HAL Id: jpa-00248060
https://hal.archives-ouvertes.fr/jpa-00248060
Submitted on 1 Jan 1994
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.
Undulation-enhanced electrostatic forces in lamellar phases of fluid membranes
Renko de Vries
To cite this version:
Renko de Vries. Undulation-enhanced electrostatic forces in lamellar phases of fluid membranes. Jour-
nal de Physique II, EDP Sciences, 1994, 4 (9), pp.1541-1555. �10.1051/jp2:1994216�. �jpa-00248060�
Classification Physics Abstracts
5.40 82.70 87.20
Undulation-enhanced electrostatic forces in lamellar phases of fluid membranes
Renko de Vries
Faculty of Clienfical Engineering, Dept. Polymer Teclin., Delft University of Technology, P. O-Box 5045, 2600 GA Delft, The Netherlands
(Received
23 March 1994, received in final form 23 May 1994, accepted 31 May1994)
Abstract. A formula for the free energy of a stack of highly charged, semiflexible mem- branes, previously derived by Odijk
(Odijk
T. Langmuir,1992),
is shown _to be the first orderterm in a formal expansion of the free energy in terms of powers of the electrostatic potential.
The formula describes the transition from a regime where the free energy is dominated by the electrostatic energy to a regime where it is dominated by bending energy and entropy.
Introduction.
The
physics
of a stack of semiflexible membranesinteracting
via an electrostaticpotential,
has been addressedby
a number of authors [1-6]. Forhighly charged membranes,
undulations areunimportant
aslong
as the distance between the membranes is either smaller than or of the order of thedecay length
of the electrostaticpotential (~). Upon adding salt,
orincreasing
the distance between themembranes,
the undulations becomestronger
and stronger, and maysubstantially
enhance the electrostaticrepulsion ill. Finally,
in the limit of veryhigh
salt concentrations or
large
distances between themembranes,
the situation is describedby
Helfrich's
theory
[7], except for aboundary layer
near themembranes, preventing
them fromtouching
eachother,
due to the electrostaticrepulsion
[2]. Theregime
to which Helfrich'stheory
does notapply,
covers anappreciable
range of membraneseparations
and salt concentrations.For membrane
separations
on the order of thedecay length
of the electrostaticpotential,
undulations will be weak. Then a harmonic
approximation
of theHamiltonian,
in whichonly
terms
quadratic
in theamplitudes
of the undulations areretained,
islegitimate.
In thisregime,
which will be called the harmonicregime,
there is a small enhancement of the electrostaticrepulsion, proportional
to the square root of the bare electrostaticpotential
[3,4].
If the(~)For
very weakly charged membranes the situation is somewhat different. Here undulations areimportant even for distances between the membranes of the order of the decay length of the potential.
amplitude
of the undulations becomes of the order of thedecay length
of the electrostaticpotential,
the harmonicapproximation
breaks down.A number of theories have been
proposed, dealing
with theregime
where the harmonicapproximation
is nolonger legitimate
and where it may beexpected
that there is a substantial enhancement of the electrostaticrepulsion
11, 4, 5]. For thisregime, only
theOdijk theory ill provides simple
closedexpressions(~). Postulating
a Gaussiansingle
membrane distributionfunction for the
amplitudes
of theundulations, Odijk
derivedsimple
closedexpressions
for the free energy, the osmotic pressure and the averageamplitude
of the undulations in a stack ofcharged
semiflexible membranes. In view of thehighly
anharmonic interactionpotential however,
it is not clear to what extent the distribution of theamplitudes
of the undulations will be a Gaussian.Here it is shown
rigorously
that theOdijk expression
for the free energy, from which expres- sions for the osmotic pressure and the averageamplitude
of the undulations can bederived,
and which is called the Gaussian
approximation,
is in fact the first order term of a formalexpansion
of the free energy in terms of powers of the electrostatic interactionpotential.
The range ofvalidity
of the first order or Gaussian approximation is determined from an estimate of the contribution of the second order term. The Gaussianapproximation
describes the tran- sition from aregime
where the free energy is dominatedby
the electrostatic energy, to aregime
where it is dominated
by bending
energy and entropy.Beyond
thepoint
where the Gaussianapproximation
breaksdown,
the convergence of theexpansion
for the free energy is poor.The electrostatic
potential
must be calculatedusing
the non-linear Poisson-Boltzmann
equation.
It is assumed that thedecay length
of the electrostaticpotential,
theDebye length,
is a great deal smaller than both thetypical
distance between the membranes and thetypical wavelength
of the undulations.Thus,
it is assumed that the orientational order ishigh
and that the averageamplitude
of the undulations is much smaller than the averagedistance between the membranes. For the interaction between two small
pieces
of membrane opposite to each other andseparated by
a distanceD,
in a stack ofundulating membranes,
wecan then use the
expression
for the electrostatic interaction energy between two flatsurfaces, separated by
a distance muchlarger
than theDebye length
11, 9]:V(D)
=8xZ(~(~Q~~)e~~~ (1)
~~~
~~ ~~~~~~~
~~~
Z =
Qa/~
is a dimensionless surfacecharge density
in terms of the two electrostaticlength scales,
theBjerrum length Q
"
e~lekBT
and theDebye length
~~~ =(8xQn)~~/~
and thenumber of
elementary charges
per unit area, a. kB is Boltzmann's constant, T is the absolute temperature, e theelementary charge,
e thepermittivity
of thesolvent,
which contains anexcess of monovalent
electrolyte
of concentration n.V(D)
is thepotential
of interaction perunit area, scaled
by kBT.
Since we areonly considering
cases in which theamplitude
of the membrane undulations isappreciably
smaller than the distance between themembranes,
short rangerepulsive potentials,
to account forpossible hydration
forces or collisions etc. will not be included inV(D).
The deformation behaviour of the membranes is describedby
asingle
elastic
bending
modulusK,
scaledby kBT.
An electrostatic contribution to K[10-12]
should be includedexplicitly.
Since it was assumed that the orientational order ishigh,
we dust haveK »1.
(2) podgornik and Parsegian [4] derive approximations for several limiting cases, but not the expo- nential renormalization of the electrostatic free energy that is found by Odijk. Unfortunately, Evans and Ipsen [5] do not present closed expressions for the free energy and the osmotic pressure.
The reference
configuration
is taken to be a stack of flatmembranes, separated
from each otherby
a distance D. The stack of Nfluctuating cliarged
membranes will therefore bedescribed
by
the Hamiltonian(scaled by kBT):
7ilzl
=L / d~x ) (Azn(x))~
+
v(D
+zn(x)
zn-i(x)) (3)
~i
where
(x, zn(x)
are the Cartesian coordinates of thepoints
on the n-th membrane and A +Ax
is the two-dimensional
Laplacian.
The undulationamplitudes zn(x)
will bedecomposed
into Fouriermodes, along
the lines of David'sanalysis
[8]:zn(x)
=~j z(kjj, ki exp(ikjj x) exp(ikin) (4)
kjj,ki
Here kjj is a twc-dimensional
in-plane
wavevector, while ki labels theperpendicular
modes.It is a
phase
difference between successive membranes.Values,
allowed for the wavevectors k +(kjj,ki)
follow fromperiodic boundary
conditionszn(zi
+L,z2)
"
zn(zi,z2
+L)
=
zn(zi, z2)
for somemacroscopically large length L,
andzn(zi, z2)
= zn+N
(xi, z2).
The range of ki is restricted to allow forexactly
Nindependent perpendicular
modes: -x < ki < x.In terms of
z(k),
theHamiltonian,
scaledby kBT
andby L~N,
can bedecomposed
into7i =
V(D)
+ 7io + 7iI, where 7io isquadratic
inz(k),
~ojz(k)j
=
I £(Kkj
+4sin2(ki/2)v"(D) )z(k)z(-k) (5)
2
~
and
7iI
contains thehigher
order terms:7iI(z(k)]
=j~ ~j
~~~~~)~~ jj(1
e'~~~)z(ki) (6)
m=3ki,k2;..,km ~' l=1
V(~')(D)
denotes the m-th derivative ofV(D).
The asterisk indicates that there is a restrictionon the set of k values to be summed over:
kill
+k2
II
+ + km
II
" °
kit+k~i+..+kmi
=2xl,1=0,+1,+2,.. (7)
For membrane
separations
on the order of aDebye length,
the undulations are weak. In this case the energy of aconfiguration
may beapproximated by
7io. This is the harmonicapproximation,
which was also usedby
Pincus et al. [6] in their treatment of the sameproblem.
The renormalization group flow
equation
for the effectivepotential
of David [8] is also derived in thisapproximation.
In the harmonicapproximation,
the Fourier transformed correlationfunction is
given by:
So(k)
+(z(k)z(-k))o
=~~~~
~-~(~~~~~~
~~~
(8)
Here, lo
denotesthermodynarlic averaging
with respect to7io(z(k)].
The correlationlength
in the harmonic
approximation,
lo =(K/V"(D))~H
will be called the bare correlationlength.
In-phase
undulations of the stack of membranes may have verylarge amplitudes,
determinedonly by
theirbending
energy. Therefore one can notsimply integrate
the correlation functionjOURNAL DE PHYSIQUE II -1 4 N' 9 SEPTEMBER 1994 58
to find an average
amplitude
of the undulations. Thisintegral diverges logarithmically,
as it should for a smectic.Instead,
the averageamplitude
of the undulations is taken to be themean square deviation of the distance between the membranes with respect to their distance in the reference
situation,
D.Taking
the limitN,
L - cxJ we haved(
=j(z~(x) z~-i(x))2)o
=
/ so(k)4sin2(ki/2)
=
£ (9)
k
where
~ ~~ ~ ~~~
/ /_~ (2~ /_~ 2x)~~
~~~~The
subscript
0again
indicates thatdo,
which will be called the bareamplitude
of the un-dulations,
is the averageamplitude
of the undulations in the harmonicapproximation.
The free energy per membrane per unit area, with respect to freemembranes,
in the harmonicapproximation,
isgiven by:
Fo
=V(D)
+/ In(Sp~(k)/Kk()
=V(D) (I
+(~do)~) (II)
k
As will be shown in the next
section,
the harmonicapproximation
is valid aslong
as~do
< I.In its range of
validity,
itpredicts
a small renormalization of the interaction,proportional
to the square root of the bare electrostaticpotential,
that is,proportional
toexp(-~D/2),
as wasalready recognized by
Evans andParsegian
[3].However,
as wasclearly
shownby Odijk ill,
if theamplitude
of the undulations isappreciably larger
than thedecay length
of the electrostaticpotential,
there is a substantial enhancement of the electrostaticrepulsion. Assuming
asingle
membrane Gaussian
amplitude
distribution for theundulations, Odijk
found an electrostatic free energy of:Fei
"V(D) exp((~d)~ /2) (12)
where
d/vi
is the rootmean square
amplitude
of the undulations. The loss of free energy duethe confinement of the
undulations,
with respect to free membranes was estimated to be?cant
=] (13)
with a constant c
equal
to about1/64(3 ).
The value of d is foundby minimizing
the total free energy. If(~d)~
becomes of order~D,
the electrostatic free energydiverges
as D - cxJ and thetheory
breaks down [2]. Therefore thetheory
will not be valid for very strong undulations.The electrostatic free energy is calculated
assuming
thatneighbouring
membranes fluctuateindependently. Also,
director fluctuations areneglected.
I find in thefollowing analysis,
that theOdijk
result can also be found withoutassuming
thatneighbouring
membranes fluctuateindependently
and withoutneglecting
the director fluctuations. TheOdijk
result is shown to be the first order term in a formalexpansion
of the free energy. There is in fact a close relation between the presentapproach
and the previous variationalapproaches
11, 4, 5] basedon the classic
Gibbs-Bogoliubov (or Peierls) inequality.
To stress this relation it will be ex-plicitly
shown that the exact first order result canalways
be found from theGibbs-Bogoliubov inequality, using
a trial Hamiltonian which isquadratic
in the undulationamplitudes z(k).
(3 ~~
jij
tb~ c~~~t~nt c was estimated to be1/32,
but a more precise calculation shows that it shouldb~
1/64
~Thich is the value adopted in [4].The
expansion
of the free energy.Consider a trial Hamiltonian
7itr[z(k)], quadratic
in the Fourier transformed fluctuationsz(k):
7itr[z(k)]
=
V(D)
+~ ~~~~
~ ~ ~~~~~ ~~~~~~~~~~~~~~~~~ ~~ ~~~~This form is obtained from the
quadratic
part of the full Hamiltonian7io(z(k)] by replacing
the bare correlationlength lo
=
(K/V"(D))~H by
an as yet
unknown, k-dependent
correlationlength A(k).
TheGibbs-Bogoliubov inequality
states that? I
?tr
+17ilz(k)1 7itrlz(k)I)tr (15)
Here F is the true free energy of the system, Ftr is the free energy of a system with Hamiltonian
7itr(z(k)]. Further, )tr
denotesthermodynamic averaging
with respect to7itr(z(k)].
The correlationlength A(k)
is obtainedby minimizing
theright-hand
side of(15)
with respect to>(k).
It will now be shown that the trial Hamiltonian 7itr in fact
yields
the first order term in a formalexpansion
of the free energy in terms of powers of thepotential V(D).
Let usexpand
the free energy in the usual way in terms of contributions of connected
diagrams.
A connecteddiagram
with 2n-external lines will be called ann-diagram,
n = 0,1,... The free energy per membrane per unit area, with respect to free membranes can be written as:F
=
v(D)
+( in(Sii(k)/Kki)
~j [contributions
of alltopologically
distinct0-diagrams] (16)
where the minus
sign
accounts for the fact that the free energy is minus thelogarithm
of thepartition
function. The rules to find the contributionrepresented by
adiagram
ofgiven topology
are:I)
associate a factor-V(~')(D) £[
~~ ~
fl]~(l
e'~~~)z(ki)
with every m-th order vertex, whereki,k2;. km
are thewile tic(rs labeling
the vertex;it)
associate a factorSo(ki)b(~
_~~ with every
line,
internal or external, where ki and k2 arethe wave vector labels
on'that line;
iii)
divideby
the symmetry number S of thediagram.
The symmetry number is theproduct
of the number of ways of
permuting
the vertices and the number of ways ofpermuting
thelines,
whilekeeping
the contractionsrepresented by
thediagram
the same.The correlation function is the sum of the contributions of all
topologically
distinct I-diagrams.
This sum can be written in terms of the sc-called properself-energy, Z(k).
Adiagram
is called proper if it cannot besplit
into twopieces by cutting
one internal line. The contribution of all properI-diagrams
toS(k)
can be written asSo(k)Z(k)So(k)
andS(k)
isgiven by:
~~~~ ~~~~~~~ ~~~
Sp~(k/- Z(k)
~~~~Any
insertion of aI-diagram
into an internal line of some otherdiagram
is called aself-energy
insertion. Let us call a
diagram
with noself-energy
insertions in its internal lines a skeletondiagram.
The contributions of all properI-diagrams
toZ(k)
are obtainedby drawing
allskeleton
I-diagrams
and thenmaking
allpossible self-energy
insertions into the internal lines.This is
equivalent
to:Z(k)
=~j [self-energy
contributions of alltopologically
distinctskeleton
I-diagrams,
withSo(k) replaced by S(k)] (18)
Generally
this is anintegral equation
forZ(k).
A
problem
with theexpansion
of the free energy is that the bare correlationlength lo
enters via the propagator So(k).
Thislength
will bestrongly
renormalizedby
the anharmonic terms of the Hamiltonian contained in7ii[z(k)]
as soon as theamplitude
of the undulations d becomes of the order of thedecay length
~~~ of the interactionpotential. Clearly
it would be better ifonly
the renormalized correlationleng~h
entered theexpansion
of the free energy. This can beaccomplished
if somehowSo(k)
can bereplaced by S(k).
To obtain the free energy in terms ofS(k),
one cannotsimply
sum over all skeleton0-diagrams
withSo(k) replaced by S(k),
since this would result in an
overcounting
ofdiagrams
[14]. Lee andYang
[13] andLuttinger
and Ward [14] have shown how toperform
the proper resummation. Because there are some differences between this calculation and the quantum one, theanalysis
in our case isgiven
in detail inAppendix
A. It is found that:~ "
~(~)
+~~(~
~(~)/~~()
+~(~)~(~)
~j [contributions
of alltopologically
distinctskeleton
0-diagrams,
with So(k) replaced by S(k)] (19)
In
Appendix
A it is also shown that~~
@
~ ~~~~Hence, the functional F has an extremum for the function
Z(k)
that satisfiesequation (18).
As is mentioned
by
Lee andYang,
this can be used as a variationalprinciple generating
theequation
forZ(k)
from the free energy functional.Now consider the correlation function of the system with the trial Hamiltonian:
~~~~
K(k(
+-4(/)4
sin~(ki/2))
~~~~For the function
A(k)
for which theright-hand
side of(15)
has a minimum, this will be anapproximation
of the correlation function of the true system.Therefore,
anapproximation
of the properself-energy
of the true system is#ven by
Z(k)
=K(Ap~ A~~(k))4sin~(ki/2) (22)
In terms of
Z(k)
thusdefined,
theGibbs-Bogoliubov inequality
can be written as:F <
V(D)
+/ ln(S~~(k)/Kk()
+/ S(k)Z(k)
2 ~ 2
~j [contributions
of alltopologically
~ distinct I-vertex0-diagrams,
withS(k)
aspropagator] (23)
where
Z(k)
is to be found from(20). Thus, using
thisinequality together
with the trial Hamiltonian7itr, naturally
leads to the first order term(I.e. only including
I-vertexdiagrams)
of the
expansion (19),
in terms of powers ofV(D).
However, the full expansion is essential forcalculating higher
order terms.The Gaussian
approximation.
From the
equation
forZ(k)
it can be seen that to first order inV(D)
the bare correlation func- tion So(k)
isonly
renormalizedby "tadpole diagrams" (I.e. diagrams
with interaction linesstarting
andending
at the samevertex).
Thismerely
amounts to ak-independent
renormal- ization of the bare correlationlength lo Thus,
in this case the free energy can be calculatedusing
a correlation functionS(k)
with ak-independent
correlationlength
A. The contributionbfi
to the free energy of the first order skeleton0-diagrams,
shown infigure I,
is:bfi =
~j ~~~~~f~ /
S(k)4sin~(ki/2)) (24)
~~
2'~m.k
~
Because the correlation
length
isk-independent,
the relation between the averageamplitude
of the undulations d and the correlation
length
isagain #ven by
d~=
A~/2xK.
The free energy to first order inV(D)
can now be calculated. In terms ofd,
F = ~
~ +
V(D) exp((~d)~ /2) (25)
In
principle
the value of d must be found from theequation
forZ(k). However,
sinceZ(k)
canbe considered to be a function of
d,
the functional derivative in(20)
reduces to anordinary
one:
dF/dd
= 0
(26)
which expresses the renormalization of the bare
amplitude
of theundulations,
do(~d)~ exp((~d)~/2)
=
(~do)~ (27)
This represents
essentially
the Gaussianapproximation,
theonly
difference between theOdijk expressions
and theexpressions presented here,
is in the constants. Because the actualrepulsion
is stronger than is assumed in the harmonic
approximation,
the renormalizedamplitude
of the undulations is smaller than the bare one. If ~d <I,
the renormalization of do can beneglected
and the harmonic
approximation
is sound.in
2
4
Fig. I. The one vertex skeleton 0-diagrams contributing to the free energy, m > 2.
It is difficult to evaluate the free energy to
higher
order inV(D),
sinceA(k)
can nolonger
be assumed to bek-independent.
Even if this were not the case,evaluating
the contributions ofhigher
orderdiagrams
in an accurate way posesproblems.
Forexample,
in order to estimate the contribution of the second order term, we make an additionalapproximation, eliminating
the difficult
ki integrations.
A stack of membranes is considered in which there areonly
twoperpendicular modes,
anin-phase
and anout-of-phase
mode. This amounts tosetting
N= 2
in equation
(3)
for theHamiltonian,
instead ofletting
N - cxJ.Then,
as is shown inAppendix B,
the second order contribution to the free energy isroughly
&>~ -~
>2v2(D)~~P)~jj(~~~
(28)
provided (~d)~
» 1. The second order contributionbF2
will beextremely
smallcompared
to the first as
long
as(~d)~
isappreciably
smaller than2/3 (~D).
We thereforetentatively
conclude that the first orderapproximation
will be valid aslong
as(~d)~
< cl(~D) (29)
with a constant cl
equal
to about2/3. Showing
that the sum of allhigher
order terms will besmall if this bound is
fulfilled,
is very difficult.Nevertheless,
inAppendix
B it is shown that if the second order contribution is nolonger
smallcompared
to the first order term, then the third order term will not be smallcompared
to the second order termeither,
I-e-,beyond
the range ofvalidity
of the first order or Gaussianapproximation,
convergence of theexpansion
of the free energy is poor. Thismight
indicate that we areentering
aregime
where the electrostaticpotential
startslooking
like an effective hard-wallpotential,
for which anordinary perturbation expansion
of the free energysimply
does not work. In view of the poor convergence of theexpansion beyond
the range ofvalidity
of the first order or Gaussianapproximation,
it does not seem worthwile to carry theperturbation expansion
of the free energy any further.Discussion.
For
highly charged membranes,
the effective surfacecharge density
Ze~ reduces to a constantx~~ and the electrostatic contribution to the elastic
bending
modulus ispredicted
to reduce to I/x~Q (10, iii. Assuming
K = Ko + I/x~Q
and Ze~ = x~~ andfixing
the value of the intrinsicbending
modulusKo,
theamplitude
of the undulations d is a function of the dimensionless variables ~D and~Qi
~~~~~~~~~~~~~~~~~
~j~~~
~~i~~o
~~~~For an aqueous monovalent
electrolyte
of concentration ns(in M)
at room temperature, ~~~=
0.30ni~/~nm
andQ
" 0.71 nm, whence
~Q
"
2.37n(/~
In table I values of ~d have beentabulated as a function of ~D and the salt concentration ns
(in M),
forKo
= I. Since ~dcan be much
larger
than unity, the enhancement of the electrostaticrepulsion, proportional
toexp((~d)~ /2),
can be substantial. Now consider the ratio of the electrostatic to the confinement free energy,given by
As
expected,
this decreases if the fluctuations increase.Only
if(~d)~ /2
« 1, does the electro- static free energy dominate over the confinement free energy. If(~d)~/2
» 1 it is the otherTable I. ~d as a function of ns
(in
M)
and ~D forKo
" I.
~D
ns
(in M)
10 20 30 40 501.0x10~~ 1.3 3.7 5.5 7.0 1.0x10~3 0.9 3.3 5.2 6.7 1.0x10~~ 0.6 2.8 4.8 6.4 7.7
1.0x10~5 0.3 2.2 4A 6.1 7A
way around. For values of
(~d)~
m2/3 (~D)
where the first order or Gaussianapproximation presumably
breaksdown,
the electrostatic free energy isonly
a small fraction of the total freeenergy. Whereas the free energy
decays roughly exponentially
in theelectrostatically
dom- inatedregime,
itdecays roughly algebraically
in theregime
dominatedby
confinement free energy. It must be noted however, that in theregime
dominatedby
confinement free energy, theamplitude
of the undulations is still very much affectedby
the electrostaticrepulsion.
Beyond
the range ofvalidity
of the first order or Gaussian approximation aregime
is entered in which the electrostatic interactiononly
affects theamplitude
of theundulations,
but doesnot any
longer
contributesignificantly
to the free energy. In thisregime,
the effects of steric- andhydration repulsion
canprobably
nolonger
beneglected. Although
it isstraightforward
to estimate the order of
magnitude
of therepulsion
in thisregime, dealing
with theinterplay
between the effects of steric-,
hydration-
and electrostaticrepulsion
in an accurate way isextremely
difficult.Nevertheless,
thestrongly fluctuating regime,
I-e- thelimiting
case of greatseparations
or veryhigh
salt concentrations, can beanalyzed qualitatively
[2]. In this limit the situation is describedby
Helfrich'stheory
[7], except for the existence of aboundary layer
of thicknessb, independent
ofD, preventing
the membranes fromtouching
each other. It wasargued by Odijk
[2] that in thislimit,
thetypical
area of interaction of twocolliding membranes,
will be of order K~~~ The thickness b of theboundary layer
can then be estimated fromK~-2v(&)
=
o(1) (32)
And the free energy will be
given by:
~
K(D b)2
~~~~where c2 is the constant
appropriate
to the Helfrich free energy, c2 " 0.10. This isexpected
to be valid down to
separations
for which ~b m1/2
~D. From the fact that thestrongly
fluctuating regime
cannotoverlap
with there#me
for which the Gaussianapproximation
will bevalid, Odijk
concluded [2] that(~d)~
<ci(~D) (34)
with cl m I, is an approximate bound on the
validity
of the Gaussianapproximation,
in fair agreement with the value cl " 2/3
obtained from the estimate of the contribution of the second order term.From a bound like
(34)
it isactually possible
to understand the structure of the second order term in aqualitative
way.Obviously, bF2
°J
-V~(D),
with a minussign
because the free energy in the Gaussianapproximation
is anupperbound
on the true free energy. To obtain the properdimensions,
this must bemultiplied by
an area A~, since the characteristiclengthscale
of the system is
given by
thecorrelationlength
>. Similar to the first order term, this will be enhancedby
adimensionless, increasing
function of(~d)~,
since thesign
of ~d is irrelevant. As in the case of the first order term, this function will be dominatedby
anexponential.
Thus wearrive at
bF2 °J
->~V~(D)exp(n(~d)~/2) (35)
As
long
as(~d)~
<ci(~D),
the first order term should be the dominant term.This, finally, implies
n = +2/ci
For I-dimensional systems,
notably
forhexagonal DNA,
theimportance
of undulation enhan- ced-electrostatic forces has now beenclearly
demonstrated [15,16]. Unfortunately, published
osmotic stress and
scattering experiments
on lamellarphases
ofcharged membranes,
have not beenperformed
in there#me
where the electrostaticrepulsion
may bestrongly enhanced,
while the contribution of the electrostatic free energy to the total free energy is still
apprecia-
ble.Nevertheless, recently Odijk
showed [2] that histheory,
if combined with the Lindemannmelting rule,
can be used tointerpret experimental
data on themelting
of a lamellarphase
ofcharged
membranes [18]. In theregime
of membraneseparations
up to about loDebye lengths,
some cases have beenreported
where it was found that the electrostaticrepulsion
wassomewhat
larger
thanexpected
[3,17].
Other osmotic stress andscattering experiments
have beenperformed
at ionicstrenghts
either toohigh
[19] or too low [20] to see the enhancement of the electrostaticrepulsion.
It would therefore beinteresting
to see whether furtherexperi-
ments can be
performed
in theregime
which has been indicated in table I. In thisregime,
the situation cannot be describedby
Helfrich'theory,
norby
pure electrostaticrepulsion
between flat membranes.However, equation (25)
shouldapply.
Acknowledgements.
would like to thank T.
Odijk
for manystimulating
discussions.Appendix
A.The
expression
for the free energy.The
proof
thatequation (19)
for the free energy in terms of the correlation function is correct,closely
follows the papers of Lee andYang
[13] andLuttinger
and Ward [14]. Tobe#n with,
a parameter is needed that counts the number of internal lines of a
given 0-diagram.
This parameter is introducedby making
the substitutionz(ki)
-»i12z(ki) (36)
in the equation for 7iI. If p
= 0, then 7i = 7io. It will be shown that for all p,
including
p= 1,
>(»)
=
>s(») (37)
where
F(p)
=V(D)
+/ ln(Sp~(k) /Kk()
+
F'(p) (38)
2 k ~
F'(p)
=~ [contributions
of alltopologically
distinct0-diagrams] (39)
and
?s(»)
=
V(D)
+/(S~~ (k, »)/Kk()
+/ S(k, »)z(k, »)
+
?1(») (40) F](p)
=
£ [contributions
of alltopologically
distinct skeleton0-diagrams,
withSo(k)
relacedby S(k,p).] (41)
At p
= 0 both free
energies
reduce to Fo. To prove theirequality
for p#
0, it is necessary to show that~~~~ ~~i~
~~~~In
F, only
F'depends
on p. Consider the contribution to F' of all0-diagrams
with u internal lines. If we open up any of these ulines,
we getI-diagrams
of u-th order in p,contributing
to the correlation function. Let us call theself-energy
contributions of thesediagrams Z~(k, p),
the total u-th order
self-energy,
which has proper as well asimproper
contributions. In terms ofZ~(k, p),
the free energy F' is#ven by:
?'(»)
=
f ) / So(k)zu(k,») (43)
The factor
1/2u
in(43)
is due to the fact thatby closing
allpossible I-diagrams
with u I internallines,
theresulting 0-diagrams
are counted 2u times. This will now be shown in a littlemore detail. Consider an
arbitrary 0-diagram having
u internal lines.Opening
up any of the internal lines leads to a number oftopologically
distinctI-diagrams. Suppose
that aparticular
one can be obtained in n ways. Then the symmetry number So of the
0-diagram
is related to the symmetry numberSi
of thisparticular I-diagram
[13]:So " nsi
(44)
where it has not been decided which of the external lines will be associated with the wave
vectors k or -k. In the
0-diagram
there must be a group of n linesplaying
anequivalent
role(since opening
up any of these lines leads to the sameI-diagram).
In theI-diagram
this group has been reduced to n I internal linesplaying
anequivalent
role. For the symmetry numbers of thediagrams
thisimplies
So/Si
=
nl/(n I)I
= n, whence relation
(44). Using
this relation it can be shown thatby closing
allI-diagrams
that can be obtainedby opening
up any of theinternal lines of a certain
0-diagram,
the latter is counted u times.Similarly
all0-diagrams
arecounted u times
by closing
allI-diagrams having
u I internal lines.However,
since for theI-diagrams
obtainedby opening
up an internal line of a0-diagram,
there are still two ways ofassociating
the external lines with k or-k,
in fact all0-diagrams
will be counted 2u times if allI-diagrams, contributing
to the correlationfunction,
with u 1 internallines,
are closed.The derivative of F' with respect to p is
given by
p°1'(»~
=f / so(k, p)z~(k, p) (45)
»
u=~ k
Let us call the sum of the u-th order total
self-energies,
the totalself-energy, Z'(k, p).
Thisquantity
can be written in terms of the properself-energy Z(k, p):
Z'(k, p)
=Z(k,p)
+Zjk, p)So(k)Z(k, p)
+..~~(~)~~~(i~j)
~~~~
Using
thisequation,
it is found thatp~jj~~~
=-j / S(k, p)Z(k, p) (47)
P k
The p
dependence
of the free energy Fs ispartly implicit,
due to the pdependence
ofZ(k, p)
and
partly explicit
due to theexplicit
interactionsoccurring
in the skeleton0-diagram
con-tributing
toF]. By
the rules ofimplicit differentiation,
the derivative of Fs with respect to p isgiven by:
~~~~ ~~~~~~Z(k,p)
~
~~~~~)~
p
~~~~~
~~~~The skeleton
0-diagrams contributing
toF]
can be obtainedby closing
all skeletonI-diagrams.
Let
Z$(k,p)
be the total u-th orderself-energy according
to(18).
Contributions toZ$(k, p)
come from all skeleton
I-diagrams
with u I internal lines whereonly
thep's occurring explicitly
are used to determine the order. Since all skeletonI-diagrams
are proper,Z$(k, p)
is a proper
self-energy. By closing
all skeletonI-diagrams
with u I internallines,
as in(43),
every skeleton
0-diagram
with u internal lines will be counted 2u times.Therefore,
we haveFl(P)
"
f ) / S(k, P) ES (k, P) (49)
u=2
and hence
»
1°ll~~1
~~~~,
=
( S(k »)~S (k »)
=
/ S(k, p)Z(k, p) (50)
~ ~
Thus,
to prove theequality
of the derivatives of F and Fs, it isenough
to show that())~~~
=)S~(k, p)Z(k, p)
+(())~~~
= 0(51)
Differentiating
anarbitrary
skeleton0-diagram contributing
toF]
with respect toZ(k,p)
amounts to
opening
up any of its internal lines.By differentiating
all skeleton0-diagrams,
all skeleton
I-diagrams
are obtained with the proper symmetrynumber,
because ofequation (44),
apart from a factor 2. The factor 2 isagain
due to the fact that forI-diagrams
obtainedby opening
up an internal line of a0-diagram,
it has not yet been determined which of the external lines wil be associated with either k or -k.Therefore,
we have~
[contributions
of all skeleton0-diagrams]
bZ(k, p)
=
)[contributions
of all skeletonI-diagrams] (52)
whence
li~)~)
,~
~~
~~' ~~~~~' "~
~~~~This
completes
theproof
ofequation (37). Finally,
note that, uponsetting
p = I in(51),
we~~~~
~ik)
~ ~~~~
which can be used as a variational
principle generating
theequation
forZ(k)
from the free energy functional.Appendix
B.An estimate of the second order term.
In the
approximation
N= 2, an
in-phase
modezi(x)
and anout-of-phase
modez~(x)
aredefined as
z;(x)
+)(zi(x)
+z2(x)) (55)
zo(x)
+)(zi(x) z2(x)) (56)
Only
theout-of-phase
mode contributes to the relevant free energy. Thus we arrive at aHamiltonian in terms of
z~(x):
~ijz~j
=/ d2x )(az~(x))2
+
v(D
+vsz~(x))
+v(D vsz~(x) (57)
To find the free energy of two
membranes,
per unit area, with respect to freemembranes,
the free energy of a free zo mode must be subtracted. In this case there is noperpendicular
Fouriertransform,
and the two-dimensionalin-plane
wavevector will be denotedby
k. Since apossible k-dependence
of the correlationlength
is not taken into account, the correlation function is givenby:
~~~~ ~~°~~~~°~ ~~~
K(k4 4A~4)
~~~~and the
amplitude
of the undulationsby
jd~
=/ S(k)
=
~j (59)
k 16
"~~~~~
j j°°
d~k(60)
k
~ -co
(2~)~
The vertex factor of an n-th order vertex
(n
= 4, 6,
8,..)
now becomes2"/~+~V~"'(D) ~
~(61)
ki,k~,.. k«
where the asterisk indicates that the k summation is restricted to the values of k for which
ki
+ k2 + + kn= 0. The second order contribution
bF2
to the free energy, per unit area, permembrane,
can be written asb?2
=-A~V~(D)) exP((~d)~)f((~d)~) (62)
The part
exp((~d)~) /2
is due to thetadpoles
of the second orderdiagrams
shown infigure
2a.The dimensionless function
f((~d)~)
is due to the middle part of the second orderdiagrams.
It is
given by:
~~~~~
2
~~~
k,
,
k~
k
~~~~~ ~~~~
1 1
2 2
nJ 2m n2
1 1 1
2 2
~J
order
diagrams
shown infigure
2b(these
are not all thediagrams contributing
to thirdorder).
Obtaining
thesediagrams
from the second orderdiagrams
offigure
2a is similar toobtaining
the second order
diagrams
from the first orderdiagrams
offigure
I.Therefore, bF3/bF~
will beroughly proportional
tobF2/bfi,
I-e- if(~d)~
>2/3 (~D),
notonly
will the second order term belarge
with respect to the first order term, but also the third order term will belarge
with respect to the second order term.
References
[Ii Odijk T., Langmuir 8
(1992)
1690.[2] Odijk T., Europhys. Lett. 24
(1993)
177.[3] Evans E-A-, Parsegian V-A-, Proc. Nat]. Acad. Sci. U-S-A. 38
(1986)
7132.[4] Podgornik R., Parsegian V-A-, Langmuir 8
(1992)
557.[5] Evans E., Ipsen J., EJectrochimJca Acta 36
(1991)
1735.[6] Pincus P., Joanny J., Andelman D., Europhys. Lett. ll
(1990)
763.[7] Helfrich W., Z. Naturforsch. 33a
(1978)
305.[8] David F., J. Phys. CoJJoq. France 51
(1990)
C7-lls.[9] Israelachvili J. N., Intermolecular and Surface Forces
(Academic
Press, Orlando,1985).
[10] Mitchell D-J-, Ninham B-W-, Langmuir 5
(1989)
l121.ill]
Lekkerkerker H-W-N-, Physica A 159(1989)
319.[12] Winterhalter M., Helfrich W., J. Phys. Chem. 92
(1988)
6865.[13] Lee T-D-, Yang C-N-, Phys. Rev. l17
(1960)
22.[14] Luttinger J-M-, Ward J-C-, PJJys. Rev. l18
(1960)
1417.[15] Podgornik R., Rau D-C-, Parsegian V-A-, Macromolecules 22
(1989)
1780.[16] Odijk T., Biophys. Chem. 46
(1993)
69.[17] Tsao Y., Evans D-F-, Rand R-P-, Parsegian V-A-, Langmuir 9
(1993)
233 [18] Dubois M., Zemb T., Langmuir 7(1991)
1352.[19] Roux D., Safinya C-R-, J. Phys. FYance 49 (1988) 307.
[20] Dubois M., Zemb T., Belloni L., Delville A., Levitz P., Setton R., J. Chem. Phys. 96