3

Electroelastic Instabilities in Double

Layers and Membranes
MICHAEL B. PARTENSKII and PETER C. JORDAN Department of
Chemistry, Brandeis University, Waltham, Massachusetts

I. INTRODUCTION

Electrical double layers and membranes have certain common features. Both are quasi-
two-dimensional structures. In both, variation of applied voltage influences the distribu-
tion of charge as well as the structural parameters (effective thickness). And in both cases
coupling between the electric field and these parameters can lead to instabilities and phase
transitions. These instabilities may be described and understood using an electroelastic
metaphor first introduced in 1973 by Crowley [1] in relation to membrane breakdown
(‘‘electroporation’’). Although this model failed to quantitatively describe the onset of
instability at small voltage with little deformation, it generated significant interest in
studies of the electroelasticity of membranes. In some aspects, the electrical response of
membranes is similar to the action of an external pressure (such as osmotic stress). But it
also has features with no direct analogy to mechanical effects, such as specific interaction
of an applied voltage with surface undulations which can cause instabilities in membranes.
Similar phenomena appear equally important in the behavior of electrochemical interfaces
where electrical stress affects structural parameters of electrical double layers. These are
most closely related to the question of whether, for electrical double layers, it is possible
for the differential capacitance, C, to be negative. It has been shown that the prediction of
C < 0 in some models of double layers has a direct analogy in the behavior of the elastic
capacitors. And, as in membranes, it is also related to possible instabilities and interfacial
phase transitions at the interfaces.
This chapter is devoted to the behavior of double layers and inclusion-free mem-
branes. Section II treats two simple models, the ‘‘elastic dimer’’ and the ‘‘elastic capaci-
tor.’’ They help to demonstrate the origin of electroelastic instabilities. Section III
considers electrochemical interfaces. We discuss theoretical predictions of negative capa-
citance and how they may be related to reality. For this purpose we introduce three sorts
of electrical control and show that this anomaly is most likely to arise in models which
assume that the charge density on the electrode is uniform and can be controlled. This ‘‘-
control’’ is a convenient theoretical construct, but in real applications only the total charge
or the applied voltage can be fixed. We then show that predictions of C < 0 under -
control may indicate that in reality the symmetry breaks. Such interfaces undergo a
transition to a nonuniform state; the initial uniformity assumption is erroneous. Most

51

Copyright © 2001 Marcel Dekker, Inc.

 52 Partenskii and Jordan

of this discussion is couched in terms of the electroelastic analogy. Section IV describes

electroelastic phenomena in inclusion free membranes with possible relation to instabil-
ities. We first treat the Crowley model (a rigid interface). We then consider the ‘‘hydro-
dynamic,’’ ‘‘electroporative,’’ and ‘‘smectic bilayer’’ perspectives on the initiation of an
electroelastic instability. Finally, we describe some attempts to modify the smectic model
to account for the basic experimental observation: membrane rupture occurs at low vol-
tages with little electrostriction. These are based on the idea that nonlocality of the elastic
moduli softens the symmetrical modes thus enhancing thickness fluctuations at short
wavelengths. Possible consequences of this hypothesis for different membrane properties
are discussed.

II. ELECTROELASTIC INSTABILITIES. SIMPLE EXAMPLES

A. Charges on a Spring (Elastic Dimer)
The nature of the instabilities related to coupling of electrical and elastic degrees of free-
dom can be illustrated by a very simple example. Consider two charges, q and q, sepa-
rated by a spring. The corresponding Hamiltonian is
q2
H¼ þ E0 ðlÞ ð1Þ
40 l
where E0 is the bond energy. In an harmonic approximation we write
K
E0 ¼ ðl  l0 Þ2 ð2Þ
2
where K is a ‘‘spring constant,’’ l is the distance between charges, and l0 is the length of the
isolated spring. For the sake of convenience we use dimensionless units
H q l
W¼ Q ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi z¼ ð3Þ
Kl02 40 Kl03 l0

so that
Q2 1
W ¼ þ ð1  zÞ2 ð4Þ
z 2
Global behavior of the system is determined by a single parameter, the reduced charge Q;
relative distance z is the internal variable, defined by the equilibrium condition
@W @H
¼0 which follows from ¼0 ð5Þ
@z @l
Before proceeding further we analyze the behavior of the system by viewing its energy
surface WðQ; zÞ (Fig. 1). At short distances, z ! 0, the elastic force cannot compete with
the electrostatic attraction and W ! 1 for any Q 6¼ 0. For sufficiently smallp jQj a local
ffiffiffiffiffiffiffiffiffiffi
minimum 0 < z 1 exists. It disappears when Q reaches its critical value Qc ¼ 4=27 and
z has contracted to zc ¼ 2=3. Both Qc and zc can be defined by solving Eq. (5) together
with the marginal equilibrium condition. As we see, ‘‘molecular equilibrium’’ at finite z, if
its exists, is always local, while globally, charges are stable only in direct contact (the
collapsed state). In reality, however, additional short-range repulsion would prevent the
system from collapsing. Figure 2 incorporates this additional interaction. We see that the

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 53

FIG. 1 Energy of elastically coupled charges.

FIG. 2 Energy of elastically coupled charges with an additional short-range repulsion term.

Copyright © 2001 Marcel Dekker, Inc.

 54 Partenskii and Jordan

‘‘molecule’’ becomes bistable, and Qc now corresponds to charge at which the barrier
between the two minima disappears.

B. Elastic Capacitors
Our elastic molecule illustrates the onset of instability. However, conditions when charges
can be smoothly controlled by the observer are hard to achieve. In applications and studies
of interfacial phenomena the controlled electrical variable is usually a voltage V across the
interfaces rather than charges involved in their polarization. The distributions of charges
composing electrical double layers of different origin can be approximated by ‘‘interfacial
capacitors.’’ As we will see in Section III, the distributions of charges are flexible and
strongly affected by the applied voltage. As a result, not only the charge q on the ‘‘plates’’
of interfacial capacitors, but also the effective ‘‘gap’’ between the plates depends on V.
This naturally leads to ‘‘relaxing gap’’ capacitors, which in many cases can be described as
‘‘elastic capacitors’’ (ECs).
The energy of an EC under fixed voltage is [2]
H ¼ H0 ðq; lÞ  qV ð6Þ
where
q2 l
H0 ðq; lÞ ¼ þ AE0 ðlÞ ð7Þ
20 A
is the energy of isolated capacitor with charge q, A is the surface area of its ‘‘plates,’’  and
l are the effective dielectric constant and the gap width respectively, and E0 is the elastic
energy per unit area. The contribution qV in Eq. (6) is responsible for charge exchange
between the potentiostat (battery) and the capacitor. The equilibrium conditions
@H @H
¼0 ¼0 ð8Þ
@l @q
@2 H
>0 ð9Þ
@2 l
define the equilibrium gap lðVÞ and equilibrium charge qðVÞ. Notice that Eq. (7) assumes
that charge distributions are uniform. The validity of this assumption will be discussed in
Section III. In the uniform case all properties can be equally well described by the surface
charge density
q
¼ ð10Þ
A
In equilibrium, as follows from Eqs. (7) and (8)
2 @E
þ 0¼0 ð11Þ
20 @l
l
¼V ð12Þ
0
Equation (12) leads directly to an expression of H through V and l:
!
0 V 2
H¼  þ E0 ðlÞ A ð13Þ
2l

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 55

Equation (13) becomes equivalent to Eq. (1) if q2 is replaced by V 2 =2. In the units of Eq.
(3), introducing the dimensionless voltage,
V
 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð14Þ
Kl03 =0

and using Eq. (2), we have

2 1
H¼ þ ð1  zÞ2 ð15Þ
2z 2
From the previous analysis it follows that the harmonic elastic capacitor collapses when 
approaches its critical value
rffiffiffi
2 2
c ¼ and zc ¼ 2=3 ð16Þ
3 3

C. General Observations
Several questions must be addressed with respect to the simple examples just outlined.
Despite their formal similarity, it is important to keep in mind that the first instability
describes an isolated system, where charge is a controlled variable. In contrast, the EC as
introduced becomes unstable only in contact with a potentiostat (battery), when V is fixed
and charge can change. Thus, the collapse of an EC leads to infinite growth of q  1=l,
accompanied by electric current from a battery to the EC.
If isolated, the EC remains stable for all values of q corresponding to l > 0, as can be
seen from Eq. (7). The major difference from the elastic dimer is that electrical energy of
an isolated EC is positive and finite for all q and l while for the dimer it goes to 1 when
l ! 0. In the next section we reexamine this result and show that the uniformity assump-
tion for the charge distribution  across the plates of a capacitor can fail, and this can
influence the stability condition for an isolated EC.
Another observation should be made with respect to the term ‘‘elastic’’ in describing
interfacial capacitors. It was originally introduced by Crowley [1] for membranes and
reflects the compressibility of lipid layers which behave in some respects like an elastic
film. Its relation to electrochemical interfaces is less obvious. Consider an interface
between a metal electrode and an electrolyte. As we will see in Section III, the effective
gap of the interfacial capacitor is the distance between the centers of mass of the electronic,
‘‘e,’’ and ionic, ‘‘i,’’ charge density distributions ne;i :
l ¼ xi  xe ð17Þ
Ð þ1
1 ni;e ðxÞx dx
xe;i ¼ ð18Þ
e;i
where e ¼ i ¼ , the surface density of electron charge. Both distributions depend on 
(and V) and l usually changes with charging. Consequently, the interface may be repre-
sented as a ‘‘relaxing gap’’ capacitor. Consider a hypothetical attempt to alter the equili-
brium electron and ion density profiles, causing the displacement of xe and xi from their
equilibrium positions. The corresponding increase in free energy can be expressed through
the effective restoring force opposing the variation of l from its equilibrium value. In turn,
the restoring force can usually be expressed in terms of an effective elastic constant. This is
where the analogy with the EC comes into play.

Copyright © 2001 Marcel Dekker, Inc.

 56 Partenskii and Jordan

The ‘‘elasticity’’ can be related to very different contributions to the energy of the
interface. It includes classical and nonclassical (exchange, correlation) electrostatic inter-
actions in ion–electron systems, entropic effects, Lennard–Jones and van der Waals-type
interactions between solvent molecules and electrode, etc. Therefore, use of the macro-
scopic term should not hide its relation to ‘‘microscopic’’ reality. On the other hand,
microscopic behavior could be much richer than the predictions of such simplified electro-
elastic models. Some of these differences will be discussed below.

III. NEGATIVE DIFFERENTIAL CAPACITANCE AND ELECTRICAL

DOUBLE LAYER STABILITY
A. Historical Review
The traditional treatment of a double layer at electrode–electrolyte interfaces is based on
its separation into two series contributions: the compact (‘‘Helmholtz’’) layer and the
diffusive (‘‘dif’’) layer, so that the inverse capacitance is
1 1 1
Cdl ¼ CH þ Cdif ð19Þ
Both terms affect whether the double layer capacitance can become negative.

1. ‘‘Molecular Capacitor’’ (MC) and the Cooper–Harrison Catastrophe

The problem of negative capacitance first arose in theoretical studies of CH based on so-
called ‘‘molecular models.’’ Molecular or ‘‘dipolar capacitor’’ models treat the layer of
water molecules adjacent to a metal electrode as a lattice of point or finite-size dipoles
[3,4]. Many studies have focused on the response of such a layer to electrode charging and
comparison with the measurements of the compact layer capacitance, CH ðÞ using the
technique suggested by Grahame [5,6]. Much of this work has been inspired by the
‘‘Cooper–Harrison (CH) catastrophe,’’ the prediction of negative CH ðÞ in molecular
models of the compact layer. Here we only briefly discuss the origin of this anomaly
(see Ref. 7 for more details and references).
(a) General Characteristics of an MC. Consider the isolated MC with a charge density
 on its plates. Focusing on the capacitance C (the index ‘‘H’’ is omitted), we first intro-
duce the potential drop
across the MC as

¼ F0 d þ
d ð20Þ
where*
F0 ¼ 4 ð21Þ
is the external field created by the plates of MC, and

d ¼ 4Px Px ¼ Ns hpx i ð22Þ
in which hpx i is the average normal-to-plane projection of the dipolar moment, and Ns is
the number of dipoles per unit area.
Using the definition of differential capacitance (per unit area),

* Unlike the rest of the chapter, here we adopt Gaussian electrostatic units which are typically used
in the literature devoted to molecular capacitors and double layer theory.

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 57

C¼ ð23Þ
d
we find
1
¼ 4ðd  4 Þ ð24Þ
4C
where
dhpx i
 ¼ Ns ð25Þ
dF0
is the susceptibility of the dipolar lattice. It follows from Eq. (24) that if it were possible
that
  d=4 ð26Þ
then it would be possible for C to become negative.
(b) The MC in a Mean Field Approximation (MFA). In the majority of papers includ-
ing the original CH paper [8], the MC was treated in Ising-type n-state models with dipoles
perpendicular to the plates: px ¼ psx where the vector sx is limited to equally spaced values
between 1 and 1. We consider the simplest ‘‘spin 1/2’’ model n ¼ 2 (Fig. 3), a choice
justified by the fact that the appearance of the CH catastrophe is independent of n.
The local field acting on a specific dipole in a lattice is
F ¼ F0 þ Fd þ F 0 ð27Þ
0
where the field F describes interaction between a dipole and its own images. Interaction
with this field does not depend on the dipolar orientation and can be omitted [7]. The field
Fd created by the rest of the dipolar lattice has been a subject of considerable controversy.
Following most authors, we restrict analysis to the MFA assuming that all dipoles except
one have moments equal to their mean value hpx i. Then,
nL Jhsx i
Fd ¼  ð28Þ
p
where J ¼ p2 =a3 is the interaction energy between antiparallel nearest neighbors, a is the
lattice constant, and nL is the effective number of nearest neighbors. For a bare lattice
where dipolar fields are not shielded by the plates of capacitor, nL equals
X 3
n0L ¼ a3 r1i ð29Þ
i

n0Lis  9 and  11 for square and hexagonal lattices respectively.

Focusing on the CH catastrophe we notice from Eq. (26) that the most ‘‘dangerous’’
point is where  is largest. In the MC models considered this obviously corresponds to
 ¼ 0; increasing  orients dipoles and makes them less susceptible to variation of the

FIG. 3 Molecular capacitor in ‘‘spin 1/2’’ model.

Copyright © 2001 Marcel Dekker, Inc.

 58 Partenskii and Jordan

external field (in the limit  ! 1 the lattice is totally frozen and  ! 0). In this ‘‘zero
charge’’ point for ‘‘spin 1/2’’ [8] we have

Ja3 Ns
 ¼ ð30Þ
kT þ JnL

from which is clear that two factors oppose the orientation of dipoles in the MC: tem-
perature and dipolar interactions. The latter occurs because in the Ising model considered,
the interaction energy between two lattice dipoles is minimal when dipoles are opposed.
Equation (30) emphasizes the importance of the effective number of nearest neighbors;
with p, Ns , and T fixed, the susceptibility becomes larger for smaller nL .
(c) Effective Number of Nearest Neighbors and the CH Catastrophe. It follows from
Eqs. (30) and (26) that C becomes negative if

4a3 Ns kT
nL <  ð31Þ
d J
pffiffiffi
For a hexagonal lattice, Nshex ¼ 2= 3a2 ; choosing for water dipoles at room temperature
J  2kT [3], and assuming a ¼ d, this leads (see Ref. 3) to

L < 14:2
nhex ð32Þ

L 11. CH made the apparently reasonable

As mentioned before, for a bare lattice nhex
assumption that screening of the dipolar field by the conductive plates of the MC should
decrease nhex
L below this value, thus insuring that Eq. (32) is satisfied, leading to the
catastrophe.
(d) Summation Over the Dipolar Images and Resolution of the CH Catastrophe. To
account for the screening of the dipolar lattice by the plates of the MC it is convenient
to use the method of images. Every original dipole in this approach creates an infinite
chain of similarly oriented image dipoles, with period d. As a result, calculation of local
fields in MFA requires summation over the infinite lattice of equal dipoles, hpx i, excluding
the images corresponding to the ‘‘original’’ dipole. Due to the conditional convergence of
the corresponding series* [9], the result depends on the order in which the sums are
calculated. This ambiguity is the cause of the CH catastrophe. Two ways of carrying
out the summation have been discussed [10].
1. The ‘‘chain’’ (or c-) approach starts with the chain of dipoles normal to the
original plane, including an original dipole located in a lattice site and the
infinite set of its images. The field of each chain decays exponentially with
distance , much faster than the  1=3 law corresponding to the bare lattice.
It leads to a decrease in the effective number of nearest neighbors and seemingly
gives rise to the CH catastrophe. It also seems to justify a suggestion that if the
lattice constant a is large enough, the CH catastrophe becomes inevitable [11].
2. The ‘‘plane’’ (p-) approach starts with the infinite plane of dipoles, with further
summation over the infinite stack of these planes (including the original lattice
and all its images). As a result,

* The dipolar contributions to the local field diverges if the terms are replaced by their absolute
values.

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 59

d
F pl ¼ F ch þ Fu Fu ¼  ð33Þ
d
Surprisingly, in addition to the ‘‘chain’’ contribution, F ch , the local field calcu-
lated in the p-approach also contains the uniform part, Fu ¼ 4Px =d. This last
term noticeably increases the effective number of nearest neighbors so that the
CH catastrophe disappears.
But does the catastrophe really exist in the chain approach? To underline the pro-
blem we used the word ‘‘seemingly’’ with respect to the appearance of the CH anomaly.
We now consider this more carefully. Notice that in the chain summation each dipole is
treated as if it were inserted between the infinite plates of capacitor. No matter how many
dipoles we have, they do not affect the potential drop across the gap of the MC. It can
only mean that the average polarization, Px , in Eqs. (20) and (22) equals zero. In its turn,
Px is proportional to Ns  Ad =A, where A and Ad are the areas of the plate and of the
dipolar array respectively. This means that the chain approach is valid when Ad =A ! 0. In
other words, even for the infinitely large dipolar lattice, its size is infinitesimally small in
comparison to A. Since in this picture Px ! 0, according to Eq. (24) the capacitance C !
4=d > 0 and the CH catastrophe disappears.
The p-approach corresponds to Ad =A ! 1 with both A and Ad ! 1. In this case
the CH catastrophe disappears because of the uniform contribution which increases the
effective number of nearest neighbors.
In other words, both summation procedures are valid.* They describe different
physical situations. In the c-summation the size of the dipolar array is much smaller
than the size of the capacitor, while in the p-approach both are equal. One can imagine
an infinite number of different summation orders corresponding to different relations
between A and Ad , leading to results between the p- and c-limits. To solve the problem
unambiguously, one must consider a finite dipolar lattice between finite plates, which
requires accounting for edge effects and makes the whole problem much more compli-
cated.
(e) Conclusion, and a Further Question. We have seen that the CH catastrophe arises as
a result of inconsistencies in the treatment of the electric field and the potential in the
molecular capacitor. Were they analyzed consistently, then, at least in the framework of
Ising models and the MFA, a catastrophe would not appear and the capacitance of an MC
would always be positive. But another question arises. Does this imply that the capaci-
tance of a double layer can never become negative? Before addressing this question we will
analyze some other occurrences of the C < 0 problem in the theory of electrified interfaces.

2. Diffuse Electron and Ionic Distributions in a Double Layer

and the Problem of C < 0
Statistical mechanical and microscopic theories of the double layer have often been inten-
sively reviewed. Our focus is only on features related to the problem of negative C.

* The following example helps in visualization of the difference between the two ways of summa-
tion. Consider the potential difference 
ðL; RÞ ¼
ðLÞ 
ðLÞ between points L and L on the
axis of a uniform dipolar disk (where Px is the surface dipole moment density) in the limit of
infinite R and L. Since 
ðL; RÞ ¼ 4Px ½1  L2 =ðL2 þ R2 Þ, limR!1 limL!1 
ðL; RÞ ¼ 0, while
limL!1 limR!1 
ðL; RÞ ¼ 4Px . The first result obviously corresponds to the c-approach, while
the second one corresponds to the p-approach.

Copyright © 2001 Marcel Dekker, Inc.

 60 Partenskii and Jordan

(a) Diffuse Models of Ionic Distributions in Electrolytes. The first model of the diffuse
ionic distribution in electrolytes was developed by Gouy and Chapman (GC) many years
ago. In this and similar ‘‘local models’’ the charge density is considered to be a local
function of the potential
ðxÞ,
nlocal ðxÞ ¼ f ½
ðxÞ ð34Þ
In such models capacitance can never become negative [12]. The GC model for a 1 : 1
electrolyte is illustrative:
Cdif ¼ aðb þ  2 Þ ð35Þ
where a and b are positive and depend on temperature, ionic concentration, and bulk
dielectric constant. However, the question of C < 0 inevitably emerges in theoretical
studies of the diffuse layer that go beyond the local approach (see, e.g., Refs. 13–18).
This anomaly is related to some specific properties of ionic distributions.
(b) Properties of the Ionic Distribution Which Can Cause a Capacitance Anomaly. Let
us examine the features of an ionic distribution that can lead to Cdif < 0. Quite generally,
the potential drop across the double layer,



ð1Þ 
ð1Þ (where x ¼ 1 and x ¼
1 correspond to the bulk electrode and electrolyte respectively) can be represented as


¼ 4 ½xi ðÞ  xe ðÞ  4Px ð36Þ
Then, using Eq. (23) we find
1 dl dPx
¼ lðÞ þ   ð37Þ
4C d d
where the effective gap of the interfacial capacitor, l, is defined by Eq. (17). Focusing on
the effect of the ionic distribution, we consider the so-called primitive model where ions
occupy a region of uniform dielectric constant . In addition we temporarily ignore the
diffuseness of the electronic distribution (the perfect conductor model) assuming
xe ¼ const: ¼ 0. Then for the diffuse layer capacitance
ð ð1
1 @ 1  
ð0Þ
¼ n ðxÞx dx n ðxÞx dx=
ð38Þ
4Cdif ðÞ @ 0 0 
Here we assume
ð1Þ ¼ 0 and choose  > 0. It is now clear that differential capacitance
becomes negative at some value of  if the centroid of the ionic distribution induced by a small
variation  becomes negative. Given that ions do not penetrate the region x < 0, this can
only happen if nðxÞ is a nonmonotonic function of variable polarity.
However, the total electrolyte charge density, nðxÞ, is not necessarily oscillating, and
it can even be monotonic. Figures 4–6 illustrate the various possibilities. nðxÞ represents a
nonmonotonous charge density corresponding to some variation of . It is a function
constructed to reflect two features of the ionic shielding of the electrode charge.
Ð1
1. Neutrality condition (required): 0 nðxÞ dx ¼ .
2. ‘‘Charge condensation’’ near the electrode (quite a typical feature, which is
necessary but not sufficient for attaining C < 0): charging is accompanied by
increased charge density near the electrode at the expense of some depletion in
the tail regions of the ionic distribution.
Figure 4 depicts a generalized Raleigh [19] model originally designed to explain
extremely high values of interfacial capacitance in contacts of metals and solid electrolytes,
silver-ion conductors (such as AgCl). This is a two-layer model satisfying conditions 1 and

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 61

FIG. 4 Induced potential drop 

ðxÞ in the Raleigh model of the ionic distribution. k ¼ 0:2
(curve 1), 0.33 (curve 2), and 0.4 (curve 3). Gray area shows the Raleigh distribution of the induced
charge with a ¼ 0:3 nm and k ¼ 0:33 (see text) which corresponds to curve 2. The total induced
charge is normalized

2. Charge densities in the layers are n1 at x < a and kn1 at a < x < 2a; the layer thickness
a and the dimensionless constant k < 1 are the parameters of the model. Using this to
mimic the distribution induced by , we find that n1 ¼ =að1  kÞ. The center of mass
of this distribution is

n1 a2 að1  3kÞ
xRaleigh ¼ ð1  3kÞ ¼ ð39Þ
i
2 1k
xi becomes negative for k > 1=3. Figure 4 demonstrates that C becomes negative when xi
is negative. This transition is obvious from the behavior of 
ðxÞ [see Eq. (38)]. For curve 1
(k ¼ 0:2) the additional barrier 
ð0Þ ð 0:75Þ has a ‘‘normal’’ positive sign; for curve 2
(k ¼ 1=3) this barrier disappears, and for curve 3 ðk ¼ 0:4) it becomes negative ( 0:5).
Figure 5 presents a more realistic picture where nðxÞ is continuous. QðxÞ is the total
ionic charge accumulated between x ¼ 0 and the point x:

FIG. 5 Model-induced charge density profile nðxÞ (curve 1), QðxÞ (curve 2), and 
ðxÞ (curve 3) in
the diffuse layer ( ¼ 0:2, dimensionless units).

Copyright © 2001 Marcel Dekker, Inc.

 62 Partenskii and Jordan
ðx
QðxÞ ¼ nðx 0 Þ dx 0
0

and
ðx

ðxÞ ¼ 4½ þ QðxÞx þ x 0 nðx 0 Þ dx 0 ð40Þ
0

It demonstrates an important feature promoting negative C: ‘‘overshielding’’ of the elec-

trode charge  in the adjacent region of electrolyte. This is the cause of depletion in the
more distant parts of the double layer causing effective displacement of the charge cen-
troid towards the electrode.* Figure 6 demonstrates the possibility that if the initial
potential profile
0 ðxÞ corresponding to some 0 is monotonic (GC-type), then a small
additional nonmonotonic variation in 
ðxÞ arising from the additional charge density of
Fig. 5 and leading to C < 0 may still leave the total potential profile
ðxÞ monotonic. The
same is true for the total charge density nðxÞ. This speculation mimics the results of
numerical experiments described in Refs. 17 and 18. The same can be true for charge
density profiles. It demonstrates that charge density oscillations are neither necessary nor
sufficient to give rise to negative C.

3. Negative Capacitance in Primitive Models of Electrolyte

A possible reason that the problem of C < 0 did not receive much attention was the
assertion [15] (BLH) that such an anomaly was forbidden. The proof was based on the
statistical mechanical analysis of the primitive model of electrolytes between two oppo-
sitely charged planes,  and . It was noticed in Ref. 10 that the BLH analysis missed a
very simple contribution to the Hamiltonian, direct interaction between the charged walls,
2L 2 (L is the distance between the walls). With proper choice of the Hamiltonian the
condition on the capacitance would be C > 2=L. It simply means that due to ionic
shielding of the electric field, the capacitance exceeded its ‘‘geometrical’’ value correspond-
ing to the electrolyte-free dielectric gap.

FIG. 6 Potential profile for 0 (curve 1) and 0 þ  (curve 2).

* Such a behavior can represent, for instance, condensation near an electrode of a monotonous
charge density n ðxÞ following a variation of  [18].

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 63

This analysis has been developed further [20–22]. At the same time attention focused
on the analogous anomaly in the theory and simulations of the double layer [17,18].
Progress was, however, slightly misdirected by the assertion [21] that there are no restric-
tions on the sign of C, regardless of the model considered or the choice of experimental
conditions. In what follows we will show that there are severe restrictions strongly limiting
the possibility of capacitance anomalies. But here we only stress that it has never been
proven that negative capacitance is forbidden for any model of an electrolyte at a uni-
formly charged interface.

4. Contribution of Metal Electrons and the Negative Capacitance Problem

In the early 1980s it became clear that metal electrons play an important role in the
electrical response at metal–electrolyte interfaces. Different aspects have been extensively
reviewed [23–27]. Here we only briefly discuss the problem of negative capacitance, and
how it arose in these studies.
Although it was clear that separation of an interface into ‘‘surface’’ and ‘‘bulk’’
components as in Eq. (19) is artificial and must disappear in a consistent microscopic
analysis, electronic effects were initially discussed in terms of a ‘‘compact layer’’ and its
capacitance CH . It was apparent early on that the electrons strongly influence double layer
properties [28–33].
In traditional models of an electrified interface, metal electrons are artificially loca-
lized within the electrode. This leads to misinterpretation of the electronic influences on
the compact layer. CH in those models would always be smaller than its ‘‘ideal conductor’’
limit (with electrode charge spread over an infinitesimally thin region at the electrode
surface x ¼ 0)

4CH IC
¼ H ð41Þ
xH
where xH is the effective width of the compact layer, and H is the effective permittivity.
Self-consistent analysis of electronic effects altered this simple view substantially. By
analogy to Eqs. (36) and (37), compact layer capacitance in the simplest model, accounting
only for electronic effects, is now defined by
1 dx
¼ xH  xe   e ð42Þ
4CH d
It was first noticed in Ref. 31 that at moderate negative  the condition
dxe
xe    xH ð43Þ
d
may be fulfilled for contacts of metals with solid electrolytes (silver-ion conductors), so
that CH ðÞ can become negative. Similar anomalies have been found in contacts of metals
with solvents [34,35], results which have not been easily accepted. Moreover, in some
calculations they have been carefully avoided, partially in the belief that they are prohib-
ited by the BLH theorem (for a discussion see Refs. 7 and 36).

5. Electroelastic Models. Proof that Negative Capacitance is Admissible for

Uniformly Charged Interface Models
Elastic capacitors (Section II) are very useful as ‘‘electromechanical’’ analogs of micro-
scopic interfacial capacitors [22,31,34]. But most importantly, they demonstrate that nega-

Copyright © 2001 Marcel Dekker, Inc.

 64 Partenskii and Jordan

tive capacitance can coexist with interfacial equilibrium and stability as long as the total
charge is fixed and the charge density is kept uniform.
Under these conditions, corresponding to so-called -control [37], the elastic capa-
citor is described by the Hamiltonian, Eq. (7). In dimensionless units this becomes
1
W ¼ s2 z þ ð1  zÞ2 ð44Þ
2
where s ¼ ð4=kl0 Þ2 , the dimensionless charge density. Other units are the same as in Eq.
(3). From the equilibrium conditions for fixed s it follows that [37]:

s2 s2 3
zðsÞ ¼ 1  vðsÞ ¼ s  C 1 ¼ 1  s2 ð45Þ
2 3 2
and C1 0 for jsj > s0 ¼ ð2=3Þ1=2 . At the same time, @2 W=@l 2 ¼ 1 > 0, and in this region
the system is stable. Thus capacitance under -control can be negative. And as this is
possible for one simple model, it should also be possible for some other models obeying
the same conditions of -control.

B. The Sign of C and Different Types of Electrical Control

In the analysis of molecular capacitors, the diffuse layer and elastic capacitors, we have
always assumed that the electrode charge density  could be controlled. Under such
conditions it is generally possible for C to become negative while the system remains
stable. For example, contraction of the gap z in an elastic capacitor proceeds smoothly
with  growing until the plates come in contact, while C becomes negative for z < 2=3. At
the same time, as shown in Section II for an EC connected to a battery, the EC collapses
after z  0:6 is reached. How can these seemingly contradictory results be reconciled? And
how can -control be related to reality? Is C < 0 observable? These questions are
addressed in this section.

1. Three Types of Control

(a) -Control. In the vast majority of studies in double layer theory, the charge density
 is considered an independent control variable. This forms the framework where the
question of negative C originated. But this assumption must be treated very carefully.
In practice there is no way to control . Therefore, -control is a purely theoretical con-
struct. It works perfectly well only if an interface ‘‘chooses’’ to remain uniform.*
(b) Isolated Capacitor, q-Control. In real experiments with ‘‘ideally polarizable’’{ inter-
faces there are basically two options: either the cell is isolated, or it is connected to a
potentiostat (battery). For an isolated capacitor the total charge on the electrode is fixed
(‘‘q-control’’). If the system remains uniform, then - and q-controls are equivalent, 
being defined by Eq. (10). However, when for any reason a uniform distribution becomes
unstable, then only q-control is a valid description of the isolated interface.

* We are not describing microscopic structural nonuniformity, which is discussed in another chapter
of this book, and assume that the electrode surface is uniform.
{ We exclude situations where there is electric current across the interface which would cause some
additional complications.

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 65

(c) Global System,

-Control. Now consider a cell connected to a battery. Usually, the
subsystem is immersed between two metal electrodes. Here electrical control of the cell is
maintained via the electronic subsystems of the electrodes [38]. The real controlled variable
is the difference between the electrochemical potentials of electrons of both electrodes,
 1
and
 2 . To avoid the complexities due to contact phenomena (‘‘Volta potential’’) we
assume both electrodes are made of the same metal and therefore have the same chemical
potentials (Fermi energies). The difference between chemical potentials equals the voltage:

 1 
 2

¼ ð46Þ
e
In this case the voltage
across the interface is the actual controlled variable and we deal
with
-control.

2. Stability Conditions
(a) C is Nonnegative Under
-Control. The transition from an isolated to a global system
is accompanied by transformation of the Hamiltonian [see Eq. (6)]:
H ¼ H0  q
ð47Þ
It is generally true (for a discussion see Ref. 39) that for Hamiltonians of the form
H ¼ H0  XF ð48Þ
the susceptibility is
1  2
@X=@F ¼ ðX  hXiÞ  0 ð49Þ
kT
where h  i denotes a statistical average with the Hamiltonian, Eq. (48).
The transition from isolated ðH0 Þ to global ðHÞ belongs to this class [see Eq. (6)],
which is why capacitance under
-control is defined as
@hi A  2
C¼ ¼ ð  hiÞ ð50Þ
@
kT
[40–43] and is strictly nonnegative.
(b) Isolated Capacitor. Suppose now that the charge q is fixed. As mentioned, if 
could be kept uniform, then C can unconditionally become negative. The elastic capacitors
we have studied are examples of systems for which - and q-controls are equivalent.
Indeed, with plates rigid, any nonuniform redistribution of charge density can only
increase the EC’s energy, and therefore cannot become stable. It would be different if
the plates of an EC were somewhat deformable. Transition to a nonuniform state could
then occur due to coupling between the electrical and ‘‘elastic’’ degrees of freedom. In
reality, the electronic and ionic ‘‘plates’’ of an interfacial capacitor are never absolutely
rigid. Therefore, a uniform distribution should always be tested with respect to its stability
under some nonuniform perturbations.
Recently [7] we constructed an example showing that interfacial flexibility can cause
instability of the uniform state. Two elastic capacitors, C1 and C2 , were connected in
parallel. The total charge was fixed, but it was allowed to redistribute between C1 and
C2 . It was shown that if the interface was absolutely ‘‘soft’’, i.e., contraction of the two
gaps was not coupled, the uniform distribution became unstable at precisely the point
where the dimensionless charge density s reached the critical value, s0 ¼ ð2=3Þ1=2 . In other
words, the uniform distribution became unstable at the point where, under  control, C 1

Copyright © 2001 Marcel Dekker, Inc.

 66 Partenskii and Jordan

would have changed sign, so that the region with C < 0 is unacceptable for such interfaces.
If, however, the Ci are coupled by a spring, then the stability region expands towards
larger .
We can also use another argument similar to one suggested by Nikitas [44]. Consider
a small area of the interface. Given that this area is much smaller than the total area of
contact, we may consider the rest of the interface a potentiostat which keeps the potential
in the chosen area fixed. Then, the selected area can be considered under
-control and its
local capacitance cannot be negative. Therefore, the selected area is likely to undergo a
transition simultaneously accepting some charge from the rest of the interface. As this
reasoning can be applied to any area of the interface, it strongly suggests that in the
isolated system, the uniform state becomes unstable in a region where -control predicts
C < 0.

C. Summary
The question of the allowed sign of C was and remains a topic of discussion with sig-
nificant contradictions. We suggest here that a major reason for these contradictions is
that theoretical calculations for electrified interfaces are more easily carried out assuming a
uniform electrode charge. Most studies have used this condition and, on some occasions,
the restriction of -control took its toll. And those were exactly the situations where
negative capacitance was predicted.
The conviction that C should be strictly positive was so compelling that it has been
typically believed to also apply under -control conditions (see Refs. 7 and 36 for a
review). In some other cases, confusion was caused by an opposite extreme, dropping
the restrictions on C < 0 even for
-control.
The discussion in Refs. 17 and 18 is illustrative. Torrie [17] presented very important
results for grand canonical Monte Carlo simulations of an electrical double layer in 2 : 1
electrolytes. Those results provide very convincing evidence that C < 0 can occur under -
control. However, instead of analyzing the consequences of this fact for real systems, the
author simply quotes the statement [21] that sign of C is not restricted, even for
-control.
Reliance on this result (see a critique in Ref. 22) simply ignored the problem and discour-
aged closer study of this system.
Similarly, Wei et al. [18] made analogous predictions of C < 0 using hypernetted-
chain theory and incorporating molecular polarizability. However, the discussion was
again restricted to the plausibility of negative C under
-control.
We suggest that results yielding C < 0 under -control have considerable impor-
tance, even though this type of control is unrealistic. They provide an important indication
that in a system studied under q-control there is a range of  where the plane uniform state
is unstable, and the system undergoes a transition to a nonuniform state with charge
density varying along the plane of electrode.
Therefore, a computational paradigm could be formulated as follows: ‘‘If C becomes
negative under -control, study the stability of this state towards ‘charge density’ perturba-
tions. It is very possible that a phase transition to a nonuniform state appears.’’
The same system treated under
-control becomes unstable in the vicinity of
c
(which corresponds to C ! 1) and can also undergo a phase transition (see Refs. 7
and 24 for discussion). But, unlike the isolated system under q-control, it can acquire
some additional charge from the battery, so that the initial and final states would corre-
spond to different values of q. The relation of transitions under q- and
-control is another
prospective area of research.

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 67

IV. ELECTROELASTIC INTERACTIONS, STABILITY, AND

BREAKDOWN IN MEMBRANES

Study of the mechanisms of membranes’ interaction with electric fields and the conse-
quences for membrane stability and electroporation are very important both for better
understanding of membranes’ properties and for different applications involving electrical
manipulation of their behavior [45–48]. By controlling the electrical setup it is possible to
influence transmembrane permeability and thus gain some control over processes such as
targeted drug delivery, DNA transport into cells, etc. This is also important for develop-
ment of ‘‘bioelectrochemical’’ sensors, where membranes are in contact with electrodes
and applied voltage is used for monitoring changes caused by the interaction with the
environment [49–52].

A. Some Perspectives on Membrane Instability

1. Crowley’s Model
The first model of membrane electroporation was suggested by Crowley [1]. In Crowley’s
model the membrane is viewed as the isotropic elastic material. The necessary background
for understanding its voltage-induced instability was discussed in Section II. Crowley’s
approximation for the elasticity energy term in Eq. (7) is
E0 ðhÞ ¼ Bh0 z½lnðzÞ  1 ¼ Ez½lnðzÞ  1 ð51Þ
where h ¼ hðÞ is the local thickness of the membrane ( is a two-dimensional radius
vector in the membrane plane), h0 is its unperturbed value, B and E are the Young’s
and ‘‘stretching’’ moduli, respectively. Using Eq. (51) with Eqs. (7) and (8), we find a
relation between the membrane ‘‘contraction’’ factor, z ¼ h=h0 (h is the equilibrium thick-
ness) and the voltage V [53]:
E lnðzÞ ¼ pE  h0 ð52Þ
where
m 0 V 2
pE ðV; hÞ ¼ ð53Þ
2h2
is the electrical pressure as a function of the applied voltage and membrane thickness, and
m is the membrane dielectric constant.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
2Eh0 1
V ¼ z ln ð54Þ
m 0 z
The critical membrane contraction can be derived in the manner outlined in Section II:
zc ¼ expð0:5Þ  0:61 ð55Þ
which corresponds to critical membrane thinning  39%. The corresponding critical vol-
tage, derived from Eq. (54), is
sffiffiffiffiffiffiffiffi
Eh0
Vc  0:61 ð56Þ
0
Equation (55) illustrates the major shortcoming of the original electroelastic model; it
predicts membrane thinning of almost 40%, a result vastly different from the experimental

Copyright © 2001 Marcel Dekker, Inc.

 68 Partenskii and Jordan

finding that critical thinning never exceeds 2–3%. The computed critical voltage Vc equals
 3:4 V and  5:9 V for representative ‘‘soft’’ [stearoyloleoylphosphatidyl-choline
(SOPC)] and rigid [SOPC-cholesterol]) (SOPC:CHOL, 1 : 1)] membranes respectively
(see Table 1 for the membrane parameters). Experimentally, the characteristic critical
voltage Vexper depends on the duration  of the applied electrical pulse. For brief pulses,
  0:01  0:1
S, Vexper  1 V, and for longer ones (  1
S) Vexper < 0:5 V. As a result,
both the extent of the membrane contraction and the breakdown voltage predicted by
Crowley’s model greatly exceed the experimental values.*

2. Hydrodynamic Models
An alternative description of membrane stability has been based on hydrodynamic mod-
els, originally developed for liquid films in various environments [54–56]. Rupture of the
film was rationalized by the instability of symmetrical ‘‘squeezing modes’’ (SQM) related
to the thickness fluctuations. In the simplest form it can be described by a condition [54]
ðd 2 Vdis =dh2 Þ0 < 22 =a2 where Vdis is the interaction contribution related to the ‘‘dis-
joining pressure,’’y  is the surface tension and a is the characteristic size of the film. At the
critical film thickness h0 , defined by this condition, the free energy cost of fluctuation due
to the disjoining pressure is negative and overwhelms the positive contribution due to the
surface tension at the wavelength  ¼ a. When this happens, there is flow in the film from
the thinner to the thicker regions leading to spontaneous growth of interfacial perturba-
tions [57].
Similar instability is caused by the electrostatic attraction due to the applied voltage
[56]. Subsequently the hydrodynamic approach was extended to viscoelastic films appar-
ently designed to imitate membranes (see Refs. 58–60, and references therein). A number of
studies [58, 61–64] concluded that the SQM could be unstable in such models at small
voltages with low associated thinning, consistent with the experimental results. However,
as has been shown [60, 65–67], the viscoelastic models leading to instability of the SQM did
not account for the elastic force normal to the membrane plane which opposes thickness

TABLE 1 Elastic Parameters for Representative Soft (SOPC) and Rigid (SOPC : CHOL, 1 : 1)
Membranes (Parameters E, Kc , and H0 are Taken from Refs. 117–119, , k1 and r1 from Ref. 53)

Membrane E (N/m) Kc (J) h0 (nm)  k1 r1

‘‘Soft’’ 0.002 0:9  1017 2.8 0.014 0.1 0.06

‘‘Rigid’’ 0.006 2:4  1017 3.4 0.0087 0.1 0.04

* The values of stretching constant E that Crowley quoted in his original paper were substantially
underestimated. As a result, his estimates of Vc were in better agreement with experiment.
However, zc did not depend on E and reached the same subnormal value. It must be noted
that critical thinning 1  zc almost universally reaches 30–40% for various models of the elastic
energy [7].
y
Disjoining pressure was attributed in Ref. 54 to the combined effect of van der Waals attraction
and long-range electrostatic repulsion between similarly charged membrane surfaces.

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 69

fluctuations. The most obvious manifestation of this neglect is the prediction of zero break-
down voltage for vanishing surface tension [61]. Thus, this approach is more appropriate
for liquid systems where thickness fluctuations arise from redistribution of molecules
between surface and inner regions of the film. In addition, this model is also descriptive
of ‘‘colored’’ lipid films. But it is hardly appropriate for lipid bilayers where steric interac-
tion between the hydrocarbon tails of lipid molecules provides the quasielastic force oppos-
ing the thickness fluctuations, a force which stabilizes the bilayer. After the viscoelastic
model is modified to account for this force [66], the original Crowley model problem recurs;
breakdown requires extensive thinning (see reference 32 in Partenskii et al. [53]).

3. The Pore Formation Model

The most developed and widely used approach to electroporation and membrane rupture
views pore formation as a result of large nonlinear fluctuations, rather than loss of stability
for small (linear) fluctuations. This theory of electroporation has been intensively reviewed
[68–70], and we will discuss it only briefly. The approach is similar to the theory of crystal
defect formation or to the phenomenology of nucleation in first-order phase transitions.
The idea of applying this approach to pore formation in bimolecular free films can be
traced back to the work of Deryagin and Gutop [71].
It was extended to describe spontaneous rupture of lipid bilayers [72,73] and electro-
poration [74] (see also Ref. 70 and references therein). The energy of the membrane with a
virtual water-filled pore of a radius r subject to an applied voltage V is [70,74]

Wp ¼ 2r  r2  0:5 
CV 2 r2 ð57Þ
where  is the energy per unit length along the circumference needed to form the pore,  is
the energy per unit area of the flat, pore-free membrane, and
 
p

C ¼  1 C0 ð58Þ
m
is the approximate change of the capacitance due to the presence of the pore; C0 is the
capacitance per unit area of a pore-free membrane, p and m are the permittivity of water
inside the pore and of the membrane respectively. The critical pore radius, rc , and critical
barrier,
Wp , are both defined by minimizing Eq. (57) with respect to r:
  2
rc ¼
Wp;c ¼ ð59Þ
 þ 0:5
C  V 2  þ 0:5
C  V 2
As has been described in Ref. 70, this approach can reasonably account for membrane
electroporation, reversible and irreversible. On the other hand, a theory of the processes
leading to formation of the initial (hydrophobic) pores has not yet been developed.
Existing approaches to the description of the probability of pore formation, in addition
to the barrier parameters , , and some others (accounting, e.g., for the possible depen-
dence of  on r), also involve parameters such as the ‘‘diffusion constant in r-space,’’ Dp ,
or the ‘‘attempt rate density,’’ 0 . These parameters are hard to establish from first prin-
ciples. For instance, the rate of critical pore appearance, c , is described in Ref. 75 through
an Arrhenius equation:
c ¼ 0 Vm expð
Wp;c =kTÞ
where Vm is the total volume of the membrane. Nucleation of a hydrophobic pore requires
cooperative motion of  8 lipid molecules and 8–10 water molecules. Given that lipid

Copyright © 2001 Marcel Dekker, Inc.

 70 Partenskii and Jordan

molecules are not rigid, the corresponding entropy contribution (or the ‘‘attempt rate’’
density 0 ) is hard to quantify with precision. Estimates [75] of 0 are very approximate
and vary over 9–10 orders of magnitude. Estimates of
Wp;c vary from 30 to 50 kT. It is
basically assumed [75] that the uncertainty in 0 is balanced by the uncertainty in the
barrier, so that taken together they result in a reasonable concentration of pores. These
assumptions are difficult to verify. Another concern is related to the fact that this
approach ignores the role of membrane undulations. These are known to be important
in membranes’ macroscopic behavior (stretching diagram, adhesion, capacitance) and as
they noticeably interact with the electric field, they are likely to also contribute in the
processes related to electroporation. At least one such relation is well known: at small
surface tension the bending modes of membrane undulations become unstable at low
voltages. In the next section we describe the membrane undulations and present a hypoth-
esis as to their role in membrane instability.

B. The Smectic Bilayer Model (SBM) in an Electric Field

The most successful continuum description of membrane elasticity, dynamics, and ther-
modynamics is based on the smectic bilayer model (for examples of different versions and
applications of this approach see Ref. 76–82 and references therein). We introduce this
model in conjunction with the question of membrane undulations.

1. Membrane Undulations in the SBM

Let U þ ðÞ and U  ðÞ be z-components of the displacement of the upper (þ) and lower ()
membrane surfaces;  is a two-dimensional radius vector in the XY plane. We express
these as

U  ¼ u0 þ u ð60Þ
where u0 is the displacement caused by external forces (such as electrical or mechanical
stress) while u describes a nonuniform fluctuation of the membrane surface. For further
analysis we decompose membrane fluctuations into a Fourier series,
X 
u ðÞ ¼ u ðqÞ expðiqÞ ð61Þ
q

(a) Elastic energy. We now treat separately the peristaltic (symmetrical, s-) and bend-
ing (antisymmetrical, a-) modes of membrane undulations,
uþ 
s ¼ us ¼ u uþ 
a ¼ ua ¼ u ð62Þ
The elastic energy (per unit area) consists of three contributions [78–80]:
1. Compression (stretching) energy, W1 , for which we choose Crowley’s form, Eq.
(51). For small undulations this becomes

E
W1 ¼ ðz  zÞ2 ð63Þ
2h20 z
with z  z ¼ uþ  u :
2. Splay (‘‘bending’’ or ‘‘curvature’’) defined by a splay constant K1 , or by a
curvature elastic modulus Kc ¼ K1 h [76],

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 71

Kc 2 2
W2 ¼ ðr uÞ ð64Þ
2 ?

where r? ¼ @=@ is the gradient operator in the plane of the membrane.

3. Surface tension contribution defined by a stress  applied to a membrane, typi-
cally expressed as [77]

W3 ¼ ðr? uÞ2 =2 þ  2 =2E ð65Þ

where the first contribution is related to undulations and the second to elastic
stretching.*
(b) Interaction with the Electric Field. In most studies of the effect of an applied
voltage on membranes, the model of ideal conductor ( ¼ 1) is used for the surrounding
solvent. Then the membrane surface can be considered an equipotential. The solution of
the Laplace equation leads to following results for the electrostatic energies of membrane
undulations [56,78,88]:
(
tanhðqh=2Þ ða-modeÞ
Welectr ¼ 2qjuq j pE 2
ð66Þ
cothðqh=2Þ ðs-modeÞ

The ‘‘ideal conductor’’ model does not account for diffuseness of the ionic distribu-
tion in the electrolyte and the corresponding ‘‘spreading’’ of the electric field with a
potential drop outside the membrane. To account approximately for these effects we
apply Poisson–Boltzmann theory. The results for the modes’ energies can be summarized
as follows [89]:
!
B2 F 2
W s;a
¼ m a0 a1 h þ a0 Bs;a þ s;a s;a
2q
! ð67Þ
 2 2   2 þ q2 2
þ s  c0 þ 2c0 c1 þ c0 Ds;a þ Ds;a
2 4

where

* The origin of the ‘‘surface tension’’ in membranes recently became a subject of intensive discussion
(see, e.g., Refs. 83–87, and references therein). To outline the problem, we note that thermody-
namic ‘‘surface tension’’ in liquids is a result of molecular exchange between surface and inner
volumes when a new surface is created. The corresponding free energy contribution is a result of
the different surroundings of a molecule in the bulk and on the surface. In isolated bilayers,
however, such exchange is impossible. Area (projected) can only change due to unfolding the
bilayer’s undulations and elastically stretching the bilayer. The tension at equilibrium is zero.
However, the situation can be different if area can grow due to molecular exchange with an
environment. Such a situation can occur, for instance, in experiments with black lipid films,
where exchange between the bilayer and the rim is possible.

Copyright © 2001 Marcel Dekker, Inc.

 72 Partenskii and Jordan

ðr  1Þ þ  Fs;a =q þ 1  1=r r

Bs;a ¼ a0 r Ds;a ¼ a0 ¼

s;a
s;a 1þ
r  c 2 B þ Ds;a ð þ Þ
a1 ¼ ð  ÞDs;a þ Bs;a c1 ¼  0  s;a
2ð1 þ Þ 4 2ð1 þ Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

s;a ¼ 1 þ rFs;a =q  ¼  2 þ q2 r ¼ m =s  ¼ rh=2

and
Fs ¼ tanhðqh=2Þ Fa ¼ cothðqh=2Þ


lD1 is the reciprocal Debye length and s is the dielectric constant of the solvent.
Through most of what follows we use the ideal conductor approximation, Eq. (66).
Ionic effects will be considered in Section IV.D.

2. Dispersion Equations
Using Eqs. (61) and (54), and combining Eqs. (63), (64), and (65), we find

Ejuq j2
ws;a ¼ fs;a ðx; zÞ ð68Þ
h20
 
1 xz
fs ðx; zÞ ¼ þ b1 x2 þ b2 x4 þ x lnðzÞ coth ð69Þ
z 2
 
xz
fa ðx; zÞ ¼ x2 þ x4 þ x lnðzÞ tanh ð70Þ
2

where
Kc 
¼ ¼ x ¼ qh0 ð71Þ
4Eh20 4E

The dimensionless constants b1 and b2 were introduced by Leikin [78]; they account for the
fact that contributions from each monolayer to the bending modulus and the surface
tension can differ for the two modes considered. Such effects are probably small so that
b1  b2  1 [78]. We note that the electrical potential enters Eqs. (69) and (70) only
through the parameter z.

3. Instability of Membrane Undulations Under an Applied Voltage

(a) Bending Modes. In some sense, the bending modes of membrane undulations at long
wavelengths (small x) are similar to those in simple fluids. The curvature contribution is
 x4 and can be neglected in this limit, while the stretching moduli (another important
peculiarity of the double layers) does not manifest itself in bending. As a result, the
stability condition for a-modes is similar to results for free liquid films (see, e.g., Refs.
56,65). It can be derived from Eq. (70). In the limit of small x where the membrane is
especially ‘‘soft’’ and the instability is most likely to occur,

fa ðx; zÞ  x2 ð þ z lnðzÞ=2Þ ð72Þ

The bending mode is stable if

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 73

1
 þ z lnðzÞ > 0 ð73Þ
2
For the physical interesting region, z  1, we find z > 1  2. Combining the last condi-
tion with Eq. (52) and using Eq. (71) we find that the a-mode becomes unstable at
sffiffiffiffiffiffiffiffiffi
h0
V Vc

a
ð74Þ
m 0

It follows from Eqs. (73) and (74) that the only stabilizing force for a-modes at long
 is the membrane tension, and critical voltage vanishes as  ! 0. In experiments with
black lipid membranes the surface tension  arises from the contact of the bilayer with the
bulk phase contained in the surrounding rim and is typically 0:002 N=m. Then choosing
h0  3  109 m, we find Vca 0:6 V. Experimentally, the characteristic critical voltage
Vexper depends on the duration  of the applied electrical pulse. For brief pulses,
  0:01  0:1
s, Vexper  1 V, and for longer ones (  1
sÞ Vexper  0:5 V [90].
Further increase of  leads to even smaller Vexper [see Refs. 70 and 91 for liquid membranes
and Ref. 92 for bilayers with inclusions (fluid mosaic cell membranes) and references
therein]. Thus the onset of BM instability is likely related to the long  pulse instability.
The increase of the electroporation threshold [91] is also in qualitative agreement with Eq.
(74). The long- instability of bending modes does not lead directly to formation of pores
because it does not affect the internal structure of the membrane. But it can possibly lead
to ripping the membrane off the rim and resulting‘‘permeabilization.’’ From our view-
point, neglect of such an instability is one of the shortcomings of the pore formation
models (see the previous discussion).
(b) Squeezing Modes. We now analyze Eq. (69). In the limit x ! 0, fs ! 1 þ 2 lnðzÞ.
As a result, the condition of instability, fs ¼ 0, leads to Crowley’s results, Eqs. (55) and
(56), derived for a plane membrane. Some softening of this condition can be expected at
x 6¼ 0 because the stabilization due to electric field grows with x [the function x cothðxz=2Þ
in Eq. (69) is essentially a linearly increasing function of x for x  1]. Indeed, instability is
first lost at finite  (for SOPC it corresponds to x  1:25 [53]). However, the corresponding
reduction of Vc and zc is insignificant (for instance, Vc drops by only  5 mV [53]).

C. Nonlocal Model of Membrane Instability

It follows from previous discussion that the destabilizing electrostatic contribution grows
in absolute value with x (with increasing ). But the influence of the nonuniform electrical
force is overwhelmed by the stabilizing bending and stretching contributions. As a result,
the traditional smectic model cannot explain how a small transmembrane voltage can lead
to membrane breakdown. The obvious solution is to abandon this approach and to
develop an alternative, such as the ‘‘pore formation model.’’ However, as we noticed
before, this approach postulates rather than proves the appearance of hydrophobic pores.
Another option is to try to modify the smectic model in order to account for the low-
voltage, low-thinning instability. It was suggested in Ref. 53 that the instability could be
explained as due to s-modes if (1) the membrane moduli depended on the wavelength of
fluctuations, and (2) this dependence resulted in substantial softening of s-modes at short
. Condition (1) is quite common and normally is described as ‘‘nonlocality.’’ In general,
nonlocality of constitutive equations means that forces acting at a point r and conjugate to
a fluctuating variable s, depend not only on the value sðrÞ at r, but also on its behavior in

Copyright © 2001 Marcel Dekker, Inc.

 74 Partenskii and Jordan

more distant regions. This leads directly to a two-point integral for the energy W so that
(in an harmonic approximation)
ð
W ¼ Kðr; r 0 ÞsðrÞsðr 0 Þ dr dr 0 ð75Þ

where, depending on the property considered, the kernel Kðr; r 0 Þ depends on a nonlocal
susceptibility or a nonlocal modulus. For instance, in nonlocal electrostatics, the role of K
is played by ðr; r 0 Þ, the nonlocal dielectric constant, while electric field strength, EðrÞ,
corresponds to sðrÞ [93–95]. In nonlocal elastic theory K is the elastic modulus Dðr; r 0 Þ
and s describes displacements uðrÞ [93,96]. The local limit in a uniform systemÐcorresponds
to Kðr; r 0 Þ ¼ K  ðr  r 0 Þ where ðrÞ is a delta-function in which case W ¼ K s2 ðrÞ dr. As
long as K is a function of r  r 0 (the uniformity condition), the Fourier transform of Eq.
(75) is
X
W¼ KðqÞjsq j2 ð76Þ
q
Ð
where KðqÞ ¼ expðiqrÞKðrÞ dr. Thus, nonlocality can be either described through the
spatial dependence of the generalized susceptibility, Kðr; r 0 Þ, or through the dependence of
its Fourier transform on wave vector q.
Condition (2) is also quite common. For instance, in crystals it results in a reduced
sound velocity, vs ðqÞ when q approaches a boundary of the Brillouin zone [93,96], a direct
result of the periodicity of a crystal lattice. In addition, interaction between modes can
lead to creation of ‘‘soft mode’’ with q 6¼ 0 and corresponding structural transitions
[97,98]. The importance of nonlocality at fluid interfaces and the corresponding softening
of surface modes has been demonstrated recently, both theoretically [99] and experimen-
tally [100].
In a phenomenological treatment [53] the parameters describing the nonlocal soft-
ening of the s-mode were chosen to fit the conditions that the membrane becomes unstable
at voltages corresponding to low thinning, z < 3% It was also assumed that at long 
(small x) the mode energy approaches Eq. (69). Two approximations for the mode energies
in the short- limit were suggested in Ref. 53; their effects on membrane undulations are
almost identical. Here we use only one of them (see also Refs. 89,101):

fsNL ðx; zÞ ¼ ½1  tðxÞ4 ðk1 þ r1 x4 Þ þ tðxÞ4 fs ðx; zÞ ð77Þ

where
expðsaÞ þ 1
tðxÞ ¼ ð78Þ
exp½sða  xÞ þ 1

The parameters a and s were chosen to satisfy the conditions that the transition to the
short- behavior occurs at   1:5h0 and that it is steep enough so that by   ð2  3Þh0
the long- limit is quite reliable. Given that membrane thinning does not exceed 2–3%, the
model parameters r1 and k1 have been selected so that the s-mode energy approaches 0 as
z ! zcr ¼ 0:97. In addition, it was required that the corresponding critical wavelength, cr
is of the order of h0 . This procedure is illustrated in Fig. 7. The thick curve corresponds to
V ¼ 0. The thinner curve illustrates the case where the applied voltage leads to instability
at   cr ðx  2Þ. The membrane parameters for the local and nonlocal approximations
are presented in Table 1.

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 75

FIG. 7 Nonlocal (dimensionless) energy fsNL ðx; zÞ of the s-mode for the transitional and short-
regions at V ¼ 0 ðz ¼ 1, upper curve) and V ¼ 1:3 V (z ¼ 0:97, lower curve). (Reprinted from Ref.
53, by permission, Journal of Chemical Physics.)

D. Elastic Nonlocality and Equilibrium Membrane Properties

It the previous section we suggested a possibility of effective softening of the s-mode as a
mechanism leading to membrane breakdown. However, this phenomenological assump-
tion requires microscopic justification. But before proceeding further, it is important to
check if this assumption is consistent with the description of such equilibrium membrane
properties as the membrane stretching diagram, thickness fluctuations, and the low-vol-
tage membrane capacitance. These properties are well described in the local approxima-
tion and introduction of nonlocality should lead to similar predictions.

1. Membrane Stretching Diagram and Thickness Fluctuations

It is well known that surface undulations strongly influence the stretching diagram of
membranes. They are responsible for the ‘‘entropic’’ behavior, the extremely steep increase
of projected area
A? at small tensions. In the analysis of these phenomena based on the
local model the contribution of the s-modes to
A? can be neglected as it is strongly
reduced due to the high values of membrane stretching constant, E. As a result, the major
contribution comes from the soft long- bending modes [76]. However, the assumption of
softening of the symmetrical modes made in the nonlocal model may enhance their con-
tribution to the projected area at short  and thus alter the stretching diagram. Such a
possibility has been analyzed in Ref. 53. Analysis very similar to that of Ref. 77 leads to
the following expression for the fractional variation  ¼
A? =A of the projected area:
ð xmax  
kT 1 1
¼ x3 þ dx ð79Þ
16Eh20 xmin f2 ðx; zÞ fa ðx; zÞ

Stretching diagrams calculated in local (L) [Eqs (69) and (70)] and nonlocal (NL) [with Eq.
(77) instead of Eq. (69)] approximations are contrasted in Fig. 8.
It can be seen from Fig. 8 that nonlocal results are in good agreement with the local
approximation and, consequently, with experiment. At small tensions ( 2  104 N=m)
the slope of the stretching diagram is very steep. This domain corresponds to the ‘‘none-
lastic’’ or ‘‘entropic’’ region of membrane stretching. For 0ð3  4Þ103 N=m the slope is
practically constant. In the intermediate region there is a sharp transition between these
two regimes. Unlike local calculations, nonlocal results depend on the choice of min (or

Copyright © 2001 Marcel Dekker, Inc.

 76 Partenskii and Jordan

FIG. 8 Stretching diagram for soft (S) and rigid (R) membranes for (a) min ¼ 1:5 nm and
(b) min ¼ 1:2 nm. ‘‘L’’ corresponds to the local model, while ‘‘NL’’ describes the nonlocal calcula-
tions (see text). The parameters of Table 1 are used (1 dyne=cm ¼ 103 N=m). (Reprinted from Ref.
53, by permission, Journal of Chemical Physics.)

xmax ). As a result, for the cut-off min  1 nm, close to the size of the lipid headgroups, the
differences may reach 10–15%, being more significant for soft membranes. However, as
noted in Ref. 53, this discrepancy might be an artifact due to the same expression,
1

A? ðqÞ ¼  A½quðqÞ2 ð80Þ
2
being used for the mode contributions in the variation of the projected area. In fact, this
expression is based on the assumption that there is tilting of the lipids such that head-
groups are always aligned parallel to the membrane–water interface, which is only valid at
long . At short wavelengths,   1:0–2.0 nm, this condition is harder to achieve and up
and down displacements of the lipids, which may consume less projected area than the
parallel alignment, are the natural ones. In other words, Eq. (80) should be modified to
account for the short  dispersion of the projected area. Effectively, it is equivalent to an
increase of min . The value   1:5–2.0 nm is a reasonable limit for the validity of the
macroscopic expression, Eq. (80).
Membrane thickness fluctuations were initially discussed in the local approach by
Hladky and Gruen (HG) [102] in conjunction with their possible effect on membrane
capacitance. They are directly related to the spectrum of s-modes:
ð
 2 kT xmax x dx
ðh  h0 Þ ¼ ð81Þ
2E xmin fs ðx; zÞ
In Ref. 53 these calculations were reproduced using the nonlocal approach. Following
HG, the results were contrasted for two  cutoffs, min ¼ 1 nm and min ¼ 10 nm. The
nonlocal results for RMS thickness fluctuations
h~ ¼ hðh  h0 Þ2 i1=2 are comparable in
magnitude, but slightly smaller than those found by HG. However, the nonlocal model
predicts that the fluctuation spectrum should peak at wavelengths  2–5 nm. Thus a

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 77

valuable test of the nonlocal hypothesis can come from investigating this range of fluctua-
tions, either by means of molecular dynamics simulation or directly, by soft x-ray scatter-
ing from real membranes.

2. Effect of Undulations on Membrane Capacitance

Initially the effect of applied voltage on membrane capacitance was attributed to the
uniform electrostriction, in the manner of the elastic capacitor model [1,103]. The effect
of undulations was first considered by Leikin [78]. In Ref. 89 the combined effect of
undulations and uniform compression is studied, including the possible influence of non-
locality. The differential capacitance Cd is presented as
C d ¼ C0d þ Cud ð82Þ
where
!
  @h
C0d ¼ m2 0 h  V ð83Þ
h @V

is the capacitance of the uniform membrane compressed by the electric field and
ð " #
kT X xmax 1 @2 fl ðx; zÞ @fl ðx; zÞ2
Cu ¼
d
fl ðx; zÞ  dx ð84Þ
Ah20 l¼s;a xmin fl ðx; zÞ2 @V 2 @V

is the contribution from undulations. At comparatively small voltages (91 V) it is suffi-

cient to use the approximation
 
h  h0 ð1  bV 2 Þ b¼ m 0 ð85Þ
2Eh0
which follows from Eq. (52). The differential capacitance can be expressed as
C d ðVÞ C d ð0Þð1 þ V 2 Þ
where C d ð0Þ corresponds to V ¼ 0 and parameter characterizes the effect of voltage. To
compare with experiment, the integral membrane capacitance C must be defined. It is [89]
 
2
CðVÞ ¼ Cð0Þ 1 þ V
3
The results are presented in Table 2.

TABLE 2 Voltage Variation Factor as a Function of Tension () for the Uniform Membrane,
and with Undulation Included in both Local (L) and Nonlocal (NL) Approximations, for Soft/
Rigid Membranes in the Metallic Approximation, Eq. (66)

 ð103 N=mÞ Uniform Undulations (L) Undulations (NL)

2 0.016/0.005 0.024/0.007 0.028/0.008

3 0.016/0.005 0.021/0.006 0.023/0.007
4 0.016/0.005 0.020/0.005 0.021/0.006
5 0.020/0.005 0.020/0.006
10 0.017/0.005 0.018/0.005

Copyright © 2001 Marcel Dekker, Inc.

 78 Partenskii and Jordan

In addition, the effect of ionic screening was analyzed [89] using Eqs. (67). The ionic
concentrations considered (1.0 M, 0.1 M, 0.01 M) correspond to Debye lengths lD ¼ 0:3; 1
and 3 nm. The results for lD ¼ 3 nm, the most dilute electrolyte, where ionic screening is
least effective, are presented in Table 3.
The results can be summarized as follows:
1. The nonlocal results for CðVÞ are consistent with local calculations. Introducing
softening of the s-modes at short  only changes the undulative contribution to
C by 10–15%.
2. Electrostriction dominates the voltage dependence of membrane capacitance.
However, at low tensions (94  103 N=m) the contribution of undulations is
also important. The latter is in good agreement with the results of [78] if the
same membrane parameters are chosen. At higher tension undulations are
damped and their role becomes less significant.
3. In rigid membranes is small; however, the relative undulatory contribution is
still significant.
4. Ionic screening has only a minor effect on the voltage dependence of the capa-
citance for the ionic strengths considered,  0:01 M.
Results at   0:002–0.004 N/m are found to be in good qualitative agreement with
the ‘‘short pulse’’ experiments [104] with exper ranging from 0.018 to 0.036.

E. Summary
We have shown that postulating short- softening of the s-modes of membrane undula-
tions, which can explain the low-voltage low-thinning instability, is consistent with
observed equilibrium properties of membranes. This finding supports the hypothesis but
obviously does not prove it.
Although partially hidden in averaged thermodynamic properties such as the stretch-
ing diagram or the voltage dependence of capacitance, nonlocality can play a more impor-
tant role when the influence of short-scale perturbations is significant. It can have a direct
relation to the anomalous roughness of a membrane surface suggested to account for the
discrepancy between calculations of the adhesion energy based on the Young equation and
on the conventional (local) theory of entropic forces between the membranes [105]. Mode
softening could be a reason for this roughness and for the corresponding hidden projected
area [105]. This suggestion complements the hypothesis of special types of structural
defects (hats and saddles) [106]. It may also be related to the finding of ‘‘remarkable
out-of-plane vibrational motion’’ of lipid molecules [107] which can contribute strongly
to short-range repulsive forces between membranes [108]. Another group of phenomena
where short- behavior could be important is related to peptide insertion into membranes.

TABLE 3 Effect of Ionic screening on the Capacitance at V ¼ 0, Cð0Þ, and on the Factor ,
Describing the Voltage Dependence of C, for Two Tensions, 1 ¼ 0:002 and 2 ¼ 0:01 N/m, for
the Representative Soft Membranes (SOPC)

C(0) (F/m) ð1 Þ ð2 Þ

‘‘Metallic,’’ Eq. (66) 6:4  103 0.028 0.018

lD ¼ 3 nm, Eq. (67) 5:9  103 0.028 0.017

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 79

Typically, the insertion induces sharp variation of the membrane profile at the distances
0.5–1.0 nm from the membrane–peptide interface [79–82]. The steepness of this perturba-
tion indicates that the short- behavior of membrane moduli must be important in the
estimates of the elastic energy. In addition, a peptide inserted in a membrane almost
certainly perturbs the membrane’s elastic moduli in the immediate vicinity of the inclusion.
Both these effects, membrane nonlocality and nonuniform modification of elastic proper-
ties by insertions, might play an important role in resolving the contradiction between the
local calculations [80] and the experimental data for the mean lifetime of a gramicidin
channel [81,109,110].*
In our approach to membrane breakdown we have only taken preliminary steps.
Among the phenomena still to be understood is the combined effect of electrical and
mechanical stress. From the ‘‘undulational’’ point of view it is not clear how mechanical
tension, which suppresses the undulations, can enhance the approach to membrane
instability. Notice that pore formation models, where the release of mechanical and elec-
trical energy is considered a driving force for the transition, provide a natural explanation
for these effects [70]. The linear approach requires some modification to describe such
phenomena. One suggestion is that membrane moduli should depend on both electrical
and mechanical stress, which would cause an additional mode softening [111]. We hope
that combining this effect with nonlocality will be illuminating.
Further progress in understanding membrane instability and nonlocality requires
development of microscopic theory and modeling. Analysis of membrane thickness fluc-
tuations derived from molecular dynamics simulations can serve such a purpose. A pos-
sible difficulty with such analysis must be mentioned. In a natural environment isolated
membranes assume a stressless state. However, MD modeling requires imposition of
special boundary conditions corresponding to a stressed state of the membrane (see
Refs. 84,87,112). This stress can interfere with the fluctuations of membrane shape and
thickness, an effect that must be accounted for in analyzing data extracted from computer
experiments.
Other possible direct probes are optical experiments similar to studies [113] of vesi-
cles but expanded towards shorter  (20–30 Å). Alternatively neutron spin-echo studies of
stacked bilayer arrays, which can probe the 10–30 Å range [114], might possibly be applic-
able here. Finally, the x-ray grazing-incidence technique has been shown to be a powerful
tool for studying short wavelength fluctuations at fluid interfaces [100]. The application of
this technique to the investigation of membrane surface fluctuations can reasonably be
expected in the near future [115,116].

ACKNOWLEDGMENT

This work was supported by a grant from the National Institutes of Health, GM-28643.

* An assumption that continuum parameters are inappropriate at the short length scale was intro-
duced in Ref. 109.

Copyright © 2001 Marcel Dekker, Inc.

 80 Partenskii and Jordan

REFERENCES

1. J. Crowley. Biophys. J. 13:711 (1973)

2. L. D. Landau and E. M. Lifshitz, Electrodynamics of Continuous Media, Pergamon Press,
Oxford, 1960.
3. W. R. Fawcett. Israel J. Chem. 18:3 (1979).
4. R. J. Watts-Tobin. Philos. Mag. 6:133 (1961).
5. D. C. Grahame. Chem. Rev. 41:441 (1947).
6. D. C. Grahame. J. Am. Chem. Soc. 76:4819 (1954).
7. M. B. Partenskii, V. Dorman, and P. C. Jordan. Int. Rev. Phys. Chem. 11:153 (1996).
8. I. L. Cooper and J. A. Harrison. J. Electroanal. Chem. 66: 85 (1975).
9. S. L. Marshall and B. E. Conway. J. Electroanal. Chem. 337:67 (1992).
10. M. B. Partenskii and V. J. Feldman. J. Electroanal. Chem. 273:57 (1989).
11. W. Schmickler. J. Electroanal. Chem. 149:15 (1983).
12. M. B. Partenskii, Z. B. Kim, and V. J. Feldman. Sov. Phys. J. 30:907 (1987).
13. L. Blum. J. Phys. Chem. 81:136 (1977).
14. D. Henderson, L. Blum, and W. R. Smith. Chem. Phys. Lett. 63:381 (1979).
15. L. Blum, J. L. Lebovitz, and D. Henderson. J. Chem. Phys. 72:4249 (1980).
16. S. L. Carnie, D. Y. G. Chan, D. J. Mitchell, and B. W. Ninham. J. Chem. Phys. 74:1472
(1981).
17. G. M. Torrie. J. Chem. Phys. 96:3772 (1992).
18. D. Wei, G. M. Torrie, and G. N. Patey. J. Chem. Phys. 99:3990 (1993).
19. D. O. Raleigh, in Electrode Processes in Solid State Iionics, Reidel Publ. Co., Dordrecht,
Holland, 1976, pp. 119–146.
20. M. B. Partenskii and L. Blum. Unpublished results, 1990.
21. P. Attard, D. Wei, and G. N. Patey. J. Chem. Phys. 96:3767 (1992).
22. M. B. Partenskii and P. C. Jordan. J. Chem. Phys. 99:2992 (1993).
23. W. Schmickler and D. Henderson. Progr. Surf. Sci 22:323 (1986).
24. V. J. Feldman, M. B. Partenskii, and M. M. Vorobjev. Prog. Surf. Sci. 23:3 (1986).
25. A. A. Kornyshev. Electrochim. Acta 34:1829 (1989).
26. S. Amokrane and J. P. Badiali, in Modern Aspects of Electrochemistry (B. E. Conway, J. O’M.
Bockris, and R. E. White, eds.), vol. 22, Plenum Press, New York, London, 1993, pp. 1–96.
27. S. Walbran, A. Mazzolo, J. W. Halley, and D. L. Price. J. Chem. Phys. 109: 8076 (1998).
28. J. P. Badiali, M. Rosinberg, and J. Goodisman. J. Electroanal Chem. 143:73 (1983).
29. J. P. Badiali, M. Rosinberg, and J. Goodisman. J. Electroanal. Chem. 150:25 (1983).
30. J. P. Badiali, M. Rosinberg, F. Vericat, and L. Blum. J. Electroanal. Chem. 158:253 (1983).
31. M. B. Partenskii and M. M. Vorobjev. Sov. Phys. Dokl. 29:746 (1984).
32. V. J. Feldman, A. A. Kornyshev, and M. B. Partenskii. Solid State Commun. 53:157 (1985).
33. J. W. Halley, B. Johnson, D. Price, and M. Schwalm. Phys. Rev. B 31:7695 (1985).
34. V. J. Feldman, M. B., Partenskii, and A. A. Kornyshev. J. Electroanal. Chem. 237:1 (1987).
35. J. W. Halley and D. Price. Phys. Rev. B 35:9095 (1987).
36. Z. B. Kim, A. A. Kornyshev, and M. B. Partenskii. J. Electroanal. Chem. 265:1 (1989).
37. V. J. Feldman, M. B. Partenskii, and M. M. Vorobjev. Electrochim. Acta 31:291 (1986).
38. R. Parsons, in Modern Aspects of Electrochemistry (J. O’M. Bockris and B. Conway, eds.),
vol. 1, Academic Press, New York, 1954, pp. 103–179.
39. R. J. Baxter, Exactly Solved Models in Statistical Mechanics, Academic Press, London, New
York, 1982, Ch. 1.
40. C. W. McCombie, Problems in Thermodynamics and Statistical Physics (P. T. Landsberg, ed.),
Pion, London, 1971, p. 459 (from [43]).
41. P. Nikitas. J. Electroanal. Chem. 300:607 (1991).
42. P. Nikitas. Electrochim. Acta 37:81 (1992).
43. J. Stafiej. J. Electroanal. Chem. 351:1–27 (1993).
44. P. Nikitas. J. Electroanal. Chem. 306:13 (1991).

Copyright © 2001 Marcel Dekker, Inc.

 Electroelastic Instabilities 81

45. T. Y. Tsong. Biophys. J. 41:135 (1991).

46. S. A. Freeman, M. A. Wang, and J. C. Weaver. Biophys. J. 67:42 (1994).
47. U. Zimmerman and G. Neil (eds.), Electromanipulation of Cells, CRC Press, 1996.
48. A. G. Volkov, D. W. Deamer, D. L. Tanelian, and V. S. Markin, Liquid Interfaces in
Chemistry and Biology, J. Wiley & Sons, New York, 1998.
49. H. T. Tien. Adv. Mater. 2:316 (1990).
50. M. Stelze, G. Weissmuller, and E. Sackmann. J. Phys. Chem. 97:2974 (1993).
51. E. Sackmann. Science 271:43 (1996).
52. B. A. Cornell, V. L. Braach-Maksvytis, L. G. King, P. D. J. Osman, B. Raguse, L. Wieczorek,
and R. J. Pace. Nature 387:580 (1997).
53. M. B. Partenskii, V. L. Dorman, and P. C. Jordan. J. Chem. Phys. 109:10361 (1998).
54. A. Vrij. Discuss. Faraday Soc. 42:23 (1966).
55. A. Sheludko. Adv. Colloid Interface Sci 1:391 (1967).
56. D. H. Michael and M. E. O’Neill. J. Fluid. Mech. 41:571 (1970).
57. C. S. Kishore, S. Salaniwal, and A. Sharma. Phys. Fluids 7:1832 (1995).
58. D. S. Dimitrov and K. R. Jain. Biochim. Biophys. Acta 779:438 (1984).
59. P. M. Bisch and H. Wendel. J. Chem. Phys. 83:5953 (1985).
60. P. M. Bisch and H. Wendel. J. Chem. Phys. 83:5962 (1985).
61. D. S. Dimitrov. J. Membr. Biol. 78:53 (1984).
62. C. Maldarelli, R. K. Jain, I. B. Ivanov, and E. Ruckenstein. J. Colloid Interface Sci. 78:118
(1980).
63. C. Maldarelli and R. K. Jain. J. Colloid Interface Sci. 90:233 (1982).
64. C. Maldarelli and R. K. Jain. J. Colloid Interface Sci. 90:263 (1982).
65. A. Steinchen, D. Gallez, and A. Sanfeld. J. Colloid Interface Sci. 85:5 (1982).
66. D. Gallez. Biophys. Chem. 18:165 (1983).
67. D. Gallez and A. Steinchen. J. Colloid Interface Sci. 94:296 (1983).
68. J. C. Weaver and A. Barnett, in Guide to Electroporation and Electrofusion (D. C. Chang, B.
M. Chassy, J. A. Sanders, and A. E. Sowrs, eds.), Academic Press, 1992, pp. 91–117.
69. Y. Chizmadzhev, in Electrified Interfaces in Physics, Chemistry and Biology (R. Guidelli, ed.),
Kluwer Academic Publishers, Netherlands, 1992, pp. 491–507.
70. J. C. Weaver and Yu. A. Chizmadzhev. Biolectrochem. Bioenerg. 41:135 (1996).
71. B. V. Deryagin and Yu. V. Gutop. Colloid J. 370 (1962).
72. J. D. Lister. Phys. Lett. 193 (1975).
73. C. Tsupin, M. Dvolaitzky, and C. Sauterey. Biochemistry 14:4771 (1975).
74. I. Abidor, V. Arakelyan, L. Chernomordik, Y. A. Chizmadzhev, V. Pastushenko, and M.
Tarasevich. Bioelectrochem. Bioenerg. 6:37 (1979).
75. J. C. Weaver and R. A. Mintzer. Phys. Lett. 86A:57 (1981).
76. W. Helfrich. Z. Naturforsch. 28c:693 (1973).
77. W. Helfrich. Z. Naturforsch. 30c:841 (1975).
78. S. Leikin. Biologicheskie Membrani (in Russian) 2(8):820 (1985).
79. H. W. Huang. Biophys. J. 50:1061 (1986).
80. P. Helfrich and E. Jakobsson. Biophys. J. 57:1075 (1990).
81. T. A. Harroun, W. T. Heller, T. M. Weiss, L. Yang, and H. W. Huang. Biophys. J. 76:3176
(1999).
82. S. May and A. Ben-Shaul. Biophys. J. 76:751 (1999).
83. D. Sornette and N. Ostrowsky, in Micelles, Membranes, Microemulsions, and Monolayers
(W. Gelbart, A. Ben-Shaul, and D. Roux, eds.) Springer-Verlag, New York, 1994, pp.
251–302.
84. S.-W. Chiu, M. Clark, V. Balaji, S. Subramaniam, H. L. Scott, and E. Jakobsson. Biophys. J.
69:1230 (1995).
85. F. Jahning. Biophys. J. 71:1348 (1996).
86. B. Roux. Biophys. J. 71:1346 (1996).
87. S. E. Feller and R. Pastor. Biophys. J. 71:1350 (1996).

Copyright © 2001 Marcel Dekker, Inc.

 82 Partenskii and Jordan

88. D. Andelman, in Handbook of Biological Physics (R. Lipowsky and E. Sackmann, eds.), vol. 1,
Elsevier Science, Washington, D.C., 1995, pp. 603–642.
89. M. Partensky and P. Jordan. Mol. Phys. 98:193 (2000).
90. R. Benz and U. Zimmermann. Biochim. Biophys. Acta 597:637 (1980).
91. G. Troiano, L. Tung, V. Sharma, and K. Stebe. Biophys. J. 75:880 (1998).
92. J. Akinlaja and F. Sachs. Biophys. J. 75:247 (1998).
93. W. Harrison, Solid State Theory, Sinauer Associates, Sunderland, MA, 1984.
94. A. Kornyshev, in Chemical Physics of Solvation (R. Dogonadze, E. Kalman, A. Kornyshev,
and J. Ulstrup, eds.), vol. C, Elsevier, Amsterdam, 1985, ch. 6.
95. P. Bopp, A. Kornyshev, and G. Sutmann. J. Chem. Phys. 109:1939 (1998).
96. N. Ashcroft and N. Mermin, Solid State Physics, Holt, Rinehart and Winston, New York,
1977.
97. M. Kaplan and B. Vekhter, Cooperative Phenomena in Jahn–Teller Crystals, Plenum Press,
New York, 1995.
98. B. Strukov and A. Levaniuk, Ferroelectric Phenomena in Crystals: Physical Foundations,
Springer, Berlin, New York, 1998.
99. K. Mecke and S. Dietrich. Phys. Rev. E59:6766 (1999).
100. C. Fradin, A. Braslau, D. Luzet, D. Smilgies, M. Alba, N. Boudet, K. Mecke, and J. Daillant.
Nature 403:871 (2000)
101. M. Partenskii and P. Jordan. Biophys. J. A54 (1999).
102. S. Hladky and D. Gruen. Biophys. J. 38:251 (1982).
103. S. White and T. Thompson. Biochim. Biophys. Acta 323:7 (1973).
104. O. Alvarez and R. Latorre. Biophys. J. 21:1 (1978)
105. W. Helfrich, in Handbook of Biological Physics. (R. Lipowsky and E. Sackmann, eds.), vol. 1,
Elsevier Science, New York, 1995, pp. 691–721.
106. W. Helfrich. Liq. Crys. 5:1647 (1989).
107. E. Sackmann, in Handbook of Biological Physics (R. Lipowsky and E. Sackmann, eds.), vol. 1,
Elsevier Science, Washington, DC, 1995, pp. 213–304.
108. J. N. Israelachvili and H. Wennerstroem. Langmuir 6:873 (1990).
109. C. Nielsen, M. Goulian, and O. Andersen. Biophys. J. 74:1966 (1998).
110. J. A. Lundbæk and O. Andersen. Biophys. 76:889 (1999).
111. V. Dorman, M. Partenskii, and P. Jordan. Biophys J. 72:A69 (1997).
112. R. Goetz and R. Lipowsky. J. Chem. Phys. 108:7397 (1998).
113. H. Engelhart, H. Duwe, and E. Sackmann. J. Phys. Lett. 46:395 (1985).
114. W. Pfeifer, S. Konig, L. J. F., T. Bayerl, D. Richter, and E. Sackmann. Europhys. Lett. 23:457
(1993).
115. C. Fradin, D. Luzet, A. Braslau, M. Alba, F. Muller, J. Daillant, J. Petit, and F. Rieutord.
Langmuir 14:7327 (1998).
116. L. Perino-Gallice, B. Amalric, E. A. Braslau, T. Charitat, J. Daillant, G. Fragneto, and F.
Graner, Colloids and Polymers (2000) (submitted).
117. D. Needham and R. M. Hochmuth. Biophys. J. 55:1001 (1989).
118. D. Needham and R. Nunn. Biophys. J. 58:997 (1990).
119. E. A. Evans and W. Rawicz. Phys. Rev. Lett. 64:2094 (1990).

Copyright © 2001 Marcel Dekker, Inc.