• Aucun résultat trouvé

Phase behaviour of an ensemble of nonintersecting random fluid films

N/A
N/A
Protected

Academic year: 2021

Partager "Phase behaviour of an ensemble of nonintersecting random fluid films"

Copied!
18
0
0

Texte intégral

(1)

HAL Id: jpa-00210735

https://hal.archives-ouvertes.fr/jpa-00210735

Submitted on 1 Jan 1988

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.

Phase behaviour of an ensemble of nonintersecting random fluid films

David A. Huse, Stanislas Leibler

To cite this version:

David A. Huse, Stanislas Leibler. Phase behaviour of an ensemble of nonintersecting random fluid

films. Journal de Physique, 1988, 49 (4), pp.605-621. �10.1051/jphys:01988004904060500�. �jpa-

00210735�

(2)

Phase behaviour of an ensemble of nonintersecting random fluid films

David A. Huse (1) and Stanislas Leibler (2, *)

(1) AT & T Bell Laboratories, Murray Hill, NJ.07974, U.S.A.

(2) Baker Laboratory, Cornell University, Ithaca, NY.14853, U. S. A.

(Requ le 17 septembre 1987, accepte le 22 decembre 1987)

Résumé.

2014

Nous considérons un modèle continu simple de films fluides aléatoires, dans lequel on néglige

toutes les interactions autres que celles du volume exclu, et on permet une topologie arbitraire de l’ensemble des films. L’hamiltonien phénoménologique des films (ou des surfaces aléatoires), qui est à la base de la

description statistique à l’équilibre de notre modèle, inclut les deux termes de courbure : celui de la courbure moyenne (extrinsèque) ainsi que celui de la courbure gaussienne (intrinsèque) ; nous considérons ici seulement le cas où les deux côtés du film sont symétriques en mettant les valeurs de la courbure spontanée et de la

différence des potentiels chimiques du volume à zéro. Nous examinons le modèle entièrement dans l’ensemble

grand canonique sans aucune contrainte sur la forme ou l’aire totale des films. En nous fondant sur les observations expérimentales dans des systèmes amphiphiliques, nous soutenons que le comportement thermodynamique de ce modèle très simple peut en fait se révéler fort riche. Nous trouvons sept phases

distinctes et étudions leur nature et les transitions entre elles. Nous discutons aussi des liens possibles avec des

modèles décrivant le polymorphisme de systèmes amphiphiliques ainsi qu’avec les modèles de surfaces aléatoires étudiés dans d’autres contextes.

Abstract.

2014

A simple continuum model of random fluid films is considered in which one neglects any interactions other than simple self-avoidence, and allows arbitrary topology of the film assembly. The phenomenological film or random surface Hamiltonian, on which the equilibrium statistical mechanical

description of the model is based, includes both the usual area term and two curvature terms (with the extrinsic

(mean) as well as the intrinsic (Gaussian) curvatures) ; the spontaneous curvature and the bulk chemical

potential difference terms are set to zero so we are considering only the balanced case of symmetry between the two sides of the film. Our treatment is fully in the grand canonical ensemble, without any constraints on the

shape and total area of film present. We argue, inspired by the experimental observations in amphiphilic

systems, that the phase behaviour of this very simple model can in fact be quite rich. We find seven distinct

phases and study their nature and the transitions between them. We also discuss some possible connections to models describisng the polymorphism of amphiphilic systems, as well as to models of random surfaces studied in other contexts.

Classification

Physics Abstracts

05.20

-

64.70

-

82.00

1. Introduction

The problem of random surfaces has recently at-

tracted a lot of attention both in the field of

elementary particle physics [1], and that of con-

densed matter [1, 2]. Connections to experiment

appear promising in the latter field, where many

(quasi) two-dimensional objects are known and intensively studied. In fact, one hopes that this

(*) On leave from : Service de Physique Theorique, C.E.N.-Saclay, 91191 Gif-sur-Yvette Cedex, France (ad-

dress after September 1st).

bidimensional generalization of the random-walk

problem can apply to such systems as polymer

networks [3], bilayer membranes [4], or microemul-

sions [5]. From a theoretical point of view it seems

useful, therefore, to describe these materials in terms of random surfaces, and explore the advan- tages and shortcomings of such a formulation.

A microemulsion is an equilibrium phase of oil,

water and surfactants [6]. Surfactant molecules are

preferentially adsorbed on the oil/water interface due to their amphiphilic nature [7]. Microemulsions have the two following properties which, among others, have attracted the attention of chemists,

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:01988004904060500

(3)

physicists and engineers over the years : i) the

existence of very low surface tensions between different macroscopic phases [8], ii) rich phase dia-

grams [9] with particularly fascinating microstructure

[10]. A substantial amount of experimental and

theoretical work has been performed to study and explain these properties. Although microemulsions appear already in ternary (or quasi-ternary) systems, these mixtures have quite complicated molecular

structures and molecular interactions. It is therefore

a challenge to catch the essential features of these

phases in a simple statistical-mechanical model

[2, 11].

Here we would like to explore a random film

(surface) model which is closely related to the

continuum description of microemulsions developed by Talmon and Prager [12], de Gennes et al. [13],

Widom [14], Safran and others [15]. It is not our aim, however, to explain the detailed properties of

microemulsions through such a model. Rather, in- spired by the experimental observations o f microemul- sions, we would like to explore a simple generalization of random surface models, and show that (despite its simplicity) it can exhibit a surprisingly rich phase

behaviour. Let us stress that throughout this paper

we discuss only two-dimensional films in a three- dimensional space. Generalizations to other dimen- sionalities are certainly interesting but are not pur- sued here.

We choose therefore to think about a microemul- sion as an ensemble of surfactant films and to

construct a statistical-mechanical model based on a

« film Hamiltonian ». This Hamiltonian can be writ- ten as a sum of three parts :

i) Hext , which depends on the « external » (as opposed to « internal », see below) degrees of free- dom, i.e. the size and the shape of the film. This part contains three important terms :

The first term here is proportional to the area of the film, the constant ro being a microscopic (bare)

surface tension. (We shall call ro the area coefficient

to avoid a possible confusion with the macroscopic

surface tension.) The second term is the curvature energy term [4], which originates in the splay and the saddle-splay of the film. Here H and K are, respect- ively, the mean and the Gaussian curvature at a

given point of the film, while K0 and Ro are the corresponding bare elastic constants. If cl and C2 are the two principle curvatures (inverse radii of curvature) at a point, then H = cl + c2 and K

=

cl C2. The spontaneous curvature Ho reflects an

asymmetry of the film : for Ho :0 0 the film prefers to

curve more towards one side. The integration in the

first two terms is performed over the entire random surface or film ; we assume that this surface is non- intersecting (self-avoiding), thus implicity taking

into account excluded volume effects. We consider only films with no edges, for which one can dis-

tinguish two sides : we will denote one side « water »

and the other « oil », as suggested by microemul-

sions. The third term is due to the chemical potential

of the « water » (side) relativer to the « oil » (side),

tL 0, and the integral in this term is over the entire volume, Vw, on the « water » side ;

ii) Jein, which takes into account the « internal »

degrees of freedom of the film, e.g. positions,

orientations and conformations of the hydrocarbon

chains of molecules, etc. It also includes all the

couplings between the internal and external degrees

of freedom, which can give rise to new macroscopic

effects such as curvature instabilities, and the appear-

ance of ordered phases within the film [16] ; iii) Hinter, which includes all molecular interactions

not taken into account by previous terms. The main

contribution to this term is the interaction between different parts of the film, such as van der Waals forces, hydration repulsion and others [17]. If the description of a microemulsion as a self-interacting

ensemble of films is adequate, then the Hamilto- nian :

should lead to the experimentally observed ther-

modynamic behaviour of the oil/water/surfactant mixtures. The Boltzmann probability of a given configuration is exp(-/3H) ; where f3

=

1/kB T, kB is the Boltzmann constant and T is the tempera-

ture. We work without any explicit constraints on

the total amount of film, « water » or « oil » present ; this is the grand canonical ensemble. The partition

function is the integral of exp (- /3 H) over all

allowed configurations. In the continuum theory we

are considering, short-distance cutoffs on the fluctua- tions are necessary to make the partition function

well-defined. Precisely how to best introduce such cutoffs is at present unclear to us. We assume that a sensible cutoff procedure can be formulated. Note that in this continuum model the molecular nature of the system is reflected in the short-distance cut-offs present in the surface integrals, as well as in the microscopic coupling constants ro, K o, ILo, etc.

The purpose of this paper is to tentatively explore

a model based on the film Hamiltonian (2). As a first

and the most drastic simplification we shall neglect

the « internal », :rein’ and the interaction, 3center,

terms. A similar approach has been taken by the

authors of several phenomenological models of

microemulsions [14, 15]. In these models, however,

the system is not entirely described as a random

(4)

surface, a phenomenological bulk « entropy of mix- ing » term being used [14]. Here we consider an

idealized version of the pure surface model in which,

for simplicity, we also set to zero the only volume

term present in (1), thus assuming that the relative chemical potential uo vanishes. Finally, we also set

the spontaneous curvature, Ho, to zero. This assump-

tion, together with JLo

=

0, makes the two sides of

the film equivalent. All of these assumptions are, of

course, oversimplifications of real microemulsions.

Note, however, that we do not make one other

common and drastic assumption, namely we do

allow the topology of our film to vary ; it can thus consist of many disconnected « components », each of them having an arbitrary number of « handles » [18]. We can therefore explore various microstruc- tures of the macroscopic phases, and we shall argue that even such a considerably simplified model

exhibits many of the different phases which one

encounters in real physical systems. Our results for this simplified film Hamiltonian are briefly sum-

marized in section 2 below. Note our reason for

simplifying JC (Eq. (2)) is not because the terms

dropped, namely Ho, tkO, acin and Winter? are incom- patible with the type of random surface treatment

we present. These terms are only dropped to reduce

the parameter space down to a size that we can fairly thoroughly explore. There are numerous interesting

effects associated with these terms that can be studied within this formulation, some of which have

already been investigated in a closely related fashion

(e.g. see Refs [2] and [12-15]).

Although the general understanding of the prob-

lem of random surfaces has progressed a lot in recent

years, many fundamental issues are still poorly

understood [1, 19]. The important question of uni- versality is not at the moment fully elucidaded [19, 20]. It appears that the random surfaces con- sidered in the context of high temperature expan- sions of lattice gauge theories [21], which do not

have a fixed internal metric [19, 22], behave diffe-

rently from « fixed-connectivity » surfaces, such as

the recently studied « tethered surfaces » [23], or triangulated random surfaces [24] (which can be

viewed as a discretization of the Polyakov string

model [25]). In fact, recent simulations of this last class of models [19], if confirmed, put in doubt the very concept of universality : the scaling properties

appear to depend on the short distance properties of

the triangulations [19]. In this paper we are mainly

concerned with fluid films. If we were to discretize

our continuum model (1) we would therefore

assume that the internal metric is not fixed, as in plaquette models [21] or triangulated random sur-

faces with a variable coordination number [26].

Plaquette models of random surfaces with fluctuat-

ing topology have recently been studied numerically [27, 28]. Even if the energy of a configuration

consists only of the area term, two distinguishable phases of different topologies, separated by a con-

tinuous transition (in three dimensions) [27, 28], are

found. By including also a topological term, which in

fact is similar to our Gaussian curvature energy in

(1), another phase of different microstructure [27] is

obtained. These results show that even very simple

« fluid » random surfaces can have quite a rich phase

behaviour if one does not restrain them to a planar topology and to have only the area energy term. In

fact, in section 4 we shall argue that the phase

behaviour found in references [27] and [28] is implicitly included in the phase diagram of our

model.

Before doing so, however, we shall summarize our

model and the main results of this paper in section 2.

Then (in section 3) we analyse in more detail the two cases of a vanishing and of a nonzero Gaussian curvature term. Finally, the last section includes the summary and some perspectives on further develop-

ments and applications.

2. The model and summary of results.

We thus consider the simplified model film Hamilto- nian :

where the integral, as before, is over a surface of unconstrained topology and total area.

Experimentally, the area coefficient ro may be

adjusted by changing the chemical potential, JLs’ of

the surfactant molecules in the film. An increase in JL s will yield a decrease in ro, and vice versa. The

(bare) curvature elastic constant, K o, might be adjusted by adding cosurfactants or varying the hydrocarbon length of the surfactant [29]. There

exist now several experimental methods which can measure in principle the bending rigidity K in amphiphilic systems. These are, for example light scattering studies of thermal fluctuations of vesicles

[30, 31] or monolayer films [32], X-ray scattering

from multilayer crystals [33], the study of the

interactions between defects [34], NMR measure-

ments of local curvatures in lamellar systems [35],

and others. Much less is understood about the Gaussian curvature term [36] and no good measure-

ments of K are, to our knowledge, available. Also it is not clear how one could independently vary the Gaussian rigidity coefficient, K0, in real systems.

Note that K o and K0 have dimensions of energy, so the « reduced » parameters KolkB T and RolkB T

are dimensionless.

Let us for the moment forget about the Gaussian rigidity term, thus setting KO

=

0. (We return to the

case K o =1= 0 later.) A possible phase diagram of this

(5)

model as a function of ro and 1/K0 is presented schematically in figure 1. We first briefly describe

the various possible phases before addressing in the

next section more detailed issues, such as the nature and locations of various phase transitions. For large positive 13 ro the system does not want to have much film present. It achieves this goal by breaking the symmetry between the two sides of the film, and forming an « oil » water ») rich phase with finite,

unconnected domains of the minority bulk compo- nent [37], « water » oil »), present. This is the so-

called droplet phase. The macroscopic surface ten- sion, a, of a film separating « oil »-rich and

« water »-rich phases is nonzero here, so we denote

the state of the film as tense in this phase [38].

Viewed in a larger parameter space including go :A 0 and Ho =1= 0 this « tense » phase of the film

represents a manifold where two bulk phases (« oil »- and « water »-rich) can coexist with the interface between the phases having a macroscopic

surface tension. Of course, the interface and its tension are only well defined on this manifold. As

soon as one leaves this manifold, say by varying

Fig. 1.

-

Proposed schematic phase diagram of our

random-film model (3) for Ko

=

0. All phase boundaries

converge at the multicritical point W at ro

=

1/ KO

=

0 as 1 / K 0 - 1 log r 0 1- 1, thus their high degree of tangency to

one another. The random isotropic phase is fully dis-

ordered « high-temperature » phase, while the ordering

increases as one moves either towards the smectic lamellar

phase or the dilute droplet phase. The various phases are

described in some detail in the text. The tense bicontinuous to dilute droplet phase transition, shown dashed, is a percolation transition with no thermodynamic consequ-

ences. The random isotropic to tense bicontinuous tran-

sition, shown dotted, is the breaking of the « oil »-« wa-

ter » symmetry. This is not a symmetry of real oil-water- surfactant systems and thus this transition will not be present generally unless one adjusts the relative chemical

potential of the bulk components, go in (1), to a special

critical value.

go or Ho, there is only one thermodynamically stable phase.

As ro decreases the droplets of the minority bulk component in the droplet phase grow in size in order to increase the total film entropy, and start forming larger connected domains. The domains of minority

component percolate before the volume fraction

occupied by it reaches 1/2. This leads to a tense, bicontinuous phase for intermediate positive values

of ro. In this phase the « oil-water » symmetry is still spontaneously broken and the macroscopic surface

tension is still non-zero, although infinite percolating

domains of both bulk components are present.

Again this « tense » phase of the film corresponds to

a manifold of two-phase coexistence in the larger

space of ILo =F 0 and H0 =F 0.

For large positive Ko and sufficiently negative

ro the system wants to put in as much film as

possible, but not to bend the film. A possible way to do this is to stack films parallel to one another and very close together. This produces the smectic lamel- lar phase, which has long-range order in the orien- tation of the films, and quasi long-range order in the

positions of the film planes along the direction

normal to the films, as in normal smectic liquid crystals [39]. This quasi long range positional order gives rise to a power law divergence (« quasi-Bragg peak ») of the scattering intensity S (q ) at the lowest- order reciprocal lattice point. There is also the

possibility that at intermediate negative values of

ro a lamellar nematic phase appears, with long-range

order in the orientation of the films, but only short-

range positional order. Note that the full rotational symmetry of our Hamiltonian is spontaneously

broken in the lamellar phases. We are using term

« lamellar » to refer to phases in which there is long-

range order in the orientations of the films, so they

are on average parallel to one another. Thus

although the lamellar nematic phase has many defects in the stacking of the parallel films and thus

no quasi-Bragg peak in its scattering, the basic anisotropy, as seen for example in birefringence

measurements [40], is established in this phase.

Finally, at values of ro near zero, between the lamellar and tense phases, there should be a dis- ordered, random isotropic phase, where no sym- metries of our Hamiltonian (3) are spontaneously

broken. The random isotropic phase is a high-tem- perature disordered phase, while the lamellar phases

are low-temperature ordered phases. The phase

transition between the random isotropic and tense

bicontinuous phase is shown dotted in figure 1

because the « oil »-« water » symmetry that is broken there is not a true symmetry of real oil-water- surfactant systems. In the bigger space of ILo =F 0 and Ho =F 0 this transition is likely to be a line of critical

points terminating a manifold of two-phase coexist-

ence. These critical points occur only at a special

(6)

value of u0. For Ho = 0, the critical point must, by symmetry, be at tk 0 = 0, so is in the special sym- metric space we are focussing on. However, in a real system with Ho =F 0, the critical point occurs at a

non-zero value of go that is a function of Ho,

ro, K0, etc. For ro > 0 the system is quite analogous

to a ferromagnetic Ising model, with ro the nearest

neighbour coupling and K0 a higher-order (4-spin or more) interaction that does not break the global Ising symmetry. Then the random isotropic phase is

the high-temperature paramagnetic phase, while the

tense bicontinuous and dilute droplet phase are the ferromagnetic phase. The symmetry-breaking field

go is then analogous to a uniform magnetic field,

while Ho is a higher-order (3-spin or more) interac-

tion that also breaks the Ising symmetry.

For large K0 and thus small curvatures a pertur- bative renormalization group calculation can be

performed [41]. At length scale L, to leading order

in kB T / K o, the renormalized parameters are :

where a is a microscopic cutoff parallel to the film

and both values a = 1 and a

=

3 have been propos- ed [2, 41, 42]. Thus we have a multicritical fixed

point (W) at the origin in figure 1. This is effectively

a zero-temperature fixed point, since it occurs for kB T / K o

=

0. In this paper we mostly focus on the

behaviour near this fixed point, where K0 is large

and the perturbation theory leading to (4) can serve

as a reliable guide. One should stress however that the results (4) are perturbative : in fact, they are in principle valid only for small curvatures, and all the effects of self-interactions are completely neglected.

,

Let us also mention that the same perturbation

calculation leads, for higher dimensional films, to

the prediction of a crumpling transition [41], which

separates two distinct regimes of the behaviour of the film, namely a crumpled film and a macroscopi- cally flat film. In fact, a crumpling transition may also take place for two-dimensional crystalline or polymerized films [43, 44]. Here, as mentioned be- fore, we are interested in the phase behaviour of bidimensional fluid films only, and thus crumpling

transitions do not occur in our model. Note that we

will use the results (4) only on length scales for which

K (L ) > kB T, in which case the renormalization of r is small, [r(L) - ro] : roo Thus, for simplicity, we

will henceforth ignore the renormalization of r,

assuming r (L )

=

ro. This can readily be corrected in what follows, leading to no substantive changes.

An important length in this system is the persist-

ence length of an isolated film, 03BEK [5, 41]. This is the

length scale at which the renormalized elastic con-

stant for the mean curvature, K (L), becomes of

order kB T. From equation (4a) we see that for Ko> kB T

For an isolated film gK is the correlation length of

the local film orientation. For the ensemble of self-

avoiding film we are considering here, 6K is one of

the important length scales, as is emphasized below.

For example, the typical local radii of curvature and

spacings between films in the random isotropic phase are of order 6K-

In the film Hamiltonian (3) we have explicitly

included the Gaussian rigidity K0. Even when the bare elastic constant for the Gaussian curvature

vanishes, K0

=

0, the renormalized one, on longer scales, may not vanish, being given by :

where the value ii = - 10/3 and «

=

0 have been

recently proposed [45, 46, 47]. Note that the physical

case of three bulk and two film dimensions is the

marginal dimension where the « reduced » curvature

elastic constants {3 K and {3 K are dimensionless ; this

results in the logarithmic renormalizations given

above. Since, in our opinion, the value of « has not

yet been well established, we shall consider (while discussing the droplet phase) various cases corre- sponding to different values of « and study some

consequences. After having first studied the case

where K can be neglected (e.g. for «

=

K0

=

0) we

will consider more general situations of non-vanish-

ing Gaussian term. However, for simplicity and

definiteness we assume &

=

0 in many of the dis- cussions that follow.

If we allow the bare elastic constant for the Gaussian curvature, K0, to be non-zero, then other

phases not shown in figure 1 can be stabilized. By

the Gauss-Bonnet theorem [18] the integral of the

Gaussian curvature is dS K

=

4 7T (nc - nh ), where

nc and nh are the number of disconnected compo-

nents of the film and the number of handles, respectively. Thus K0 only couples to the topology of

the film : Ko > 0 favours more handles and fewer components, while Ko 0 favors more components and fewer handles.

Figure 2 shows a possible phase diagram for our

model with K 0 =F 0 as a function of the area coef- ficient ro and the rigidity K o at K o > kB T fixed. It we

first move on the ro axis of this diagram, putting K o

=

0, we obtain again the same sequence of phases

as in figure 1, namely : lamellar smectic

-

lamellar

nematic

-

random isotropic (relaxed)

-

tense

bicontinuous

-

droplets. However, as K o is in-

(7)

Fig. 2.

-

Proposed schematic phase diagram of our model (3) as a function of ro and Ko, for K0 > k, T. The abbreviation denote : NL, nematic lamellar ; RI, random isotropic ; and TB, tense bicontinuous. The « elastic constant »

K0 couples to the topology of the film, favouring many handles and thus the plumber’s nightmare phase for Ko sufficiently positive and favouring many droplets which crystallize for Ko sufficiently negative.

creased from zero we expect that each of these

phases will transform itself into a new ordered

phase, which we call here by the generic name

« plumber’s nightmare » [48, 49]. The ideal plum-

ber’s nightmare phases are fully connected, periodic

and topologically nontrivial surfaces that have zero mean curvature H everywhere. For this last reason

they are often called minimal surfaces [49], but this

name is (in our opinion) misleading since the total

area of such surfaces is not in general minimal (as opposed to the case of finite surfaces spanning a

closed loop, which also have zero mean curvature

everywhere [50]). Four unit cells of a simple cubic plumber’s nightmare phase are depicted in figure 3.

Such structures have many handles (of order one per unit cell) and, ideally, only one component, so for K o > 0 the Gaussian energy term favours them. We

Fig. 3.

-

Four unit cells of an ideal simple cubic plumber’s nightmare phase. The film is shown, the interior of the

shape shown is occupied by one bulk component, say

« oil », while the exterior, which has the same shape as the

interior but displaced by the vector 1 11 1 1 2 2 2

,

is

occupied by the other bulk component, « water ». See reference [49].

do not address here the question which of the many

possible [49, 51] plumber’s nightmares are chosen by

the system, although this question will be quite interesting to study both analytically and numeri-

cally. In fact, even our simple model may have some structural phase transitions (inside the region marked

Plumber’s Nightmare in Fig. 2) between periodic

structures of different symmetries. Such different phases and the transitions between them have indeed been observed in some amphiphilic systems [52].

Note, however, that in some real systems the plumber’s nightmare phases appear to arise as the result of competition ( frustration) between the curva-

ture energy term and the energy of stretching [51, 53]

of the hydrocarbon chains, the latter being an

« internal » degree of freedom not explicitly treated

here.

When K o is sufficiently negative the surface does

not want to have many « handles », as in the plumber’s nightmare phase, but rather prefers to

form many disconnected components. This can be done in presumably the most efficient way by forming many roughly spherical droplets. When

these droplets are very closed-packed they should

freeze into a droplet crystal. We expect that this

crystalline droplet phase is stable at sufficiently negative Ro. Since it arises from the system’s desire

to have as many droplets (components) as possible

in a given volume, it will presumably be one of the close-packed crystals, either f.c.c. or h.c.p.

We now proceed to discuss in more detail the

various phases and phase transitions shown in fig-

ures 1 and 2. Note that in drawing figures 1 and 2 we

have assumed what we feel is the simplest scenario,

with most of the phase transitions being either

continuous or weakly first order. Some phases, particularly the nematic lamellar and tense bicon-

tinuous, could be absent over some or all of the

(8)

phase diagrams if they are preempted by strongly

first-order transitions. Of course, strongly first-order transitions occur in many real experimental systems either due to effects not included in our model (3) or

due to effects within our model that we have overlooked.

3. Phase behaviour of the model.

3.1 THE DROPLET PHASE.

-

Let us first consider the dilute limit of the droplet phase. This occurs for sufficiently large ro, where, in order to avoid having large amounts of film present, the system spon-

taneously breaks the « oil »-« water » symmetry, forming a phase with dilute droplets of the minority

component and the remainder of space occupied by

the majority component. We will begin by discussing

the statistics of dilute near-spherical droplets.

We consider droplets of radius R such that the

mean curvature elastic constant renormalized to

length scale R, equation (4a), satisfies K (R) > kB T.

In terms of the persistence length (5) this means R , 6 K [54]. We also assume that ro R 2-c K (R) so

that the bare surface tension, ro, can be neglected in estimating the fluctuations of the droplet about a spherical shape by amounts of order

as discussed in some detail by Helfrich [47].

AR is the measure of the « roughness » of the fluctuating film. The radius and the position of the droplet are naturally viewed as uncertain to this

extent. Therefore a sensible partition function for this droplet is obtained by integrating over all configurations with the average radius within AR of R and the centre-of-mass position within AR of a

given point. The resulting Boltzmann weight for the droplet should be :

where C is a constant, and Ko

=

0 is assumed [55].

Assuming the droplets are dilute so we can neglect

excluded volume effects, the density of droplets with

average radius within AR of R is then

where C is another numerical constant. Let us now

introduce a new lengthscale

It is a characteristic length associated with ro > 0 ; droplets of size g have area energy

The total volume fraction occupied by the droplets

is then

Now, if 2 « + a were negative, which has not been proposed in the literature, this integral would con-

verge for small ro (large g r), yielding

cP ~ (Ko/kB T)2 exp (- 8 -iT /3 K 0) and thus only a

small volume fraction occupied by spherical topology droplets for K o > kB T. However, as we shall discuss later droplets of other, more complicated topologies

with handles would proliferate if a 0. There is every indication that 2 a + a > 0, in which case we

have :

as the volume fraction occupied by near spherical droplets. This expression is valid only for K (03BEr) > kB T, or 03BEr 03BEK , otherwise AR becomes of order R or greater for the the droplets of size

R -- ç and they are far from spherical. In terms of 6K, the total volume fraction occupied by near spherical droplets for 6K >> 6, is :

Let us now examine this result for various values of a and see what would be the consequences for the behaviour of the droplet phase.

i) a > 0. (This case has not been proposed.) The

volume fraction 0 becomes of order unity for ç r ç K. This means that the droplets become dense

and thus interacting when their radii are still well below the persistence length and they are therefore

all near spherical. Thus the system retains a very strong preference, due to K (6,) 0, for spherical droplets, even when the droplets are dense. This scenario sounds somewhat implausible because one

would think that the film could gain much entropy by reconnecting into more complicated, topologically

nontrivial shapes, rather than just simple spheres.

(Remember that we neglect here the molecular interaction term Jeinter which in real systems could stabilize dense droplet phases [56].)

ii) « 0 (as proposed, for instance, in Ref. [45]).

(9)

The system at long length scales prefers in this case topologically nontrivial droplets with handles, e.g.

tori and more complicated objects. The minimal curvature energy of a droplet with N handles and

linear size R is 41T[e(N) K + (1- N ) K], where e (0) = 2 and we expect e(N) - N2/3 for large

X. The minimal curvature energy droplets with large JV are presumably pieces of a plumber’s nightmare structure, closed (capped) on the surface.

Only the surface « caps » contribute to the mean curvature energy, the interior having zero mean

curvature, this results in the proposed 2/3-power

law. The coefficient of K is precisely (1 - X) by the

Gauss-Bonnet theorem [18].

Let us define s (X) such that the total surface area

of the minimum curvature energy shape with

X handles and average local radius of curvature R is 4 1TR2S2(X), where s (0) = 1 and s (N) - Xl/2 for large N. Then such a droplet with total area energy

kB T has R = gr/s(X). If we then apply the same analysis as given above for near-spherical droplets,

we find that the total volume fraction occupied by

N -handled droplets is (ignoring prefactors of (K (gr)/kB T)) : ·

Therefore, if Ko and Çr are large enough, the total

volume fraction occupied by droplets with one or

more handles could greatly exceed that occupied by near-spherical droplets, even when that total is much less than unity and 03BEr 6K so that the individual droplets are not very rough. This is a rather surpris- ing situation and to our knowledge has not been suggested.

iii) a

=

0, in this case, proposed by Helfrich [47],

the volume fraction occupied by each type of droplet

becomes of order unity when ç becomes of order of

ÇK. Thus when K0

=

0 the system does not develop a preference for a given topology and all types of droplets proliferate as K (ç r) decreases. This scenario strikes us as the most plausible and we will

assume a

=

0 in most of the remainder of this paper.

An actual calculation of the free energy as a function of size for droplets of various topologies (e.g. spheres, cylinders, multi-handled torii, etc.) is clearly needed to resolve the issue of « and

« and to check the above expectations for the

statistics of dilute droplets.

It is important to stress that in the above discussion

we have neglected the role of the spontaneous

curvature Ho. A nonzero spontaneous curvature will clearly favor spherical droplets of that curvature.

There have been a variety of treatments of the

droplet phase within « mean-field » models of the

sort originated by Talmon and Prager [12]. A recent

version that includes the renormalization of K has been proposed by Safran et al. [15]. Their model has, within the droplet phase, monodisperse droplets (modeled as cubes) of linear size g and density

n = 0/03BE3 . The model free energy per unit volume of such phase is, for W « I and K =&=ro=O:

where the first term is a phenamenological entropy of mixing. For the choice a = 1, taken by the

authors [15], the minimum free energy is attained by making g as small as possible, g

=

a, and thus only having microscopic droplets with density na3 = exp (- 8 ’7TB K o). Thus, for a : 3/2 and

K o > kB T, the droplet phase remains dilute even for

ro - 0 in this model (16). This results in a strongly

first-order transition from the dilute droplet phase to

the random isotropic phase. Such strongly first-order transitions do occur in some microemulsion systems.

This is very similar to what would occur in our model if the coefficient a were negative. If «

=

3, on the

other hand, the model free energy (16) does not

have a dilute droplet phase for ro -> 0. Moreover, if

ro =1= 0 is considered, one finds that the droplet size is

of order 03BEr, and the volume fraction occupied by droplets is 4> = (g r/ g K)2 a. This differs from our

result (14) only by the factor of (R/ åR)4 =

[K (6,)IkB T]2 originating from the roughness of the droplets. Thus the behaviour of the droplet phase

for a = 1 is rather different from what one obtains if

«

=

3 is instead used in their model (16). The

difference with our results for « 3/2 arises because the monodispersity assumption requires approximat- ing the droplet size distribution (9) N(R) with a

delta function at R = g. Now we have

n (R) -- R2 a - 3, so for a 3/2 the most probable droplet size, which dominates the total free energy

(16), is microscopic, 6

=

R

=

a. The monodisperse

model naturally chooses this size. However, if one calculates, within our approach, the total volume fraction 0 occupied by droplets (12), this is domi- nated by the large droplets for all a > 0. Thus the

assumption of monodispersity leads to misleading

results for 0 « 3/2.

3.2 BICONTINUOUS TENSE PHASE.

-

As 6r increases

and becomes of order 6K, the volume fraction occupied by the minority component approaches

1/2. Assuming 0 increases continuously with 6r r (arguments supporting this are given below), there is

a percolation transition in the minority domains at

some volume fraction Op less than 1/2. This is not a thermodynamic phase transition only a change in the

statistics of instantaneous connectivity of the min-

(10)

ority domains. This transition is presumably in the universality class of three-dimensional percolation,

with §K serving as the microscopic length at the

transition. For large 6K, the transition should occur

at a value of ø = CP p which is independent of 6 K. For 1/2 :> ø:> øp there is an infinite connected

domain of minority (as well as majority) component present so the system is now bicontinuous. The

« oil »-« water » symmetry is still broken, since 0 =A 1/2, so the surface tension a is still non-zero.

To differentiate this phase from bicontinuous phase

with 45 = 1/2 and unbroken « oil »-« water » sym- metry, this phase is denoted as tense. As mentioned above, this tense phase represents a manifold in the larger parameter space including ILo -:1= 0 and H0 =F 0

on which the bulk phases coexist.

3.3 THE TENSE-RELAXED TRANSITION.

-

As gr is

increased still further (or equivalently, as ro is

decreased) the « oil »-« water » symmetry must be restored at some point. At this transition the macro-

scopic surface tension [38] vanishes. We assume that

in absence of attractive interactions this transition is continuous and, like percolation, occurs at g r of

order §K, though the percolation transition clearly

occurs first as 6, is increased. These assumptions are supported by the following suggestive arguments :

Let us consider putting our model for small K0 on a simple cubic lattice, so that the film must lie

on the plaquettes of the lattice. If K0

=

0 the model is then equivalent to a ferromagnetic Ising model

with nearest neighbour couplings J

=

ro a2/2, where

a is now the lattice spacing. The tense phase is the ferromagnetically ordered low-temperature phase,

where there is a surface tension, and the relaxed phase is the paramagnetic high-temperature phase,

with no surface tension. This system then has a continuous transition at i3J =-= 0.22, or gr ==

(4 Tr P ro)- 1/2 -= 0.4 a. The curvature, of course, is not well defined on a lattice, but we can argue that the typical local curvature of the interface confined to a lattice like this is of order 1/a. Furthermore, the persistence length of the interface normal is of order of a. Thus if we consider K0>0, the curvature

energy per unit volume near critical point is of order K o, and should not radically change the nature and position of the transition for K0 kB T. If we keep

K o of order kB T, and remove the system from the lattice, short-wavelength fluctuations that were pre- vented before by the lattice constraint are still

suppressed but now by the curvature energy term, Ko. Thus we argue that for K o of order kB T ÇK of order a), the system has a continuous transition in the Ising universality class at g r of order a. In similar way, for larger K0 the curvature energy furnishes an effective cutoff at CK and the transition

should occur at g of order g K and should still remain in the Ising universality class.

This transition separates the bicontinuous tense

phase, where the « oil »-« water » symmetry is broken, from the random isotropic phase, where no symmetries of the system are broken. The Ising

order parameter is (1/2 - 0 ). For large )K, the

volume fraction occupied by the minority phase

should scale as

where the scaling function, P (x), behaves as

for x - xc . The transition occurs at 03BEr/ 03BEK

=

xc, and

(3 =--x 0.31 is the usual Ising order parameter ex- ponent. For x > xc, P (x ) = 1/2. For x -+ 0, by (4), (5) and (14) we have (assuming here a

=

0)

Similarly, the surface tension, a, should scale as

with the scaling function behaving as

for x --> xc , with tL 1.26, and S(x) -> 1 for x -> 0.

The scaling of the correlation functions is determined

by noting that the energy is « stored » in the films so

that the surfactant concentration is an energy-like variable, while the « oil » and « water » play the role

of « up » and « down » spins. Thus surfactant-surfac- tant correlation functions scale as energy-energy correlation functions, while « oil »-« oil », « oil »-

« water » or « water »-« water » correlations scale as

spin-spin correlations in usual Ising model. In these

scaling functions ÇK serves as the microscopic length.

The correlation length ÇI which diverges at the

transition scales as

where

for x - xc, v = /-t /2, and X(x ) -- x, for x -+ 0. For

example, the truncated « water »-« water » corre-

lation function should, for R > ç K, scale as

where G(y)-yle-Y for y --+ oo, G (1 )

=

0(l) for

y - 0, and q = 0.03.

We have argued that this tense-relaxed transition

is continuous in our model (3). However, many

microemulsion systems are known to have a strongly

Références

Documents relatifs

Let u, compare our,iiiiulation re,ult, i&.ith the experinient, reported in II), (. The author, iii lo have urea,ureLl the average tilt angle in iree ,tanding iiliii, ;i, ;i

Light scattering from the free surface near a second order nematic to smectic a phase

Pour des valeurs particulières des paramètres de l’ellipsoïde le modèle présente une transition du second ordre isolée de la phase isotrope à la phase

Due to the presence of a large amount of ordered mate- rial, nucleation of the ordered phase was not neces- sary, and growth could proceed on the boundaries of the not

The temperature dependence of the critical stress and strain required to induce mechanically one phase from the other, allows the Landau parameters of the transition to

tion is condensed (namely the first harmonic p2qo), SA, a smectic A phase in which both the fundamental P qO and the first harmonic p2Qo of the density modulation are

Critical behaviour of second sound near the smectic A nematic phase transition.. L. Ricard,

on elastic interaction, recalling that its average energy is proportional to 1 /r per unit length of straight and parallel edge dislocations. Simple considerations