• Aucun résultat trouvé

2D crystalline order and defects in a stack of membranes

N/A
N/A
Protected

Academic year: 2021

Partager "2D crystalline order and defects in a stack of membranes"

Copied!
17
0
0

Texte intégral

(1)

HAL Id: jpa-00247852

https://hal.archives-ouvertes.fr/jpa-00247852

Submitted on 1 Jan 1993

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.

2D crystalline order and defects in a stack of membranes

David Morse, T. Lubensky

To cite this version:

David Morse, T. Lubensky. 2D crystalline order and defects in a stack of membranes. Journal de

Physique II, EDP Sciences, 1993, 3 (4), pp.531-546. �10.1051/jp2:1993149�. �jpa-00247852�

(2)

Classification Physics Abstracts 87.20E, 6130J, 64.70M

2D crystalline order and defects bla stack of membranes

David C. Morse

(~)

and T-C-

Lubensky (~)

(~) Corporate Research Science Laboratories, Exxon Research and Engineering Co., Annandale, NJ 08801, U.S.A.

(~) Department of Physics, University of Pennsylvania, Philadelphia PA 19104, U.S.A.

(Received

18 November 1992, accepted 23 December

1992)

Abstract. We consider the stability of two-dimensional

(2D)

crystalline order within the membranes of a lyotropic Iamellar phase,

concentrating

on the effects of 2D crystal dislocations and thermal undulations. At lamellar spacings d less than a critical value, the 2D melting

transition is found to be essentially unaffected by the flexibility of the membranes, and thus

occurs at a melting temperature Tm near that of a single membrane on a rigid substrate. At

larger spacings, the interactions between membranes become weak enough to allow buckling of the membrane in a finite region around each thermally excited dislocation. This leads to a partial screening of the elastic interaction between dislocations, and acts to depress, but not destroy, the

melting transition. Thermal undulations act to soften the membrane and thus further depress Tm, which is predicted to vanish continuously in the limit of large d. We discuss implications

for recent experiments on biological membranes.

The

spatial

fluctuations of a membrane

are

strongly

aRected

by

its

degree

of internal order:

crystalline [1-3],

hexatic [4], and fluid membranes [5] all exhibit

qualitatively

dilserent confor- mational statistics.

Conversely,

the presence of

spatial

fluctuations

might

be

expected

to alsect the

thermodynamic stability

of

crystalline

or hexatic order. In the Kosterlitz-Thouless-Nelson-

Halperin-Young (IITNHY) theory

[6,7] of 2D

melting, crystalline

and hexatic order both melt via an

unbinding

of

topological defects,

via dislocations in the

crystalline phase

and disclina- tions in the hexatic. In a flat

membrane,

the elastic interactions between such defects grow

logarithmically

with distance. In a flexible

membrane, however,

the form of these interactions

can be

modified,

both

by

thermal renormalization of the elsective elastic constants and

by

the mechanical

tendency

of the membrane to 'buckle~ in response to defects. An isolated

crystalline membrane,

for

instance,

will buckle so as to confine the

crystal

strain field to finite

regions surrounding

each dislocation [8]. This elsect leads to a finite dislocation energy, and thus pre-

vents formation of

quasi-long-range (QLR) crystalline

order at any nonzero temperature. In

an isolated hexatic

membrane,

the

long-range

interactions between disclinations are

essentially

unaffected

by

the membrane's

flexibility

[9],

leaving

hexatic order stable at

sufficiently

low temperatures.

(3)

Experiments

to

probe

the internal order of a membrane

have, however,

never been conducted

on isolated membranes. More

typical

are

experiments

on lamellar

phases,

such as are considered here. Oriented stacks of surfactant

bilayers

have been observed to

undergo phase

transitions between the

two-dimensionally

disordered La

phase

and various ordered

Lp, phases. X-ray scattering

data for the

Lp, phases

is thus far consistent with either hexatic or

crystalline

2D order

[10].

In a recent

study

of water-diluted stacks of

bacteriorhodopsin protein membranes,

Shen et al.

[ll]

have observed

sharp

2D

quasi-Bragg peaks (as expected

for

QLR crystalline order)

that vanish at an

apparent melting

transition. The observed transition temperature in-

creases with

decreasing

lamellar

spacing, suggesting

that the intramembrane order is stabilized

by

intermembrane interactions. There is no

experimental

evidence of 3D

crystalline

order.

Motivated

by

these

experiments~

we thus consider a model

describing

a

lyotropic

lamellar

phase,

and examine the effectiveness of intermembrane interactions in

stabilizing

2D intramern~

brane order. We find that such

interactions,

when

strong enough,

can

effectively

suppress both dislocation-induced

buckling

effects and the

softening

effects of thermal

undulations, leaving

the 2D

melting

transition unaffected

by

the

flexibility

of the membranes. When interactions

are

weaker,

these elsects become

important

over an intermediate range of

length

scales but

are

always suppressed beyond a'flattening length' L~.

The

logarithmic

form of the dislocation interaction is thus restored for dislocations

separated by

distances

greater

than Lt. The renor- malized elastic moduli felt at these distances are,

however,

reduced from their bare

values,

due to

softening by thermally

excited dislocation

pairs

and

by

thermal undulations at shorter distances. The

crystal

melts whenever the renormalized moduli become too small to prevent dislocation

unbinding.

1. The model.

The

configuration

of a stack of

crystalline

membranes can be described

by

3D

position

vectors

B~(x), specifying

the the

position

of a mass

point

labeled

by

a 2D internal coordinate x and a membrane index n. The membrane conformations are then described

by height

fields

hn(x)

e I

B~(x).

Fluctuations of the fields

hn

are controlled

by

a smectic

elasticity

Hs =

~ / d~~lK(fi~hn)~

+

$(hn+i hn)~l (i)

n

where It and B are the smectic elastic

moduli,

d is the average lamellar

spacing,

and

fi~hn

=

fi~hn(x)/fi~~.

The

compression

modulus B is

generally

a

decreasing

function of the lamellar

spacing d,

and

can be

primarily

either

entropic [12-15]

or electrostatic

[15,

16] in

origin.

Both of these

assumptions

about the

origin

of B lead to a

algebraic dependence

on d,

B(d)

o~ d-v

,

(2)

at

large separations,

where v

= 3 if B is

entropic

and v = 2 if it is electostatic. The

bending

modulus It

originates

in the

bending rigidity

~ of an individual

membrane,

with K

=

~/d.

To discuss intramembrane

order,

we construct an effective

single-membrane

free energy for

height fluctuations,

which we define

by formally integrating

out all fields

hn

with n

#

0 in

equation (I).

Because Hamiltonian

(I)

is

quadratic

in

hn,

this process can be carried out

exactly, giving

an effective Hamiltonian which is

quadratic

in ho- This

single-membrane

energy

can thus be obtained

by simply inverting

the intramembrane correlation function

~~~~~~~ ~~~

_~~ ~ e(qzj~~+1(q4

' ~~~

(4)

where

h(q)

is the 2D Fourier transform of

ho(x), e(qz)

e

2[1- cos(qzd)]/d~,

and q is the

magnitude

of the 2D wavevector q.

Equation (3)

contains a natural crossover

length Lp

~w

(d~K/B)~/~,

referred to as the

patch length.

At

large wavenumbers, qLp

>

I, equation (3) yields

an

essentially single-membrane

spectrum

ih(q)h(-q))

=

$ (4)

where ~ is the

single

membrane

bending rigidity.

At smaller

wavenumbers, qLp

<

I, equation (3) yields

a spectrum associated with the collective undulations of the

smectic,

(h(q)h(-q))

=

( (5)

where r %

24i

is

an effective

field,

due to intermembrane

interactions,

that tends to

align

the membrane normals

along

I. The

patch length

is

conveniently expressed

in terms of the

single-membrane

parameters ~ and r as

Lp

~

W (6)

The crossover between limits

(4)

and

(5)

is somewhat

complex, and,

since it

depends

in detail

upon the

high-qz

behavior of

e(qz),

is

apparently

nonuniversal. A convenient

approximation

is

given, however, by simply taking (h(q)h(-q))

ci

(~q~

+

rq~)~~. Adopting

this

approximation

for the correlation

function,

and

adding

the internal elastic energy of a 2D

crystal ill,

we obtain

an effective

single-membrane

Hamiltonian

H "

d~~'l~(fi~h)~

+

~(fih)~

+

>"la

+

2~"lbl (7)

In the

above,

nab +

(fialL fiblL bob)/2

is the nonlinear strain tensor, and I and ~ are 2D Lam4 coefficients. A more

rigorous

way of

deriving

of

equation (7)

would be to first use the full anharmonic

theory

for the stack [17] to calculate renormalized parameters

I(R(q), lR(q),

~R(q),

and to

only

then calculate the fluctuation spectrum of a

single membrane, identifying rR(q)

e 2

1(R(q)B

and

~R(q)

+

I(R(q)/d,

where B has been assumed to be unrenormalized

[17].

Note that the elsect of intermembrane

interactions, represented by

r in Hamiltonian

(7),

is

qualitatively

different from that

produced by simply confining

the membrane between inflex- ible walls [14]. The

r(fih)~

term

explicitly

breaks rotational symmetry, but not translational

symmetry, as would the presence of

confining

walls. The effect of this term is

formally

identical [2, 3] to that of a tensile stress,

making

all of our conclusions

regarding

a membrane within a

lamellar

phase equally applicable

to an isolated membrane under tensile stress.

The

strength

of the elastic interaction between

crystal

dislocations is controlled

by

the 2D

Young's

modulus

y =

4»(1

+

~)

l +

2~

~g~

Dislocation

unbinding

occurs when the renormalized

long-wavelength

modulus satisfies

YR(q

"

0)b~

<

161rT,

where is the

magnitude

of a dislocation

Burger's

vector. The presence of

thermally

excited dislocations and

height

fluctuations

always

lowers the renormalized

Young's

modulus,

so the membrane can never remain

crystalline

at temperatures T >

Yb~/161r,

where Y is the bare modulus.

(5)

In the model

presented above,

we assume that the 2D translational order within a membrane is uncorrelated with the translational order in

neighboring membranes,

and that the system,

therefore,

does not resist

parallel

relative motion of the membranes. In reference

[13], however,

Toner showed that a shear

coupling

between membranes is a relevant

perturbation

of this de-

coupled phase,

in the renormalization group

(RG)

sense, whenever the individual membranes exhibit

QLR crystalline

order.

Physically,

Toner's argument

implies

that the intermembrane

correlations between

weakly-coupled

untethered

crystalline

membranes will exhibit a nonzero

long

distance

limit,

and that a stack of such membranes will thus behave at very

long length

scales like a true 3D

crystal.

The

crystalline decoupled phase

considered here is thus associated with an unstable

(but potentially physically important)

fixed

point

of the RG. Our

underly- ing physical assumption

when

considering

this

phase

is that the intermembrane

correlations, though long-range,

can be

exceedingly

small in

magnitude,

as is

suggested by

the

complete

absence of observable 3D order in Shen et al.'s

experiments.

The presence of a small

degree

of 3D

order,

and a

correspondingly

small shear

modulus,

is

expected

to

change

elastic behavior

only

at

length

scales

beyond

some

large

crossover

length.

Because the

melting

transition is

relatively

insensitive to the very

long-wavelength

behavior of the

stack,

we expect the

melting

temperature of the 3D

crystal

to

smoothly approach

that of the

decoupled

system as the

degree

of intermembrane

registry

vanishes. Formation of a weak 3D

crystal

can also be inhibited

by non-equilbrium effects,

such as an initial randomness in the

crystal alignment

of

neighboring membranes,

which

might

prevent 3D order from

forming

on

experimental

time scales

ii Ii.

The present model of

decoupled

membranes should thus

provides

a useful

description

of the

phase behavior,

if not the

equilibrium elasticity,

of

a real system of very

weakly coupled

membranes.

2. Dislocation mechanics.

The elastic interaction between

dislocations,

in the absence of

screening by

either

height

fluctu- ations or other dislocations, is

given by

the total, mechanical

equilibrium

energy of a membrane with two dislocations. This

equilibrium

energy can be obtained

analytically

[18]

only

when the

membrane is constrained to lie in a

plane,

but a

qualitative description

of the interactions in a

nonplanar

membrane can be obtained from a combination of

scaling arguments

and numerical simulation.

Consider first the behavior of an infinite membrane with a

single

dislocation of

magnitude

b at the

origin.

To describe behavior at distances r «

Lp

away from the

dislocation,

where

we expect ~ to dominate in

H,

we set r = 0. The

resulting single

membrane

problem

was

considered in detail

by Seung

and Nelson [8], who showed that the

equilibrium

values of the

height

and strain fields can be

expressed

in the

scaling

forms

h(r)

=

/7 i()), ua~(r)

=

( fia~(), «), (9)

where me

I/(1+ 2p)

is the Poisson ratio and

L~

-~

</(yb) (io)

is referred to as the

buckling length.

The

corresponding

energy is a function E

=

Yb~ll(Lb/b)

of

~ and

Yb~,

and is

independent

of the Poisson ratio. The

physical interpretation

of the

buckling length

is as follows: inside a

region

of radius

Lb

around the

dislocation,

it is

energetically

unfavorable for the membrane to buckle

sharply enough

to

substantially

reduce the strain. At

larger distances, Lb

< r <

Lp, (assuming

Lb <

Lp) buckling

becomes

favorable,

and

nab(r)

drops rapidly compared

to its value in a flat

crystal.

(6)

At

length

scales L >

Lp,

intermembrane interactions

represented by

r are

dominant,

allow-

ing

us to

ignore

the

bending rigidity

~. A

straightforward

modification of

Seung

and Nelson's arguments to the case ~ = 0, r

#

0

(see Appendix) yields

the alternative

scaling

forms

h(r)

=

WI()), ua~(r)

=

) ia~(), «), (11)

where

yb

(12)

Lf -~

j

will be referred to as the

flattening length.

Note that

Lr

increases as

d(~+")/~

with

increasing separation.

At distances

Lp

< r « Lr from the

dislocation,

intermembrane forces dominate in H but are still too weak to suppress

buckling.

At distances r >

Lf,

the membrane

finally

flattens out due to

interactions,

and becomes

increasingly planar

with

increasing

r

(see Fig.

lb).

The

prediction

of a buckled

region

that can extend to

lengths

well

beyond

the

patch length Lp

is a result of

treating

the

neighboring

membranes as a flexible

medium,

rather than

as inflexible walls.

A buckled

region

can thus exist around each dislocation if either Lb %

Lp

or

Lp

j~ Lt.

Note, however,

that these two conditions are

equivalent,

since LbLr

-~

L(.

The flat state thus becomes

mechanically

unstable in the presence of a

dislocation,

and

buckling

effects becomes

important,

whenever Lb £

Lp.

More

precisely,

dislocation-induced

buckling

can occur

only

when the dimensionless ratio

g +

Yb/@

-~

Lp/Lb (13)

exceeds some critical value gc. This

implies

that

buckling

will occur in a

given

system of membranes

only

when d exceeds a nonuniveral critical value

dc,

where

dc depends

upon ~ and Y and upon the function

B(d).

For d <

dc,

the 2D

melting

temperature is thus

essentially

unalsected

by

the finite

flexibility

of the membranes and becomes

roughly independent

of d.

To test some of these

scaling

arguments, we have carried out numerical simulations of dislocation-induced

buckling

on a finite

triangular

lattice. Our discrete Hamiltonian contains strain and bend contributions

Hy =

~~~j((r;j(-a)~

4 Iii)

~~ ~(~'

~~

' ~~~~

where r;j are vector

connecting nearest-neighbor particles,

fi; and

hi

are unit normals on

neigh- boring triangular plaquets,

and a is a

preferred

lattice

spacing.

We represent the additional

r(fih)~

contribution in H

by

an energy

HT "

/~ ~j(I

~ij)~

(15)

iiji

where I is the unit normal to the

preferred plane

of

alignment.

The numerical

prefactors

in

equations (14)

and

(15)

are chosen so as to

reproduce

the continuum limit of

equation (7),

with I

= p =

(Y.

Mechanical

equilibrium

conformations and

energies

have been calculated for

hexagonal

mem-

branes with L

particles

between the center and each corner,

using

L <

96,

each with a

single

dislocation of

magnitude

= a at its center

(see Fig. I).

If the membrane is forced to lie in a

(7)

la)

16)

Fig,I.

Dislocation-induced buckling: figure la shows the equilibrium conformation of membrane of size L = 16 with elastic parameters ~ = o,ois Yb~ and r

= o, representing an isolated membrane.

Figure

16 shows a membrane with the same value of

~/Y,

but r

= 0,00 is Y, representing a membrane confined by interactions with its neighbors. The dislocations

are visible as neighboring 7-fold and 5-fold coordinated particles in the center of each membrane. For r # 0

(i'b),

the presence of a bending rigidity primarily affects the membrane conformation within a small region, of radius r

~J

Lp,

around the dislocation, while the presence of a nonzero r controls the behavior at large distance§, r jS

Lp, causing

the membrane to flatten out at distance

r jS Li >

Lp.

For r

= 0

(la),

the

height

field apparently diverges at large distances from the dislocation. The vertical

(I)

scale is exagerrated in both pictures by a factor of 2.

plane (I,e., h(x)

=

0),

the strain energy associated with the dislocation scales with the system size L as

~

Eflat

ci ~~

~ln(Lla)

+

const]

,

(16)

81r

where the constant represents a dislocation core energy.

Simulations of

buckling

in isolated

membranes, corresponding

to the limit r = 0 in the

above,

have been carried out

previously by Seung

and Nelson [8].

Using

a discrete Hamiltonian identical to our

equation (14),

these authors found that the energy E of a buckled dislocation

approaches

a finite value as L goes to

infinity,

and that this

limiting

value is well

approximated by

that of a dislocation in a flat

crystal

of radius

Lb

lim E ci

~~

[ln(Lbla)

+

const] (17)

r-O

8~

As is clear from its

form,

this energy is due

primarily

to the strain energy contained within a radius r

-~

Lb

around the dislocation. An unconfined buckled membrane is shown in

figure

la.

When r is nonzero, as it is for a membrane within a

stack, scaling

arguments

suggest

that the

membrane should flatten out at

large

distances. This should

produce

an energy that

diverges

as

In(L)

in the limit of

large L,

but that remains less than

Eflat(L) by

an amount that increases with

increasing flattening length

Lt. In

figure

2, we show the calculated energy E as a function of

Y/r

~K

Lr/b

for membranes of fixed size L

= 48 and several values of

~/Yb~

~K

Lb16.

For

sufficiently

small values of

Li (I.e., large r),

the membrane lies

entirely flat, giving

an energy

E =

Eflat.

The membrane

begins

to

buckle,

and E

begins

to

decrease,

when

Lr

exceeds some critical value Lrc that

depends

upon the

magnitude

of Lb- For Lb < a,

buckling begins

at a value Lfc

-~ a. For a j~ Lb

L,

this

instability

occurs at Lfc

-~

Lb,

and for Lb >

L,

the

membrane remains flat at all values of

Lf.

Once

buckling begins,

the energy decreases

steadily

with

increasing Lf

until

Lf

2

L,

and then levels

of,

as r becomes too small to alsect the energy

(8)

a a O O & &

a ° &

~ O &

j

o &

~

o

<

~

~/Bn

O ~

Wo O

2

O

3

w

O

a

3

o

o-o m 4.o o-o e-o lo-o

In~i$/7) Fig.2.

Dislocation energy E of membranes of radius L

= 48 as a function of

Y/r

cc

Li16.

Squares represent membranes with ~ = 0, circles represent ~ =

o.o2Yb~,

and triangles ~ = o.o4Yb~. The solid line has a slope of

1/8~,

as indicated.

in this

regime.

For ~ = 0 and L~c j~ Lf j~ L, E is found to exhibit a

logarithmic scaling

behavior

lim E m

j~~ ~In(L/L~)

+

const]

,

(18) reflecting

the release of strain in area of size r ~K Lf. As ~ is increased from zero, the range

over which Lrc < Lr < L narrows

rapidly, making

it

impossible

to observe a clean

logarithmic scaling regime

in membranes of this size for ~ 2

0.01Yb~.

In

figure 3,

we show the

dependence

of the critical

flattening length Lrc16

~K

Y/r

upon

Lbla

-~

~/Yb~.

The

slope

of this

logarithmic plot

is

expected

to

approach unity (corresponding

to a constant value of the critical ratio gc =

Ybll%)

in the continuum elastic

limit,

a < Lb < L.

To check for finite-size

effects,

we have determined Lrc for the three

largest

values of

Lb/b

in membranes of two

sizes,

L = 48 and L = 96. The results obtained from these two sizes differ

significantly only

at the

largest

value of

Lb, indicating

that the variation from

linearity

at lower

Lb

should be attributed to finite

lattice-spacing,

rather than

finite-size,

effects. We can obtain

a

rough

estimate of the continuum value of gc

by extrapolating

the data for size L = 96 onto a line of unit

slope, giving

gc m 80

,

(19)

as indicated

by

the solid line in

figure

3.

(9)

a

a

p~ ~~

w

a

D lD

2+lJ~ ll~ ll~

K/Yb~

Fig.3.

Critical value of

Y/r

cc Lic16 as a function of

~/Yb~

cc Lb16. Squares represent membranes of size L

= 48 and triangles L

= 96. The solid line is a line of constant gc =

Ybll%,

with an

extrapolated value of gc

= 80.

3. Dislocation mediated

melting.

In this

section,

we consider the effect of membrane

flexibility

upon the 2D

melting

temperature.

We first discuss the effect of dislocation-induced

buckling,

and then include the additional effect of undulations.

The presence of

thermally

excited dislocations softens the renormalized

Young's

modulus YR, and thus tends to destabilize the

crystal.

The

perturbative

correction to Y due to

thermally

excited

pairs

on a

triangular

lattice is calculated in reference [7], and can be written as

b2 oJ ~2~

~ 2

~~

~' ~

T

/~ a( ao~ ~~~~~

~~~ ' ~~~~

where E is the excitation energy of a

pair

of dislocations with

Burger's

vectors + b and

separation

r. The

angular

factor

A,

which is

specific

to the

triangular lattice,

is

given by A(9)

= [~ +

( sin~

0], where 0 is the

angle

between b and r. The energy E is the mechanical

equilibrium

energy of a dislocation

pair

and should thus include any effects of

buckling.

At

separations

r j~

Lb,

where the membrane is forced to lie flat

by bending rigidity,

E is well

approximated by

the

corresponding expression

for dislocations in a flat

crystal,

E m

(~ [ln(I)

+

sin~

0] + 2Ec

,

(21)

ao

(10)

E(r)

2Eb

2E~

ao Lb Lp Lf

~(r)

FigA.

Schematic plot of the dislocation pair energy E

as a function of

In(r),

where r is the dislocation separation. See text for definitions of various length and energy scales.

where Ec is a

microscopic

core energy. At

separations

Lb % r j~ Lf, this interaction is screened

by buckling, making

E

essentially independent

of r.

Here,

the

pair

energy can be

approximated

as a sum of

single-dislocation energies, E2

"

2Eb,

where the effective

single-dislocation

energy,

Eb " Ec +

j~~ In(Lblao)

,

(22)

is

roughly

the core energy

plus

the strain energy contained within a distance

Lb

around the defect. At

separations

r 2 L~, E must

again

take on the

logarithmic

form

(21) appropriate

to a

flat

crystal,

but with an effective core size ao

- Lf and core energy Ec - Eb. The

dependence

of E on

separation

is shown

schematically

in

figure

4.

The effect of

buckling

upon the

stability

of the

crystal

can be understood

using

a

simple coarse-graining procedure. First,

estimate the renormalization of Y~~

arising

from dislocations with

separations

r < Lf. Then use this renormalized value of Y~l

as the

input

in a coarsw

grained theory

valid at

length

scales L > Lf

beyond

which

buckling

is irrelevant.

Following

Nelson and

Halperin's

[7] treatment of the flat

crystal,

we define

a

coarse-grained Young's

modulus

YR(a),

and dislocation

fugacity yR(a)

e

exp(-ECR(a)/T),

where

ECR(a)

is a coarse-

grained

core energy, as functions of a renormalized core size a. It is also convenient to define a reduced modulus

VR(a)

e

YR(a)b~ IT.

At

length

scales a j~

Lb

and a 2

Lf,

where the

crystal

is

flat,

the renormalization group

(RG)

flow of

VR(a)

and

yR(a)

is well

approximated by

the KTNHY recursion relations for a flat

crystal.

Here we focus on the RG flow in the

regime

Lb % a j~ L~, where

buckling strongly

modifies the elastic interaction.

Following

now standard

procedures [19],

the

coarse-grained Young's

modulus

YR(a)

is defined

by integrating

the dislocation contribution in

equation (20)

up to a maximum

separation

r = a, while the core energy

(or fugacity)

is defined so as to restore the form of

equation (20)

to the

remaining integral. Treating

the dislocations as

completely non-interacting,

and

using

the

approximation

E m

2Eb

for the

pair

energy, we obtain

YRla)

Ci

e~~~/~l«lao)~ (23)

(11)

Vi~(a)

Ci

n~(Lb)

+

)lYi(a) Yi(Lb)1 (24)

appropriate

at

length

scales Lb a j~ Lf. For the sake of

simplicity,

we have included

only

the effects of dislocation

pairs,

as described

by equation (20), ignoring higher

order effects due to

triplets

that were included in reference [7].

The

quantity Yp~(Lb) appearing

in

(24)

can be obtained

by integrating

the KTNHY re- cursion relations up to a =

Lb, thereby taking

into account the contribution to

§~

from

dislocation

pairs

with

separations

r <

Lb-

This contribution can,

however,

be shown to be small at any temperature that is

significantly

below the

melting

temperature of the flat

crystal

[7].

Temperatures

near the

melting temperature

Tm of the flexible membrane will

satisfy

this

condition whenever

buckling significantly depresses

Tm.

In

equation (23),

the

integration

of a up to a = Lb has been carried out

explicitly, using

the

assumption

that

YR(a)

ci Y for a < Lb-

Allowing

for a renormalization of

YR(a)

at a <

Lb

would

simply

renormalize the effective excitation energy

Eb, corresponding

to the use of a

slightly

renormalized modulus in

equation (22).

As we see from the form of

equation (23), yR(a)

can be

interpreted physically

as the

probability

of

finding

an

unpaired, thermally

excited

dislocation of energy Eb within a core

region

of radius a. This

fugacity

increases

monotonically

because the strain energy contained within a radius a around the dislocation has been assumed to be

independent

of a. In a flat

crystal,

this strain energy increases

logarithmically

with a,

giving

a

fugacity

which can either increase or decrease with a

depending

on the balance of the

increasing

strain energy and the

increasing

translation entropy of the dislocation within a

redefined core

region.

The RG

trajectories given by equations (23)

and

(24)

are shown

by

the solid lines in

figure

5. The RG flow at

lengths

a j~

Lb

and a 2

Lg

is obtained

by grafting

the

trajectories

for this buckled

regime

onto the KTNHY relations for a flat

crystal, represented by

dashed lines. The

crystal

melts if the

coarse-grained

parameters

gj~ (Lf)

and

yR(Lr)

fall outside the flat

crystal's

limit of

stability.

Consider the

stability

of the

crystal

in the limit of low temperature, where

Pm

I and

Eb/T

>

I,

and

large Lt.

In this

limit, §~(Lb)

and

yR(Lb) always

lie

near the

origin

of

figure

5, and then follow

a

trajectory

which crosses the flat

crystal's

limit of

stability

somewhere near the

point

P. The

limiting low-temperature trajectory,

which connects the

origin

and the

point

P, describes the

physical

limit in which the dislocation contribution to

Yj~

dominates in

equation (20),

so that the

Young's

modulus

YR(Lr)

is determined

primarily by

the dislocation

fugacity

rather than

by

the

microscopic

modulus Y. In this

limit, V(~(Lr)

and

yR(Lr)

take on universal values

(corresponding

to the coordinates of

point P)

at the

melting temperature Tm. Setting

either

yR(Lr)

or

V(~(Lr)

to constants at temperature Tm then

gives

a

melting

temperature that varies as

)

ci

)[2 In(Lr lao)

+

const] (25)

m b

with variations in Lt. This

approximation

is valid in the limit of

large Lr/Lb>

where Tm becomes

significantly

lower than the

melting

temperature of the flat

crystal.

Since the

length

Lr increases

algebraically

with

d,

this

gives

a

dependence

Tm ~K

In~~(d) (26)

in the limit of

large

lamellar

spacing.

We now consider the effect of thermal

height

fluctuations. In an isolated

membrane, height

fluctuations cause an anomolous renormalization of the wave-number

dependent

elastic param-

eters

~R(q)

and

YR(q)

[1, 2]. These renormalized

quantities

become

significantly

different from

(12)

v~ia)

P

'

,$.

,

, ,' ,

'-~-" ,' ,'

,W' ,'

'

1/161r pjl(~)

Fig.5.

Schematic flow diagram Solid lines represent the RG trajectories for the buckled regime Lb < a < Li, where the dislocations are non-interacting. Dashed lines represent represent the KTNHY

trajectories for a flat crystal with logarithmically interacting dislocations, appropriate for

a < Lb and

a > Li. The shaded area is the the region of a stability of the flat crystal. The connected set of dashed

- solid

- dashed fines represent the full RG trajectory of a particular system. The trajectory shown represents a membrane slightly above its melting temperature.

their bare values ~ and Y

only

at

wavelengths greater

than a non-linear

length

L~i ~

~lli (27)

For

q~~

2

Lj~,

the elastic parameters in an isolated membrane [2] scale as

~R(q)

~W

(qLni)~°~~

YR(q)

~W

lqLni)+°"Y (28)

where rotational invariance

gives

[2, 3] the relation q~ +

2qh

" 2. Most numerical simulations

[20] have

given

qh Ci 0.8 + 0.I. In a

lyotropic smectic,

this

single-membrane scaling

behavior is

obeyed only

at

lengths q~l

shorter than the

patch length Lp.

At

lengths q~l

2

Lp, ~R(q)

and

YR(q)

become

roughly q-independent,

except for much weaker

logarithmic

corrections

[17].

Significant softening

of Y

by

thermal undulations can thus occur in the smectic

phase only

when

L~j

«

Lp.

A useful constraint on the

magnitude

of L~i is obtained if we compare L~i and

Lb, giving

Lni ~ Lb Yb2

IT.

As we noted

earlier,

it is

theoretically

consistent to

postulate

the existence of

crystalline

order

only

at temperatures T <

Yb~/161r, implying

that

Lni 2

Lb (29)

at all temperatures of interest. This

inequality

has several

implications; first,

it indicates that it is inconsistent to consider a situation in which undulation efTects are

important (Lnj

j~

Lp)

but

(13)

defect-induce

buckling

is not

(Lb

j~

Lp).

This

justifies

our earlier statement that the

melting

temperature remains near that of a membrane on a

rigid

substrate whenever dislocation- induced

buckling

is

suppressed, I.e.,

whenever

Lp

j~ Lb-

Second, equation (29) implies

that thermal undulations will have little effect on the dislocation interaction at

separations

r j~

Lr,

since this interaction is

already

screened

by buckling

at any

length

which is less than Lr and at which undulation effects are

important.

Undulations can,

however,

soften the effective

interaction felt at

separations

r

2 Lr,

where the

logarthmic

interaction is

restored,

and

thereby

further

depress

Tm.

The effects of thermal fluctuations upon the elastic constants of a membrane can be treated

by

various

analytic expansions

and

approximations

[1,

3, 21].

The best estimate of such elsects is

probably given by

the self-consistent

screening approximation

for tethered membranes

[21].

A

straightforward adaptation

of this

approximation

to include the

symmetry-breaking r(fih)~

term in H

gives

a renormalized

Young's

modulus

j$~R

~~~~ ~ ~

~~ /(~~2 ~R($i~~~~~k)k2]2

' ~~~~

where

P$

= bob

qaqb/q~.

The

practical

value of this

expression is,

of course, limited

by

its

dependence

on the renormalized

quantities rR(q)

and

~R(q),

which

should,

in

principle,

be determined

self-consistently

with

YR(q).

We note,

however,

that the dominant contribution to the

integral

comes from wavenumbers k £

Lpl,

where

~R(q)

and

TR(q) depend only weakly

on q. We will thus estimate the

integral by replacing

these parameters with their q = 0

limits, rR(0)

and

~R(0), giving

a contribution

~~~

~~°~ ~

4'rTR~~~R(0)

~~~~

The

long-wavelength

parameters that appear in

equation (31)

are difficult to calculate from

microscopic

parameters, but can, in favorable

circumstances,

be measured

experimentally [15].

When the undulations elsects are

small,

~R and rR can be

replaced by

their bare values in

equation (31), yielding

the lowest-order

perturbative

correction to YR. This undulation

con-

tribution to

Yj~

can, as a first

approximation, simply

be added to the

previously

calculated dislocation contribution.

It is

possible

for the undulation correction to YR to dominate over the correction due to

thermally

excited dislocations. This occurs for

Lp

2 Lni and

large

dislocation excitation energy

Eb,

where dislocation elsects are

suppressed by

an

exponentially

small factor

e~~~/~

When undulation effects are

dominant,

the

melting

temperature for dislocation mediated

melting

may be calculated

simply by setting VR(q"0)

to its critical value of161r. This

yields

a

melting temperature

which decreases

algebraically

with d at

large separation.

4.

Experimental implications

and conclusions.

Given estimates of the elastic parameters Y, ~, and B, our results can be used to estimate the effects of membrane

flexibility

on

crystal stability

in various

experimental

systems. Consider the

Bacteriorhodopsin (Br) membrane,

which is the

only

membrane in which 2D

crystalline

order has thus far been observed. This membrane forms a

triangular

lattice with

a rather

large

lattice

spacing,

a ci 63

I,

and a trimer of

protein

molecules at each lattice site

[22].

A

prelimary analysis

of Shen et al.'s

X-ray scattering

data

gives

an estimated

Young's

modulus of Y

~J

500-600Tla2

at T ci

65°C,

which is

roughly

10

degrees

below Tm. These

experiments

have thus far been carried out

using

pure water

dilution,

without added

salt,

and the intermembrane

(14)

interactions are believed to be

primarily

electrostatic in

origin.

The smectic

compressibility

B

can thus be estimated from the theoretical

prediction [15,

16] for an

electrostatically

stabilized lamellar

phase

with

large separations

and a

single sign

counterion:

B =

j

,

(32)

where dw is the water

separation

between

membranes,

e is the counterion

charge,

and le e

lre~/(eT),

with

le'~

19

I

for

a dielectric constant e = 80

appropriate

to water. Thin approx- imation is

appropriate

for

ledwela 2

1, where a is the surface

charge density

of one side of the

membrane,

and should be

appropriate

in the Br membrane for

separations

dw > 30

I.

We

have no

experimental

measure of the

bending rigidity,

but will assume that the

protein

lattice

is at least as

rigid

as a

typical single-species lipid membrane, suggesting

~ 2 30T.

We can use these estimates of the elastic parameters to estimate a critical water

separation

dwc for the onset of dislocation-induced

buckling. Using

~ = 30T and the estimate gc m 80 for

g, we find that g =

Yb/j%

reaches gc at a critical water

separation

of

dwc ~w 700

I (33)

Using

a

larger

estimate for ~ in the above

produces

a

proportionately larger

estimate for dwc.

While

extremely rough,

this estimate does indicate that dislocations can be forced to

lay

flat in this system even at

fairly large

water

separations, leading

to a rather stable

crystalline phase.

The

large

observed value of

V,

which is believed to be

roughly

ten times its critical value at temperatures

only

10 20 °C below

Tm,

also

suggests

that the

melting

transition in Br may not be

explainable

in terms of

a

simple

dislocation

unbinding

mechanism, but may involve some

change

of structure within the Br unit cell. The critical

separation

dwc of an

electrostatically

stabilized system can

readily

be reduced

by adding salt, thereby screening

the electrostatic interaction and

reducing B(d).

The minimum achievable value of B at a

given separation

is obtained when the membranes interact via steric

repulsion alone, leading

to the

entropic repulsion

described

by

Helfrich

[12].

What are the

implications

of these results for

simpler lipid membranes,

such as those studied in reference

[10]?

One

important

difference between

a

lipid membrane,

such as

DMPC,

and

a lattice of

large proteins,

such as Br, is the difference in lattice

spacing:

a is a full order of

magnitude

smaller in

lipid

systems, where a ci 4-7

I,

than in the Br lattice. The effect

of

decreasing

the lattice

spacing

in our

theory

is

(all

other

things being equal)

to decrease the

importance

of

buckling effects, by decreasing

the ratio g =

Ybll%,

and to

drastically

increase the

importance

of the conventional dislocation

unbinding mechanism, by decreasing

the

quantity V

= Yb~

IT.

If we assume that the bare

Young's

modulus Y of a 2D

lipid crystal (by

which we mean the modulus at

sufficiently

low temperature so that thermal excitation of dislocations is

suppressed)

is the same order of

magnitude

as that measured in the Br

membrane,

then we would obtain a critical

separation

in water of several thousand

I

and a

room temperature value of

V

that is

roughly

an order of

magnitude

below the critical value of161r. A much

higher

value of Y would thus be

required

to

produce

a room temperature 2D

lipid crystal

that is either stable with respect to

simple

dislocation

unbinding

or that is

mechanically

unstable with respect to dislocation-induced

buckling.

Since we see no reason for the intrinsic moduli of a

lipid crystal

to be

drastically higher

than those in

Br,

we

speculate

that any such

crystal

would

probably

melt via dislocation

unbinding

or some other

purely

2D

mechanism,

rather than via dislocation-induced

buckling.

In

conclusion,

we have shown how

crystalline

intramembrane order can be stabilized

by

intermembrane interactions within a lamellar

phase.

We have discussed the various mechanisms

Références

Documents relatifs

The thermal-induced changes in the shear modulus of ex vivo bovine muscles using ultrasound are consistent with theorical changes of myosin and collagen microstructure.. Given

They showed that the problem is NP -complete in the strong sense even if each product appears in at most three shops and each shop sells exactly three products, as well as in the

Aware of the grave consequences of substance abuse, the United Nations system, including the World Health Organization, has been deeply involved in many aspects of prevention,

The restriction on the dimension of the ambient space in the flat case comes from the problem that to start the flow, we have to replace a bounded set U ⊂ R n+1 with smooth boundary

Let σ be the scattering relation on a compact Riemannian man- ifold M with non-necessarily convex boundary, that maps initial points of geodesic rays on the boundary and

We introduce a family of DG methods for (1.1)–(1.3) based on the coupling of a DG approximation to the Vlasov equation (transport equation) with several mixed finite element methods

First introduced by Faddeev and Kashaev [7, 9], the quantum dilogarithm G b (x) and its variants S b (x) and g b (x) play a crucial role in the study of positive representations

The second mentioned author would like to thank James Cogdell for helpful conversation on their previous results when he was visiting Ohio State University.. The authors also would