HAL Id: jpa-00246425
https://hal.archives-ouvertes.fr/jpa-00246425
Submitted on 1 Jan 1991
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Fluctuations of a polymerized membrane between walls
G. Gompper, D. Kroll
To cite this version:
G. Gompper, D. Kroll. Fluctuations of a polymerized membrane between walls. Journal de Physique
I, EDP Sciences, 1991, 1 (10), pp.1411-1432. �10.1051/jp1:1991211�. �jpa-00246425�
J Phys. I Fmnce 1
(1991)
1411-1432 OmOBRE1991, PAGE 1411Classification
PhysicsAbs+acts
68.45Gd-05.70Jk-64.60Fr
Fluctuations of
apolymerized membrane between walls
G.
Gompper(~)
and D.M.Knoll(21*)
(~ Sektion
Physik
derLudmig-Maximilians-Universit3t
Monchen, Theresienstr. 37, 80W Monchen 2,Germany
(~) AHPCRC,
University
of Minnesota, 11WlAhshington
Avenue South,Minneapolis,
MN 55414, U.S.A.(Received3 June 1991,
accepted
3Ju~yI ml)Abstract. The fluctuations of a
self-avoiding, polymer12ed (or tethered)
membrane which is re- stdcted by the presence of twoparallel
hard walls are studiedusing
Monte Carlo simulations andscaling
arguments. tilescaling
behavior of the area, the pressure, the normal vectorsusceptibility,
as well as the mean curvaturesuceptibility
areinvestigated.
lAh calculate thescaling
functions and show that the critical behavior follows the predicted scaling laws vith an exponent q = 0.55 + 0.10.1. Iutloductiou.
OrR of the
major
currentchallenges
in theoreticalphysics
b to understand the statbtical me- chanics of surfaces and membranes(see,
e.g. Refs.[1,2],
and referencestherein).
This field hasattracted considerable attention
recently
because of itsimportance
inunderstanding
suchdiversq problems
as the structure of lamellarliquid crystals, microemulsions,
blockcopolymers
and otherself-assembling
structures[3,4],
cell membrane interactions inbiology
[5] and biomimetic systems, and world-sheetdynamics
instring theory [6].
The membranes of
physical
interest in chemical andbiological applications
can beregarded
astwo-dimensional
self-avoiding
su~fiaces imbedded in three-dimensional space.They
are built up ofmonomers which may be either free to diffuse in the membrane
surface,
or are frozen inplace
and thus form a network with fixed coordination. In the first case one
speaks
offluid
membranes[7-14],
and in the second, of tethered orpo~ymeri2ed
membranes[11,15-33].
Anotherimportant possibility
are hexaficmembranes,
with extended bond orientational order[34,35].
Because theirmicroscopic
surface tension is small or vanishesaltogether,
thestrength
of theout-of-plane
fluc- tuations in membranes h controlledby
the curvature orbending
energy[7,8].
Since the corre-sponding
elastic constant is often of the order ofkBT,
membranes exhibit wild fluctuations. Theimportance
and effect of these fluctuations b determinedby
the internal state of the membrane.(*)
Permanent address; Institut forFestk6rperforschung,
KFA Jolich, Postfach 1913, 5170 J01ich, Germany1412 JOURNAL DE PI IYSIQUE I N°10
Polymerized
self-avoiding
membranes have been shown to be flat atlong length
scales[23,25-27],
whereas fluid membranes are believed to be
crumpled objects
with a nontrivial fractal dimension[7-14].
The fact that membranes areself-avoiding
surfacescomplicates
any theoreticalinvestiga-
tion of even the
simplest questions regarding
the conformation ofsingle
bolatedmembranes,
and is one of theprimary
reasons that Monte Carlo simulations haveplayed
aparticularly important
role in this field of research
[23,25-27].
Since a tethered membrane is
always
flat, itsphysical properties
are veryanisotropic.
In order not to looseimportant information,
it is therefore necessary tostudy
orientedsamples.
The sim-plest
way to achieve thisobjective
is to restrict the rotationaldegrees
of freedom of the membraneby placing
it between twoparallel
walls. However, this notonly
orients the membrane, but also restricts theout-of-plane
fluctuations. This tums out to bequite useful,
sinceby varying
the wallseparation,
one can control thestrength
of fluctuations and thusstudy
their effect on differentlength
scales[36].
The
primary
motivation of the present paper b to obtain a betterunderstanding
of the elasticproperties
ofpolymerized
membranes as well as the interactions of membranes with surfaces orother membranes. This is necessary if future
experiments
are to beinterpreted correctly. Indeed, X-ray scattering experiments
on lamellarphases
of these membranesprovide
the mostpromising
method for
measuring
thebending rigidity
andin-plane
elastic constants, as well asdetermining
the form and
strength
of the steric interaction between membranes[37-50].
TheX-ray scattering
factor in this case is characterized
by power-law singularities,
and the exponentdescribing
this behavior is related to the membranes' elastic constants; simulations arerequired
to determine the universalamplitudes [44,45]
which appear in thisexponent.
In the
present
paper wepresent
a detailedanalysis
of thelong-length-scale
behavior of thebending rigidity
and varioussusceptibilities.
Some related work has been described in a recent letter[50].
The outline of the paper is as follows. In sections 2 and 3, we introduce the continuum modelcommonly
used in the theoreticalanalysis
ofpolymerized
membranes and summarize the details of the simulationprocedure.
Section 4 contains a discussion of various effects which makea direct
comparison
of simulation data withtheory
difficulL Inparticular,
we have found thatboundary
effects areparticularly pronounced
for the freeedge boundary
conditions weemploy,
and that care must be taken to
separate
the 'bulk' behavior fromedge
orboundary
effects. Thetechnique
of submembraneaveraging
weemploy
is discussed in this section. Our results for the critical behavior of the areafluctuations,
the pressure, thesusceptibility
of normal vectors, as wellas mean curvature
susceptibility
are thenpresented
in sections 5-8. We close the paper with acomparison
of our results with those of other simulations ofpolymerized
membranes.2. Continuum model for
nearly
flat membranes.In the flat
phase,
deviations of the monomers from the flat zerotemperature
reference state can be described in terms ofin-plane phonon displacements u(z1, z2)
andout-of-plane displacements z(z
i,z~),
where(z
i,
z~)
are internal membrane coordinates. In terms of thesedisplacements,
the free energy reads[15]
F(z, u)
= 2~/d~z(V~z)~
+ 2/d~z[2pu(
+lu);], (1)
where the strain matrix u;j is related to z and u
by
u;j =)[0;uj
+0j
u; +(0;z)(0j z)].
The firstterm in
(I)
describes the elastic energy ofbending,
and the second the elasticstretching
energy,with the Lamd coefficients I and p. Renormalization group
theory
shows[18]
that the interactionN°10 FLUCTUAIIIONS oF A POLYMERIZED MEMBRANE BETWEEN WALLS 1413
of the
in-plane phonon
modes and theout-of,plane
undulation modes leads to renormalized wave,number-dependent coupling
constants pR,lR,
and KR:pR(q), lR(q)
~
q°~
(2) KR(q)
~q~~,
with
"
((2 Qi). (3)
Using
theseresults,
we can calculate thedependence
of severalquantities
on thedistance, 2d,
between the walls
[47]. By comparing
<
~~(~)
>C"kBT /
~ ~
°'f(~~, (4)
q>j~l
R qwhere < >c denotes the
cumulant,
with < z~ >c~wd~,
we obtain theparallel
correlationlength
fjj~w
d~/(~-n)
This can be inserted back into the renormalized elastic constants togive KR(d)
~wd~n/(~-n),
andpR(d)
~wd-~n>/(~-n).
These elastic constants describe the elastic re-sponse of stacks of
polymerized
membranes onlength
scaleslarger
than the correlationlength fjj [47].
3. Monte Carlo simulation of tethered membranes.
We consider a
triangular
network of Nspherical
beads of diameter a=
[51]. Neighboring
beads in the network are linked
by
tethers oflength
lo. Self-avoidance isgenerated by
thepair-
whe hard-core
repulsion
of all beads,together
with a choice of tetherlengths
lo <V$
and asufficiently
small'stepsize'
s for each trial move(we
use lo = 1.6 and s < o.15 in our simula-tions).
Forsimplicity,
we do not include anexplicit bending
energy term into the Hamiltonian.This does not
imply,
however, that no such term appears in thecorresponding
model(I):
the excluded volumeinteractions, together
with thetethering constraints,
generate this term[31]
onintermediate
length
scales(larger
than a, but much smaller than the correlationlength
fjj).
Theglobal shape
of the network ishexagonal,
with a diameter of L monomers, and a 'radius' R suchthat L = 2R + 1. Such a membrane consists of N =
(3L~
+ 1)/4
=
3R(R
+I)
+ I monomers, andof Na =
(L
1)~= 6R~
triangles.
We have simulated membranes from size L= 17 to L
= 49,
and for 10~ to 2 x 10~ Monte Carlo
steps
per monomer(MCS),
thelonger
timescorresponding
to thelarger
membrane sizes orlarger
wallseparations.
The walls are orientedperpendicular
to the z-axh and restrict thez-component
of the center of each bead to lie in the intervall [0,2dj.
A fewtypical
membraneconfigurations
are shown infigure
I.Although
thelong wavelength
undulationmodes are
clearly suppressed by
the walls, for thelarger
wallseparations they
lookquite
similar to theconfigurations
of a free membrane shown in reference[31].
The results of reference
[26]
indicate that the relaxation time rR of a membrane of size L withtether-length
lo =vi
andstepsize
s = 0.20 can beapproximated by
m = To +TiL4~n,
with To
= 1585, Ti = 0.79, and q = 0.7.
Assuming
that thin relation remainsapproximately
valid for the somewhat smaller tether
length
andstepsize
used in our simulations and that wecan
extrapolate
this result to thelarger
system sizes considered here, wc find rR = 300, 000 MCS for thelargest
membrane considered(L
=49).
For a membrane between walls, we would have to insert theparallel
correlationlength
fjj for L in this relation. Forlarge fjj
we therefore have rR ~wd~(4-n)/(~-n)
so that the relaxation time decreasesextremely rapidly
withdccreasing
d. This1414 JOURNAL DE PHYSIQUE I N°10
Fig.
I.Configurations
of aself-avoiding
tethered membrane between two walls ofseparation
2d = 9.0(a)
after 12 x 10~ MCS and(b)
2 x 10~ MCS later. In both cases, theprojections
are on the zy,, the zz- and theyz,planes. lbthering
bonds are drawn betweenneighboring
monomers, whose hard core size is notshown.
N°10 FLUCTUATIONS oF A POLYMERIZED MEMBRANE BETWEEN WALLS 1415
Fig. 1.
(continued).
makes us
reasonably
confident that we have notonly
reachedequilibrium
in oursimulations,
but have alsoaveraged
over a sufficient number ofindependent configurations.
4.
Boundary eTects,
finite sizeeTects~
and corrections toscaling.
In our
simulations,
the finite extension of the model membrane manifests itselfthrough
bothboundary
and finite size effects.Bounda~y effects
influence the monomers in thevicinity
of the1416 JOURNAI- DE PI IYSIQUE I N°10
membrane
edges,
where, due to themissing neighbors,
the fluctuations are muchstronger
thanin the interior of the membrane. We will see several indications of thin effect below. The width of this
boundary region
has beenconjectured [31]
to be of the order off(~~~l/~
~w d. One way todistingubh
bulk fromboundary
behavior in the simulations is to consider submembranes[11]
of
hexagonal shape
with a dhmeter L,(radius
R, which are smaller than the diameter L(radius R)
of the simulated membrane. Theboundary
effects can then be identified from achange
inbehavior as L,
approaches
L.The
finite
sizeeffects,
on the otherhand,
are due to the lowercutoff,
q~n;n= 2x
IL
in all Fourierintegrals.
Onlength
scaleslarger
than L, the membrane acts as arigid body. Finally,
there b the effect that for wallseparations
of the order of L the orientationaldegrees
of freedom are restored.This,
however,
is not veryimportant
for thepresent
simulations because finite size effects set in for much smaller wallseparations,
and we make noattempt
tostudy
our systemsbeyond
thatpoinL
Because it is
possible
to fulfill theinequality
d « fjj «L,
the wallsprovide
an effective way forkeeping
theboundary
effects under control.There is another limitation on the range over which
scaling
laws can be observed. Thescaling
laws
only
hold when the correlationlength
is theonly
relevantlength
scale in the system. Thb canonly
be the case if it b muchlarger
than allmicroscopic length scales,
like the tetherlength.
Fbr finite correlationlengths
we therefore get corrections toscaling.
Allthermodynamic quantities,
like the variance of the
z-distribution,
must have thegeneral
form< Z~ >C"
f~' "(L/ill, '°/ill, ~/ill ), (5)
with
(
=(2 q)/2,
and ascaling
function E which will ingeneral depend
on theboundary
conditions. The results of the continuum model
(I)
canonly
becompared
with the simulations in the limit lo/fjj
- 0 anda/fjj
- 0.In
general,
there will benon-negligible correction-to-scaling
contributions in our simulation data so that we expect, forexample,
< z~ >c= Eo
d~'/(~-n)(I
+Eid-~
+), (6)
with a
correction-to-scaling
exponent w. For localquantities
such as < z~ >c, these correctionsoriginate
from the the finiteness of thearguments
of thescaling
function E or the presence of irrelevantoperators
in our discretization.The corrections to
scaling
are mosteasily
identified for aquantity
which has a knownscaling
behavior for
large d,
like the smallesteigenvalue, 13,
of the moments of intertia tensor Nlap
=p Llra(I) fallrp(I) fpl, (7)
where r is the coordinate vector of monomer I, a,
fl
E(z,
y,z),
and f is the center of mass of theparticular configuration
under consideration. <13
>obviously
scales as d~.However,
when we
plot
<13
> d-~ versus d-I,
seefigure 2,
we do notget
a constant, but anapproxhnately
straight
line with finiteslope:
<
13
>d-~
= ao +
aid-~, (8)
with ao
= 0.0177, al = 0.0824. A
comparison
with(6)
shows that thecorrections-to-scaling exponent
w ci I.Another
quantity,
which should also scale asd~,
is the variance of the distribution of z -values. It can be seen fromfigure
3 that corrections toscaling
are infact almost identical with those found inN°10 FLUCTUA3IONS oF A POLYMERIZED MEMBRANE BETWEEN WALLS 1417
o-1
_~ «
<l~>d
°.°~
+ 17
X 25
o 33
D 49
o
o,o 5 1-o
1/2d
Fig. ~ The scaled average of the smallest
eigenvalue,
13, of the moments of inertia tensior I,equation (7),
versus1/2d.
m6
~
O.05m
A L
+ 17
m~ X 25
~ ~
u 49
2 4 6
1/(2d)
Fig. 3. The variance, < z~ >c, of the distribution of height variables, as a function of the inverse wall
separation.
0.05 + + + + + + + + + + +
~i
~ ~
cQ
o o o o
=
° o o o o
~
D D D D D D D D D ~
~
D
My
2d
/
+ 2.5
M
~ ~~
V Q
~
D 9.o
~° ~ 40
Fig.
4. The variance, < z~ >c, of the distribution ofheight
variables, as a function of submembrane size, obtained from a simulation of a membrane of diameter L= 49.
1418 JOURNAL DE PHYSIQUE I N°10
equation (8).
This isperhaps
not toosurprising,
because <13
> and < z~ >c are both measures of the width of the membrane. It isinteresting
to note that < z~ >c, when calculated for circularpieces
of membrane of diameter Ls, isessentially independent
of Ls(see Fig. 4).
Submembrane averages suppress
boundary effects,
but are different fromperiodic boundary
conditions. In both cases, the finite size
scaling
function E can be calculated. At the level of Gaus- sianfluctuations,
with thephonon degrees
of freedomintegrated
out, an effective HamiltonianHlzl
=(° L(q~-n
+fj[~~-~~)zqz-q, (9)
q
can be used
[52],
where q; =fm;
with m; = 0,1,..,
N;, I = 1,2. In the case of
pmodic bbundaJy
conditions we have< z~ >c=
~~~(~))~ £
_~~_~~
(10)
~°
9 ~~ ~ ~
~ll
For q = 0, the calculation of the
scaling
function ispossible analytically,
with the resultE(y,0,0)
=j~~
+(ii ? arctan(j~~)j, (II)
where
§
=)
=
L/(2xfjj),
and we have absorbed a factor no/kBT
in thescaling
function.Here we have treated the q
= 0 mode
separately,
andreplaced
the sum over all other modesby
an
integral [53].
In the submembrane case, on the other hand, we consider a smallpart
of a verylarge
membrane of size L(infact,
we willusually
take thethermodynamic
limit L-
cc).
For localquantities,
like < z~ >c, there areno finite size effects associated with
Ls,
butonly
with the size L of the wholemembrane;
these will appear forLs/L
- I, if at all. This can be seenclearly
infigure
4.5. Real and
projected
area.In the
simulations,
both the real area, A, and the areaprojected
onto thewall,
Ah, fluctuate. The infinitesimal area element isd~z/@,
whereg(x)
is the determinant of the metric tensor, sothat the mean area, < A >, and the mean
projected
area, < Ah >, aregiven by
< A >=
/ d2z
<
AR
>(12a)
< Ah >=
f d~z
<
li nz(x)
>,(12b)
where nz is the
z-component
of the unit surfacenormal,
n. In theMonge
gauge, theconfigu-
rations are
parameterized
in terms of asingle-valued
functionz(zi, z2)
of the Cartesian coordi- nates of a referenceplane.
In this case n has the form n=
(-01z, -02z, 1)/
+(Vz)2,
so that< Ab >=
f d~z.
For small undulations thesquare-root
can beexpanded,
toyield
~ ~
A~>~~
~
"
~~~~ /))~ ~~KR(~)q~'
~~~~with qm;n
=
2x/L
for free membranes, and qm;n =2x/fjj
for mcmbranes between walls(for
fjj <
L).
Since theintegral
contains an upper momentum cutoff qmax which kproportional
to the inverse size of a monomer,(13) implies
that for a free mcmbrane the ratio ofprojected
toN°10 FLUCnJAIIIONS oF A POLYMERIZED MEMBRANE BETWEEN WALLS 1419
real area
approaches
a finite constant as L- cc. This should be
compared
with the case of fluidmembranes,
where the ratio(13) diverges logarithmically [9,10]
with system size(for
L «fp,
where
fp
is thepersistence length [54]).
It follows from(13)
that theasymptotic
ratio of < A >and <
Ab
> isapproached
with the power lawfin
~w
d-~n/(~-n)
fora membrane between walls.
Biological
or artificial membranes are believed to have a fixed area peramphiphile
orlipid headgroup,
so that the real membrane area does not fluctuate. In thesimulations, only single
monomers are moved at a
time,
so that the area of the membrane cannot be constanLHowever,
the average area fluctuates very
little,
as shown infigure
5.Furthermore,
the average area and its fluctuations are almostindependent
of the wallseparation. Therefore,
we will consider themembrane area to be constant in the
following analysb
of our data.<1.O
Z~
~$ +».mwm . m . o X D
5 +
"A
~~~*X8
. e . O + X
<
i ~
< L1725 3349
m~ V + X o D
I
o.o ~ °'~o 5 to
2d
Fig.
5. Average membrane area per elenlentarytriangle,
< A >/Na (upper
part), and its fluctuations,(<
A~ > < A>~) /
< A > (lowerpart),
as a function of wall separation, 2d.o. 5
+
~ o X
<n
~ +
< L
A ~
+ 17
<
V X 25
o 33
u 4g
o,o
o 5 to
Fig. 6. Relative deviation of the average
projected
area, < AI~ >, from the real membrane area, < A >,versus the wall
separation,
2d, for membranes of various sizes. The curve shows theexpected
power law,v4th q = 0.60
(see text).
The relative deviation of the average
projected
area from the real area is shown infigure
6.For
larger
values of d there are substantial finite size effects.Nevertheless,
thelarge
L data are1420 JOURNAL DE PIIYSIQUE I N°10
in
quite satisfactory agreement
with the power law behavior(<
A > <Ab >)/
< Ab >=ko
ki(2d)-~n/(~~n),
with ko= 0.40,
ki
= 0.30, and q = 0.60.
6. The pressure.
The pressure, p, which the membrane exerts on the walls, can be obtained from the
entropic
interaction
[37,42,52,28,47]
where
[5~28]
T =4/(2 q).
One hasp =
-oini/od
~-
d~~~~ (15)
In the simulations, we determine the number
density
of monomers at the surface, nw, andemploy
the well-known sum-rule forhard-sphere
systems near a hardwall,
flP
= nw,(16)
where
fl
= I
/(kBT),
to calculate the pressure.Explicitly,
we use Nnw =
~ ~
~~
~Lib< (z;)
+b<(2d z;)1, (17)
where
with c = 0.1.
10° ~
~P
~j
~
i
l+~eff
~
i
+ 3
o 25
lo
x 33 ~
D 49 °
lO°
~~ lO~ 0 0.05
~,~ Ol
N'
Fig.
7.(a)
The pressure flp, which the membrane exerts on the walls, as a function of the wall separation.(b)
The effective exponent T~« obtained from(a)
as a funtion of membrane size. tile full line represents anaive
extrapolation
of the data tolarge N-1/~
~w L.
N°10 FLUCTUAIIIONSOFAPOLYMERIZEDMEMBRANEBETWEENWALLS 1421
The pressure is
expected
to decrease with the power law(15)
as the wallseparation
increases.Thb is indeed the case, as shown in
figure
7a. The effectiveexponent
T~jj(L)
for membranes of different size b shown infigure
7b. Anextrapolation
to L = cc is difficult, aslong
as there bno
guidance
as to what theL-dependence
of the correction term should be. Fromfigure 7b,
weconclude that T = 3.2+0.5, which
implies
via thescaling
relation T=
4/(2-q)
that q = 0.75+0.20.We can get a better estimate of T when we
try
to eliminate theboundary
effects and the effect of the finite monomer size. The first can be avoidedby calculating
averages for submembranes ofincreasing size,
Ls, obtained from a simulation of asingle large
membrane of size L = 49. Weinclude
again
the corrections toscaling
to minimize the later, I.e. we writep=
pod~~(I+pid~~+.. ), (19)
assuming
that thecorrections-to-scaling exponent
is close tounity,
as in(8).
The pressure as a function ofLs
for four different wallseparations
is shown infigure
8a. We take the values of Ls = 29 to calculate the criticalexponent,
becausethey
should be least affectedby boundary
orfinite size effects. A fit to the form
(19)
for 2d =2.5,
4.0, and 6.0yields
r = 2.54, po = 6.04, and pi = -0.52(see Fig. 8b).
Thbimplies
q = 0.43. The pressure for 2d = 9.0 b somewhatlarger
than
expected
from thefit;
we attribute thb to the onset of finite size elects.lO°
iP
++ + + + + + +
(~l~[
io-~10'
~ 4.0
tip
o o o o o o o o o o
6.0
~Q-Z X X X X X X X X X X ~
g-o
° D a a u a a a u a a
~~-3
O 20 40 10~
L~ 2d
Fig.
8.(a)
The pressure flp, as a function of submembrane size Ls, for various wallseparations,
as indi- cated.(b)
The pressure flp for Ls= 29 as a function of the wall
separation.
The curve is a fit of the data for 2d = 2.5, 4.0 and 6.o to equation(19)
(see text).7. Nornlal,nornlal correlations.
Another
quantity
we considered is the normal-normalsusceptibility
x"Na[<it~>-<fi>~], (20)
where
fi =
) En; (21)
b
JOURNAL DE PHYSIQUE i T I,M 10, OCTOBRE 199I 56
1422 JOURNAL DE PHYSIQUE I N°10
is the average of all normal vectors of a
particular configuration
of the membrane. Thb quan-tity
b very easy to calculate in the simulations. For a membrane orientedparallel
to the walls(perpendicular
to thez-direction,
so that < fi~ >=<by
>=0),
thesusceptibility
can besplit
into the two contributions X
= Xz + xii, where
Xz =
Na[< fi)
> < hz >< fiz>],
~
(22)
xjj = Na < fijj >,and fi =
(fijj,
fiz).
In thefollowing,
we willonly
consider theparallel
component of x,namely
Xii
In the continuum
limit,
Xii becomesXii =
/ d~zGjj(x), (23)
where Gjj
(x)
=< njj(x)njj (o)
>. lbleading
order, we can write Gjj in theMonge
gauge asGi(x)
Ci<Vz(x) Vz(°)
>,(24)
so that for a
portion
of membrane of area D(with
a linear dimensionll~),
we haveXii "
/ d~z
<
Vz(X) Vz(O)
>(25)
The
leading
contribution to theintegral (25)
comes from the behavior of the correlation func- tion atlarge
dhtances. For a freemembrane,
the correlation function in(25) decays asymptotically
as
<
$7z(~) $7z(Q)
>= ~~j' ~~~eiq'X
~ ~-q
(~~)
/
(2x)2 KR(q)q~
Therefore,
for a membrane between thewalls,
Gjj must have thescaling
formGll(~)
" ~ ~Bll(~/ill). (27)
The
scaling
function8(y) decays exponentially
for y » I. Thisimplies
thatXi "
/d~ZGj(X)
"f(~~*(l~s/fj ). (28)
For
large
d(large fjj),
Xii becomesindependent
of d, so thatil(y)
-y~-°
for y - 0. Thescaling
function lP canagain
be calculated in the Gaussianapproximation.
Forperiodic boundary conditions,
Xii =
Gii(q
=0)
% 0.(29)
For submembranes of circular
shape
with radius ll~, one hasXii =
(2x)~R, /~ dqGjj (q)Ji (ql~s), (30)
o
where
Gjj(q)
=kBTq~/[Ko(q~~°
+f[~~~°~)]
and Ji denotes a Bessel function. For q= 0 the
integral
can be doneanalytically,
with the resultN° lo FLUCTUAIIIONS oF A POLYMERIZED MEMBRANE BETWEEN WALLS 1423
~V(v) =
(2K)~v kef(v), (31)
where
kei'(z)
=
£kei(z),
andkei(z)
denotes a Kelvin function.Here,
as in(11),
we have ab- sorbed a factor no/kBT
in thescaling
function. From(31),
we obtain the behavior oflP for small andlarge argument.
For y - 0, we have*(v)
=-(2K)~v~lCE
+In())1, (32a)
with the Euler constant GE =
0.577215;
for y - cc,*(v)
=-2K~/~v~/~ exP(-
fi)ICOS( fi) Sin(fi)1. (32b)
The results of a numerical evaluation of
(30),
for Q = 0.7, and(31),
for q = 0.0, are shown infigure
9a. Note that thescaling
functions for both values of q becomenegative
forlarge
ar gum ent, in agreement with theasymtotic
behavior(32b).
q
~ o.i
o 5 z-
~
2d
a + 2.5
(
o 4.o,y
(
X 6.o7J O 9.0
JD
,
X
&
O. 'i °
Q
°~ + + + +~ ~
R~ /(~~
° ~
n
s/ Iii
~m
~
0,I~ 2d
fl~ + 2.5
$ ~
o 4.0I
~ ~ ~~' 005
f~D
o 9.0~ fii~
x'i
~~°~o~
°~ 0 ~~
- 0 5
n~/(j,
Fig. 9. The scaling function fit of the susceptibility Xii of normal vectors: (a) calculated from equation
(31)
for ~= 0.lJ (dashcd line), and from
Eq.(30)
for ~= 0.70 (full
linc).
(b) MCscaling
function, for co = 0.667, cl= 1.1.5, and ~
= lJJ5. (c) MC
waling
function, for co= fi.40, cl
= lJ.43, and ~
= lJ.30. In
(b)
and(c),
thcscaling
variablc 15 proportional toRs/(jj.
1424 JOURNAL DE PIiYSIQUE I N°10
The
susceptibility
Xii also contains conUibutions from the short distance behavior of the corre-lation function. The local contribution should scale as
where co =
O(I)
and ci =Oil)
are constants. For d- cc, this contribution
approaches
aconstant, and therefore does not
change
thescaling
behavior of xjj.However,
for theinterpre-
tation of the simulation data, it b
important
to take(33)
into acccunL Another correction scales like the energydensity,
and therefore goes like [eo + ei(2d)-~(~+°)/(~-°)]
forlarge
d. Theleading
correction term is therefore
given by (33).
The
scaling
function, obtained from the databy subtracting (33),
b shown infigure
9b. All data fall onto asingle
curve for q= 0.65, co = 0.667 and ci
= 1.15.
However,
the same data also scale for q=
0.30,
co = 0.40 and ci=
0.43,
seefigure
9c. Thecollapse
of the data onto asingle
curve bobviously
not very sensitive to the exact value of q.However,
the behavior of thescaling
function forlarge
argumentchanges
with q: lvhereas thescaling
function bpositive
overthe whole range for q
= 0.30, it becomes
negative
forlarge
y for q = 0.65. Since thescaling
function for the effective Hamiltonian is also
negative
forlarge
argument, we believe that the estimate q= 0.50 + 0.10 is reasonable.
By comparing
thescaling
function calculated from the effective model(9)
and the MCdata,
we can get an estimate for the
nnipl1tlide
of the correlationlength
fjj =to (2d)~/(~~°) Figure
9implies to
= 0.30 + 0.07.8. Renonnalized
bending rigidity.
The renormalized
bending rigidity,
KR, b of fundamentalimportance
for anunderstanding
of the behavior offluctuating
membranes. We would therefore like to calculate itdirectly (as
an inversesusceptibility).
Therequired expression
must involve a correlation function[50,55]
of the meancurvature,
H(x)
=)lt[K(x)]
=)[I/Ri(x)
+1/R2(x)],
where K is the(local)
curvature tensor, and RI andR2
are theprincipal
radii of curvature. We therefore definejjj
_
j j ~~~i / d2z'i
<
H~x~~~~'~
>~ ~~~~where
g(x)
is the determinant of the metric tensor.Using
theMonge
gauge andexpanding
for smallundulations,
one obtains, toleading
order,H(x)
m)V~z(x)
so that~~~
=
/d~z / d~z'
<
V~z(x)V~z(x')
>c(35)
K~fl < A >
in this limit. Consider now the contribution to K~a from a circular
piece
of membrane of radiusR,
« R. We can use Gauss' theorem to write theintegrals
in(35)
as lineintegrals
around theperimeter
C of this domain. Since nopoint
on theboundary
of a disc bsingled
out, we have/d~z d~z'
<V~z(x)V~z(x')
>c= 2xll~j~
ds <er(s) Vz(s) er(s') Vz(s')
>c,(36)
N°10 FLUCTUAIIIONSOFAPOLYMERIZEDMEMBRANEBETWEENWALLS 1425
where s and s' are located on C, and er b the 2-dimensional radial unit vector in the space of internal coordinates.
Using
thescaling
form for the correlation function derived above, we arrive at)~
=
<~f(~°r(R, /fjj), (37)
where
r(y)
- const. for y - cc. Fbr a free membrane, wherefjj
- cc, theright-hand
side of (37~ must becomeindependent
offjj,
so thatr(y)
-vi
-n for y- 0. This
implies K~a(l&)
~w II~Qfor a free membrane.
We can
again
calculate thescaling
function r in the Gaussianapproxhnatioll~ Ignoring
thecomplication
of the circularboundary
in(36),
andreplacing
itby
a squareboundary
of sidelength Ls
=2xll~,
we haveIn the case of
periodic boundary conditions,
it h easy to show(compare Eq.(29))
thatKj~, equation (35),
vanikheshientical~y. By
virtue of(36),
thisimplies
that the lineintegral
around theperimeter
also vanishes. Nevertheless, a non-zero result is obtainedalong
other closed contours as well asportions
of theperimeter.
lb determine r in the latter case, we consider a membrane ofquadratic shape
and linear dimension L = 2xll~.
The contribution from oneedge
of thb squarecan be obtained
by replacing
the sum over the non-zeroq-modes
in(38) by
anintegral
with lower cutoff qm~=
2x/L
=I/ll~.
In thb way we find~2 (~~~
~(Y)
"~~i~~~ ~;,
~~
q~~~ ~
~(i~~~~~'
For q= 0,
r(y)
=jjInj
+(Y
+( 2arctanj ~
i)
2arctanj ~
+
i)
+2xj. j40)
2
2y
+ y YFor q > 0, the
integral
has to be evaluatednumerically.
Thescaling
function b shown infigure
10a.
For the submembrane case, on the other
hand,
we obtain from(38),
with R >ll~,
r(y)
= 4dqi )
dq~~~~~~~'Y~
~
~
(41)
%~ ~
qi q
Q~+
I For q
=
0,
theq2-integral
can be carried out, so thatThis function b
plotted
infigure
10b. For q > 0, we find it convenient to write theintegral (41)
in the form
r(v)
= 4K/~~
da/~ dqqGjj (q)Jo(aq), (43)
1426 JOURNAL DE PI IYSIQUE I N° lo
~~i
rlo~~ lo° lo' 10~~
10°
R~/(~~ R,/(~~
Fig.
lo. The scaling function r,(a)
calculated fromequation (39), (40)
for the case ofperiodic boundary
conditions; ~b) calculated fromequations (42), (43)
for the subnlembrane case.where Jo denotes a Bessel function. In thb
form,
theintegrals
areeasily
evaluatednumerically.
The result b shown in
figure
lob for q = 0.7. Note that in contrast to theperiodic
case~fig. 10a),
the
scaling
function is non-monotonic in thb case.We have used three different discretizations of
(34)
to calculate K~«. The first involves express- in g the curvature H in terns of correlations of thegradient
of the surface unit normal vector. Ifn; is the unit vector normal to
triangle
I, we define[50]
$j~~~
"~
ILL
~'J
(~'~ ~J)llL L
~km
(ilk ~m)1
>,(44)
e& I j(I) k m(k)
where e;j =
(Il~ Rj Ii
Il+Rj
b the unit vectoralong
the linejoining
the centers of mass, Il~of
trhngles
I andj.
In order to better understand thisexpression,
note that if we choose a metric in which all intemallengths
areequal,
I.e. alltriangles
on the surface areequilateral
and of area one, e;j(n;
nj isjust
the normal curvature in the direction e;j in the continuum limit. Now ifki,
,
k3
are three normal curvatures at apoint
z of the surface in directions which intersect at2x/3 radians,
thenki
+k2+k3
= H[56]. Thus,
with thbmetric, (44)
is the natural dhcretization of(34).
Itzykson
describes how to discretize fields ontriangulated
random surfaces in reference[57].
Discretizing
the mean curvature in thin way, one has~~~'
j~~
~ =<ii
fi;i j'J (r;
rj)iii
fi~j j~~ (r~ rm)1
>,(45)
~efl I j(I) 'J k m(k) ~~'
where the sums over I,k are over all monomers, and the sums over
j(I), m(k)
over theirneigh-
bors.
Furthermore, I;j
is the dbtance between the two monomers I,j,
and a;j b thelength
ofa bond in the dual lattice.
Finally,
i1; is the surface normal at monomer I, which is obtainedby averaging
over the surface normals of allUhngles
which have I as a corner.If we assume that there are no
overhangs,
we canagain
work with theMonge parametrhation,
and
expand
toleading
order in derivatives of z. This leads to thesimpler expression
~
~"
i~ ~~~
=<IL L I'( (z; zj)I IL L )~"'(zk
zm)1 >(46)
~e& I j(I) 'J k m(k) ~~'