DENSITY FLUCTUATIONS FOR A NON-INTERACTING ANCHORED CHAIN. We will denote

by lt$

the anchored end of each chain a and let

r[

label the s segment of the chain from this

end,

hence

R[

=

R$

+

r[. (See

also

Fig. 2).

For a

non-interacting chain,

with one end

(7)

anchored at the

origin,

the

unperturbed

structure factor

A(

is

given by Al

= w

(ll~ql~lo- l~q>ol~)

where

~~

=

£

exp

iq r[ (3, I)

s

and q~ is now used to denote the number

density

of chains.

For the annealed case, where the

single

chain that can be

anywhere

in space, the second tern

(~~)~

= 0 for q # 0 as a result of translational invariance of the

system.

The non zero

value of this term for the

unperturbed

anchored chain will account for the

principal

difference between the melt and the anchored

« melt » of

interacting

chains. It will be useful to have

explicit

model

expressions

for

(~~)~

and

([~~[~)~.

These are

readily

found

using

the Gaussian chain model as follows :

z (exp iq r))

~ =

z

exp q

~(r)~) ~/2

s s

N

= ds exp q~b~ s /6

o

~

~2 ~2

p~

= j

(I

exp

(3.2)

q b 6

and

similarily

z (exp iq (r[ r[, )) N

~ =

ds ds' exp

q~ b~ Is s'[

/6

ss. o

=

~~

ii

~

~ (l

exp

~~~~~ (3.3)

q~b

q b N 6

Using (3.2)

and

(3.3)

in

(3,I)

for

A( gives

o

12Nq~

6

j2j q2b~N q~b2Nj

~q

~

2

~2

+ 9'

fi

~ ~ ~~P ~ + eXP ~

(3.4)

The

principal

features of this result are

given by

the

following

limits : at q values

large enough

to

explore

the intemal structure of a

single

coil I,e,

q~b~N WI,

there is no

significant

difference between the fixed chain and one which is free to be

anywhere

:

o 12

Nq2 ~o (3 5)

~q

"

~2 ~2

q

The difference between

A(

and

S(

first appears at

larger

scales

beyond

the size of a

single

coil

(I,e,

q values :

q~ b~

N «

I).

The first two

leading

terms in

(3.4)

cancel and

tends to zero as

A(

m cN

(q~ b~N/9 (3.6)

whereas in the melt

S(

= NC at q = 0

(3.7)

c is the concentration of monomers, c

=

Nq2.

(8)

For later use it will be necessary to have a

simple analytic

forms that

parameterise

both

A(

and

(~~)~.

Based on

((3.5), (3.6)

and

(3.2))

suitable forms are

~~

2 cN

(q~Nb~/6)

~

[(4/3)~~~

+

q~Nb ~/6]~

(~~)~

=

~

(3.8)

1 + q Nb /6

In order to facilitate the discussion of our results it is

slightly

inaccurate but

qualitatively

convenient to

replace

the numelical factor

(4/3)~'~-

l and to define

a scaled variable

Q

~

= q~N b

~/6,

then

~o

~

2

CNQ

~

~

li

+

Q~l~

and

(~~)~

=

~

~.

(3.9)

+

Q

In the next section we will consider the effect on the

unperturbed

structure factor

A(

of both the interactions between the chains and the presence of anchored chain ends. In this case the chain ends must be handled

by

the

replica

method first before the usual random

phase approximation

can be

applied.

3.2 REPLICA CALCULATION FOR THE DENSITY FLUCTUATIONS OF ANCHORED CHAINS.

The situation is shown

schematically

in

figure

2.

« chain end fixed at

it

R~

V(R) R)

ii,

s

j

chain end fixed at

R)

Fig. 2. A melt of

interacting

polymer chains with each chain a having one end anchored to a fixed

point R(.

The free energy for the

interacting

chains is

given by

F

(R()

= In Z

(Rl)

(9)

where the

partition

sum Z (ROY) is

given by

n I

Z

(R()

= exp

z z V~

exp ik.

(R(

+

r[ R( r( ) (3.10)

~

k a, s 0

fl,s'

The

averaging represented by (. )~

is done over the

unperturbed

chain

configuration

variables

(r)),

while the

(Rl)

remain as

quenched

variables. The

(r))

averages are

evaluated

using

a method devised

by

Edwards

[3, 4], whereby

a transformation is made from the

polymer

chain variables

(r))

to the collective variables

(p~) representing

the

polymer density fluctuations, given by (2.3).

This is

accomplished using

the

identity

element

I

=

fl ldp

~ 8

(p

~

£

exp ik.

(R~

+

r) )) (3.

II

)

~ ~ ~

Furthermore in order to find the free energy

fl averaged

over the

quenched position

of the anchor

points (RY )

,

fl

= F

(RY)

the

replica identity

is used for the In term tern, I,e.

fl

=

In Z

(RY)

=

lim

jj [Z "(RY) Ii.

(3,12)

n~o ~

«=i

The details of the calculation are

presented

in the

Appendix

where it is shown that the

quenched

free energy F can be written as :

I + nF

=

j fl dpf

exp

~

z pf(Qij,(k)) pi' (3.13)

2

k««,

where

Qij,(q)

is a

replica

space matrix with the

following

structure :

Qil,(q)

=

[Aj

l

C~ T~ U]~~> (3,14)

1is a unit matrix in the

replica

space and U is a matrix with all the matrix elements set

equal

to I. The coefficient

Aj

is

given by

Aj

=

A(~

+

V~ (3,15)

A(

is the

unperturbed

structure factor for an anchored

chain, given by (3.I), V~

is the interaction

potential

between the

chains, C~

is the

quenched

structure factor of the chain ends

given by

Cq

" ~l

i

eXP

iq [(R~ R~')]. (3,16)

(For

a random distribution of chain ends

C~

= q2) and

T~

=

(~~)~/A(. (3.17)

(10)

From

(3,13)

the structure factor

S~ (R~)

for the melt of anchored chains is

given by

the relica method as

S~

(R~)

=

n

p~ ~)

=

lim

z Q

««

(q ) (3,18)

n~0 ~

~i

The

replica

space matrix

Q

is

readily

inverted

(the

details are

given

in the

Appendix)

to

give S~ (R~ )

= A~ + A~

T~

~ C

~

(3. 19)

The result

(3,19)

is a

major

result of this paper and will be discussed in detail in the next section.

3.3 DENSITY FLUCTUATIONS IN A MELT OF ANCHORED CHAINS. The result

(3,19)

for the

structure factor of a melt of anchored chains is

formally

similar to that obtained in the last

section for the molten

polymer

in an environment of

quenched impurities, through

the

replacements

:

S(

~

A(, Sf

~ A~ and

U~

~

T~ (3.20)

In the result

(3,19)

there are

again

two kinds of

density

fluctuations : those associated with the

« molten »

degrees

of

freedom,

described

by

the tern A~, and the others

directly

related to the

distribution of the

quenched

chain ends

through

the chain end structure factor

C~

but

enhanced

by

the ternl

[A~T~[~.

The denominator of the term A~ has more structure as a

function of q than the

corresponding

melt case

given by Sf

and

gives

rise to two correlation

lengths.

To see this the

analytic

fornl

(3.9)

is used for

A(

and from

(3.15)

A~ is

given

as

~ 2

CNQ~

(3.21)

~

jl

+

Q~]~

+ 2

cV~ NQ

~

The denominator can be factorised as

(Q~+

a

~)(Q~+ p ~),

where for

cV~

N »1

a

~

= 2

cV~

N and

p

~

=

l/(2 cV~ N)

= a ~

(3.22)

In a melt the

screening leigth

f

(Eq. (2,18))

is of the same order as the monomer

length

b. Hence 12

cV~

m

I,

and

a

~

m N and

p

~

m N For the range

Q

~ ~

p

~

m N A~ can be

written as

A = ~ ~~

=

~~ ~~~~

(3 23)

~

(Q~

+

« 2)

(q2

+

f~~)

This is the same as the melt structure factor

Sf (Eq.(2,17))

where f,

given by

equation (2,18),

is the usual

screening length

found in

polymer

melts

[4].

The condition

Q~~ p

2

implies

q

~

(Nb )~

i,e, the interactions between the anchored chains establish near melt conditions

(A~

m

Sf)

out to scales of the order of the

length

of a

fully

extended chain.

This is in contrast to the

unperturbed density

fluctuations

A(

and

S(

which differed on scales of the order of or greater than the coil radius

(~

N

~'~b).

The

specific

contribution of the anchored chain ends to the

quenched

structure factor S~

(R~)

is

given by

the second ternl of

(3.19),

I,e.

C~

A~

T~

~. For a random

distribution,

the

(11)

chain end structure factor

C~

= q2, while

T~

is

given by (3,17),

hence

C

~ A~

T~

~ =

~ ~~~

~

(3.24) (1

+

V~

~

Using

the

parameterisations (3.9)

for

~~

and

A(

~q'~q~q'~ ~~/2)~~))j~~~p2)j2

~~'~~~

as

Q~0 C~[A~T~[~~Nc

whereas the first ternl of S~

(R~)

I,e. A~ - 0 in the limit q

- 0. Hence the

specific

contribution

(2.25)

from the anchored chain ends dominates the

density

fluctuations in this limit. This is in

contrast to the

quenched impurities

in a

polymer

melt. The two contributions to

S~(R~)

are

comparable, using (3,19)

when

A~ =

C~[A~ T~[~

This occurs for a q value q~

~~ ~~~~2

~~~

i~~2

~

fi~

~2

~2 ~~'~~~

L0Gfli

N=100

ANCHORED CHAIN

MELT 0

extended chain

)

single

call

<~

-2

4 0 4

LOG(

q~b2N16

q ~

b1

~ ~

l/l

b)-1

~~3/4 ~~-1

q~ixui-i

Fig.

3. The structure factor S~

(R~)

of a melt of

interacting

anchored chains calculated

using (3.21)

with N 100. For comparison the

corresponding

melt case is also shown.

(12)

q~ defines a scale internlediate between the size of a

single

chain coil

(N

~'~

b)~

and the

fully

extended chain

~

(Nb)~

The full behaviour of S~

(R~),

based on

(3,19)

and

using (3.21), (3.25)

for N

=

100 is shown on a

log-log plot

in

figure

3 and confirnls that the melt-like conditions for the

density

fluctuations in an anchored chain system extend a factor of N ~'~

beyond

the size of a

single

chain even

though

each chain is anchored at one end.

The

spatial

behaviour of the

quenched density

fluctuations

S(R R~)

is

given by

the Fourier transfornl of S~

(R~) using (3.21),

the

leading

tennis can be written as

S

(R

;

R~)

=

~~

) i exp(- R/f

+ c exp

(-

6

Rf/(Nb~)) (3.27)

b

~

R N b

If,

as in the

quenched impurity problem

of section

2,

we choose

f~= b~/12,

and set

c =

Nq2,

where q2 is the concentration of chains/chain

ends,

then

The result is very similar to that obtained for

quenched impurities

in a

polymer

melt

(cf. (2.20)).

The first ternl is the pure melt ternl, while the second ternl is due to the fixed chain ends instead of the

impurities.

In this case the form factor associated with a chain end has a range of the order of the

length

of the

fully

extended

chain,

whereas in the

impurity

case

it was of the order of the

screening length f.

The full

spatial

behaviour of S

(R

;

R~)

based on

equation (3.28),

is shown in

figure 4,

LOG( SIR;R~

lf3Jc)

~ =joo 4

1 ~

i-1

°

~ R~l

exp(- R/j

-I ANCHORED CHAINS

~

exp(- R/lNf)

MELT

-4

-1 -1 o i 1

I

L0©(

1/f)

~

i =

Xl

g =

j

fl z fil~l

( )

coil

) extended chain

Fig. 4. The structure factor S

(R

;

R~)

of a melt of

interacting

anchored chains calculated using

(3.28) with N

= 100. For

comparison

the

corresponding

melt case is also shown.

(13)

This

completes

the discussion of the

homopolymer

melt. In the next section we consider the

case of a blend of two distinct

polymers (A/B).

The unusual structure of the A~ term and the appearance of two correlation

lengths

manifests itself in this case with the

possibility

of

microphase separation

over

spatial

scales with an upper and lower bound. This can be

anticipated

since for

incompatible polymers

the fixed chain ends will prevent a

macroscopic phase decomposition.

We will show that the

stability

criterium is very similar to that of an

unrestricted melt of an A-B diblock

polymer.

4. A blend of anchored chains.

The methods of the

preceeding

sections are

easily adapted

to an A/B

polymer

blend where

one end of each of the chains is fixed in space. Collective

density

variables

pA~,

p~~

for the two

species

of

polymer

chains can be defined in a similar way to

(2.3)

and the interactions between the chains written as

( Z Iv

~~q p~q p~

q +

vBBq pBq

pB

-q + 2

v~~q p~q

p~

qi (4. i)

q

The free energy of the system is calculated

using

the

previously developed

methods and is

given by

a

generalisation

of

(3.13)

as

' +

nfl

=

In dPi~ dpi~

exp

( Z iPi~ toil« (k)) pit

k««'

+

Pik (QB~«'(~)) Pii

~ ~'AB

P~k Pik ~««'l (4.2)

where

QA~«'(k )

"

(d~q )

+

~AA ~««'

~

Ak

~Ak

~

~««'

and

similarily

for

Qj/~,(k).

For

simplicity

we can consider an

incompressible system

and set

p(~

= p

~j.

Then

+

nfl

=

lfl dp (~

exp

~

£ (p (~ (Qjj>(k))

p

it

2

k««'

where

~"~'~~~ ~~~~q)

+

(dBq)

2

XF/C) 8~~,

~~

Ak

~Ak

~ +

~Bk TBk ~l U««, (4.3)

and

X~

is the usual

Flory

interaction parameter

given by

X~/c

=

V~~ (V~~

+

V~~)/2

and c is the total concentration of monomer units : c

= c~ + c~. The

quenched

structure factor

l~

for the blend is determined as

~~ j~ ~j~ ~""~~~

j(A(~ )~

+

Aj~)~

2

X~/c)

CAk

~Ak

~ + ~

Bk

~Bk

~

(

~_

~)

~

(A(~ )~

+

(A(~ )~

2

X~/c)

~

(14)

The limits of

stability (compatibility)

are then determined

by (A(~ )~

+

(A(~)~ ~X~

= 0.

(4.5)

c

The

simple

case where

N~

=

N~

=

N will be considered

together

with the

approximate

fornl

(3.21)

for

A(. Equation (4.5)

becomes

~~~f~~

+

-~x~=o

NQ

~A CB C

or

(1+ Q~)~-4 q2(1- q2)NX~Q~=

0

(4.6)

where q2

=

c~/(c~

+

c~)

The roots

Q~(± )

of this

equation

deternline the

spatial

extent of the

stability

of the

quenched system. They

can be written as

Q~(± )

=

(xl 1)

±

~ (4.7)

where

X/=2q~(1-q~)NX~.

Various ranges of

Xl

must be considered :

(a) Xl

< 0 : in this case

Q~(± )

are

negative

and the fluctuations are

damped

and the

system

is stable on all scales.

(b)

0

<

Xl

< 2 : the

Flory

interaction energy is

repulsive,

however

Q~(± )

are

complex

and the

quenched

blend will show

damped

but

periodic

fluctuations of concentration but will still be stable on all scales.

(c) Xl

~2 :

Q~(± )

are real and

positive

and for

Q~(-

<

Q~<Q~(+ )

the

quenched

blend will be unstable and

microphase separation

will occur over the

rangle

set

by

Q(±

)_

Microphase separation

first occurs on a scale

given by

the size of a

single chain,

since when

Xl

=

2, Q~(± )

=

q~Nb~/6

=

1.

This situation and the criterium

(4.6)

are very similar to an annealed A-B diblock

polymer

melt.

Briefly,

if

S(~

and

S(~ represent

the

unperturbed single

chain structure factors of the two separate blocks and

S(B

the inter-block structure

factor,

then the random

phase

approximation

deternlines the full structure factor

S~~

of one block

[20, 21]

as

~Al

~

~~~

~

~~B

+ 2

S(

~~A S(~

SO

~ 2

X~

~~~

(4.8)

For a diblock

polymer

of blocks A and B with identical structures I,e.

S(A

=

SIB,

then

(4.8)

can be written as

Sil

=

2

(Dp X~) (4.9)

~'~~~~~

~~

"

S~A S~B

(15)

has

exactly

the same structure as

At

the structure factor of

a

single

anchored chain. The

stability

criterium

(I X~

D

°)

= 0 is then identical to

(4,5)

obtained for the

quenched

blend.

5. Anchored chains in a melt.

For

completeness

the

following

situation will also be considered : one

species (A)

of

polymer

chains are anchored in space while the other

species (B)

fornl an unconstrained melt. In this

case all the tennis

relating

to the annealed B chain variables will be

diagonal

in

replica

space.

The method is

entirely

similar to the last section on it is

only

necessary here to quote the result for the

quenched

free energy of an

incompressible

blend :

I +

nfl

=

lfl dp ~j

exp

~

z (p ~j (Qpj>(k )) p11 (5. I)

~

k««>

where

Qp/,(k)

is a

straightforward

modification of

(4.3),

I,e.

Qpj,(k)

=

(A(~ )~

+

(S(~)~

2

cX~) 8~~> C~~ T~~

[~ U~~>

(5.2)

The structure factor is

deternlined,

as

before, by

§q

# lint

£ Q~~(q)

#

n~0 ~

~l

j(A(~)-

i +

1(~)-

i

~x~j

~

j(A(~)-

i[iii l

~x~j

~'

~~'~~

c c

The limits of

stability (compatibility)

are then deternlined

by

(diq)

+

(S~q) XF

" °

(5.4)

Again

the

simple

case where

N~

= N

B = N will be considered

together

with the

approximate

form

(3.21)

for

A(. Equation (5.4)

becomes

('+Q~)~ (l+Q~)

2

CAfifQ~

~

C~N C~~

~'

(5.5)

The roots are determined

by

an

equation

very similar to

(4.7) Q~(*

"

~~Nb16

=

(Xl iPB)

±

~/(X/ iPB)~

4 q2B

(5.6)

Recall that q~B is the concentration of the unrestricted melt

component. Again

the blend shows three

regimes

of behaviour

depending

on the value of

Xl (or X~)

:

(a) Microphase separation

occurs when the square root in

(5.6)

is real I,e.

XI~q~B+2q~('~

(b)

The blend is

everywhere

stable but shows

damped periodic

concentration fluctuations when

iPB~~9'#~*~/*iPB+~9'~~~

(16)

(c)

The fluctuations are

completely damped

when

X/<q2~-2q2('~

6. Discussion.

In a

polymer

melt or blend where each chain has full

configurational

and translational freedom the

density

and concentration fluctuations are

satisfactorily

accounted for

by

the

random

phase approximation (RPA).

The translational freedom of the chains

plays

an

important simplifying

role in

ensuring

that

density

or concentration fluctuations

belonging

to different wavevectors are not

coupled.

However there are many

polymer systems

of interest where the translational freedom is absent. To

investigate

the effect of the loss of this

degree

of

freedom,

we have considered a

simple

case of a melt of

polymer

chains where one end of each chain is anchored to a fixed

point

in space. The average over the distribution of the

quenched

chain end variables was achieved

using

a «

replica

space » calculation

together

with the usual RPA for the annealed

configurational degrees

of freedom.

An

explicit analytic

result has been

obtained,

which shows that there are two distinct contributions to the

density

fluctuations : one, which does not

depend

on the distribution of the fixed chain

ends,

is identical to the pure melt situation out to scales of the order of the

length

of a

fully

extended chain I,e. Nb. The other contribution

depends explicitly

on the distribution of the chain ends and is similar to the

scattering

fornl factor in the

quenched impurity polymer

melt

problem.

The

density

fluctuations from this ternl dominate those of the first ternl on scales q <

(N ~'~b)~

~.

We have also considered a

binary

blend of chains with anchored chain ends and shown that when the

Flory

interaction

parameter

exceeds a critical value

(similar

to that in the nornlal

blend) microphase separation

occurs on a scale of the order of R.

Again

the situation was

shown to be very similar to an unconstrained melt of an A/B diblock

copolymer.

The

major

conclusions from this work are that the

density

and concentration fluctuations in the

quenched systems

we have considered are

virtually

identical to those in the unconstrained

melts, despite

the loss of the translational

degree

of freedom. Whilst this

might

have been

anticipated

on scales of the order of the size R of a

single coil,

the

surprising

result from this work is that the conclusion extends to much greater scales N

~'~R. Only

on these scales does the loss of the translational freedom become

apparent.

Appendix.

Using

the

identity (3. ii) parameterised

as

=

fl ldp

~ dq~~ exp I

£

q~

~ p~

£

exp ik.

(R~

+

r[ ) (Al)

k k a, S

in

expressions (3,10),

the free energy for a

given

distribution

(RY)

of chain ends can be written as

F

(RY)

= In

lfl dp~

dq~~ exp I

£

q2~ p~

£ V~ p~ p_~

x

k k

~

q

x

exp

I

£

q2~

£

exp ik.

(R"

+

r)) (A2)

k

a,

s

~

JOUR~AL DE PHY~IQLE 2 ~ 5. MAh 1992

(17)

In order to average the free energy

F(RY)

over the

quenched position

of the anchor

points (RY),

I.e.

fl

= F

(RY) (A3)

the

replica identity

is used for the In ternl, I-e- In Z

=

(Z" I)

In as n - 0 to

give

+

nfl

=

fi lfl dpi dq2i

exp

z pi pi z V~ pi p5~

x

«=I k k

~

k

exp

I

£ pi £

exp ik

(R~

+

r)") .

(A4)

~

~

~

The second moment

approximation

is used to evaluate both the annealed and

quenched

averages in

(A4).

This consists of

replacing

(exp

iX

)

exp

j (X~) (A5)

where we

identify

X with

X =

£ q2f(£

exp ik.

(R~

+

r[~)) (A6)

~~ ~ ~

Then

(i~)

=

z q2f q2l' z exp[ik.R~

+ ik'. R~

](exp

ik.

r)"

ik'.

r$'"') (A7)

a,s

k'«' a',s'

Translational invariance of the

quenched

system ensures that

only

the ternl k

=

k' in the sums in

(A7)

will contribute. The annealed average is taken over

non-interacting chains,

the

single

chain, single replica

term a =

al,

«

=

«' is

separated

out and the average written as

(exp

ik.

(r)" r) "'))

~ =

(exp

ik,

r)") ~(exp

ik.

rl'"

~ +

+

((exp

ik.

(r[" rl'"'))~- (exp ik,r)")~(exp

ik.

rl'"')~) 8~

~

8~

~.

(A8)

The averages in

(A8)

do not

depend

on either the chain

(a )

or

replica («)

labels.

Using (A8)

in

(A7) gives

(i~)

=

z

q2

~

~

In ~z

exp

[ik. (R" R"')] z (exp

ik

r~)

~ ~ +

k««' aa' s

+ q2

z (exp

ik

(r~ r~,))

~

(exp

ik

r~)~ (exp

ik

r~,) ~)

8

~ ~

(A9)

ss,

n is the volume of the

system

and q2 is the concentration of chain ends.

The

density

fluctuations

At

of a

single

chain fixed with a fixed end are

given by

At

=

z (exp

ik.

(r~ r~,))

~

(exp

ik,

r~)~ (exp

ik

r~,) ~) (A10)

ss,

(18)

The

quenched

structure factor C

(k)

of the chain ends is

given by

C

(k)

=

fl~

~fiik. (R~ R~')i. (Al I)

Finally

set

£ (exp

ik

r~)

= ~k

~~~~~

~

Then

(Xi

can be written from

(A9)

as

(X~)

=

£ [q2~[~[C(k)[~~[~U~~,+A(8~ ~] (A13)

k««'

U~~

is a

replica

space matrix with each element set

equal

to I.

Using

the results

(A5)

and

(A13)

in

(A4),

the free energy for the

quenched system

is

given

as

1+nfl=- fi lfl dpfexp-~£V~pfps~

x

«=i k

2

k

i l~~'i

~XP

i 9'k'~~~ (k)

~k

'~ ~««'

+

di ~«, al

+ I

I Pi Pi (J~14)

k k««' k

The

integration

over the q2~ is a standard Gaussian and

gives

[det

M

]~

~'~ exp

z pf Mp/, pi' (A15)

~

k««'

where M is the

replica

space matrix

M~~, (k )

= C

(k )

~~ ~

U~~,

+

At 8~

~

(Al 6)

The matrix U

=

(U~~ )

has the property

U~

=

nU,

which enables matrices of the form h4=Al+BU

to be

inverted,

I.e.

_, , B

~

(Al?)

~~'~ "

I iA

+ nB In the limit n

- 0

~i«~'

~

d(

~

~, «'~ C

(k )

~

k

~

~~

~

~««'

and

[det

M

]~

~'~

= l

(Al 8)

Finally fl

=

fl ldPi

exp

Z Pi iQp1,(k)j pi'

k««'

Wh~~~

Q al'

~

(d~

+

~k ) ~«,

«' C

(k)

~k ~

Al

~

U~~, (Al 9)

Which is the result used in section 3.2.

(19)

References

[ii DE GENNES P. G.,

Scaling Concepts

in

Polymer Physics

(Comell

University

Press, Ithaca, 1979).

[2] FLORY P. J.,

Principles

of

Polymer Chemistry

(Comell

University

Press, Ithaca, 1971).

[3] EDWARDS S. F., Proc.

Phys.

Soc. 88 (1966) 265.

[4] DOI M., EDWARDS S. F., The

Theory

of

Polymer Dynamics

(Oxford

University

Press, Oxford, 1986).

[5] FISCHER E. W., Frontiers of Macromolecular Science, T. Saegusa et al., Eds. Blackwell Science Publications, Oxford (1989).

[6] STEPANEK P, et al., J. Chem.

Phys.

94 (1991) 4171.

[7] METER G., MOMPER B., FISCHER E. W.,

preprint.

[8] SARABAN A., BINDER K., Macromolecules 21

(1988)

711.

[9] DEUTSCH H. P., BINDER K.,

preprint.

[10] FELISBERTI M. I., STADLER R., Mackro. Chem.

Symp.

44

(1991)

75.

[11] BINDER K., FRISCH H. L., J. Chem.

Phys.

81 (1984) 2126.

[12] MILNER S. T., WITTEN T. A., CATES M. E., Macromolecules 21(1988) 2610.

[13] RAPHAEL E., DE GENNES P. G.,

preprint.

[14] EDWARDS S. F., MCLEISH T. C. B., J. Chem.

Phys.

92

(1990)

6857.

[15] BASTIDE J., BUzIER M., BouE F.,

Polymer

Motion in Dense

Systems,

D. Richter and T.

Springer

Eds.

(Springer Verlag) (1988)

p. 112.

[16] EDWARDS S. F., ANDERSON P. W., J. Phys. F 5 (1975) 965.

[17] MEzARD M., PARISI G., VmAsoRo M.,

Spin

Glass Theory and

Beyond

(World Scientific,

Singapore, 1988).

[18] EDWARDS S. F.,

Polymer

Networks, S. Newmann Ed. (Plenum Press, N-Y., 1971).

[19] PANYUKOV S. V., JETP 69 (1989) 342.

[20]

LEIBLER L., Macromolecules 13

(1980)

1602.

[21]

VILGIS T. A., BORSALI R., Macromolecules 23

(1990)

3173.

Références

Documents relatifs

Namely, we show that for any given Lasso regularization parameter λ and any observed data y belonging to a set of full (Lebesgue) measure, the cardinality of the support of a

Moreover, even if it is not stated here explicitly, Theorem 4 can be extended to unbiasedly estimate other measures of the risk, including the projection risk, or the estimation

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

This paper considers concentrated polymer systems where the translational degree of freedom of all the chains to move has been quenched, ln particular we consider the

Generally speaking, the partition function equation (5) gives the complete statistical mechan- ical description of the deformed lightly cross-linked polymer network.. Nevertheless,

Thus the binary collisions play a prevailing role in a polymeric coil; if B = 0 then the contribution in all the characteristics of the chain from these collisions vanishes

Vaiter S, Deledalle C, Peyr´ e G, Fadili J, Dossal C (2012b) Local behavior of sparse analysis regularization: Applications to

In the special case of standard Lasso with a linearly independent design, (Zou et al., 2007) show that the number of nonzero coefficients is an unbiased esti- mate for the degrees