You are on page 1of 101

Home Search Collections Journals About Contact us My IOPscience

Defects in liquid crystals

This content has been downloaded from IOPscience. Please scroll down to see the full text.

1989 Rep. Prog. Phys. 52 555

(http://iopscience.iop.org/0034-4885/52/5/002)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 203.110.242.23
This content was downloaded on 26/08/2017 at 05:26

Please note that terms and conditions apply.

You may also be interested in:

Effect of Frustration in Liquid Crystals and Polymers


Maurice Klman

Defects in liquid crystals: homotopy theory and experimental studies


Mikhail V Kurik and O D Lavrentovich

The liquid-crystalline blue phases


T Seideman

Surface effects and anchoring in liquid crystals


B Jerome

Liquid Crystal Polymers and Applications


J C Dubois

Disclination unbinding transition from the hexatic N+6 phase to the nematic phase in discotic
liquid crystals
C Giannessi

Liquid crystals in micron-scale droplets, shells and fibers


Martin Urbanski, Catherine G Reyes, JungHyun Noh et al.
Rep. Prog. Phys. 52 (1989) 555-654. Printed in the UK

Defects in liquid crystals

Maurice KlCman
Institut Curie, Section de Physique et Chimie, 11 rue Pierre et Marie Curie, 75231 Paris CCdex, France and
Laboratoire de Physiques des Solides, UniversitC Paris-Sud, 91405 Orsay, France

Abstract

Defects are local breakings of symmetry in an ordered medium. The physics of defects
has long been reduced to the study of dislocations in solids, and to the main physical
phenomena they are responsible for, like plastic deformation. Dislocations break
translational symmetries. Disclinations break rotational symmetries and are the basic
defects of media with continuous symmetries, like liquid crystals. In this review, it is
stressed how their study has contributed to a renewal of the physics of defects. Static
and dynamic properties in the nematic, cholesteric, blue, smectic and columnar phases
of liquid crystals are described in detail, up to the most recent results, for small-molecule
thermotropic liquid crystals as well as for lyotropic and polymeric liquid crystals. We
also discuss the homotopy classification of defects, including point defects, and compare
it with the Volterra classification; finally we present the curved-crystal description of
frustration in the light of the relevant situation in liquid crystals. The usefulness of
the concepts introduced for liquid crystals for the study of other systems (structural
problems in biology, dissipative structures, magnetic domains, etc) is also emphasised.

This review was received in its present form in December 1988.

0034-4885/89/050555 + 100$09.00 @ 1989 IOP Publishing Ltd 555


556 M Kle'man

Contents
Page
1. Introduction 557
1.1. An overview of the theory of defects 557
1.2. Liquid crystals: materials, symmetries 559
1.3. Liquid crystals: free energy 563
1.4. The scope of this review 566
2. Static properties of defects. General considerations 567
2.1. The Volterra process for solid media 567
2.2. The Volterra process for partially ordered fluids 568
3. Static defects in smectic phases 569
3.1. Simple disclinations in smectics 569
3.2. Dislocations in smectics 570
3.3. Focal domains 577
3.4. Layer instabilities and defects 582
3.5. Defects in layered media other than SmA: generalities 583
3.6. Defects in SmC phases; bulk specimens 584
3.7. Defects in SmC* phases; bulk specimens 588
3.8. SmC and SmC" free suspended films 589
3.9. Lamellar phases with long-range and short-range 2D positional
order 589
4. Plasticity of smectic phases 590
4.1. SmA phases 590
4.2. SmB phases 596
4.3. Hexatic smectics 597
5. Defects in columnar phases 597
5.1. Developable domains and disclinations 598
5.2. Dislocations, walls and grain boundaries 602
6. Defects in nematics and cholesterics 605
6.1. Disclinations and singular points: a summary 605
6.2. Recent results on disclinations 606
6.3. Disclinations: dynamics 607
6.4. Chiral nematics (cholesterics) 609
6.5. Biaxial nematics 613
6.6. Polymeric nematics 615
7. The classification of defects by algebraic topology methods 62 1
7.1. The essentials of the topological classification of defects, and
examples 62 1
7.2. Other results concerning the topological theory of defects 625
7.3. Defects with partial topological definition (semi-defects) 628
7.4. Surface defects, cores and the use of the exact homotopy sequence
in various problems involving defects 629
8. Defects in media with frustrated symmetries 632
8.1. Generalities 632
8.2. Frustrated liquid crystals 635
8.3. Curved space representation of double twist 64 1
References 646
Defects in liquid crystals 557

1. Introduction

Research in liquid crystals has been a very active field for almost twenty years, despite
several decades of somnolence; it has contributed considerably to our understanding
of a number of important topics in condensed matter physics. The large variety of
different phases displayed by thousands of liquid crystalline materials synthesised by
the chemists has led physicists to investigate at least two major fields in which major
advances have thus been made: phase transition theory (global symmetry breakings)
and defect theory (local symmetry breakings). This review will try to cover the major
developments in the field of defects in liquid crystals over the last ten years; however,
most of the concepts are not restricted to these materials and consequently we shall
be led, in some instances, to consider other condensed systems. It will repeatedly be
apparent that defects are not only some sort of mischievous object that perturbs
perfection and which should be avoided as much as possible, as the name defect would
seem to indicate: they also have unique and beautiful properties which means they
are essential to the understanding of phase transitions, disordered systems, frustrated
media, convective instabilities, some biological shapes, etc. Historically, liquid crystals
have been the mould in which many of these new features of defects took shape. But
in their own right defects in liquid crystals constitute a beautiful field of research which
is far from being exhausted.

1.1. An overview on the theory of defects


The concept of a defect was first postulated in order to explain the processes of plastic
deformation in metals: the relevant physical object is the dislocation of Burgers,
Orowan, Polyanyi and Taylor (see Friedel 1964), which is an elastic singularity of the
medium akin to what Volterra long ago defined as a distorsione, and at the same
time a topological singularity of the ordered medium. The existence of defects in
metals was definitively ascertained in the early 1950s when the electron microscope
came into use. Dislocations and their arrangements in various kinds of textures (grain
boundaries, pile-ups) are the only kind of relevant defects in solid crystals; this is not
so in liquid crystals, where one finds dislocations, disclinations (which, like dislocations,
are line defects), point defects, global defects (which affect the whole topology of the
specimen and are non-singular) also called configurations, etc. This rich variety of
objects explains why their study was so profitable to the general theory of defects in
ordered media.
The theory of defects in liquid crystals is recent but disclinations were discovered
in liquid crystals well before World War I. They are indeed much easier to observe
in these materials than in metals; they very soon fascinated a number of scientists
because of the rich and mobile textures they display in the polarising microscope.
A huge number of observations were gathered together by Lehmann (1904) at the very
beginning of the century, and the true textures were correctly described by Friedel and
his collaborator Grandjean. The review by Friedel (1922) still remains a most fascina-
ting and motivating introduction to the subject. However, our knowledge of the first
class of defects ever encountered in condensed matter physics stopped with these first
558 M Kleman

(very clever) attempts, until Frank brought this subject afresh to the attention of
physicists in his much quoted 1958 paper; he coined the word disinclination (now
disclination) and indicated the kinship of this subject to that of dislocations.
Briefly stated, while dislocations are topologically related to symmetries of transla-
tion of ordered phases of condensed matter physics, disclinations are related to their
symmetries of rotation. Such a connection implies a number of similar questions in
both fields of research (in that respect the relevance of metallurgical concepts in the
physics of disclinations has to be recognised) while the differences pave the way for
problems of a somewhat different, and unexpected, nature.
As dislocations, disclinations are sources of internal strains, and are therefore in
general metastable entities. Comparing dislocations and disclinations energies, the
latter appear to-be as relevant as or more relevant than dislocations not only in liquid
crystals but also in spin systems, phases of biological materials and various other
systems often characterised by the presence of continuous symmetries. Typical ques-
tions like the nature of the core, the presence of kinks on the defect lines, their
interactions, growth, movement, multiplication, splitting, can be asked. The answers
are not mere replicates or extensions of the known answers for dislocations. In
particular there is a complex and beautiful topological interplay between dislocations
and disclinations (when they coexist in the same material) which reverberates between
splitting and mobility mechanisms. Also, since fluidity is an essential property of liquid
crystals, concepts pertaining to metallurgy and hydrodynamics must be used simul-
taneously, in a very original way. In contradistinction to the case of liquid crystals,
the basic mechanisms that rule the mobility of spin disclinations (called Bloch lines)
in ferrimagnetic garnets are well understood (Malozemoff and Slonczewski 1979), but
much remains to be done in investigating their interactions with the lattice defects
(dislocations, precipitates, etc).
The doubled-sided nature of the defect concept (topology on one hand, strains and
stresses on the other) is best reflected in the Volterra process of defect fabrication,
which yields the so-called Volterra classification of line defects (each defect type
corresponds to an element of symmetry (translation and rotation) it breaks), and opens
the way to a discussion of the energetical stability of a defect. In that respect the
Volterra process has some advantage over a more recent approach to the classification
of defects which uses the so-called homotopy groups of algebraic topology (Toulouse
and Kliman 1976), and deals therefore only with the topological stability of defects.
This is, however, treated in a somewhat deeper way in this approach, since difficulties
like the non-commutativity of the symmetries of rotation, or the relationships between
dislocations and disclinations, do disappear completely. Another great advantage of
the homotopy approach is that it applies to defects in all ordered media and can be
generalised to point-like, planar defects and configurations. It also extends to vortices
and other defects in superfluids and all sorts of isotropic materials that have gauge
symmetries.
Phase transitions are often driven by defects, which within this perspective can be
considered as condensates of fluctuations. Kosterlitz and Thouless (1972) have
described in this way the transition in a 2 D X - Y model, which invokes magnetic
vortices (disclinations); more recent models that use the splitting of dislocations in
two disclinations, describe a new 2 D phase, the so-called hexatic phase (Halperin and
Nelson 1978) whose existence is ascertained in some smectic materials. Evidence of
the same type of phenomenon in 3D is less clear; however, the vortical rotons of
Hiritier et a1 (1979) in superfluid 3He and 4He, which are condensates of solid-like
Defects in iiquid crystals 559

fluctuations, may play a role in the liquid-solid transition. Similarly the martensitic
transformation involves the creation of defects, as well as the classical description of
the solid-liquid transition (for a review of this last topic, see Halperin (1981)).
Finally disclinations are met in configurations resulting from competing interactions
which are geometrically incompatible at molecular scale; examples are, amorphous
metals, where local pentagonal packing of atoms cannot extend in space homo-
geneously, and blue phases of liquid crystals, whose complex structure results from a
balance between incompatible cholesteric and nematic ordering. In all these cases this
'frustration' can be given a geometrical representation by a 'crystal' in a 3D curved
space (KlCman and Sadoc (1979)); disclinations are the tools which are useful to
decurve this space and obtain structures in real space. This is indeed no surprise for
dislocation researchers, who have known for a long time that dislocation densities can
be represented by torsion in a non-Riemannian manifold, and disclination densities
can be represented by curvature (Kondo 1955, Bilbey et a1 1955, Bilby 1960, Kroner
1981). But the interesting fact is that we have at our disposal physical realisations of
curvature.

1.2. Liquid crystals: materials, symmetries


There are many excellent books and review articles that describe in detail the liquid
crystalline state and the chemical natllre of the materials involved. We shall therefore
make this section rather short. In fact the development of the theory of defects does
not necessitate, at the stage we have in mind, much more than an elementary knowledge
of the molecular properties of liquid crystals (LC) and a general understanding of their
symmetries. For more details, the reader is referred to de Gennes (1974a) and
Chandrasekhar (1980).
1.2.1. Materials. It is usual to distinguish between (i) thermotropic and (ii) lyotropic
liquid crystals.
(i) Thermotropic liquid crystals, made of pure organic molecules, either elongated,
like ~ C (4-n-octyl-4'-cyanobiphenyl)
B

which is a standard amongst smectic A phases at room temperature, or plate-like


(discotic), like hexapentoxytriphenylene:

o /
/ o
HllC5 /
HnC5
which orders in a columnar LC phase between T = 69 "C and T = 122 "C. One will
easily recognise that a character common to all molecules is the coexistence of a rigid
560 M Kle'man

central part (usually benzene rings) and of flexible exterior parts. These ingredients
seem indeed necessary to obtain a LC phase. Most of the LCS studied today consist
of elongated molecules. Some amphiphilic compounds, like block copolymers (e.g.
polystyrene-polyethylene) give rise to thermotropic phases. A number of polymers
give thermotropic LC phases. They fall into two classes: main-chain polymers, made
of mesogenic monomers in line and side-chain polymers, i.e. a non-mesogenic backbone
along which mesogenic molecules are grafted at periodic repeat distances (figure 1).
Some of these new LC phases, like Kevlar, are most important industrially; their study
is expanding rapidly (cf Cifferi et a1 1982, and the Proceedings of the First Conference
on Liquid Crystal Polymers, Bordeaux, 1987, published in Mol. Cryst. Liq. Cryst.
February 1988).

Figure 1. ( a ) A monomer in a main-chain polymer; ( b ) side-chain polymer. The box


represents the rigid central part of the mesogenic monomer, the zigzag the flexible part.

(ii) Lyotropic liquid crystals, made of organic elongated molecules associated with
other molecules like water, oil, alcohol, etc. Rigid rods in solution (e.g. tobacco mosaic
virus TMV) or semi-rigid polymers in solution (e.g. molecules of biological origin like
PBLG, xanthan or DNA), display LC phases which are mostly nematic or cholesteric.
Another, wide class, of lyotropic phases is obtained from amphiphilic molecules, with
a polar head (hydrophilic) and a paraffinic tail (hydrophobic) in water and/or oil.
Typical molecules are lecithin, Na laurate, sodium decylsulphate ( SDS), pentaethylene-
glycol dodecylether ( c ~ ~ E o a~ non-ionic
), surfactant. With water, they act as surfactants
and build liquid or solid interfaces with fascinating spatial arrangements, a number
of them having the same symmetries as thermotropic LCS, others extending the realm
of LC order.

1.2.2. Symmetries. The symmetries of LC phases are intermediate between those of


liquid and those of crystalline solids, hence their name of mesophases. They divide,
according to Frank's recent classification (1979), into nematic, smectic, columnar and
blue phases. The nematic and smectic phases were known to Friedel, who characterised
them for the first time. Since symmetries play such a central role in the properties of
defects, a detailed account of those met in LCS is presented here in order.
(i) Nematic phase N. The building blocks (thermotropic molecules or anisotropic
aggregates of amphiphilic molecules) have their centres of gravity disposed at random
(no positional correlations), as in a liquid, but display long-range orientational order
(figure 2). The nematic phase N is optically uniaxial ( N A ) if these orientational
correlations have cylindrical symmetry; if not, the medium is biaxial (NB). These
orientational correlations give rise to continuous symmetries in uniaxial nematics (any
axis perpendicular to the cylindrical directions is a twofold axis and the symmetry
group is H = RODmhwhich is the semi-direct group product of the group of continuous
translations R by the point group P = D m h )and
, discrete symmetries in biaxial nematics
NB (if the local correlations can be represented by parallel ellipsoids disposed at
Defects in liquid crystals 561

Figure 2. A schematic representation of the nematic uniaxial order.

random, the point group is P = D 2 ,but many other possibilities exist, with P being a
subgroup of the group of rotations O(3)t). While uniaxial nematics are most common,
biaxial phases have been discovered only recently, in lyotropic systems and in polymeric
nematics (Yu and Saupe 1980a, b, Windle et a1 1985, Hessel and Finkelmann 1986).
A cholesteric N* is a twisted nematic, i.e. the building blocks rotate with a constant
pitch along some direction. It is made of chiral mesogenic molecules or can be obtained
by introducing in a nematic some quantity of chiral molecules. Most recent studies
+
of N * phases concern lyotropic materials (surfactants water: Yu and Saupe (1984b),
Figueiredo Net0 et a1 (1984, 1985); cholesteric mesophases of biological origin like
DNA in water: Livolant and Bouligand (1986) etc).
(ii) Smectic phases. These are lamellar phases, made either of molecules stacked
parallel one to another (thermotropic) or of bilayers of amphiphilic molecules separated
by water or oil (for instance, soaps, from which they take their name). Smectics are
therefore one-dimensional solids (in the direction perpendicular to the layers); different
types of ordering in the layers lead to various smectic types, either liquid-like (SmA,
SmC; see figures 3(a) and (6)) or solid-like (SmB sensu stricto, etc). In SmB hexatic,
SmI or SmF, the order inside one layer does not propagate coherently (figure 3(c));
we have 2D short-range order and 2D or 3D bond orientational order with some
unusual phase transitions where defects play an important role; since it is easy to
achieve monolayer samples, or at least samples containing a few layers, the physics
of these phase transitions has been quite active in recent years (Prost 1984). For a
review on 2D order in smectics, see Helfrich and Heppke (1980). All smectic phases
are reviewed from the point of view of order and textures in the beautifully illustrated
book of Gray and Goodby (1984). SmA phases have been the standard medium for
the study of defects relevant to the layered structure only (focal conics).
As there are twisted nematics N*, smectic phases made of chiral molecules may
have a helical pitch superimposed on the layer periodicity (SmC*, for instance).
(iii) Columnar phases (also called canonic phases). They arise either with discotic
molecules (which stack in columns which arrange in a 2D ordered lattice), or with
amphiphilic molecules arranged in infinite cylinders either direct (water outside) or
inverse (water inside), forming 2D hexagonal lattices (figure 4). These latter phases
have long been known in the soap industry. Discotic phases are a more recent discovery;
their study has again emphasised the importance of geometry in defect theory, a fact
already known since Friedels analysis of focal conics in smectics.

t We follow the Landau and Liftshitz (1962) notation for the groups.
562 M Kle'man

(C)

Figure 3. Smectic order: ( a ) SmA; ( b ) SmC; (c) a representation of the non-coherent


propagation of order in a SmF lamella (adapted from Gray and Goodby (1984)).

Figure 4. ( a ) Discotic hexagonal phase; ( b ) and (c) lyotropic hexagonal phases (direct
and inverse).

(iv) Blue phases. Blue phases were first characterised in myristate cholesterol by
Saupe (1969); it is a bluish (hence the name), optically isotropic phase, with cubic
symmetry, made of chiral molecules, stable in a small temperature range between the
isotropic (liquid) and the cholesteric phase. The lattice parameter is of the order of
the pitch. Intense research on blue phases started only in the early 1980s, and has
proved remarkably rewarding. There are in fact three blue phases, noted BP I, I1 and
I11 (temperature increasing). BP I and BP I1 are cubic ( B P I is probably simple cubic
and BP I1 BCC) and grow with beautiful crystalline shapes (Barbet-Massin et a1 1984);
BPIII (blue fog) is isotropic but with a short-range order extending over a length
comparable with the pitch. The importance of blue phases stems from the fact that
they are frustrated phases: the molecules have the tendency, in their ground state, to
twist in all directions (this is called double-twist; in N" there is only one helicoidal
axis), but this geometry is non-crystallographic and thus forbidden to extend infinitely.
Defects in liquid crystals 563

There is therefore some characteristic length on which the unfrustrated geometry can
extend, but not beyond; the medium is then made of domains of size 6, separated by
boundaries which can be analysed topologically in terms of defects. However, the
whole structure is thermodynamically stable. The term frustration was first introduced
for spin lattices (Toulouse 1977b), then applied to spin glasses, amorphous media and
to the complex alloys (phases of Frank and Kasper (1959)) which are the analogues
of BPS in metals. Blue phases offer an example where the geometrical content of
frustration shows up clearly (figure 5 ) . It has been proposed that the blue fog is an
icosahedral liquid quasi-crystalline phase (Hornreich and Shtrikman 1986) similar to
the aperiodic quasi-crystals of A1-Mn alloys first discovered by Shechtman et a1 (1984).

1.3. Liquid crystals: free energy


The remarkable anisotropic properties of liquid crystals in an electric or magnetic field,
as well as the considerable birefringence ( A n = 0.3) of usual thermotropic systems, are
all related to the anisotropic shape of the building blocks. Reference should be made
to de Gennes (1974a) or Chandrasekhars (1980) excellent books for an introduction
to the subject. It has always been a subject of amazement for scientists that this strong
anisotropy is accompanied by a liquid-like macroscopic behaviour. The viscosity
constants are anisotropic, have typical values of 1 poise in a nematic phase, and reach

Figure 5. ( a ) Cholesteric phase N * . One axis of helicity (simple twist). We use the nail
convention (Frank) to represent a director at some angle from the plane of the drawing,
assuming that the head of the nail (which is along the principal molecular axis) is on the
side of the observer. Here the cholesteric is right-handed. ( b ) A molecular local arrange-
ment with double-twist (along the axis of the cylindrical region); ( c ) frustration of
double-twist between the cylinders, with the defects as in ( d ) ,along the ternary axis. This
geometry is believed to occur in blue phases.
564 M Kleman

1000 poise in LC polymers; but the measurement of viscosity constants by standard


methods (Poiseuille, Couette, etc) has always to be considered critically: the rheology
of LC is dominated by defects, and the samples cannot be considered as homogeneous
and perfectly ordered when flowing.
The Landau-Ginzburg free energy of these anisotropic media contains a symmetric
traceless tensor Qij which represents the nematic part of the order parameter, which
is always present for any LC, whatever its symmetry, since the molecules are anisotropic.
At some local length any LC looks therefore like a more or less distorted nematic.
Qij is either the anisotropic part of the dielectric tensor:
Q~~= constant ( c o p - &&x) (1)
or of any second-order anisotropic property. Frank has rightly noticed that the
eigenvectors of these different tensors can differ slightly. But the eigenvector n, the
so-called director, corresponding to the largest eigenvalue in modulus, is always roughly
in the direction of the long axis of the building block. If the nematic order is uniaxial,
Qij reads, when diagonalised:

S is a scalar, the Maier-Saupe order parameter, depending on temperature and


vanishing in the isotropic liquid phases:
s = t(3 COS e - 1) (2)
where 0 is a polar angle with respect to the local n ( r ) , and the brackets stand for a
thermal average. n ( r ) is equivalent to - n ( r ) . Observations of defects in LCS are nearly
always carried out using thin films of material sandwiched between glass plates. n ( r )
can be determined by polarising microscopy. It is common to speak of homeotropic
alignment, when n is anchored perpendicularly to the glass plates, and of parallel
alignment, when it is anchored tangentially.
The free energy density reads, up to second order, when written in terms of S, n
and grad n :
+ +
pf = aS2- pS3 yS4 $ K ,(div n)+ $K2(n curl n + qo)2
+$K,(nxcurl n ) 2 - $ K 2 , d i v ( n d i v n + n x c u r l n ) (3)
where the part depending on the scalar order parameter is omitted when studying the
distortion of a medium at constant T. qo = 2 n / p is related to the pitch p in a cholesteric
N* and vanishes in N. The geometrical meanings of the gradient terms which govern
the four elastic constants are well known: div n is the splay density; K1, the splay
constant, is the only non-vanishing gradient in figure 6 ( a ) ; n * curl n is the twist density;
K 2 , the twist constant, is the only non-vanishing gradient in the N* ground state
(figure 6 ( b ) ) . Mauguin (1911, 1913) has introduced in the experimental practice of
nematics the use of plages tordues (twisted areas), i.e. samples obtained by rotating
with respect to each other two glass plates bounding a nematic single crystal. If the
molecules are anchored to the glass plates, they follow the rotation and the final sample
has acquired twist:
t=n.curln=-a/d
Defects in liquid crystals 565

Twist

Figure 6. The basic distortions of a director field: ( a ) splay only; ( b ) twist only; ( c ) bend
only.

where cy is the angle of rotation and d the thickness; n x curl n, the bend density ( K 3
bend constant), is the only non-vanishing gradient in figure 6 ( c ) .
K24;the constant of saddle-splay, governs a gradient term which, by Stokes' theorem,
amounts to a surface term. This is why it is not always considered. However, it plays
an important role when new boundaries are created, or when the boundary area scales
faster that v213,v being the volume. A large positive K24stabilises blue phases (KlCman
1985b). The geometrical meaning of the gradient term itself is very simple when n is
normal to a surface (as in SmA, where it is the normal to the layers, or in water-surfactant
systems, where it is the normal to the interfaces). We have
= f div(n div n
c1c2 + n x curl n ) (4)
where cl and c2are the principal curvatures of the surface, and e1u2its Gaussian
curvature.
In a smectic, the free energy has to be supplemented by a term which describes
the layer elasticity. In a SmA it is:

where U is the elastic displacement of the layer, measured along its normal. It is tacitly
assumed in equation (5) that when the SmA is deformed n remains normal to the
layer. d u / d n is the derivative of U along the normal. In addition, note that the Frank
free energy simplifies in a SmA, where n is along the normals to the layers. For this
geometrical condition we can write:
n curl n = 0
so that twist is identical to zero (except in the cores of defects, where the layering is
broken). It can also be shown that the bend contribution is small compared with other
contributions, except when the radius of curvature of layers is small compared with
A , = (K1/B)'I2.
566 M Kleman

If the medium is biaxial, QIj is still traceless but with three different eigenvalues;
Hence the Landau-Ginzburg energy is more complicated than equation (3).
Other free-energy densities have been derived when the order parameter has
components other than a pure nematic one. We shall not have to refer to them in
general, although our approach to the topology of defects requires a detailed knowledge
of the order parameter. But it will be apparent that our approach to the toplogy of
defects in fact provides details of the concept of order parameter.

1.4. The scope of this review


It is no easy task to review a subject which has so many branches and which has
evolved so much in the last ten years (the period which the editors asked the author
to cover), mainly because the subject of defects is far from being common *knowledge
amongst solid state physicists (as opposed to physical metallurgists). For those who
want to take stock of the basic knowledge of defects in liquid crystals before the early
198Os, a reading of de Gennes (1974a) and Meyer (1976) is recommended. IUCmans
book (1983a, 1978 for the French edition) is devoted almost entirely to defects. The
recent review of Chandrasekhar and Ranganath (1986), which is at an introductory
level, is abundantly and usefully illustrated, and is recommended for non-experts as
a starting point before reading the present update. However, we have always tried to
always tried to make the link with basic knowledge, and have even sometimes expanded
some topics of general knowledge (like focal conics). Sections 3, 5 and 6 are devoted
to static defects in layered phases, columnar phases and nematic phases, including
chiral nematics (cholesterics). However, we have made the generalities on this latter
topic relatively short, although the subject is active and promising (for example on
the side of biological polymers). But cholesterics are constantly mentioned, especially
in 9 8 (which deals with frustration and structurally inhomogeneous liquid crystals).
Section 4 discusses the rheological (plastic) properties of layered media; this subject
has advanced in the last ten years (the role of dislocations), although the fundamental
questions about the mobility processes of focal domains, disclinations, have not yet
been formulated. The rheology of nematics (short-molecule nematics and polymeric
nematics) has still to be developed, and is discussed in 0 6. Sections 7 and 8 are
essentially devoted to problems which go beyond liquid crystals (0 7, homotopic
methods of classification of defects; 8, frustration). These sections are long, because
of the novelty of the subjects. We have tried to quote all the papers of some significance
in the last ten years that have come to our knowledge, even when we do not develop
the particular topic. Of course, the particular topics that are developed further reflect
the particular tastes and fancies of this author.
However, questions strongly related to defects in liquid crystals, such as phase
transitions driven by defects (for a review, see Halperin (1981)) and the presence and
role of defects in liquid crystal biological analogues, have not been fully covered. For
the biological aspects, the reader is referred to the review of Bouligand (1981) or to
the work of Harris, who has been among the first to recognise the interest of studying
static defects and their mobility in surface crystals, which are models for surface coats
of unicellular organisms, or the bacterial flagellum, etc (see, for example Nabarro and
Harris (1971) and the many subsequent papers). Finally, the analogies and differences
between the physics of defects in liquid crystals and magnetism have not been discussed
at all. The interested reader is referred to Malozemoff and Slonczewski (1979) and to
KlCmans review (1982).
Defects in liquid crystals 567

2. Static properties of defects. General considerations

We shall introduce static defects in liquid crystals without using the full power of the
homotopy classification; we shall rather present experimental results (observations)
and theoretical analysis essentially at the same time as they appeared historically. This
presentation has the advantage of emphasising the extraordinary variety of the physical
properties of defects in relationship with the molecular properties. For example,
disclinations in N phases appear quite different whether they are studied in short
molecules (usual) nematics or in polymeric nematics, even when they belong to the
same class of local breaking of the symmetry group. Such differences show up in the
core properties and in the relative importance of splay, bend and twist distortions
around the defect. The natural guide in such an approach is the well known Volterra
classification of defects, which works perfectly well for dislocations and to some extent
for disclinations. The necessity of introducing topology will become apparent during
this presentation, as we want to take inspiration from the methods devised for physical
metallurgy; this path has also the advantage of providing us with texture analysis
which is widely used to characterise the nature of a phase by simple observation of
its defects (Demus and Richter 1978, Gray and Goodby 1984).

2.1. The Volterra process for solid media


The Volterra process addresses the question of the origin of internal stresses in a
continuous, homogeneous, elastic medium; the answer is that internal stresses are
caused by elastic singularities which can be analysed as distributions of lines of
discontinuity of the stresses and the strains, each line being characterised by some
invariant, as follows (for details, see Friedel (1964)).
Consider a closed line L in an unstressed solid (figure 7), and a surface Z bounded
by L; cut the solid along Z, and displace the two lips Z' and 2'' of the cut with respect
to each other without deforming them. This relative displacement d is at any point
M of Z the sum of a pure translation b and a pure rotation a . O M where Cl is the
matrix of rotation. The empty space produced in this way is filled with unstressed
matter (or unstressed matter is removed in regions of space which are doubly covered),
the lips Z glued along, and the medium left to relax elastically. According to
Weingarten's theorem (see Nabarro 1967), we are left with a state of stress which is
continuous everywhere, except on the line L ; the result does not depend on the precise
choice of Z.

I
Figure 7. The Volterra process for a line defect (6, n) in a solid.
568 M Klkman

In an ordered solid, continuity through requires a vector b defining a tranlational


symmetry, and for Cl a tensor defining a rotational symmetry. By definition, a (perfect)
dislocation is associated with a pure translational symmetry invariant (b, 0)and a
(perfect) dischation is associated with a pure rotational symmetry invariant (0, Cl).
One speaks of a defect in the sense that the corresponding symmetry is broken along
the line L. A screw dislocation is a straight line defect parallel to the Burgers vector.
An edge dislocation is a line which is everywhere perpendicular to the Burgers vector.
A wedge disclination is such that the rotation vector fl associated with Cl is along the
line; in a twist disclination, fl is perpendicular to the line. Imperfect dislocations
(disclinations) are associated with (b, C l ) invariants which do not belong to the symmetry
group of the medium; the resulting defect carries symmetry breaking on the cut surface.
The Volterra process applied to ordered solid media puts the emphasis on the
invariants and on the topological properties; at the same time it gives clues to the
strains and stresses which accompany the defect under consideration, a result which
is not reached with the topological classification.

2.2. The Volterra process for partially ordered jluids


The Volterra process just described acts in fact on the frozen structure of the liquid;
we also have to consider in a liquid the viscous relaxation of the structure (Friedel and
Klkman 1969, Friedel 1988), whose specific properties can be described in terms of
injnitesimal dislocations db and infinitesimal disclinations dfl; these defects relax most
of the stresses set up by the (solid) Volterra process.
An infinitesimal dislocation db of such a nature belongs generally to the continuous
symmetries of the medium (in a nematic, db can be any vector, since the translations
are not quantised; in a SmA db is parallel to the layers). Consider in the first instance
a straight disclination line L in a nematic, of wedge type, and displace it parallel to
itself by a distance S (figure 8 ( a ) ) . Assume also that fl does not move during this
displacement. In terms of the Volterra process, this is equivalent to displacing the lips

I
1 6 *
I I

Figure 8. ( a ) The displacement of a wedge disclination (0, a)can be accomplished by


the emission or absorption of a dislocation ( b , 0).
( b )Densities of dislocations are attached
to a curved disclination. (c) The curvature of a set of parallel layers can be interpreted
in terms of dislocation densities.
Defects in liquid crystals 569

of the cut surface with respect to each other by a supplementary quantity fl x 6,i.e.
introducing a dislocation of Burgers vector b = SL x 6 perpendicular to the disclination.
Such a process would require a large energy in a solid. In a nematic, the dispersion
of the dislocation into infinitesimal defects db relaxes this energy and is in fact the
process by which fl follows the line L to its new position. Observe also as a consequence
that the movement of a disclination in a liquid is accompanied by a general process
of emission (absorption) of dislocations.
Consider now a curved line L in a nematic. fl will vary from point to point along
L in order to decrease the stresses, but at the expense of the presence of static
infinitesimal dislocations attached to the line. Consider, as in figure 8(b), two points
P and Q = P + tds infinitesimally close on L. The dislocation introduced by the
variation of position of SZ from P to Q is, by reasoning on the cut surface as above,
equal to:
d b = n x Q M -fl x PM=fl x t ds (6)
where M is any point on the cut surface.
Note that the angular variations of the director n about the disclination line can
be understood as being caused by this density of dislocations attached to the line.
This point of view is developed in Klkman (1983b, ch. 6 ) . Such an effect is quite clear
in a smectic, where a static dislocation db located between two layers creates a mutual
glide of the layers, and consequently a curvature of the layers. A density n of
dislocations of such a type leads indeed to a curvature 1/R measured in a plane
perpendicular to the dislocation lines, and which is proportional to n (figure 8(c)).
This approach is in fact much akin to the approach of curvature in bent crystals made
by Nye as early as 1953; in this approach the density n of dislocations introduces extra
matter to an extent that compensates exactly for the strains; accordingly the resulting
bent crystal has a constant density and is unstressed at a scale of the order of the
distance between dislocations.
Finally, while infinitesimal dislocations translate fl along the disclination line,
infinitesimal disclinations change its direction. For example they can be invoked to
describe the wedge character along the entire length of disclinations in cholesterics.

3. Static defects in smectic phases

3.1. Simple disclinations in smectics


Consider a SmA phase; any straight line in a layer is a twofold axis. Corresponding
elementary disclinations are obtained through the Volterra process by introducing
(removing) a sector of angle multiple of T. We illustrate here the cases of a +T wedge
disinclination (or S = + i ) (figure 9(a)) with one sector removed and of a - 7 ~wedge
disclination (or S = -f) (figure 9(b)) with one sector added. In figures 9(d) and (e)
we show how the hodograph of a closed circuit surrounding a line defines the strength
(here S = - 1) of the line: the angular aperture of the hodograph is the angle, the sign
is positive if the orientation of the image circuit in the hodograph is the same as the
orientation of C ; otherwise the sign is negative. Disclinations of opposite strengths
can annihilate by collapse (figure 10) or pair into dislocations (see below). The geometry
+
of wedge disclinations S = +fand S = 1 (figure 9 ( c ) ) does permit parallel layers;
they are therefore of smaller energy than wedge disclinations S = -4, - 1, where
variations in thickness of the layers, or walls, are necessary. Twist disclinations are
570 M Kle'man

la)

Figure 9. Wedge disclination lines in a SmA. ( a ) S = +&; ( b ) S = -4; ( c ) S = 1. ( d ) and


( e ) Strength of a wedge S = - 1 disclination (method of the hodograph).

Figure 10. Disclinations A and B can smoothly approach one another and annihilate.

objects conceptually difficult to visualise (for solids, and R = ~ / 4 see


, Harris (1977))
and of a very high energy. As indicated in P 2.2, the twist component can be relaxed
by dislocations; this situation has been discussed in detail in smectics by Williams
(1976) and Bourdon et a1 (1982).
In fact, disclinations in SmA media appear usually in two types of pairings, forming
either dislocations or focal conics. In both cases they take curved shapes, and the
small energy of these configurations can always be interpreted in terms of infinitesimal
dislocations (or disclinations) attached to the two elements of the pair and possibly
extending from one disclination to the other.

3.2. Dislocations in smectics


Dislocations in smectics can be defined by the Volterra process as in usual crystals.
The Burgers' vector is necessarily of the form b = dong where z is an unit vector along
the normal to the layers, and do the repeat distance of the layers. Dislocations are
either of edge type, screw type or mixed type.
The strain and energy of a dislocation in a SmA phase vary to a larger extent than
in usual crystals with the orientation of b with respect to the line. This is of course
related to the anisotropy of the order parameter, which is reflected in the equations
Defects in liquid crystals 571

of elasticity obtained by minimising the sum of the terms of equations (3) and ( 5 ) .
The dominant terms read:
K , AC\:u= B a 2 u / a z 2 (7)
where
A, = a2/ax2-+ a2/ay2.
One will notice by the presence of fourth-order derivatives how different this equation
is from the equation of standard elasticity.

3.2.1. Edge dislocation. The straight edge dislocation (figure 1 1 ) has been considered
by de Gennes (1972) and KlCman and Williams (1974). The strain a u / a x and the
energy per unit length of line is:
d u l d x = $&[ b/(Az)] exp( - x 2 / 4 A z )
W = Klb2/2Ar,+ W,
where A = (K1/B) is a penetration length, W, the contribution of the core to the
total energy and b = ndo. Different core models will be discussed as a function of
A / d o . This quantity is usually of the order of unity, but can be much smaller in some
materials where the layers are very flexible ( K1small), like non-ionic surfactants + water
(Oswald and Allain 1985) or lamellar phases close to microemulsions (de Gennes and
Taupin 1982).

Figure 11. A birds eye view of a straight-edge dislocation.

The strain 13 = a u / d x is strong inside two parabolas x2 = *4AZ, and weak outside.
Its variation with x and z is special to the elasticity of equation ( 7 ) .
The energy varies as the square of the Burgers vector, which should favour
dislocations of small Burgers vector ( n = 1 ) . However, some observations indicate
that dislocations of the same sign gather into dislocations of huge Burgers vectors.
This might be explained with a core model which is split into two disclinations L1 and
L2 of opposite signs, as in figure 1 2 ( a ) (such a splitting was proposed long ago by
Friedel (1964) for dislocations in graphite). In this model, r, is of the order of the
distance between L , and L 2 , and scales like b ; the elastic energy is then proportional
to b, and the core energy reads:
W,= ( 1 r K , / 2 )In b / 2 d o + w, (9)
where w, is a new core energy for the disclinations which does not depend on b. The
first term on the right-hand side of equation ( 9 ) is pure curvature energy (of L 2 , where
the layers are parallel). Strain energy of L , can also be minimised (at the expense of
curvature energy) by some instability along the line, which can also be described as
some (longitudinal) splitting of the line into focal domains (see 0 3.4).
572 M Kleman

Figure 12. Splitting of an edge dislocation into disclinations of opposite signs. ( a ) Usual
model, A = d o ; ( 6 ) anisotropic core in a low K , lyotropic smectic.

The core model which has been discussed has to be modified when A/do is very
small (Allain and KlCman 1987); the core extension & along the z axis is still the
distance at which the width of the parabolas x2= f 4Az reaches do (in this region the
curvature of the layers is larger than l / d o and the linear analysis of the distortions
fails), but this quantity & = di/A is now much different from & = rc = b and the core
is anisotropic (figure 12(b)). Note that .$ does not depend on n. Such anisotropic
configurations of cores probably exist in small K 1 layered systems, and they have
already been observed in 2D layered systems formed by the roll electrohydrodynamic
instabilities of nematics (Ribotta and Joets 1984); the quantity analogous to A is here
the frequency of the AC voltage which drives the instability. When A is small, the
extended dislocation core plays a major role in the shear of the varicose electrodynamic
instability.
Edge dislocations are present in homeotropic wedges where they gather along the
mid-sample tilt grain boundary. Their mean distance A is then simply related to their
Burgers vectors and wedge angle (cf figure 26):
A = b/a. (10)
In this geometry energetic reasons (Pershan and Prost 1975) would rather favour
dislocations of small Burgers vector, as is found experimentally (Meyer et al (1978),
Defects in liquid crystals 573

Chan and Webb (1981a), Bourdon (1980), see also the recent experiments of Nallet
and Prost (1987) in a lyotropic SmA with a very large repeat distance -1000 A).

3.2.2. Screw dislocations. An obvious displacement field for a screw dislocation is, as
in usual dislocation theory:
U = b9/21r. (11)
The line energy, calculated to second order in a SmA, is zero, the smaller the Burgers
vector, the better this approximation. Thus, to this order, screw dislocations are objects
whose energy is restricted to a core energy: screw dislocations of small Burgers vector
have indeed a very small energy and are experimentally extremely frequent; they have
been observed with a density of 1O8cm- in different lyotropic SmAs by electron
microscopy of freeze-fractured specimens, which cleave easily along the interfaces of
the paraffinic chains (KlCman et a1 1977, Allain 1985). In these conditions screw
dislocations piercing the cleavage surfaces determine cleavage steps which gather in
rivers (figure 13) as in usual crystals (Friedel 1964, ch. 12). They have also been
observed by optical microscopy in homeotropic samples of SmA, near the SmC
transition (Meyer et a1 1978, Bourdon 1980). It is interesting to notice that in the latter
observations, edge dislocations finite segments are often visible (figure 14): at each
end of the segments there is a vertical screw dislocation segment continuing the
dislocation line, with of course the same Burgers vector. The absence of a smooth
variation (at least on an optical length) from the edge to the screw part of the line is
a further indication of the small energy of the screw dislocation; at the same time any
fluctuation in shape of the screw segment is difficult, which can be ascribed to a large
configurational line tension of the screw line. A calculation by Bourdon et a1 (1981)
gives:
T= 0.004Bb2 In(21rdAlrf) (12)
for a line of length d pinned at its extremities. A screw line resists curvature by
developing a configurational forcef= T/ R, where R is the radius of imposed curvature;
curvature arises (see D 4.1.2.) when the layers experience a compressive stress a,which
induces a Peach and Kohler force on the line f = a b (for this concept, see Friedel

Figure 13. Screw dislocations, steps and rivers on a cleavage surface of a non-ionic lyotropic
surfactant (freeze-fracture; courtesy M Allain).
574 M Klkman

Figure 14. Edge dislocation segments in a SmA thermotropic phase. Polarising microscope
photography, courtesy L Bourdon.

(1964) pp 31 and 34; some subtleties concerning its use in the case of liquid crystals
have been emphasised by Lejcek (1986)).
Taking into account higher order elasticity, screw dislocations are well understood
in two limits:
(i) in the vincinity of a SmA-nematic second order transition, A = ( K 1 / B )
becomes very large since B = 0 in the nematic phase; in the limit of B = 0 we must
have a vanishing density of splay, i.e. any geometry must obey the condition div n = 0,
which means, for a system in which the director field n is perpendicular to a surface
(the layers), that the mean curvature of the layers vanishes everywhere (see KIiman
(1983a ch. 5); the layers take the shape of minimal surfaces. For an isolated screw
dislocation the solution is a helicoid, whose equation reads:
cp= -b8/27r+z (13)
where cp = cp(r) is a phase variable constant on each layer, z is the coordinate along
the dislocation, and 8 a polar angle. Clearly cp + cp - b when 8 + 8 + 27r at constant z.
Equation (13) is in fact the same as (1 l), with U = z - cp, but now the solution is valid
to any order. The energy reads:
W = Bb4/ 128(r, - R - 2 )+ w, (14)
where R is the external radius of the sample. W does not diverge when R + C O and
the b4 behaviour ensures that only dislocations of very small Burgers vector are stable,
which differentiates them greatly from edge dislocations. Also, the order of magnitude
of (10) is 1000times smaller than that of the edge dislocation of the same Burgers vector.
(ii) When B is large compared with K1 (i.e. any interesting length in the problem
is larger than A ) , the layers prefer to stack with constant thickness do, i.e. parallel to
each other. Then & / a n must be taken to be equal to zero. Such a geometry has been
studied by Frank and Kliman (see Kliman (1983a) ch. 5). It consists of a parent layer
having the shape of a helicoid, with periodicity b, and a set of layers parallel to the
parent layer and stacked at constant distance do from each other. Let us emphasise
the difference with the former model: when div n =0, all the layers are helicoids;
Defects in liquid crystals 575

therefore, if b = ndo there are n such helicoids that fill the space; in the continuous
limit we have ( 1 1 ) . Clearly the axis r = 0, on which all these helicoids meet, is a
singular axis, and a core model has to be devised. However, when & / a n = 0, there is
only one layer in the shape of a helicoid, passing through the axis r = O , and no
singularity. The normals to the set of parallel layers are straight lines: note the analogy
between the set of layers and a set of wavefronts, between normals and light rays.
There is therefore a focal surface enveloped by the normals; in fact, the normals meet
two helices C, and C2 of pitch b, which are singular lines of the geometry and cusp
lines of the focal surface. C,and C, are the only visible singularities: the layers fold
around these lines as around disclinations (see figure 15). This is another example in
which two disclinations of opposite rotations couple to give a dislocation. This double
helix geometry has been observed by Williams (1975) in some SmA samples, with
dislocations of huge Burgers' vectors (figure 16). An estimation of the energy of the
screw dislocation split into a double helix leads to:
W=1.81K1+10-2Bb2+w, (15)
per unit length of screw line. The first term refers to the energy of the configuration
of the disclination type around the helices; the second term refers to those long-range
deformations created by the screw line outside the cylindrical region enclosed by the
helices: in this region splay is unimportant, and this term is in fact akin to the energy
calculated above (equation (14)), with rc = b/2 ( w , is a core energy). It can be shown
that the splitting in a double helix is favoured only for large enough Burgers' vectors
(i.e. b = 25do and above); this is why it is possible to observe this beautiful geometry
quite easily in experiments.

b/2n
Figure 15. Screw dislocation split into two helical disclinations C , and C2 of strength
S = f : transverse (above) and meridian cuts of the geometry of the layers. 5 = b/27r.
A measures the distance of the layer to the parent layer A = 0, which is a helicoid.
576 M Klkman

Figure 16. Screw dislocation of huge Burgers vector split into a double helix. Courtesy
C Williams.

Pleiner (1988) has recently calculated in their full generality the role of non-linear
terms in the energy of screw dislocations. His conclusions are that there is a large
contribution of bend elasticity near the core, and that two parallel screw dislocations
of same sign attract through a potential which varies logarithmically with distance.
The standard core model for a screw dislocation of Burgers vector unity in a
lyotropic is indicated in figure 17. Its advantage is to maintain a clear separation
between water and aliphatic chains, either water in the core (as in figure 17), or,
alternatively, aliphatic chains on the core. The core is in fact a sort of extended pore
in a unique membrane (there is only one layer, connected by the helical geometry). It
a unique membrane (there is only one layer, connected by the helical geometry). It
explains rapid diffusitivity in the bulk of amphiphilic molecules, which do not have

Figure 17. The core model for an elementary screw dislocation in a lyotropic medium.
Defects in liquid crystals 577

to jump through water to pass through the medium, and of water (pipe diffusion). If
n > 1, one can imagine n superimposed helices. But a definitive model of the core has
still to be worked out in this case.
Morawietz and Demus (1987) have argued that the growth of the SmA phase in
the shape of b6tonnets (Friedels term) occurs with a screw dislocation, very similar
to the mechanism of growth in usual solids.

3.2.3. Interaction between a microscopic impurity and a dislocation. The static deforma-
tion field of an impurity in a smectic phase and its interaction with a dislocation has
been studied by Lejcek (1986).

3.3. Focal domains


This is by far the most well known defect in SmA, thanks to Friedel (1922) who
described it completely in geometrical terms, and gave the main properties of the
textures made of focal conics. Other very valuable descriptions have been published
after his, in particular by Bragg (1933) who gave an analytical description without
being aware of Friedels work, and by Bouligand (1972a).

3.3.1. A quick reminder of the geometry of focal domains. Domains built on cofocal
conics are the most frequent non-trivial arrangements of parallel layers met in smectics
at the scale of the optical microscope observations. The condition of parallelism
implies curl n = 0, n being the normals to the surface: there is no twist (necurl n = 0)
or bend ( n x curl n = 0) deformations in focal domains, and n does not change as it
moves parallel to itself. It is the fundamental geometry which best relaxes imposed
boundary conditions in a smectic phase with liquid layers (i.e. where one layer easily
glides upon another, at no compressive elastic expense, only curvature energy). Thus
geometries of parallel layers are observed in SmAs and SmCs; this is essentially under
the form of focal conic domains (Friedel and Grandjean 1910), where the layers have
negative Gaussian curvature. Note however that spherical layers are also observed,
either between focal domains, or in free droplets, in both cases at small scale lengths
(Sethna and KlCman 1982, Lavrentovich 1986a).
Consider a set L of parallel (but non-planar) layers: as a general result, this set is
singular on the centres of curvature C1and C2of the layer surfaces; all parallel layers
have the same normals and the same C, and C, on a given normal; C1and C2generate
the evolute F of L, formed of two sheets F1and F 2 . In geometrical optics, F is called
the focal surface of L; it is indeed also the envelope of the normals to L. A pictorial
discussion of some topics in the theory of surfaces, quite useful for a thorough
understanding of what follows, can be found in Hilbert and Cohn-Vossens book
(1952); the interested reader who wants to go into more depth should read Darbouxs
remarkable treatise (1888, reprinted 1954).
The singularities of a set of parallel layers can be two- one- or zero-dimensional.
Two-dimensional singularities of parallel layers have never been observed; a onc-
dimensional subset of a 2D focal surface F is sometimes visible, the rest of the focal
surface being virtual. We have met such a situation when previously discussing the
splitting of screw dislocations into two helices (cf also KlCman (1983a) ch. 5).
A zero-dimensional singularity occurs only when the layers are spherical.
If both sheets of the focal surface are lines (one-dimensional singularity), then it
can be proved (Darboux 1954) that the two focal lines are cofocal conics (one ellipse
578 M Klkman

and one hyperbola, or two parabolas), and that the layers are Dupin cyclides (figure
Wa)).
Special cases of Dupin cyclides occur when: (i) the hyperbola is at infinity and the
cofocal circle E is a straight line; the layers form a set of parallel circular cylinders
about E. (ii) The hyperbola H is reduced to a straight line and the circle E is of finite
radius, R, centred on H ; the Dupin cyclides are nested tori centred on E, which is a
singularity of the same type as in (i) (figure 18(b)). The singularities of H consist in
conical points of layers, with a solid angle at the apex which tends to 27r at long
distances, hence the singularity on H becomes negligible at some distance of the plane
of the circle. (iii) Parabolic focal domains (Rosenblatt et a1 1977). The limiting case
of PFDS has not been studied by Friedel and Grandjean. The distortion of the layers
is extremely weak, except in the region where the parabolas cross (see the excellent
computed plots of the distortions in Rosenblatt et a l ) . PFDS are of small energy, and
are easily formed (figure 1 8 ( c ) ) .
The generic case has been well described quite often, and we refer the reader to
the classical reviews on the subject, starting with Friedel (1922).

3.3.2. Textures made of focal domains. The huge majority of observed focal domains
contains only those parts of the Dupin cyclides with negative Gaussian curvature, i.e.
those parts where the surface locally looks like a saddle. A few exceptions have been
observed (Meunier and Billard 1969, Zasadzinsky et a1 1985), where the surface is

I ,

Figure 18. Focal domains. ( a ) General case; ( b )toric domains; (c) parabolic focal domains:
a frequent association of PFDS, as it occurs in strongly dilated homeotropic samples.
Defects in liquid crystals 579

locally spherical. Most of them relate to free surfaces. The reason for the predominance
of saddle surfaces is probably because they have less curvature energy, since the splay
term reads 4Kl(uI+ u2)2 for a surface of principal curvatures u1= 1/ R1 and u2= 1/ R 2 .
It is therefore an advantage to have uluz< 0. Conditions of growth can also play a role.
In an actual sample of SmA, focal domains are the rule, rather than the exception,
except if great care is taken to produce a single crystal, generally in the form of a
homeotropic sample. Focal domains appear either complete (as described above), or
fragmented (the ellipse or the hyperbola being incomplete) or limited between two
cones with apexes on the hyperbola, etc. But they cannot tile space: this is forbidden
by their shape; in fact, in most instances, they associate according to rules that have
been described by Friedel and which can be satisfied entirely in an often-encountered
geometry (Friedel 1922, Williams 1976)-the geometry of a tilt grain boundary; this
we will now go on to describe.
Consider, as in the scheme of figure 19, two single crystals of SmA which meet
along a plane bisecting the normals to the layers on both sides; the plane is necessarily
a region where the order parameter experiences a strong variation. Assuming that this
variation occurs elastically, this requires an energy ( K,/B)12f( a ) with f(a )= a 3 for
vanishing angles (de Gennes 1974) but tending to a constant for not too small angles.
This is large compared with two other possibilities: introducing dislocations in the
plane, or focal domains. Let us consider these two possibilities.

I
Figure 19. A grain boundary in a SmA with smooth variation of the layers through the
boundary.

(i) Dislocations. A grain boundary can be split into dislocations (Friedel 1964).
This is a usual situation in solid crystals. Call a the angle between the layers on both
sides, and assume that the parallel edge dislocations of Burgers vector b are located
at a distance d = b / a from one another. The energy per dislocation unit length being
of the order of K l b 2 / A r , ,the energy per unit area is approximately (K1B)12when
rc = b. This solution is chosen in low-angle grain boundaries, and occurs in homeotropic
samples where the boundaries are not always strictly parallel (figure 26).
(ii) Focal domains. In this geometry, the ellipses are in the plane of the grain
boundary and the hyperbolas have asymptotic directions perpendicular to the layers
on both sides of the boundary; therefore they belong to planes perpendicular to the
boundary and the axis of rotation, and all have the same eccentricity e H . The ellipses
580 M Klkman

therefore have eccentricities eE = e i ,and have parallel major axes. Furthermore, they
are in contact, in order that the focal domains they carry fill space best. These focal
domains are tangential along the molecular directions which join these contacts to the
points at infinity along the asymptotic directions (figure 20), therefore they match
together with a vanishingly small energy, so that they tend to multiply, and focal
domains of smaller and smaller size could hierarchically appear between larger focal
domains and fill in the gaps, down to some size where the multiplication of singularity
lines should no longer be favoured.

Figure 20. A grain boundary in a SmA crystal. The layers are perpendicular to the long
singular lines which are hyperbolas belonging to focal domains whose ellipses are in the
grain boundary. Courtesy C Williams.

This iterative filling of the boundary with ellipses all of the same elongation and
the same eccentricity does satisfy the first of Friedels laws of association (if two focal
domains are in contact, they are tangential to each other along the common generator
of their boundaries). The second of Friedels laws (law of corresponding cones) states
that the cones of generators which have a common apex M belonging to two conics
must coincide. This is achieved in the geometry of the grain boundary if the contacts
M between ellipses are either along the major axes, or along the minor axes. This
seems to be the case in the observations of Williams and Kliman (1975) in CBOOA.
In lecithin, some observations from freeze-fractured specimens suggest a minimum
size of the order of 1000 A (KICman et a1 1977a, b).
Bidaux et a1 (1973) have considered the (simpler) iterative problem of circles; it
corresponds to the case of a zero disorientation angle a with hyperbolas reduced to
straight lines; Friedels second law is satisfied for any contact between circles. The
problem of the iteration of circles is known in mathematics as Apollonius problem.
Let g ( p ) be the number of circles of radius R > p, when the iteration reaches circles
of radius p. We have a scaling law g(p)--p- with t of the order of $ (numerical
calculation). Bidaux et al show that the energy per unit area then scales as:
U - ( K , B ) * (I/L)*- (16)
where L is the size of the largest circles, 1 the size of the smallest ones. Note that,
when a # 0, because of Friedels second law, the perimeter P ( p ) should increase more
slowly when p decreases, i.e. t should increase with a. Hence U decreases with a,at
constant L and 1. This supports the fact that grain boundaries, except for very small
angles, are split into focal domains (figure 20).
Defects in liquid crystals 581

In a grain boundary, the hyperbolas meet at infinity. The case when they all meet
at a finite distance is no different in principle: the planar layers between the focal
domains are replaced by spherical layers centred on the point of intersection of the
hyperbolas (Sethna and KlCman 1982).
The principal textures made of focal domains are described in Friedel (1922).
Amongst them let us quote the textures d bbtonnets, which appear at the isotropic-SmA
transition: (these b4tonnets are complex nuclei of SmA phase), the polygonal textures,
the fan textures, the Grandjean terraces (often decorated with small focal domains),
etc (for illustrations, see Gray and Goodby (1984)).
3.3.3. Energy of focal domains. The energy of the region of size A surrounding the
focal lines is difficult to calculate, since it is a singular region. It is reasonable to take
it of the order of K1 per unit length of line. The curvature energy of the bulk domain
has been calculated by KlCman (1977, 1983a), using Darbouxs method of local axes
at each point M of each layer (Mx, and Mx2 along the lines of principal curvature,
Mz along the normal). The fundamental quadratic form of a Dupin cyclide reads in
these axes:
ds2 = A du2+ B2 du2 (17)
with
A = ba2/ ((+I - ~ 2 ) B = kbq1 /((TI - ~ 2 )
(18)
(TI = 1/R1= ( C COS U - z)- a , = l / R , = ( a cosu-z)-l
where c = a - b and e = c / a ( a and b are the major and minor axes of the ellipse); U
and U parameterise each layer z, where z is measured along any normal to the layers.
Formulae are different for parabolic focal domains.
For small eccentricities, the bulk energy of a focal domain reads:
w = ~ T K , (-I e 2 ) K ( e 2 ) aIn a / r c (19)
where K ( e 2 ) is the complete elliptic integral of the first kind and rc is a core radius;
W is a decreasing function of e (for small values of e, as in equation (19) but also for
large es) for a given value of the major axis a (the domain volume scales like ab).
Hence, the situation in which the Dupin cyclides degenerate into tori is not stable:
the tori which are visible in real samples must be stabilised by some external factor,
generally surface anchoring forces, or topological forces (the transformation from a
torus to a generic cyclide requires a change in the nature of the core; this topological
barrier is indeed equivalent to the transition from a wedge disclination (the circle) to
a twist disclination (the ellipse) and requires the absorption or the emission of a
dislocation line (see 90 6.3.1., 7.1.2); it might be high).
For a parabolic focal domain confined to a cylinder of radius R, the bulk energy
reads:
W = 47rfK, In( R 2 / 4 f r , ) . (20)
Parabolic focal domains have smaller energy, for a given volume, than any other
type of domain. In fact, PFDS appear in all sorts of geometries. They build instability
in the strong non-linear stage of the undulation which appears when homeotropic
samples are submitted to large dilations (Rosenblatt et a1 1977, Oswald et a1 1982),
exist down to microscopic sizes (Clark and Hurd 1982) and have been observed in
thermotropic and in lyotropic smectics (Asher and Pershan 1979, Scudieri et a1 1979)
(see figure 18 above).
582 M Kliman

Equations (19) and (20) refer to isolated focal domains, which do not carry any
elastic energy of compression of layers. The iterative fillings of space by focal domains
necessitates some energy of that sort, which is included in a calculation like the one
leading to equation (16).

3.4. Layer instabilities and defects


SmA layer instabilities occur frequently, either as localised defects in the SmA phase,
or driving some phase transition through topological changes of the layers (see (iii)
below; also the smectic- nematic transition in the surfactant DACl/ NH4/Hz0, which
occurs through the formation of micelles out of the bilayers (Holmes and Charvolin
1984, Mihailovic 1987)).
(i) Layer dilation instability. The first stage of the non-linear regime of the undula-
tion instability of Delaye et al (1973) is made of edge dislocations parallel to the
undulations, of alternate signs. They relax easily back to the homeotropic configuration
(Ribotta and Durand 1977). These dislocations transform to PFDS in the strong
non-linear regime, as already indicated. The relation of dislocations to focal domains
is in fact quite general (Rault 1975).
An edge dislocation of large Burgers vector or a dislocation at a free surface
(Grandjean terraces) are most often decorated by some sort of longitudinal static
instability, which in many cases can be analysed in focal domains, whose ellipses are
transverse to the line defects, when b >>A. This is of course the result of a strong
tendency towards parallelism of layers; the instability length scales like b (KlCman
unpublished) and obtains as a compromise between the gain of energy (with respect
to the pure dislocation) because of the focal domains, and the extra energy of the
focal lines. Such instabilities have been observed for microscopic Burgers vectors ( L a
lecithin free layers in electron microscopy, Colliex et a1 (1976)) and for dislocations
of huge Burgers vectors produced in shear experiments (Williams and KlCman 1975).
(ii) Curvature instability. The curvature energy of the layers surrounding a screw
dislocation of large b is relaxed when it splits into two S = i disclinations, as described
above (8 3.2.2.), since lvl+ u21= 0 on the central surface. Oily streaks, on the other
hand, are probably related to the pairing of edge dislocations. They were first described
by Friedel in homeotropic lecithin-water systems ((Freidel (1922), see also Nageotte
(1936)) and since discovered or rediscovered in many other materials (thermotropic
smectics: Steers et a1 (1974); DMPc-water: Powers and Pershan 1977; binary and
ternary block co-polymer-solvent systems: Ballet et a1 (1981), Candau et a1 (1982),
Wittman et a1 (1982); quaternary systems containing anionic surfactants: Hirsch et a1
(1982), and also in cholesterics (Chistyakov 1963, Gerritsma and van Zanten 1972)).
In this geometry, the energy of the edge dislocation is relaxed by a longitudinal
curvature instability, probably in the same manner as the bead instability observed
along the axis of homeotropic tubes (Cladis and White 1976). Recent very careful
observations of oily streaks (Schneider and Webb (1984), see also Benton and Miller
(1983)) have led to the scheme of figure 21(b), which reveals clearly the saddle character
of the layers near to the two S = i singularities; existing theories indicate that the
periodicity would scale like where t is the thickness of the samples or the
radius of the tube (de Gennes and Pincus 1976). However, experiment seems to be
closer to a scaling law proportional to t, as in the instabilities described in (i).
(iii) Saddle-splay instability. Although the K24 term (equation (3)) can be trans-
formed to a surface term, it plays a role when there is a change in the layers topology,
Defects in liquid crystals 583

P L Pi

'Lx
Figure 21. Paired dislocations of same sign in a smectic (model for oily streaks); ( a )
transverse cut; ( b ) longitudinal cut; with instabilities displayed.

and has as such been invoked to explain the stability of pores in layers, micelles, etc
and of various lyotropic phases. For example, micelles are favoured by K24> 0
(Turkevich et a1 1984) in certain ranges of concentration of water versus surfactant;
recentiy Helfrich (1986) has advocated a < 0 to explain the stability of cubic phases.
This is related to the fact that the saddle-splay term factorises the Gaussian curvature
(equation (4)), hence K24> 0 favours positive Gaussian curvature, i.e. surfaces which
are locally spherical while K24 < 0 favours saddle-shaped structures. Such structures
with negative Gaussian curvature occur locally in pores (which have the geometry of
the inner part of a torus) or in the cores of screw dislocations of lyotropic materials.
Saddle-shaped structures were first discussed by Scriven (1977) for microemulsions,
and Helfrich (1981) argued that is negative in lecithin, where lattices of pores in
layers have been observed (Harbich et a1 1978). If this is true, this might also explain
the strong tendency of lecithin to grow in water in myelinic shapes which easily curve
into helical shapes, like flexible polymers (Friedel 1927, Sakurai et a1 1985).

3.5. Defects in layered media other than SmA: generalities


The SmA phases can be considered as the archetype of layered media: the layers are
simple 2D isotropic liquids, and the associated defects (dislocations, focal domains)
do not involve anything except the notion of liquid layers. Two degrees of complexity,
at least, appear with the other smectic phases: (a) with anisotropic liquid smectics,
like SmC or SmC", the classification of defects of SmA phases is still relevant, but in
addition the singularities of the projection t of the optical axis on the layers provide
for new defects; (b) 3D smectics like SmB or SmG show up some textures (mosaic
texture, grain boundaries) which are expected here as in usual crystals, in addition to
the layer dislocations discussed for SmAs, but these 3D smectics also show up, more
interestingly, paramorphotic textures of liquid smectics, like focal domains, which
involve complex relations of epitaxy between layers and are described in the simplest
cases in terms of the specific defects which break the internal symmetries of the 2D
layers.
Except for smectics with anisotropic liquid layers, most of the corresponding
textures, although thoroughly observed and well described (Gray and Goodby 1984)
have not been deeply investigated; the same is true for the SmB hex, SmI and SmF
phases, which are intermediate between liquid smectics and 3D smectics in the sense
584 M Kleman

that they have short-range (in-plane only) positional correlations of the molecules, but
keep long-range ( 2 D or 3D) bond-orientational order. Finally SmA and incommensur-
ate thermotropic smectic phases (for a recent review of their structural properties, see
Hardouin et a1 (1983) and the more theoretical paper by Prost (1984))are still practically
not investigated for their textures; however, some lyotropic phases of similar symmetries
( P p equivalent to SmB, and ribbon phases) have been studied by freeze-fracture
techniques (see, for example Ruppel and Sackmann (1983)). We shall therefore
highlight the essential problems raised by the smectic new phases, rather than describe
in detail the textures and defects (for such a purpose see Gray and Goodby 1984).
Finally some smectic phases of chiral nature show up phenomena believed to be
typical of frustration (Carlson et a1 1988). We shall discuss them in 9 8.

3.6. Defects in S m C phases; bulk specimens


We shall characterise the local states of a SmC phase by a tripod n, t and N, where n
is the normal to the layer ( n = -n), and t is parallel to the layer and such that the
plane (n,t ) is a plane of symmetry. Furthermore N = t x it. Since the medium is biaxial,
the optical axis, the diamagnetic axis, etc are not necessarily aligned and it is not
possible to define a director unambiguously (Frank (1983)). The free energy has been
worked out by the Orsay Liquid Crystal Group (1971) and Helfrich (1974). For
extension to large distortions see KlCman (1983a ch. 7). In this section we examine
disclinations and dislocations typical of SmCs in terms of the Volterra process; similarly,
we present some specific results concerning focal domains in terms of simple
geometrical constructions, in particular the presence of disclinations inside focal
domains. However, some of these problems will benefit greatly from the topological
classification of defects, which we postpone until 7.

3.6.1. Disclinations of the tripodjield. The Volterra process proves useful in investigating
the configurations corresponding to the breaking of the symmetries of rotation: (i) the
n axis is a 2rr axis of rotation for the field t (and the associated field N ) ; (ii) the N
axis is a twofold axis for the n field and the t field (figure 22). We study the associated
defects in turn. More details and figures can be found in Allet et a1 (1978), Bouligand
and KlCman (1979) and Bourdon et a1 (1982).

/ f

Figure 22. The local tripod in a SmC.

(i) Defects of the t field were observed a long time ago (Taylor et a1 1970) and
constitute the elements of the schlieren textures of SmCs. The distribution of t about
the corresponding disclination lines, which are perpendicular to the layers, is similar
to the director distribution about wedge lines of integral strength in nematics. We use
the notation m(2rr), m(-27r) for these m lines.
Defects in liquid crystals 585

c f-

Figure 23. Disclinations in a SmC phase: ( a ) I ( + ? r ) wedge line, perfect; ( b ) I(+.n)


wedge/twist line, imperfect; (c) Planar representation of one layer of the disclination in
2 3 ( b ) ; ( d ) NCel walls and m lines separating them.
586 M Klkman

(ii) Defects which simultaneously affect the n field (i.e. the lamellar structure) and
the t field (i.e. the director field) will be noted l ( * ~ ) 1(*257),
, and called 1 lines. In
the simplest case, the distortions of the t and n fields are coupled as in figure 23(a),
which represents an I ( + T ) wedge line, which is a S = ++for the t and the n fields
separately. Figure 23(b) represents an imperfect line in the sense that it has wedge
character for the n field and twist character for the t field; it can be obtained as follows:
assume that N is at some angle from L and create a disclination of the n field about
L, letting the t field attach unperturbed to the layers. A surface discontinuity of the t
field appears on the layer bound by L. Each molecule located on one side of the
discontinuity is then rotated by an angle 2 a about n and allowed to relax; there appears
a twist disclination for t along L, which can also be described as a NCel wall of angle
2a; if any layer surrounding L is unwound on a plane, the directions t(m) and t ( - 0 0 )
make an angle 2 a (figure 23(c)). Such NCel walls in l ( + v )and l ( - ~ )lines have been
observed and calculated (width, energy) by Allet e? a1 (1978) in DOBCP. Hinov (1984)
has calculated the effect of a magnetic field on such lines, and shown that the partial
twist character can disappear above some threshold field.
NCel walls have a definite chirality; NCel walls of opposite chiralities are separated
by m lines (figure 23(d)) transverse to the 1 lines, which possibly form loops by joining
two transverse m lines (wedge) by two twist m lines. When cooling a specimen from
the nematic phase to the SmC phase such disclination loops often remain: they delimit
regions of opposite chiralities.

3.6.2. Focal conics. Focal domains exist in SmC phases for the same reason they do
in SmAs: layers can easily glide one upon another, and have a minimal energy when
they remain a constant distance apart. But focal lines can also be defined topologically:
they are disclination lines of mixed character (wedge and twist), with rotation vector
vector along N (Bourdon et a1 1982). In the present context of SmC phases, focal
ellipses are I ( T ) lines of mixed character. Many situations can occur for the t field,
the two extreme ones being, for a toric domain, the following: (i) N is parallel to the
singular circle everywhere: we have in any section the configuration of a perfect wedge
I( 5 7 ) ; the axis of the focal domain is an axial S = +1 line for the t field (i.e. an m(257)
line); it is this axial disclination which curves N along the singular circle.
(ii) N has a constant direction at infinity: the local I ( T ) line along the singular circle
is of mixed, variable, character, with a varying with the polar angle cp of the circle as
shown in figure 24(a). 2 a is discontinuous for cp = 5712 and cp = 3 v / 2 , which fact
necessitates the presence of two wedge m lines, which are of strength S = -1 (see
figure 24( 6 ) ) . The two m lines are attached to the singular circle; therefore they pass
through all the layers surrounding the I ( v ) line and reach the axis of the toric domain,
either at the same point (this is energetically the most probable situation, and it has
been observed and analysed first by PCrez et a1 (1978)), or in two different points
which must therefore be the extremities of a S = + 1 , m ( 2 n ) line segment, along the axis.
The case of a general focal domain is more difficult. It has been discussed at length
when N is uniform at infinity (Bourdon et a1 1982).

3.6.3. Dislocations. Edge dislocations form low-angle grain boundaries in homeotropic


samples (since the slides can never be exactly parallel); they are visible in this situation
near the SmA-SmC transition temperature To, especially on the SmA side (Meyer et
-La
Defects in liquid crystals

R aN
587

Figure 24. Line singularities in a focal domain of a SmC phase; see text.

a1 1978), because of the compressional stresses near the core which trigger the transition
in this region at a slightly higher temperature TL> To: the medium is then locally
biaxial, which is enough to provide for a strong optical contrast of the lines. Many
observations of the physical properties of the lines (relationship with m lines, visibility
of finite segments of edge lines continuing along screw lines, different mobility in the
SmA and SmC phases, etc (Bartolino and Durand 1978, Bourdon 1980)) have been
made with the help of this technique. Screw lines seen head-on are also visible.
Some calculations of the line energy and speculations on the core of the screw
dislocations have been made (KlBman and Lejcek 1980). The energy per unit length
of line reads:

W = [(A12A21)2/2/a](b2/4~)j32r;2 (21)

where, with

K =$All(A12A21)1/2

we have

and r, is some length related to the core radius. For the definition of the involved
elastic coefficients see KlBman (1983a ch. 7). W goes to zero when j3 + 0, i.e. in the
SmA limit of the anisotropy of the coefficients. The same authors propose that the
core is liquid, and that the molecular configuration in its vicinity is an m ( 2 ~ line
)
centred on it and screened at a short distance by an m( - 2 ~ line.
) The same model
has been proposed independently, on experimental grounds, by Ishikawa et a1 (1984).
588 M Kliman

3.7. Defects in SmC* phases; bulk specimens


Chiral SmC phases have attracted much interest because of their ferroelectric properties
(Meyer et a1 1975, Meyer 1977, Handschy et a1 1983). Each molecule bears a permanent
dipole moment P parallel to the plane of the layer. P rotates with the pitch of the
SmC*, so that the mean dipole moment vanishes over a pitch. However, all the dipoles
can be oriented by an applied electric field, which effect is at the origin of electro-optic
switching display devices of industrial importance (Clark et a1 1983). The elasticity
of these systems (Meyer 1977) involves spontaneous twist and spontaneous bend of
the director.
These devices consist of thin specimens (with thickness d smaller than the pitch
by the order of a few microns) with layers oriented perpendicular to the glass plates
and strong anchoring conditions (or surface stabilised ferroelectric liquid crystals
(SSFLC)). In these conditions all the dipoles orient in the direction of the applied field
(whichis in the direction of the anchoring) above some threshold E,= K / d P . The
reversal of the orientation, when E changes sign, is by motion of walls which are either
of the Bloch type or of the NCel type, according to their orientation. This bistability
process has not been observed in films obtained with ferroelectric SmC* mixtures of
very large pitch (Cladis et a1 1983).
In thicker specimens ( d > pitch) of the same orientation, still with strong anchoring,
and in zero field, there is a transition region between the surface and the bulk, with a
row of disclinations (Martinot-Lagarde 1976). These dechiralisation lines were first
studied in detail for their topology by Brunet and Williams (1978) (see figure 25); the

Figure 25. Dechiralisation lines in a SmC* phase (courtesy Monique Brunet). Planar
anchoring on both lamellae.
Defects in liquid crystals 5 89

unwinding of the twist under an electric field proceeds by motion and annihilation of
these disclinations, and has been studied in great detail, both experimentally and
theoretically, by the Czech group (Glogarova er a1 1983, Lejcek 1985, Fousek and
Lejcek 1986, etc).
It has also been observed that focal conics in a SmC" always contain disclination
lines (Perez et a1 1981, Bourdon et a1 1982). This phenomenon has the same origin
as the similar phenomenon in a SmC.

3.8. SmC and SmC" free suspended films


Free suspended films of SmC and SmC" phases have been much studied because of
their interest as physical realisations of X Y two-dimensional systems (Rosenblatt et
a1 1980). Such free films contain from one to a few layers; at such a scale SmC and
SmC" phases essentially do not differ, except that it is easier to act on SmC" films
through an electric field. Macroscopic changes of orientation and defects can be
observed with the help of polarised reflection microscopy (Rosenblatt er a1 1980,
Pindak er a1 1980). In particular, it is worth noting the results concerning singular
points (2D X Y vortices), walls and their collapse under an electric field, etc. Recent
observations of the mobility of vortices (van Winkle and Clark 1985) have stimulated
numerical simulations (Loft and DeGrand 1987) and theoretical studies for the stochas-
tic motion of these disclination points (Pleiner 1988a, b).

3.9. Lamellar phases with long-range and short-range 2 0 positional order


SmB, H, E, etc are three-dimensional plastic crystals with a lamellar structure. As
indicated above, they display textures which have so many analogies with those of
usual liquid smectics, that it explains why they have long been classified as liquid
crystals. In fact it would be wrong to go the opposite way and remove them from this
classification, inasmuch as the differences in texture between them and genuine liquid
crystals (in SmBs, transition bars on focal domains, mosaic texture (Schulze and Kunz
1976), germs and broken fan textures (Demus et a1 1984 etc); in SmEs arced focal
conic textures; see Gray and Goodby (1984) for more) involve subtle and interesting
mechanisms of epitaxial dislocations between layers (KlCman 1983a ch. 7), and of
disclinations which control the curvature of the layers (Nelson and Peliti 1987).
Moreover the classical method of determination of phase diagrams by the use of
miscibility criteria gives quite often ambiguous results; Goodby and Pindak (1981)
have observed miscibility between a SmB 3D plastic crystal, and what is called an
hexatic SmB (SmBhex),i.e. a phase which still retains some short-range positional order
in the layers, and probably also some short-range positional order between layers.
However, it is usual to consider hexatic phases as true liquid crystal phases.
Studies of defects in 'hexatic' phases (SmBhexrSmI, SmF) are not very numerous.
They comprise descriptions of the textures at the transition with other liquid crystal
phases (Goodby and Pindak 1981) and observations of thin films by depolarised laser
reflection microscopy (Dierker et a1 1986); there are also some plasticity studies (see
below in 0 4.3) which have been interpreted in terms of defects typical of hexatic
phases. Dierker's observations bear on tilted hexatic films and relate the existence of
five-armed star defects (which nucleate from point disclination defects at the SmC" +
SmI* transition) to a competition between bond orientational order and bending of
the tilted director in the layers. Also one can interpret these arms as loci where the
590 M Kle'man

excess dislocations of the 2D pattern (a disclination can be thought of as a cluster of


dislocations of the same sign, see 2) coalesce, in a sort of polygonisation process
which in fact is similar to that one which happens in the crystalline SmG phase, at
the transition. The temperature dependence of the bond-orientational elasticity
deduced from the textural patterns in the SmI" seems to agree with the 2D melting
theories of Halperin et a1 (1978).

4. Plasticity of smectic phases

Plasticity of smectic phases shows up a quite complex behaviour because of their


anisotropy and the different roles played by the various types of defects and textures.
We first discuss the much studied case of SmAs, which reveal a unique mixture of
hydrodynamical and metallurgical properties. SmBs, like anisotropic solids, act at not
too large applied mechanical stresses. Hexatic phases have an original intermediary
behaviour.

4.1. SmA phases


The mobility of dislocations in SmA phases can be analysed, as in solids, under the
categories of glide and climb, and the configurational forces which act on them can
also be referred to as Peach and Kohler forces. The situation is, however, more
complicated: the stresses have two origins, viscous and elastic; dislocations as well as
other defects like focal domains can be dragged along the flow and suffer as such
various kinds of instabilities more typical of liquids than of solids; defects nucleate
and multiply easily under shear flow, even in a perfect sample, etc. Concerning the
mobility of dislocations, the situations that have been best analysed are at low elastic
stresses and low shear flows; they concern individual defects but the results should be
useful to understand annealing processes. Understanding of rheological properties at
large stresses, large velocity gradients, and/or large frequencies necessitates the
inclusion of interactions between defects, their instabilities and multiplication proces-
ses. Interesting observations and related theories have been reported on the nucleation
and growth of focal conics in homeotropic and planar geometries. Very complex and
little understood behaviours have been obtained for very small displacements (a few
A) imposed on the layers (Durand 1983). More generally, any rheological experiment
performed on a sample whose defect content is not controlled leads to results which
cannot be analysed in simple terms (Kim et a1 1976) except possibly with the help of
standard theories for nowNewtonian flows (Porter and Johnson 1983, Duke and
Chapoy 1976). But the hydrodynamics of SmAs is fundamentally Newtonian, and the
non-linearities are caused by defects. This is the perspective in which we analyse below
the rheology of smectics.
4.1.1. Liquid viscosity and instabilities. Navier-Stokes equations in SmAs read:
do
p x = - V p + d iv u ' + g

where u' are anisotropic viscous stresses and g = -apf/au is the force per unit volume
exerted on the layers by their elastic response to any deformation field U of the layers;
in linear elasticity g reads
g = (0, 0, B a2U/dz2-K,A:u). (23)
Defects in liquid crystals 59 1

This force is relaxed by permeation (a diffusive motion of the molecules through the
layers (Helfrich 1969))
A,g = (0, 0, U - 0,) (24)
a process which is also much akin to the percolation of a fluid through a porous
medium, and obeys indeed Darcys law in the stationary and inviscid case
U - vZ = A, aplaz. (25)
The Newtonian viscosity p2 corresponding to the glide of the layers past each other
has been measured on perfect homeotropic samples (Horn and KlCman 1978, Oswald
1986b) at very low shear rates. At larger shear rates the samples do not remain
homeotropic and one always observes a progressive development of a regular array
of parabolic cofocal domains (see figure 18). They are related to the, unavoidable
undulation instability which always sets in, since the equidistance between the plates
cannot be maintained better than the penetration length A during shear. This array
develops in the direction of shear (Oswald and Ben-Abraham 1982). In Horns
experiments, the shear stress U necessary to produce a shear rate y during the
development of domains can be written as
u=A+vy (26)
where the viscosity v is low, of the order of 1 poise, and varies but little with the
proportion of regions filled with focal domains; but the solid friction A increases linearly
with this proportion, up to a value K,/12,where 1 is the size of the domains which fill
the sample. The existence of this yield stress implies that the domains, which are
probably anchored to the surfaces, impede the flow and are distorted under it.
Anomalous viscous behaviours and differences in the Couette profile have also been
reported in the regime before the nucleation of PFDS (Oswald 1986a), which is due to
the coupling of the flow with the forest of screw dislocations.

4.1.2. Climb of dislocations. The main mechanism of motion of edge dislocations is


by climb: i.e. the line moves in a layer, perpendicularly to its Burgers vector. This is
in contrast with the main mechanism of motion of dislocations in solids, where
conservative glide is dominant. Climb is non-conservative, i.e. requires the displace-
ment of matter by diffusion or flow process of molecules through the layers (per-
meation). In lyotropic systems, such a process of diffusion can take many forms (Chan
and Webb 1981b); there are also reasons to believe that in many instances climb is
speeded up by pipe diffusion of matter through the cores of screw dislocations (Oswald
1987).
Let us define the mobility of a dislocation line as m = u / u , where U is the velocity
and U the stress at the origin of the displacement. For an edge dislocation located
along the Oy direction, we have configurational forces
fx = - ~ : 3 b fy=o fi = Ul3b (27)
where U is the sum of elastic stresses U and hydrodynamic stresses u.Consider first
elastic stresses; in linear elasticity, near static equilibrium, we have
CT,=~ - K , a3u/ax3 ~ 3 =3 B d u / d z (28)
with the equilibrium condition
(+13,1 + u33,3 =0
592 M Kleman

where permeation has not been taken into account. This relationship implies that (+I3

is much smaller than (+33. We have indeed:


sx A
u,3 = u33-= u33-
sz SX

since any perturbation of length Sx in the layer propagates along Sz = ( S x 2 / A ) along


the z direction. We therefore expect that elastic stresses favour only climb, as topology
does: glide necessitates a periodic change in the core structure, with breaking and
re-forming of the layers, i.e. in most cases, except in low K , materials (A << do,KlCman
1988)), glide experiences a large Peierls friction (see below).
Viscous stresses of the type
(+I3 =/ + av,/ax)
~ 2 ( d ~ , / a z U33 =p 3 avz/az (29)
would play a role comparable to elastic stresses only for very large gradients, i.e. large
velocities or small wavelengths. This remark justifies a metallurgical approach of the
mobility of edge dislocations in climb (KlCman and Williams 1974), which leads to
the relationship
m = -Dlla
- = A p1/ l .
k,T 1
It is assumed in this model that the molecules jump from one layer to the next near
the tip of the supplementary layers, thought of as having an infinite density of jogs of
one molecule each. There is a constant probability of adding (removing) molecules
along this ragged tip if the driving force is constant and the line moves parallel to
itself. In lyotropics, this diffusion consists of a flip-flop from one monolayer to the
next or a shift from one bilayer to the next, followed, in each case, by a rapid diffusion
Dlialong the layer (Dll> D,; see Chan and Webb (198la)).
Equation (30) gives a mobility which is independent of the Burgers vector. Models
which consider that the permeation process takes place all along the layer, not only
at the tip, lead to a mobility proportional to the Burgers vector (Orsay Group on
Liquid Crystals 1975). The study of the flow of matter along the dislocation in a model
where the central layer of the dislocation is treated as a rigid boundary, shows that
only a thin parabolic zone (akin to a boundary layer) is affected by permeation in the
wake of the climbing dislocation
IZI (K-lX)

with
K- =(Vh,) 1/2
(31)
( K - ~is of molecular dimensions) and that the force per unit length necessary to produce
a velocity v is multiplied by a factor K b compared with a classical viscous force pv
F =Kbvv. (32)
(The same type of relationship has been obtained for the climb of dislocations in
Rayleigh-BCnard cells, see Dubois Violette et a1 (1983).) The real situation might well
be in between these two models.
A measurement of the mobility of edge dislocations in climb leads to a direct
measurement of the order of magnitude of A,. This is achieved in Bartolino and
Durands experiments (1977a, b), where a sudden strain is applied on a homeotropic
Defects in liquid crystals 593

sample; a lattice of parallel edge dislocations is created in the non-linear stage of the
resulting dilation instability, which relaxes with a relaxation time
r = ( d / m )Bpdo (33)
where p is the linear density of parallel dislocations, and do the Burgers' vector, which
experimentally is observed to be equal to one layer. A, can also be measured by the
'lubrication wedge' method (Oswald and K16man 1982), which is described in figure
26: when a liquid is sheared between two planar surfaces making a small angle a, it
is well known that a large pressure is set up, which is at the origin of the lubrication
process; by replacing the liquid by a homeotropic SmA, one expects an even more
spectacular effect since, apart from the easy flow of the layers past each other, there
is a genuine solid-iike resistance perpendicular to the layers. The lubrication set up
with shear y = v / d is also equivalent to a compression process with a vertical velocity
v, = cyv imposed on one of the plates. One observes then that the dislocations move
with constant velocity in the plane of the subgrain boundary where they sit, after a
transient time:
T= (d/cym)B
Typical values obtained this way are given in table 1. One will notice that the A,s are
all of the same order of magnitude but that the activation energies U differ widely.
Note, however, that in lecithin, the measurement of typical annealing times in wedge
specimens leads to A,- (Chan and Webb 1981a), much smaller than in the other
materials.
When a homeotropic sample is submitted to a constant load (compressive creep),
the edge dislocations climb under the u33stress, but also multiply, probably by a
mechanism where edge segments form Frank and Read mills (Friedel 1964) about
their screw parts (which are not affected by uS3stress). Also, as we shall see now,
screw lines pinned on the glass plates can suffer helical instabilities which generate
edge loops.
Helical instabilities of screw lines have been predicted by Bourdon er a1 (1981)
and seem to have been observed only in thermotropic smectics (Oswald and KlCman
1984). The mechanism is described in figure 27: the U,, stress is relaxed (above some
displacement threshold of the order of d o ) by the production of a helical turn which

Figure 26. The lubrication wedge method for smectics.

Table 1.

Compound Mobility (CGS) A, (cm' poise-') N (eV)

80CB 3 x lo-' at 25 "C 10-is 1.8


C,,EO, 6 x lo-' at 50 "C 4 x 10-1~ 0.4
594 M Kleman

/
/
--
I A
1 -

A
/
Figure 27. Helical instability of screw dislocation lines in smectics.

is equivalent to the formation of an edge dislocation loop; in figure 27 a layer is missing


inside the loop. Successive helical turns show up when the stress is increased; the
relaxation times are smaller when the number of helical turns increases. The experi-
ments seem to agree well with the theory.

4.1.3. Glide of dislocations. Screw dislocations move parallel to themselves under the
action of configurational forces f , = ba,, or fx = ba,, ,
A screw dislocation moving parallel to itself with velocity vo in the x direction
provokes a backflow very akin to the backflow set up by the motion of a cylinder of
radius 5 in a liquid; .there is however, in addition, an important 21, component which
brings matter from one layer to the next, by a helical motion which avoids permeation;
as a result, permeation is practically non-existent outside the core (Pleiner 1986b).
Hence
V, = ti - v0 auldx. (34)
There are therefore two contributions to the drag force which acts on the dislocation
line, a classical Stokes force proportional to p2v0,and a (larger) force due to the U,
components, which reads (Pleiner)

To produce such a large force by the action of mechanical stresses would require
a large ~ 2 3 ,which could exist only near a kink on an edge line. However, if the
dislocation is pinned, and the sample sheared, the velocity field which is set up can
be large enough to couple to the dislocation. This produces an anomalous flow
behaviour and an increase in apparent viscosity which has already been mentioned
(Oswald 1986a). It is not possible to invoke the non-conventional hydrodynamics of
Mazenko er a1 (1982) to explain these anomalies, since p 3 does not show up any l / w
divergence in Mazenkos theory.
Glide of edge dislocations involves a periodic variation of the molecular structure
of the core (Friedel 1975), and the dislocation will tend to lie longer along the most
stable of those structures. Glide involves therefore a finite Peierls stress cr,,b. This
stress should depend on the nature of the core of the dislocation. For elementary edge
dislocations in usual thermotropic smectics, where the core is not expected to be split
into disclinations, this stress should not be too large and the process of formation of
double kinks of screw character (figure 28) would be thermally activated, and limited
by permeation. For multiple edge dislocations in usual thermotropic smectics, and for
edge dislocations of any strength in lyotropics, where the core is split into disclinations,
glide is expected to be very difficult, or even impossible, since such a motion would
require the exchange of dislocations between the two disclinations.. The case of
lyotropics has, however, to be considered apart when K , is very small, i.e. the penetra-
tion length h = ( K , / B ) 2 small compared with the layer repeat distance d o , as is the
Defects in liquid crystals 595

Figure 28. A double kink of screw character on an edge dislocation.

case in non-ionic polyoxyethylene surfactants C,EO,, with m = 12 and n = 5 or 6 (Allain


et a1 1985, Allain and KlCman 1987). As stated in 0 3.2.1, the core is split along a
plane normal to the layer on a length
52=dilA (36)
and the core layers are stretched under what could be described as an efectiue surface
stress
uc= h K , / A (37)
which facilitates the formation of pores, by reducing their nucleation energy by a
quantity ucA, where A - Ird: is the surface of a pore (KlCman 1988). This is not a
negligible quantity, since it is of the order of 0.3 eV for K,do= k,T and A = do/lO.
The double kinks we alluded to above necessitate the nucleation of such pores, which
transform easily to the cores of the screw dislocation segments.
Glide of dislocations has not been observed directly except in Williams and
KlCmans (1975) continuous low shear experiments performed on planar samples. The
layers tilt during shear and break near the surface at some critical angle e,., for which
curvature contributions and layer compressibility contributions compensate in the bent
boundary layer of thickness 1 = A, according to the analysis of Marignan and Parodi
(1979, 1983), which yields BC = 20 . The breaking of the layers is relaxed by an array
of edge dislocations near the surface, with average distance A = b / ( 1 -cos OC); but b
is very large (-50 to 100 do) and clearly the elementary dislocations gather together
by glide. And as the shear proceeds, the dislocation wall is progressively replaced by
a Grandjean wall of focal conics, which can also directly result from an instability
in alternative shear (see below).

4.1.4. Focal conics. The motion of a disclination requires the emission or the absorption
of dislocations (Friedel and KlCman 1970). This is probably true also for its nucleation
and growth. This makes the rheology of disclinations of the focal conics quite difficult
and in fact largely unknown. However, the nucleation of focal conics, resulting from
instabilities of the flow, and their motion, have been carefully observed. We refer (i)
to the numerous observations of the appearance of parabolic focal domains in homeo-
tropics under dilation and shear (Asher and Pershan 1979, Oswald et al 1982); (ii) to
the observations of permeative flow around an obstacle made by Clark (1978) and (iii)
in planar geometries, to the experiments of Marignan et a1 (1978), Marignan and
596 M Kleman

Parodi (1979, 1983) in alternative shear; focal conics are periodically produced in the
volume of the sample, with their ellipses parallel to the surface. When the amplitude
of the shear is increased, these ellipses move surprisingly easily through the sample
and reach the surface, where they form geometries similar to those observed by Williams.
The motion of the focal conics is worth studying in more detail, inasmuch as it can
be observed directly. The threshold of appearance of the domains has been studied
theoretically (Marignan et a1 1983), and the results seem to be corroborated by
experiments; they can be described as a function of two quantities, N and M :
N = BOg/wT = 1 Og/wr+ (38)
where r,,, is the relaxation time of the order parameter and Os the critical inclination
M = TH (39)
where T~ is the hydrodynamical relaxation time ( rH= L2p/D, where L is the thickness
of the sample and p the density), w is the shear frequency. There are two regimes.
(i) When M < lo-, i.e. at low frequencies of shear Hz to lo2 Hz) Os is of
the order of a few degrees and varies like w l * , i.e. N is constant. The ellipses do not
change eccentricity at lo-* Hz when they move under small-amplitude shears, which
indicates a large amount of fairly easy permeation, while the layers bend elastically.
This is probably possible because the layers are firmly anchored at their extremities
on the glass plates. At frequencies of the order of 10*Hz, the geometry seems to
involve more plastic relaxation of the focal conics, which change shape when moving.
(ii) When M is large, N = M-4; but this regime has not been studied.
The study of the threshold at low frequency enabled Marignan to measure T+ as
a function of temperature, and obtain this way the critical behaviour of B in ~ O C B .
This yields B -c. T23as expected from de Gennes theory.

4.2. SmBphases
The crystalline nature of the SmB phase consists in a compact hexagonal structure
with a stacking of the layers of the ABAB type (Levelut er a1 1974,1977). The anisotropy
is very large ( a / c = 6), which explains the existence of a very small energy of coupling
between layers, and the presence of numerous small-energy stacking faults. The
dislocations of the basal plane, with Burgers vector along a, are therefore easily
dissociated into Shockley partials. The anisotropy reflects also in the elastic coefficients;
C44=lo6 erg cmP3,while the other elastic coefficients are lo3 to lo4 times bigger. At
very low applied deformations ( E < Cagnon and Durand 1980, 1981) under a
constant shear rate i. in the layers the SmB phase of 40.8 (butoxyl benzylidene
octylaniline) has a Newtonian response with an effective viscosity of the order of lo7
poise, i.e. reacts as a plastic solid. The corresponding shear stress which develops is
very small ( U < 200 dyn ~ m - ~ this
) ; suggests a regime of microplasticity. In a creep
experiment under large applied shear stresses ( U > 2000 dyn ~ m - the ~ )creep rate varies
in an exponential way, and suggests according to Oswald (1985) a model of thermally
activated glide of the basal dislocations. In this model the activated process is due to
the crossing of the forest of screw dislocations perpendicular to the layer ( b s = c). The
Peach and Kohler force drags indeed the basal dislocations whose Burgers vector b,
*
has a component f b along the shear, with a force F = ub along the shear direction;
the activation energy is the energy of the jogs formed at the crossing ( U = 0.6 eV,
including the energy of constriction of the basal dislocation lines). Comparison of
Defects in liquid crystals 597

experiments with theory leads to a density p d of active basal dislocations of the order
of 10 cm-, a dissociation width of 40 A and a stacking fault energy of lo- erg cm-.
Microplasticity experiments are explained in this frame by assuming that the basal
dislocations move so little that they have many crossings back. The density of disloca-
tions is very high and can probably be explained by the curvatures of the layers, which
introduce matching dislocations of the type described long ago by Nye (1953) for
curved crystals (see figure 8(c)). Nyes relationship between p and R, namely p = 1/ Rb
leads to R = 1 cm, which indicates that very small curvatures, which are not unthinkable
in such an anisotropic liquid-like material, can be indeed at their origin. There is also
a possibility that (largest) spontaneous curvatures are relaxed by vacancies and inter-
stitials, which would give rise to a Nabarro creep at much smaller characteristic lengths.

4.3. Hexatic smectics


Shear viscosity parallel to the layers in the hexatic B phase of GOBC and in the hexatic
F phase of 90.4 have been measured by Oswald (1986c), who has found that it varies
like .$i ( n = 2.5), where is the coherence length of the hexatic phase, of the order
of a mean distance between the free dislocations of the hexatic phase (Halperin and
Nelson 1978). He explairis this experimental result by a model in which shear is
relaxed by the motion of dislocations, assisted by a 3D atomic diffusion; this leads to
n =: in the framework of a Nabarro-Eyring theory (Nabarro 1948).

5. Defects in columnar phases

While numerous chemical syntheses of new discotic molecules have been performed
in recent years, the study of the physical properties of the mesomorphic thermotropic
phases associated with these molecules is as yet largely undeveloped. The book edited
by Helfrich and Heppke in 1980 still provides a relevant bibliographical introduction
to the chemical and structural properties of discotic molecules (see Billard (1980), also
Luz et a1 (1985), who puts emphasis on NMR methods of investigation). This slow
start is explained by the difficulties met in these syntheses, and the fact that obtaining
well aligned samples is often a question of luck. Some discotic materials, which have
an easily alignable high-temperature nematic phase, should give good single crystals
of columnar phases also at low temperature, oriented at will, but these materials are
not easily available. Freely suspended discotic strands have been obtained from
triphenylene hexadodecanoate (van Winkle and Clark 1982), where the columns are
along the strand, allowing for precise x-ray study. Finally, mesomorphic lyotropic
hexagonal phases have also been little studied, but are of fundamental interest, since
they show up phase transitions (hex e= isotropic or nematic micellar; hex e= lamellar)
which necessitate changes in the topology of the layers through mechanisms involving
defects (Gruner et al 1985).
While we know up to now only hexagonal lyotropic columnar phases, thermotropic
phases differ from each other by the nature of the columnar stacking, the tilt of the
molecular discs with respect to the column axis (Frank and Chandrasekhar 1980), the
rotation of the molecules and finally the density modulation along the columns (cf
de Gennes 1983, Godrhche et a1 1985). But we shall limit ourselves to a description
of the defects in the case of flexible, liquid columns, with untilted molecules; the
conclusions apply, largely unchanged, to hexagonal lyotropic phases.
598 M Klkman

We will discuss in order: (i) the developable domains, whose deformation field
implies only curvature elasticity and which are the analogue for these 1 D liquid phases
of the 2D liquid smectic A phases focal domains; (ii) defects which imply positional
elasticity (dislocations, walls). The most conspicuous defects in the polarising micro-
scope are straight disclinations (they are special developable domains), grain boun-
daries probably analysable in dislocations, walls analysable in dislocations and small
developable domains (figure 29).

5.1. Developable domains and disclinations


5.1.1. General notions on developable domains. We assume in this section that columns
can glide along each other, in such a way that there is no elastic distortion in any
surface II perpendicular to the columns. As stated first by Saupe (1977) this implies
that these surfaces are planar (figure 30). The columns are along curves C which are
parallel. Consider therefore any planar line 8 in a plane II, passing through a subset
of intersections of the columns C with the plane n; the surface S which contains the
entire columns of the subset (this is called a surface of translation, a concept first
developed by Monge (1807)) is generated by a planar curve invariable in shape, equal
to 8, and moving in such a way that the velocities of all its points are normal to the
plane which contains it. It is possible to partition any plane II with parallel 8 curves,
and hence to partition space with parallel Monge surfaces, hence building a layered
structure. This property is reminiscent of smectic phases, and one is led to think that,
by a suitable choice of 2, geometries analogous to focal domains would show up in

Figure 29. Typical textures in a discotic hexagonal phase. Polarising microscope observa-
tion, courtesy P Oswald. ( a ) Planar texture; ( b )homeotropic texture, with grain boundaries,
and straight disclinations evidenced by their shadows.
Defects in liquid crystals 599

a columnar phase with only curvature deformations, inasmuch as planar textures (figure
29( a ) ) are very similar to the fan textures of smectics. In fact Monge surfaces are not
Dupin cyclides, except in the special case of the sphere and of the cylinder. Therefore
the analogy is mistaken (but is it not possible, a contrario, that fan textures could
involve Monge surfaces?). The simplest singularities involving only curvature energy
in a columnar phase are developable surfaces, and not focal conics, as we show in what
follows.
The columns C form parallel lines on a Monge surface; their normal trajectories
X are therefore geodesics. One can show that X and C curves are also lines of curvature
of the Monge surfaces. Two Monge surfaces are either parallel or intersect along a
C, which is a line of curvature on both of them; according to Joachimstahls theorem
(cf Darboux 1954), they therefore cut with a constant angle along C ; this is precisely
the condition of non-angular distortion we wished to satisfy in planes II perpendicular
to the Cs. Since the columns are also parallel, there are no local elastic distortions in
the planes II at all.
Figure 30 represents two infinitesimally close planes II; D is their line of intersection,
which envelops a space curve L when II moves. The infinitesimal motion of II can
be divided into a motion of pure rotation about 0, namely dq, and a motion about a
normal v to II passing through the point of contact of D with L, namely d$; II is the
osculating plane to L, and v is the binormal.

Figure 30. Local arrangement of molecular columns C in a developable domain.

Consider any column C : its centre of curvature is on D, and its osculating plane
perpendicular to D. The curvature of the columns reads:
l/R=ltxcurl tl=dq/ds (40)
where t is a unit vector (director) along the column, and s the curvilinear abscissa on
C. Similarly, the curvature l l p of L in M (see figure 30) reads:
l / p = d$/du (41)
where u is the curvilinear abscissa on L. Properties of reciprocity exist between L and
C, namely:
p/T= -T/R (42)
where T and T are the torsions of C and L, oriented by the motion of II. The curvature
1/R of C becomes infinite on D, whose locus is consequently the singularity of the
600 M Kleman

set of parallel Cs. D describes a developable surface (figure 31) which limits the
domain of existence of columns. We call it a developable domain. L is the cuspidal
edge of the developable surface. For a complete theory, see KlCman (1980b).
Numerous examples of developable domains can be found in Bouligand (1980).
The curvature energy of a columnar phase is the same as the curvature energy of
a nematic, with the director t defined as above. In a developable domain we have:
div t = 0 t.curlt=O (43)
since there is conservation of the column flux and since the columns are parallel. The
only contribution comes therefore from bend and can be calculated as some integral
on L. We have:

where W, is a surface energy (of the developable surface) and where the y axis is
along the straight line D ( n ) in each plane n. Call 1, and 1, some lengths typical of
the domain; we have:

where b is some molecular length (a cut-off length near the developable surface);
=I
R I , Idc/pI, the average torsion of L, is a quantity which measures the strength of
the developable domain and depends on the torsion of L. If L is a planar curve,
1/ r = 0 and W = 0: this is indeed what we expect, because if L is planar the associated
developable surface is the plane containing L, and the columnar phase is perfect.
A remarkable fact is that the bulk term of W does not depend on the curvature of L ;
5
R, = Idalp1 is, however, an interesting quantity, which measures some intrinsic twist
of the columnar crystal about t along L.
Developable domains are classified in Klkman (1980).

5.1.2. Simple types of developable domains. We discuss disclinations and twisted points.
(i) Disclinations. If the developable surface is a circular cylinder K, we get the
simplest type of defect, the S = +1 wedge disclination (figure 3 2 ) . The intersections
of a plane containing a circular section of the cylinder with the half plane n are the
half tangents to K. The columns C are the evolutes of K. Assuming now that the

Figure 31. A developable domain: global geometry. L, the cuspidal edge, is the envelope
of the straight lines D which generate the ruled surface which limits the developable domain.
Defects in liquid crystals 601

Figure 32. The S = + 1 disclination in a columnar mesophase; the columns are along the
evolutes of the cylinder K.

developable domain is in a hexagonal columnar crystal, we conclude that the axis of


the cylinder, which must be a direction of symmetry of the 2D stacking, is along a
direction L1 or L2 (figure 33); in fact, the two types of disclinations have been observed
in the C,-hexaalcoxy derivative of triphenylene (Oswald 1981). It appears also that
the S = f lines are far more numerous than S = 1 lines, but that their geometry is that
of a half-developable domain, with cylinder radius a of the order of 0.5 pm, core
radius r, slightly larger. This extremely large core, quite unexpected from optical
observations, can be understood either as a hollow core filled with a perfect columnar
phase aligned along the disclinations (Oswald and KlCman 1981) or as a liquid core,
thus implying a large coherence length due to a very large, unusual, K , (lo6 times
larger than usual; Cagnon et al (1984)). The origin of such a large K3 is not yet
understood, and there are no observations at a smaller than optical size.
Disclinations have been observed in planar and quasihomeotropic textures; in free
droplets and planar textures, they are perpendicular to the glass plate, and the free
surface shows up strong variations, with peaks at the disclination locations (Oswald
and KlCman 1981).
(ii) Twisted points. This is a very unexpected geometry whrch happens when the
developable surface is a circular cone (KlCman 1980a, b, c). The spheres centred on
the apex of the cone are Monge surfaces; the columns inscribe on them helicoidal
paths, either left or right handed (two types of twisted points). We therefore have

ILz

Figure 33. The two possible positions L, and L, of wedge S = f disclinations in a hexagonal
columnar mesophase.
602 M Kleman

-
here a geometry whose twist has a definite sign, while the twist density t curl t is
vanishing by virtue of equation (43). Twisted points have not yet been observed.

5.2. Dislocations, walls and grain boundaries


Dislocations introduce elastic distortions of the 2D lattice; in the hexagonal case, the
elastic energy density reads (Kats 1978, KlCman and Oswald 1982):

pf = -
e
1 -+-
au au
B(dx a
+c-
~ ) ~ [:(:; I;:::
-+-
2)::
---

where U and U are the x and y components of the elastic displacement in the plane
perpendicular to the columns (along z ) and B and C two positive stiffness coefficients.
This equation assumes that the columns are liquid-like and relax to a constant density
in a finite time; however, in some geometries (when the stress a,, differs from zero)
this relaxation requires exchange of molecules between columns, a process which is
not elastic and is not supposed to take place quickly: in such case, a more general
expression of pfel is required, with a displacement w explicitly introduced.
For small distortions, the curvature energy density reads:
(-+ -)
pfc = f K l a2u a2u + f K,( - +fK,[
a2U - A) +(
(e) $)2].
(47)
axaz ayaz ayaz axaz ax2
We discuss below dislocations and walls of small disorientation in the frame of
this second-order elasticity for small distortions; this is of course not a good approxima-
tion for large curvatures (as they appear in developable domains); but even small
distortions might require in some cases a more elaborate free energy density, including
third-order elasticity (second-order terms display cylindrical symmetry; hexagonal
symmetry appears only in third-order terms). For example, the shape of the free
droplets which grow around disclinations (Oswald and KlCman 1981) or the growth
of instabilities above their threshold (dilation instability, buckling instability, KlCman
and Oswald (1982)) are probably governed by those terms (Frank 1982); the onset of
buckling instability should also be very sensitive to the plane of bending of the columns
forced by the experimental set-up.

5.2.1. Dislocations. The Burgers vector is necessarily perpendicular to the columns.


We distinguish three types of elementary dislocations on geometrical grounds. Corre-
sponding observations are unfortunately still lacking.
(i) Longitudinal edge dislocations, whose axis is along the columns; their distortions
are confined to the x, y plane and do not differ from those of an ordinary edge
dislocation in a solid. Therefore their energy scales like Eb2 In R / r , + w c , where w, is
a core energy. The core should have a molecular size.
(ii) Transverse edge dislocation, whose axis is along a direction of the x, y plane.
Such dislocations necessarily involve elastic and curvature distortions. Taking the line
along the y direction, we have to solve the equation (cf figure 34):
a4v a4v a4v
El,= K1-
ay ay2 az2+ K 3 G 4

which is akin to the equation for an edge dislocation in a smectic and can be discussed
similarly. At a distance from the core larger than A , = ( K , / B ) , the K , term is
Defects in liquid crystals 603

Figure 34. Transverse edge dislocation in a columnar mesophase.

negligible, and we recover exactly the smectic equation, governed here by the charac-
teristic penetration length h3= ( K3/B)/; if K3 is very large (as claimed by Cagnon
et a1 (1984)), the elastic distortions extend in a large parabola (of width h3) whose
axis is along the y direction. At distances from the core smaller than A I ,the displace-
ment field reads approximately:

b b
u ( y , z ) =-+-tan- zhl/xh3. (49)
4 2

This core structure is quite novel; one expects a large core energy, since A l is large,
possibly relaxed by a Peierls-type splitting, which would introduce the columns free
ends. Direct microscopic observations are clearly needed to go further.
(iii) Screw dislocation. Taking the line along the x axis, the only relevant displace-
ment is U , which obeys the equation:

c d2U
- ?= K 2 - d4U 2 + d4 U
K3-z.
2 ay ay2 az dz

The calculation goes as above, with two characteristic lengths A;= ( 2 K 2 / C ) 2and
A: = (2K3/C)l2. In the region in the vicinity of the line (in a range of order Ah), there
is essentially twist deformation; the core region has the remarkable feature that there
is no necessity to introduce any physical singularity on the columns (figure 35). A jog
of molecular size on a screw dislocation introduces one free end; this should be in
practice a way of relaxing stresses which require a large divergence term div i.

it
Figure 35. Screw dislocation in a columnar mesophase.
604 M Kleman

5.2.2. Walls. Let us first distinguish two broad classes of walls (Bouligand 1980). Walls
of the first class are singularity walls; walls of the second class are planes II common
to two developable domains (figure 36). Together they form characteristic fan textures
which have been observed in lyotropic hexagonal phases, where they are due to
instabilities following desiccation (Rogers and Winsor 1969, Saupe 1977) and in
thermotropic phases, where they are also probably due to some elastic dilation instabil-
ity. They were first explained by Merucci (1976).
Walls of the first type divide in: curvature walls (cw), walls with developable
domains ( DDW), and discontinuity walls (DW), in ascending order of disorientation
angle. They can be symmetric or asymmetric, in which latter case they probably contain
dislocations. Low-angle grain boundaries are probably always split into dislocations,
as well as grain boundaries (always observed in homeotropic specimens). We will
discuss briefly the curvature walls (for details see Oswald and KlCman (1981)).
We assume as in figure 36 that the columns are planar and call 0 the angle of the
director t with the z axis perpendicular to the wall. The free energy density reads:

pj= I(K , sin20 + K3 cos 0)(d@)* +;l3 (1-- .


cos 0 0

Minimising this expression and restricting ourselves to small angles 0 yields the first
integral :

($)2[ K, + ( K1- K 3 ) 0 2=] $(@: -

where we can safely assume that ( K ,- K3)0*<<


K 3 , so that the solution reads:
0 = 0,tanh Oox/2A1. (53)
We therefore expect a wall width of the order of A I /ao,and an energy per unit length
along x:

W = $0:(
K3E). (54)
The O 3 behaviour indicates that such walls would be stable only for small values
of 0,;they are indeed observed for 06 15 O in thermotropic and lyotropic specimens.

Wall of the second t y p E /

\ cylinders

Figure 36. Walls of various types in a columnar mesophase.


Defects in liquid crystals 605

For these cw, as well as for DW and DDW, the inner structure of the wall deserves
more detailed microscopic studies.

6. Defects in nematics and cholesterics

The subject of disclinations in nematics has been comprehensively reviewed from


fundamentals by Chandrasekhar and Ranganath (1986), we therefore make the
introductory section rather short. Then, after an overview of new results on line and
point defects in conventional nematics, we discuss defects in chiral nematics
(cholesterics), in the recently discovered biaxial nematics and in polymeric nematics.
A number of promising topics require a knowledge of the topological theory of defects
(see S7).

6.1. Disclinations and singular points: a summary


The Volterra classification of disclinations in uniaxial nematics is well known (Frank
1958, KlCman 1980a), but ignores their topological properties, which have been worked
out in the early 1970s, before the homotopy theory of defects. We summarise these
new features without referring to homotopy, which yields straightforwardly the
existence of only two kinds of lines (thick and thin, Nehring (1973)) and not an infinity
(as the Volterra process would suggest) and the existence of singular points.
*
(i) Lines of integral strength ( S = 1, *2) do not require a singular core, as is well
known. The corresponding experiment was achieved by Williams et a1 (1972) and
revealed at the same time the presence of singular points, since the director can escape
in the third dimension (Meyer 1973) either up or down the cylinder axis. An argument
given by Bouligand (1972b) goes deeply into the topology of this phenomenon: consider
a loop surrounding a disclination in a nematic and enlarge this loop into a ribbon of
constant width by materialising at each point of the loop the director which is met at
that point (the loop must be chosen in such a way that it is nowhere tangent to the
director): if the surrounded disclination is of half-integral strength, the ribbon twists
along the loop an odd number of times and is therefore a Mobius ribbon with one
side, and cannot be reduced to a single direction; on the contrary, if the disclination
is of integral strength, the ribbon is an ordinary ribbon with two sides, and can be
reduced to a single direction. It is this process of reduction which is the aforementioned
escape in the third dimension. There are therefore two kinds of lines: thin lines which
are necessarily singular, and thick lines which can be continuous.
(ii) Since lines of strength S = n ( n is an integer) can be made continuous, any
line of large half-integral strength might be reduced continuously to a line of strength
( S ( = * f . Since S = f and S = - f differ by an integer, they must be equivalent. This
has been illustrated pictorially by Bouligand (1981), who shows how one could go
continuously from a wedge S = f line to a wedge of S = -f line, the mid-situation being
a twist line. Similarly all integral lines are equivalent. Since the lines of smaller S are
presumably energetically favoured, this explains why one sees essentially lines of
strength /SI = 1 and (SI= f. The classification of lines by the set of all values S = p / 2
( p is an integer) makes sense only for wedge lines.
(iii) Singular points differ by their strengths N, which can be defined as follows.
Surround the disclination point by a manifold homeomorphic to a sphere entirely
606 M Kle'man

situated in a region with no singularity of R in the so-called 'good crystal'; the integral

taken on this manifold is an integer. 4 v N represents the flux through a sphere of unit
radius centred at the origin of the vector n at any point of this closed surface, transported
to the origin; therefore N is the algebraic number of times this unit sphere is covered
in the considered mapping. Since the integral is a third power of n, the sign of N
depends on the (arbitrary) sign given to the director. Now assume that the singular
point makes a complete turn around a half-integral line: n changes sign after completion
of the loop, and N also. Therefore the singular point which has changed sign would
now annihilate with another singular point of strength N. This mutual collapse would
not have occurred before completion of the turn.
Singular points were first considered in ferromagnets by Feldtkeller (1964, 1965),
who introduced the above mapping on a unit sphere; soon afterwards Doring (1968)
calculated their energy. These are now of great importance in the physics of domains
and walls (Malozemoff and Slonczewski 1979). Equation ( 5 5 ) was first given by KlCman
(1970) for nematics and rediscovered by Blaha (1976) in his study of vortons (which
are singular points at the terminations of singular lines) in 3He-A.
Singular points can also be classified as vector field singularities (PoincarC sin-
gularities) as shown by Nabarro (1972). They have been observed and studied by
polarising microscopy in nematic-filled capillaries, first by Williams et a1 (1972, 1973)
and more recently by Melzer et a1 (1977a, b).

6.2. Recent results on disclinations


The zigzag instability of S = f lines observed in round capillaries by Cladis (1974) has
been studied by Mihailovic et a1 (1988) and Galerne et a1 (1988), in uniaxial ther-
motropic and lyotropic nematics. This instability shows up when K 1 is large compared
with K 2 , K 3 . Therefore the non-singular S = 1 line of the round capillary is not
4
favoured, and splits into two S = lines, through an activated process where singular
points appear along the line.
The core of disclinations has been studied very little experimentally. Various
theoretical studies (Lyuksutov 1978, Schopohl and Sluckin 1988) predict that the core
should be biaxial, with an order parameter varying smoothly from outside the core to
the central line.
The role of the anisotropy of the Frank coefficients is at the present time a quite
important question, in view of the study of defects in nematic polymers and in biaxial
nematics. The 2D case splay-bend anisotropy K , # K 3 has been considered by
Dzyaloshinskii (1970) and more recently by Ranganath (1980, 1982, 1983b) and
Nitayanda et a1 (1980). Ranganath has also demonstrated the anisotropy of the
interaction between disclinations and shown that the stability of wedge versus twist
disclinations is increased by the anisotropy. But few other results are known.
It is worth mentioning that the nature of the equations which govern nematic
distortions, with fixed boundary conditions for the director, has attracted a lot of
attention from the mathematicians (see the recent review by Ericksen (1987) on the
subject). Very sophisticated methods have enabled BrCzis (1989) to calculate exactly
the energy of a set of point defects in isotropic elasticity, with specific boundary
conditions.
Defects in liquid crystals 607

6.3. Disclinations: dynamics


This field is far from being well known: the hydrodynamics of nematics, even in the
absence of defects, is extremely complex since it involves rotation of the molecules
and movement of their centres of gravity simultaneously (de Gennes 1974b, 1976).
This hydrodynamics has been studied only for uniformly oriented specimens (homeo-
tropic, planar) and very simple cases of Couette or Poiseuille flows. Deformation
modes and instabilities are therefore well known (e.g. Pieranski and Guyon 1974,
Guyon and PiCranski 1975, Wahl and Fischer 1973, Wabl 1979, Hilltrop and Fischer
1976). There are also beautiful dynamic properties under thermal gradients and
magnetic and electric fields. Defects nucleate in all these geometries for fields not far
above the thresholds, or at the transition between the various ordered textures they
display (umbilics, walls, defects of the ordered textures, etc). We shall not study them
(see Chandrasekhar and Ranganath 1986) and will rather discuss (i) the motion of an
isolated disclination line in a nematic, under low applied forces, (ii) the effect of large
shears, under which disclination lines nucleate in number. Very few theoretical studies
have been devoted to simple flows in the presence of defects (see, however, Sarkar
(1983) who claims that the Ericksen scaling laws are not obeyed in Poiseuille flow if
there are disclinations transverse to the flow). We discuss the dynamics of polymer
nematics in 0 6.7.

6.3.1. Mobility of an isolated disclination. A local analysis of the mobility can be made
in terms of the Peach-Koehler force (KlCman 1983b). Eshelby has argued that the
force on a disclination in a nematic LC is, unlike its namesake for a dislocation in a
solid, not a fictitious configurational force, but a real force. Formally, this is a
consequence of the near identity of the Ericksen stress and the appropriate energy-
momentum tensor.
A global analysis of the mobility is easy if we assume that the motion of the line
does not involve any backflow, but only a rotation of the director. The theory has
been worked out by Imura and Okano (1973) and de Gennes (1976), along very similar
lines. In the simple model of Imura et al, let

be the entropy production due to a disclination moving with constant velocity uo along
the x direction, y = a 3 - a 2 being a viscosity coefficient and cp some angle which
describes the orientation of the director about the line. By definition of the friction
coefficient J; we write U =fui.
Since there is no backflow, and assuming isotropic Frank elasticity, cp obeys the
equation

Then, for distances r from the disclination larger than r, = K / yuo, the second term in
the left-hand side of equation ( 5 6 ) is negligible, and the orientation of the director
about the moving line is the same as about the line at rest, namely
cp = cpo+ s tan-' [ y / ( x- u o t ) ] . (57)
608 M Kleman

This yields, by an easy calculation

(+
= 2 L
= -yuo In -
2 10

f = yW/2K (59)
where W = r K S 2 In L/ro is the energy per unit length of the line, L the size of the
sample and ro a core radius which has no reason to be equal to the core radius at
rest and which is necessarily.equa1 to or larger than rl . This theory assumes dissipation
for r > ro.
Geurst et a1 (1975) have studied experimentally the velocity U of a straight twist
Jisclination under the action of a magnetic field. It is easy to show by scaling argumellts
that uo depends on the magnetic field as follows

where xa is the diamagnetic susceptibility, d the thickness of the sample and CY a


geometrical coefficient which also depends on some core radius, and can be estimated
by equating the entropyproduction (as in equation (58)) to the gain of energy per
unit time due to the motion of the line under the magnetic field ( uo= xa H 2 L ) . Again,
it is assumed that there is no dissipation in the core. Comparison with experiment
yields ro = 0.04L i.e. a core region of gigantic dimension and depending on the sample
thickness. This result is corroborated by similar experiments in a cholesteric Can0
wedge (Malet et a1 1976). Further estimations which take into account other dissipation
processes (backflow, isothermal relaxation of the order parameter in the core, etc) lead
to even larger cores (Parodi et a1 1982). The situation is very confused, since there
are no experiments at the present time that enable us to observe directly the molecular
distribution in the core.
In fact, this experimental situation displays a number of characteristic lengths
(i) the analysis done above would be valid only at large lengths (IClCman 1984),
where backflow is negligible, i.e. for r > r2; r z - u o y / x a H 2is the coherence length of
the magnetic field; for r < r2 backflow is due to the non-uniformity of the magnetic
torque (Parodi et a1 1982). But above r2 the orientation of the director depends strongly
on the magnetic field, and the line energy which enters in equation (59) is not the
energy at rest.
(ii) r, = v/uo measures the length on which any flow impulsion or flow vorticity
created by the line motion vanishes. v is the kinematic viscosity for fluid motion (of
the same order of magnitude as y l p , the kinematic viscosity for molecular rotation).
We can visualise this production of vorticity as due to the change of configuration in
the core itself related to the emission of a density of dislocations b of rate
db -
_ - uo.
dt
A disclination in a nematic is indeed a sum of dislocations of infinitesimal strength
(0 2, Friedel and KlCman 1969, KlCman 1980b). The region of size r3 extends in the
wake of the disclination, if our image of emitted dislocations is true, like a vortex sheet
of effective thickness = (vr/ uo)12,where 0 < r < r, . We therefore expect that the core
of a moving disclination is very different from the core of a static one; furthermore,
as suggested by Friedel (1979), the core might show a longitudinal instability. An
Defects in liquid crystals 609

intuitive model for the dynamic splitting of the core has been suggested by Kltman
(1984; see figure 16). Note that r 3 / r ,= v y / K lo4; hence the splitting affects a quite
2-

large region, in agreement with the experiments on the core radius.


(iii) The above remarks relate to the stationary regime; we also have to consider
characteristic times. At the onset of motion, there is first a reorientation of the director,
which relaxes with a characteristic time ~ ~( y / K ) L 2 ;then flow relaxation takes place
2 -

in a time T ~ =yL2. The ratio r f / r ,goes like r 3 / r , . In the presence of a magnetic field,
it is the largest of the two times, Tf or r H =y / x a H 2 which dictates the relaxation
behaviour. Generally, T H >> T ~ .
(iv) The study of the dynamics of line defects in nematics has recently been extended
to the effect of electric fields (Cladis et a1 1987).

6.3.2. Nucleation and relaxation of disclinations in large shear $ow. Graziano and
Mackley (1984b) have experimentally put into evidence several regimes in a homeo-
tropic sample of MBBA under shear. At very low shear rate ( y < 1 s-l) the molecules
align along the velocity flow, in the manner described by Hilltrop and Fischer (1976),
up to saturation of their orientation. At slightly higher shear rate, disclinations nucleate
(see also Cladis and Torza 1976) in the form of closed loops of thicks and thins. Thins
nucleate in the boundary layers. These loops elongate in the flow and tumble, due to
the gradient of the velocity field. At y 10 s-' the entire field of view is covered with
2-

a complete mesh-like network of interacting lines. After cessation of the shear, they
relax by collapsing to points (this is the case for thins and nodeless thicks) or by
breaking at the nodes and shortening (thicks). An interesting observation is that the
contrast of the thins broadens when the shear stops; this observation is consistent with
the idea that the dynamic core is different from the static one, but now the static core
is larger, probably because the lines considered by Graziano et a1 are along the flow,
and not transverse as in 6.3.1.

6.4. Chiral nematics (cholesterics)


There is very little difference in structure between a twisted conventional nematic and
a cholesteric, at a scale smaller than the pitch p . No wonder therefore that the optical
observations of Bouligand (1974a) in very large pitch cholesterics show up, quite
noticeably, thicks and thins of the same nature as in a nematic. However, the situation
is considerably more involved, and quite original, when the defects scale regions larger
than p . We summarise their main aspects which have already been reviewed in some
detail in Kltman (1983a). Most of the results which follow date back to the early
1970s. In fact research on cholesterics in the last ten years has been less active, since
blue phases have been researched at the expense of chiral phenomena. This apparent
neglect does not invalidate the importance of the concepts acquired in the study of
cholesterics, which open up a wealth of new problems and ideas in the theory of defects.
There are three physical ingredients that take a part in the building of defects and
textures in real cholesterics: the layered structure, the local tendency to double twist
and the tendency towards continuous cores. We will develop this latter factor below.

6.4.1. ClassiJication of defects and general properties. As in a SmC, there is a local


trihedron which here rotates with the pitch along the cholesteric axis x. Call A a local
direction defined by the molecule, with A . x = 0; call T = A x x the third axis. Each
of those three axes is twofold, and carries a director. We shall note, as in Friedel and
610 M Kleman

KlCman (1970), A(S), T ( S ) ,x ( S ) , the corresponding disclinations of strength S. The


escape of the singularity along the core is here a more difficult problem than in a
nematic, since it not a director which has to escape, but a trihedron. It turns out that
S must now be an even number S = 2 n , not simply an integer. The demonstration
requires the use of the topologial theory of defects, but it is in fact possible to convince
oneself of this phenomenon by outlining a drawing of the disclinations S = 1 and S = 2
(see Anderson and Toulouse (1977) for a similar problem in 3He-A). Also the
phenomenon is akin to the orientation entanglement relation of a cube mobile with
respect to a fixed frame, discussed in Misner et a1 (1973).
In fact, A, ,y and T have quite different physical meanings. If the cholesteric is
locally uniaxial or weakly biaxial, ,y and 7 are immaterial directors, and it suffices
that A escapes in order to make the core continuous. This condition can be achieved
partly, as we shall see, in S = 2 n + 1 lines. No escape at all, of course, for S = n + i
lines. Brand and Pleiner (1985) have suggested the existence of strongly biaxial
cholesteric phases, and indeed some biaxial nematic phases made of very anisotropic
entities like micelles become N $ in the presence of a twisting agent (Figueiredo Net0
et a1 1985). In that case the singularities on the three directors would have some
meaning. In principle, even usual cholesterics should show up a small (but non-
measurable) biaxiality (Pleiner and Brand 1985).
There is no distinction between dislocations and x disclinations, although the
concept of dislocation is often quite useful. This is due to the equivalence between a
symmetry of translation m ( p / 2 ) , y ( m is an integer) and a symmetry of rotation m v
about the x axis.
Finally, since the various rotation invariants involved in the disclinations are not
commutative, there are some special effects of interaction, like the obstruction to mutual
crossing of the defects, or the change of invariants when they rotate around each other.
This will be discussed in 0 7.

6.4.2. Edge dislocations. These are observed in many situations, and in particular in
the Grandjean-Can0 geometry (Cano 1968) in which a cholesteric is inserted in a
wedge with parallel anchoring. The number of cholesteric layers increases therefore
from the edge of the dihedron outwards; this increase is relieved by a number of
discontinuities of the twist, which can be analysed in terms of x twist disclinations
parallel to the dihedron (de Gennes 1968). As shown by Bouligand (1974b), the line
closest to the edge has the configuration of a classical nematic twist line (this is a
region where the thickness of the specimen is smaller than the pitch), while the following
ones are dislocations split into two opposite wedge disclinations of strengths 14,either
a 7- and a A + coupled at a distance p/4 (figure 37a), forming a dislocation of Burgers
vector b = p / 2 (also a ~ ( i )or
) ,a A + and A - coupled at a distance p / 2 (figure 37(b)),
forming a dislocation of Burgers vector p (also a ~ ( 1 ) ) Note
. that other couplings (a
A - T + at d =p/4, a T-T+ at d = p / 2 , figure 37(c)) would lead to the same Burgers
vectors, but they are not observed. No doubt this is related partly to the gain in energy
due to the continuous cores.

6.4.3. Screw dislocations. These can be generated, in an equivalent manner, either by


the classical Volterra process, or by a helicoidal motion of a pattern contained in a
meridian plane passing through the helical axis. The pitch P of this motion must be
necessarily a multiple of the periodicity p / 2 of the cholesteric: P = m p/2, m being
Defects in liquid crystals 61 1

.................. .. --

, - - - -- -
___-- - ,l .......................

Figure 37. Splitting of edge dislocations in a cholesteric. ( a ) b = p / 2 , splitting F A + ; ( b )


b = p , splitting A-A+; (c) b = p , splitting ~ - 7 (~d ); b = p / 2 , splitting A-T+.

any positive or negative integer. One can show (Bouligand and KlCman 1970) that
the strength S of the ,y disclination generated along the axis is S = 1 - m / 2 , and that
b = -pS (figure 38(a)). Rault (1973, 1974a) has shown experimentally that this simple
model should be modified in most cases by letting the molecules 'escape' along the
axis. For a ,y(+l), this escape is possible everywhere except in singular points distant
from p / 2 (figure 38(b)). For a ,y(-l), the escape of the A-director can be complete
(figure 38(c)), so that the only remaining singularities are those of the T and of the y,
directors, but they are immaterial. The same remark is true of course for the ,y(-1).
The generating pattern is interesting in that it can be modified at will in order to
introduce another singularity at some distance from the axis (figure 39). This is a way
to study in particular the configuration of helical pairs of disclination lines (note that
if m = 2, there is no singularity on the axis itself).

6.4.4. Pairing of disclination lines. As indicated in 0 2, a curved disclination is attended


by a density of infinitesimal defects, whose role is to relax a large part of the stresses,
by transporting R along the line. While curved lines of any shape can be easily
stabilised by such a process in a nematic, because of the existence of infinitesimal
symmetries of translations along any vector db (hence R can be easily translated along
the curved line), and also of infinitesimal symmetries of rotations along the director
(hence R can rotate in a plane perpendicular to the director), this is not so simple in
a cholesteric, since the only infinitesimal defects topologically possible are the disloca-
tions with a Burgers' vector d b perpendicular to the ,y axis. Therefore (see equation
(56)), only the ,y' disclinations can take any shape at little energy cost. The same
process is in principle possible for the T and A lines, but the attached infinitesimal
defects would not belong to the symmetries of the medium (imperfect defects) and
612 M Kleman

e
-4 ,;p
m =4
.
I-
0 .

Figure 38. Helical defects in a cholesteric. ( a ) Defect along the helical axis (see text); ( b )
Rault model of a x ( + 1) with partial escape; ( c ) Rault model of a x( - 1 ) with total escape
of the h director.

----------

Figure 39. Helical pairing (see the text).

there exists a fault plane in the medium marking their cut surface, after the manner
of stacking faults in solids. However, it is possible to pair two disclinations of opposite
signs (a, -Cl) at a constant distance in order that they cancel their effects at a large
distance. They are then linked by a small ribbon of infinitesimal defects (Friedel and
KlCman 1969, 1970). Note furthermore that such a coupling (a, -a)is a mode of
splitting of a dislocation, of Burgers vector b = at x d, where t is tangent to the line,
i.e. also a x disclination. No wonder therefore that the pair can curve at will, since
this is a property of a x line (and of a dislocation).
These pairings have already been illustrated; they are observed also in the kinks
of the dislocations and in closed loops. Bouligand (1974a) has given a complete
analysis of such observations, which can be quite often understood in physical terms
Defects in liquid crystals 613

by the presence of A continuous cores, as in Rault's observations and models (1973,


1974).

6.4.5. Textures and double twist, Cholesterics display large textures of the focal conic
type, since they are layered media, with, however, more flexible mutual arrangements
than in smectics, since the pitch can be modified elastically at lower cost than the
thickness of the layers in a SmA (see Bouligand 1972b, 1973a, b).
Another important ingredient in the description of the textures of cholesterics is
the role of double-twist, which can happen locally as a helicoidal instability of small
energy (Lequeux 1986, Lequeux and KlCman 1988). This role of double twist is
probably enhanced by a large positive KZ4(KlCman 1987).

6.5. Biaxial nematics


Biaxiality is a measurable effect in some lyotropic mesophases made of anisotropic
micelles (Yu and Saupe 1980a, Figueiredo Net0 et a1 1985c), and probably in some
thermotropic phases (main chain nematic polyesters, Windle et a1 (1985)) and in some
small-molecule nematogenic Cu complexes (Chandrasekhar et ul 1988). In both cases
the anisotropic shape of the basic entity (micelle, molecule) induces orientational
correlations. The second-order elastic free energy density of biaxial nematics has been
established by Trebin (1981), Saslow (1982) and Brand et a1 (1982) in terms of the
gradient of the rotation which fixes the position of a moving frame with respect to a
reference frame. The moving frame can be defined by a set of three orthogonal vectors
n, m,t = n x m carried by the basic entity of the medium. Each of these vectors has in
fact the symmetry of a director R = - n, etc. The whole symmetry of the second-order
elasticity is therefore orthorhombic, and lower symmetry terms can appear in higher-
order elasticity. For a recent review on all aspects of biaxial nematics, see Galerne
(1988).
There are very few detailed observations of defects in NBs. The disclinations are
similar, at first sight, to those met in uniaxial nematics, as indeed expected. Galerne
and LiCbert (1985) have noticed that S = 4 lines take quite often a zigzag shape, which
they explain by a coupling of the core (which they claim is uniaxial nematic) with a
surrounding applied field. Since zigzag instabilities also exist on lines in uniaxial
nematics, this phenomenon cannot be used to characterise NBs. In the absence of
more experimental studies, it is worth trying to make some conjectures as to the possible
specific properties of defects.
Each of the directors n, m,t can carry disclinations of the uniaxial nematic type;
there are therefore three types of S = *4 lines, which we shall denote n(i), m ( f ) ,t(4),
and, as can be concluded from the homotopy theory of defects (see 0 7), there is no
possible smooth transition between the three types. It is, however, possible to go
continuously from n( +t) (respectively m, t ) to n( -4) (respectively m, t ) , as in a uniaxial
nematic.
There is complete isomorphism between the classification of line defects in N$s
and in N*s (see also P 7 ) since there is also a local frame of reference in a cholesteric.
Although the physical differences are large, this isomorphism can help in the analysis
of the defects of a NBphase, inasmuch as in lyotropic NBsthere seem to exist lamellar
correlations between the building entities (Figueiredo Net0 et a1 1985a).
Let us therefore identify n with the A axis of the N" phase, m with the T axis and
t with the x axis (the cholesteric axis). Our identification assumes, in some way, that
614 M Kliman

n is the strong director, that the biaxiality is weak ( m and t are weak directors) and
that t is perpendicular to the lamellar correlation. We can therefore imagine, as a
possible application of this isomorphism, that n and m lines couple like A and 7 lines
in order to form wedge disclinations of the t type, as in figure 40, according to the
algebraic rule n(f) m(f) = t ( 4 ) .
Two n(f), two m(5) or two t(f) couple to form lines similar to the S = *1 lines of
the uniaxial nematics, which cannot have a continuous core, since an escape in the
third dimension of n would not allow m and t to escape at the same time. There is
+
of course topological equivalence between n( 1) and n ( - l ) , as there is between
S = + 1 and S = -1 in a uniaxial nematic. As in cholesterics, we have to go to n(*2),
m(k2), t ( ~ t 2lines
) to get complete escape of the local frame, and a continuous core.
A considerable difference between NBs and N s is that singular points are not
topologically stable in NBs (see 0 7). Therefore a defect loop L of the type pictured
in figure 41, which is obtained by a rotational symmetry of the pattern of a n(+f), say,
does not exist without another defect along the axis of revolution M, because the loop
cannot shrink to a singular point as in a uniaxial phase. M is a defect of the type
n ( + l ) . The whole pattern is of course reminiscent of a focal conic in a SmA. In other
words, we expect that some kinds of disclination loops couple systematically with
S = + 1 defects.

Figure 40. ( a ) The three directors m, n and t of a biaxial nematic having the anisotropy
of a parallelepiped. ( b ) The coupling of two lines of types n and m leads to a line of
type t.

Figure 41. A disclination loop L in a NBcoupled to a disclination M,


Defects in liquid crystals 615

6.6. Polymeric nematics


As already stated (0 l ) , there are two main polymeric molecular architectures which
give nematic phases, the main-chain polymers and the side-chain polymers. In this
second class of LCP the mesogenic parts interact to some extent as in a usual small
molecule nematic fairly independently of the polymeric backbone, if the spacer is long
enough; static defects, therefore, are expected to show up the same properties as in
usual nematics. They should, however, greatly differ from usual defects in their dynamic
properties, which are mainly controlled by the backbone (Fabre et a2 1984). But the
subject of defects in side-chain polymers has been little studied and we shall discuss
only the other class of LCPS, which includes (i) various thermotropic melts like homo-
and copolyesters, (ii) semi-flexible polymers in solution like polyamide or polytmeph-
thalamide in sulphuric acid (nylon enters this category), (iii) colloidal suspensions of
rod-like polymers like tobacco mosaic virus (TMV), (iv) semi-flexible polymers in a
solution of chiral molecules, most of biological interest, like poly(benzy1 -L or D-)
glutamate ( PBLG, PBDG), hydroxypropylcellulose ( HPC),collagen, DNA, etc. Some of
the polymers in the last category give nematics N or cholesterics N*, depending on
the solvent, the concentration, the pH, etc.

6.6.1. Thermodynamic properties; material constants. The specific properties of defects


in LCPS are due to a number of factors, like the scarcity of chain ends, their anisotropic
viscoelastic properties, etc, which are all related of course to the great length of the
molecules and to the character of rigidity or flexibility of the chain. Statistical models
of the order parameter and of the nematic isotropic transition have been recently
developed (see, e.g. de Gennes (1984), Wang et a1 (1986) and ten Bosch et al (1983)).
The variation of theFrank constants and of the viscosities as functions of the chain
length L, chain radius d, volume concentration of the polymers cp, and of the persistence
length I, have been established by a number of authors: Straley (1973), in the line of
the model of Flory for rigid rods in solution, Odijk (1986), for semi-flexible chains in
solution, Meyer (1982), de Gennes (1982), for thermotropic semi-flexible polymers,
etc. Measurements of these viscoelastic constants have been made, mostly using the
Freedericks transition under a magnetic field (Meyer et a1 1985, Sun et al 1984) or
under an electric field (Pate1 et al 1979), or using Rayleigh light scattering (Meyer et
a1 1985).
For rod-like polymers in solution, the theory predicts that the ordering of the chains
affects primarily the bend deformation constant K 3 ,which should be the largest, while
K , = 3K 2 . Fraden et a1 (1985) find K 3 /K 2= 42 and K1/ K 2= 2.5 for TMV. They also
obtain large anisotropies for the viscosity which go in the same direction as the theory.
For semi-flexible polymers in solution, and for thermotropic polymers with large
enough concentrations cp, the theory predicts, in contrast with rod-like polymers, a
large K1 to which the major contribution comes from a single-particle entropy effect:
splay and constant density requires segregation of bottom from top ends of molecules
in the region of splay (Meyer 1982).
4 LkBT
K,=;cp--
d d
K 3 is limited by the persistence length, and does not diverge
2 lp kBT
K3=;cp--
d d
616 M Klkman

K 2 keeps a low value, as in small-molecule nematics. The orders of magnitude expected


from this theory have been obtained in polyesters (Sun et a1 1984), and tested with
success versus cp in PBG (Taratuta et a1 1988).
The same Freedericks transition experiments yield Frank constants and viscosities,
but the Freedericks transition in main-chain polymers has more than practical interest.
In the geometry for measuring bend (homeotropicboundary conditions), the first mode
of deformation, at a critical field H,,is extremely slow because it corresponds to a
large effective viscosity which is easily understood in the infinite chain limit (Meyer
1982); but at twice the critical field, the effective viscosity of the second mode of
deformation is small and the distortion appears quickly. The twist geometry shows
up another phenomenon, which is the appearance of a pattern of transient stripes
whose corresponding flow has a small dissipation, while the usual twist viscosity is
very large. Finally the geometry for measuring splay does not display splay at all,
when K , / K , > 3.1, if K , is extremely large! Hurd er a1 (1985) have indeed shown that
it appears to be a periodic structure composed mostly of twist distortion, which has
a lower threshold field than the simple splay transition (figure 42).

Figure 42. Pure twist stripes appearing in the Freedericks transition for splay in a nematic
solution of PRLG (courtesy D Srajer, F Lonberg and R B Meyer). (Applied field 3.4 kG;
periodicity 125 p m . )

6.6.2. Textures and defects-thermotropic specimens


(i) Observations in uniaxial nematics. Thermotropic nematic polymers belonging
to a series of polyesters differing only in the length of the flexible alkyl group -[CH,]-"
have been investigated by KlCman et a1 (1983). For large n ( n = 14, say) the viscosity
is extremely high and the sample does not anneal at all during the optical microscopy
observations; the local orientation does not vary even after long annealing in the
isotropic phase. However, x-ray diffraction gives evidence for the existence of a nematic
phase; the molecules are probably tightly correlated by coiling around one another,
and these coilings subsist in the isotropic phase while the extent of the correlations
decreases. For n = 5 the texture varies with the degree of polymerisation. For shorter
Defects in liquid crystals 617

chains (A4 = 1000) one observes a typical small molecule liquid crystal (SMLC) thread
texture, with thins (IS1 =$) and thicks (IS/ = 1). In the course of time the threads have
a tendency to disappear and be replaced by a well resolved texture of Friedel's nuclei
(plage ci noyaux, with only integral lines and singular points). However, for longer
chains ( M = 10 000 and more) there are very few integral lines or nuclei, and most of
the defects are half-integral lines, either loops floating in the bulk and tending to
collapse in a few minutes after their appearance (these loops have a twist character)
or lines attached to either the glass covers or (in free droplets) the free boundary. In
this last case one can observe directly that the size of the cores of S = +:wedge segments
is much larger than the size of the cores of S = -4 wedge segments (Mazelet and
KlCman 1986).
The interpretation relies on two related effects: (i) K1 is large in this product
( K ,= 3 x dyn; Sun et a1 1984), because of the scarcity of chain ends, while K 2
has its usual SMLC values ( K 2 = 3 x lo-' dyn). K , is probably small too. (ii) The cores
of S = + i wedge lines are clusters of chain ends (a large part of the necessary splay
deformation tends to be concentrated here); chain ends accumulate differently near a
S=-'l ine, where the chains align parallel to the core of a three-arm star. Fayolle et
a1 (1979) have observed a slightly different homopolyester and corroborate some of
these observations.
The fact that K , is large compared with K 2 and K , should lead, by an argument
of continuum elasticity, to a spreading on large distances of splay deformation,
according to Meyer (1982) who explains this way the formation of splay domains of
width R separated by disclinations where twist-bend deformation is highly concen-
trated. This argument is certainly true for a one-dimensional deformation, where we
expect R = ( K ,,I K23)d ( d being some radius, where twist-bend concentrates, of the
order of the transverse size of the specimen, K2, some elasticity modulus which involves
only K 2 and K,). But it may happen that there are local molecular mechanisms of
relaxing splay deformation which do not enter in the framework of continuum elasticity,
as for example the formation of hairpins (de Gennes 1977) or the concentration of
chain ends in dislocation cores. In the first case, the effective splay modulus could be
strongly reduced if the activation energy of formation of hairpins is small; this is the
only possible mechanism in the infinite chain limit, and for very long molecules, like
DNA. We discuss the second mechanism in more detail below.
The same polyesters also display 90" Bloch walls (we use a terminology borrowed
from the study of magnetic walls), in which the molecules rotate about an axis
perpendicular to the wall, with practically pure twist. This indicates that K 2 is small,
and provides an example of 1 D geometry of the type alluded to above. The domains
on both sides contain bend and splay spread all over.
Very thin films (a few hundred .$) of the same polyester, spread over a convenient
substrate and afterwards quenched (at a moderate rate) show up in high-voltage electron
microscopy an alternation of nematically ordered and disordered stripes, with a
periodicity comparable with the molecular length (Mazelet et al 1988). The stripes
become quickly amorphous under electron irradiation, but keep their individuality;
they are still visible by thickness contrast in electron microscopy. This effect had
indeed allowed Thomas et a1 (1985) to make the first observation of such stripes, but
in a random copolyester.
Finally, let us indicate the observation of anisotropic droplets in a nematic-nematic
phase separation, in a mixture of a side-chained polysiloxane with a SMLC nematic
(Casagrande et a1 1987). These very unusual objects seem akin to the tactoids quite
618 M Klkman

frequently observed in lyotropic polymers of rigid rods (Bernal et a1 1941). Here they
are explained as resulting from the competition between a low interfacial energy and
elastic energy of the nematic matrix.
(ii) Models for wedge disclinations S = ib. It has been proposed by Mazelet and
Kleman (1986) that the large-scale off -core observations of half-integral wedge segments
in free droplets have to be explained by the anisotropy of the Frank coefficient, for
example on the basis of the models of Nityananda and Ranganath (1980) as in figure
43(a). Such models are derived in the frame of the standard elasticity of SMLC nematics;
but in the limit of very long chains, we have to take into account a contribution of the
compressibility E-' (Taratuta and Meyer 1987), which dominates splay contribution
on lengths smaller than
A, = 2 ~K1(/ E)'/* (64)
E has not been measured in the polyesters we consider, but one expects that E scales
like the inverse length of the molecule, as long as there are no entanglements (L< Le);
taking then E = lo7 erg ~ m - K ~1,= dyn we get A, of the order of a few hundred
A. We show that it is also a typical size of the core when K 1 is large, so that the
model of the core has to be 'elastic' essentially. Of course the calculations are valid
as long as the core is not much larger than the molecular length L.

Figure 43. (a) A core model for the three-arm star observed in S = -4 wedge lines. ( b )
Core model for a S = +fwedge line.

For a S = +i (figure 43(b)), we assume that the core elastic energy is minimised
by a clustering reminiscent of the geometry of a crystal bent under the effect of a
density of dislocations ( 0 2.2 and figure 8(c)): the polymer chains are represented by
half straight lines, which are the analogues of the extra atomic rows in Nye's model;
d, the mean distance between chains, plays the role of the lattice parameter in the
transverse direction; the chain ends ate dislocation segments of the lattice, the whole
dislocation lines being along the core. The total (elastic) energy of the core is in this
geometry essentially restricted to the interaction energy between chain ends. If 1 is
-
the mean distance between the corresponding dislocations, we have 1 (drm)1'2,where
Defects in liquid crystals 619

rm is the size of the cluster. Hence for the total core energy
E dr, r,
W,=- In - (65)
8 d
and for the total line energy
W = W,+ K23 In A/ r,
where K z 3 is a bend-twist constant and A a typical distance between disclinations.
Minimising with respect to r,, one finds
-1
r m = -8(K 2 3 ,+In?) A23
Ed
where A23 = 2 7 ~K( 2 3 /E). This is of course a very large core, because of the ratio
A23 / d, which is anyway large, even if is much smaller than A,.
The model of an S = -4 (figure 43(a)) does involve chain ends in a quite different
way, with chains probably escaping vertically along the core and the three-arm-star
walls.
Non-singular integral lines would have an elastic energy larger than the elastic
energy (core included) of the singular half-integral lines, because of a large K , . This
is why they are not observed.
(iii) Observations in copolyesters. Observations of disclinations seem to be at vari-
ance with those reported above for homopolyesters. Mackley et a1 (1981) interpret
theirs as demonstrating the existence of numerous disclinations with strengths S = 1, *
in the classical plage & noyaux (schlieren texture). The core itself is rather large, and
the whole contrast is much fuzzier than in a usual plage Ci noyaux; this could be due
to a gradient of chain ends on large scales. However, little more can be said, because
we ignore even the order of magnitude of the molecular weight. Noel et al (1984)
have also reported schlieren textures with integral and half-integral lines in another
copolyester.
Graziano and Mackley (1984a) and Aldermann er a1 (1985) have reported on the
changes in optical textures under controlled shearing experiments together with the
subsequent relaxation behaviour at the cessation of flow. Some observations, like
the nucleation and multiplication of defects, are reminiscent of those observed in SMLC
nematics by the same authors, at least at low shearing rates. But others appear unique
to LCPS, in particular the fact that the shear has a significant and specific effect on
the microscopic texture, and presumably the microstructure. More precisely, Graziano
et a1 describe: (i) the appearance of a worm texture, a kind of disordered polydomain
texture at the scale of a few micrometres, i.e. inferior to the size of the specimens and
independent of it; the worm density increases with increasing shear rate, (ii) a sharp
transition between the worm texture to a birefringent texture (with optical axes along
the direction of shear), called ordered texture, made presumably of a very high density
of suboptical oriented defects of the type seen in the worm texture. The shear rates
at which these transitions occur decrease with decreasing temperature and decreasing
thickness of the sample. Finally the relaxation behaviour is very peculiar, and consists,
in high enough molecular weight nematics (A4= 20 000), of a banded texture perpen-
dicular to the shear direction, after cessation of shear in the ordered texture; the banded
texture relaxes to the worm texture and back finally to the quiescent original state with
few disclinations. The banded texture itself seems to be a specific uniform texture
which, like all other textures observed, always occurs with the same specific characteris-
tics under a given set of shear and thermal conditions.
620 M Klbman

Banded structures (of another sort?) were observed in thin films of random CO-
polyesters by diffraction contrast in electron microscopy (Donald and Windle 1984a, b).
Annealed specimens show up walls which tend to be either parallel or perpendicular
to the original shear direction, with a characteristic rotation of the molecules out of
the plane, interpreted by the authors as the splay-splay compensation of the type
alluded to above (Meyer 1982). Comparison of electron and optical microscopy
pictures, and x-ray diffraction patterns have led the authors to infer that the copolyesters
they have observed are biaxial, with orientation correlations about all three axes
(multiaxiality). It is quite possible that the copolyesters used by Mackley et al (1981)
in their large-scale shearing experiments are also biaxial. This property has implications
for new observations of the disclination lines, and it might well be that the important
differences observed between the textures in copolyesters and those in homopolyesters
are related to the different natures of the molecular correlations.

6.6.3. Textures and defects-lyotropic specimens


(i) Nematic phases. A thread-like texture, made of disclinations with a very thin
singular core, has been observed in a series of semi-flexible polymers in solution in
pure sulphuric or chlorosulphonic acids (Arpin et al 1977, Millaud et al 1978). The
authors claim that most of the lines are S = $1 disclinations, on the basis that small
micrometre-size loops are seen disappearing by progressive attenuation of their optical
contrast. However, for largest loops, the core is singular, with a radial splay geometry
and no escape of the molecules along the core. The calculation performed above for
the core of a disclination in a thermotropic polymer does in fact apply here, and leads
to a core of molecular dimension, since the Burgers vectors of the Nye distribution
in the core is of the order of the intermolecular distance, which is large and comparable
with AZ3, Thin cores are also documented, by freeze fracture electron microscopy, in
TMV solutions in water, but the strengths are IS1= 4 (Zasadzinski et a1 1986).
When the quiescent thread-like texture just described is mechanically disturbed, it
appears a very dense thread-like texture, similar in many aspects to the worm texture
of Graziano et al (1984a). It is remarkable that the relaxation of the worm texture
does not depend on the viscosity or the persistence length of the molecules, but on
the molecular weight and the concentration, by their ratio cp/ M, i.e. likely only on the
density p of free ends, by the relation cp/M =:p/mN, where m is the mass density of
the mixture and N is Avogadros number. These free ends cluster near the defect
cores. The relaxation takes more time when the density of chain ends is smaller.
Furthermore there is a certain critical density pc below which the worm texture never
relaxes to the ordinary quiescent thread-like texture (Millaud et a1 1979). All these
phenomena are the analogues of those met in the thermotropic copolymers.
(ii) Cholesteric phases. DNA, PBLG, xanthan and other polymers of biological
interest in solution display characteristic cholesteric textures whose defects have been
extensively studied recently (Bouligand et a1 1984, Livolant and Bouligand 1986).
Most of these defects are similar to those observed in usual thermotropic SMLC
cholesterics. This is the case of spherolites with disclination radii or disclination
diameters, first observed incidentally in PBLG (Robinson 1961, 1966). There are some
textures which are characteristic of polymeric phases, like tactoidal shell defects, and
precholesteric states.
Tactoids have been observed under a magnetic field in PBDG (Filas 1978); they are
akin to cholesteric spherolites, with a radius of disclination S = + 2. Tactoids seem to
be a preferred type of nucleation in polymers.
Defects in liquid crystals 62 1

Precholesteric states and hierarchical textures: banded textures have been observed
at the isotropic cholesteric transition of DNA in solution, when the transition is driven
by a slight desiccation of the sample. These banded textures show up also in concen-
trated solutions. Their geometry has been studied by Livolant (1984, 1987); since they
appear between the isotropic and the cholesteric state, they have, as expected, some
resemblance to the blue phases which will be discussed later, i.e. they result from
phenomena of the frustration of the cholesteric order. Essentially these bands are
complex supermolecular arrangements of bundles of molecular lines twisted along
their length. Geometrically, equal bundles can pack together and form ordered (parallel
bundles or twisted bundles) or disordered arrangements on high scales. It is therefore
feasible that bands are high-scale hierarchical arrangements of bundles of lower scale.
Livolant has also observed self-assemblies with very large pitch of large hexagonal
packings of PRLG.
Hierarchies of organisation have been observed in collagen, which has a cholesteric
phase (Gathercole and Keller 1982).
(iii) Rheological properties of cholesteric polymers under shear and relaxation. They
are a subject of constant study since the pioneering work of Kiss and Porter (1980)
who provided evidence for the existence of three regimes under shear in a solution of
PBLG in meta-cresol: (i) at low shear, appearance of defects elongated in the direction
of shear; (ii) at intermediate shear, multiplication of the defects and existence of a
negativ'e first normal-stress difference N , = u l l- u22 (measured in a plate-cone
geometry); (iii) at high shears, N , is again positive. Bands perpendicular to the shear
direction appear in relaxation in the first regime, but show up directly in the second
regime, during shear. These experiments have been done subsequently, with the same
results, in hydroxypropylcellulose liquid crystals in water (Navard 1986). The suc-
cession of regimes and the appearance of bands bear some resemblance to the
phenomena observed when shearing a nematic copolymer (see above). Note finally
that Donald et a1 (1986) have observed that the bands in HPC can relax subsequently
to a texture of parabolic focal domains.

7. The classification of defects by algebraic topology methods

7.1. The essentials of the topological classijication of defects, and examples


There are many excellent reviews on the classification of defects in ordered media by
the methods of algebraic topology (Mermin 1979, Michel 1980, Trebin 1982);therefore
we shall outline the general principles very briefly, and illustrate them on some simple
examples of liquid crystal phases, first in order to bring out the concept of topological
stability (the essential contribution of this theory to the physics of defects), and secondly
to return in some detail to the effects of non-commutativity of the symmetries, which
are taken into account in a systematic way in this theory.

7.1.I. Uniaxial nematics and abelian homotopy groups f o r line and point defects. Consider
a Burgers' circuit r surrounding a disclination line L (as in figure 9 ( d ) ) ; if the
configuration of the director is planar, the mapping of the director on a unit circle C
(see figure 9 ( e ) ) is enough to tell us about the nature of the singularity. If the
configuration is three-dimensional, an evident generalisation is to map the director on
a unit two-dimensional sphere S 2 : the image of an S = =kiline is therefore a path y
on S 2 between two antipodal points, which represent in some way the same director
622 M Kleman

n = -n. Take the convention of identifying antipodal points on S2;one obtains a new
manifold, called the projective plane P 2 , representative of the topological properties
of the order parameter of a nematic, and whose topology has been much studied by
the geometers (see for example Hilbert and Cohn-Vossen 1952). On P 2 the path y is
closed. If the Burgers circuit r is displaced in space without ever traversing L, its
image y moves on P 2 in a continuous manner; it cannot be reduced to a point and
cannot vanish in a continuous motion. The contrary would indeed mean that there
would be a continuous way to transform a singular ib line into a non-singular one.
Therefore topologically stable defect lines are represented on P2 by classes of closed
paths which are non-homotopic to zero. It is not very difficult to convince oneself
(figure 44) that there is only one class of such paths on P 2 : any oriented closed path
which is not in the null class (like yo) is homotopic to y = AB. For example y- = BA
is homotopic to y (rotate the line y by moving its end points A and B along the
circle, always keeping them diametrically opposite, until A comes in B and B in A.
This is a continuous motion). Also the product of two oriented paths, both non-
homotopic to zero, is in the null class.

B
Figure 44. Types of loops on the projective plane P 2

The classes of homotopy of paths on P 2 form a group, the so-called fundamental


group rIl(P2). It is denoted 2, and composed of two elements, the element unity
corresponding to the null class, 1, and another element, a, such that
a = a-1 a 2 = 1. (68)
Hence (i) all integer disclination lines, whatever S might be, are in the null class, since
any integral line can escape to a continuous configuration, (ii) half-integer lines are
all in the same class a. It is then possible, by a continuous motion in 3D space, to
transform an S = +$into an S = -f (see Bouligand (1981) for a pictorial realisation of
such a process). Clearly a similar continuous motion does not exist in this 2D space
in which we have described wedge lines; in 2D space the representative manifold is
a unit circle, and the fundamental group is the group of integers 2 whose elements
measure the number of times the path winds around its perimeter.
This approach extends to other types of singularities, like points and walls. For a
point, the Burgers manifold which generalises the Burgers circuit is homeomorphic
to a sphere (see B 6.1.3); its image on P 2 covers P 2 an integral number of times N
(equation ( 5 5 ) ) . One can also define classes of homotopy for these mappings of 2D
spherical Burgers manifolds, and they also form a group, I12(P2).
Note that IIl(P2)= 2, is an abelian group. This is a very simple situation in the
topological theory of defects. Most manifolds V representative of order parameters
Defects in liquid crystals 623

of ordered media have a fundamental group R I (V) which is non-abelian. This character
introduces a new degree of complexity in the non-linear topological behaviour of line
defects(see next section).
The ordered media whose corresponding fundamental group ITl is abelian are in
fact very few. Apart from nematics and amorphous ferromagnets, there are the various
superfluid phases of 3He and 4He, whose order parameter is defined through a gauge
group. For the discussion of their disclinations, see Volovik and Mineyev (1977) and
Bailin and Love (1978).
Topological stability, just discussed, and energetical stability, are different concepts.
Consider the simple example of usual nematics, made of short molecules, the energetical
stability of defects meets the criteria of topological stability, and in particular integral
lines have generally a continuous configuration. The situation is different in some
nematic polymers, where the configurations required by the escape of the core are very
high in energy, and are rarely observed.

7.1.2. Efects relevant to a non-abelian first homotopy group. A non-abelian first


homotopy group originates in many cases in a non-abelian symmetry group, because
IT,( V ) is the inverse image of the symmetry group H = T 0 P in the mapping which
brings the universal covering E = R30 SU(2) of the euclidean group E = R3 0 SO(3)
onto E itself (KlBman and Michel 1978b). IT, is therefore the semi-direct product of
the group of discrete translations T of the crystal by the inverse image P of the point
group P. P contains twice as many elements as P. However, l l l ( V ) might be non-
abelian even if P is abelian. This is the case of the biaxial nematics, in which P = D 2 ,
and P = Q, the group of quaternions.
The consequences of the non-commutativity of H I , in spite of the counter-example
of the biaxial nematics, can be quite well understood by investigating, on specific
examples, the relationship between disclinations and dislocations. Let us consider
then again layered media (cholesterics, smectics), whose corresponding rotation and
translation invariants are non-commutative.
The modes of splitting of a dislocation into two disclinations have already been
investigated (see figures 12, 15 and 37). More generally, two wedge disclinations L1
and L2 of strength *$ like to pair at a distance d to form an edge dislocation of
Burgers vector

b=2d (69)
the only constraint being on d, which must be such that b is a permitted Burgers
vector. This relationship does express the group theoretical law of multiplication of
two opposite twofold rotational symmetries at a distance d ; they result in a translation.
Figure 45 illustrates a different case, but of the same nature. The same type of group
theoretical relationship applies for a kink on a disclination (figure 46). It is visible
indeed that the difference between figure 46(a) and ( 6 ) is just in the displacement of

. ...
Figure 45. Splitting of a dislocation into two disclinations (with respectively five-membered
and seven-membered rings around their cores) in a 2D hexagonal lattice.
624 M Kleman

ti
L, L2
t
L1

Figure 46. To illustrate the relation between kinks on disclinations and dislocations (see
the text).

one of the wedge disclinations of the pair to another position, but with its end still
attached to the twist segment, which we call here a kink. The scheme of figure 46(b)
applies here to the case when b is a permitted Burgers vector, but it can be easily
generalised to a curved disclination (figure 8( b ) ) ; equation (69) then generalises to
db = Cl x d s (70)
a relation which holds for a nematic, as we already know.
The presence of dislocations at kinks is well illustrated experimentally in hexagonal
columnar mesophases: we know from 5 that S = +$wedge disclinations form easily
in these media, while twist disclination lines are never observed. But kinks might be
present, to which dislocations are attached, forming, rather than a well defined grain
boundary, a pair of lines which disperse far away from the kink, as observed by Oswald
(see the illustration in KlCman (1983a)).
Let us now notice that opposite disclinations, when merging, do not necessarily
form a dislocation (figure IO); they can also anneal. It appears indeed that the final
result depends in fact on the path followed by the two disclinations which are to meet.
More generally, the relative motion of two defects has topological implications. For
example the Burgers vector of a dislocation changes sign when this dislocation makes
a complete turn about a half-integral disclination (Frank 1969) (see figure 47). Consider

Figure 47. A dislocation changes sign wher, circumnavigating about a half-integral disclina-
tion in a layered medium.
Defects in liquid crystals 625

also a dislocation which crosses a disclination in a plane perpendicular to the disclina-


tion (we illustrate the phenomenon (figure 48) in a 2D hexagonal lattice); according
according to our latter remark, the Burgers vector of the dislocation changes during
its travel, and reaches values which are dijferent on both sides. This means that a new
defect has been created when the dislocation and the disclination have crossed
(de Wit 1971, Poenaru et a1 1977, KlCman 1977b); this defect is a dislocation, which
forms a triple node (obeying the Kirchhoff law) with the two parts of the moving
dislocation, and terminates on the w,edge disclination, where it creates a (wedge) kink.
This is a single example of a phenomenon of obstruction to crossing which is to be met
frequently with intersecting disclinations and with intersecting dislocation and disclina-
tion. It is well known that dislocations can cross, the crossing leaving jogs on each of
them (Friedel 1964); if instead of crossing, they form a junction, the reason for this
behaviour is energetical, not topological. But with defects of a more general nature,
the obstruction to crossing may also be topological.
Obstruction to crossing and topological implications of the relative motion of two
defects result in a fundamental way from the non-commutativity of the group. The
same phenomena should happen in biaxial nematics (see the details in 7.2).

7.2. Other results concerning the topological theory of defects


7.2.1. Order parameter space and homotopy groups: general comments. The manifold
of internal states, or order parameter space V , is the starting concept; by definition it
is a space whose points are in one to one correspondence with each value of the order
parameter of the ordered medium. V is a sphere S 2 for a system of spins in an
amorphous ferromagnet (each direction in space is equivalent), and V = P 2 for a
nematic phase, as seen above. More generally, V is the quotient space G I H , where
G is the thermodynamic group which leaves all the physical properties of the ordered
medium invariant, and H the group of symmetry of the medium. For crystals, G is
the euclidean group R 3 0 O ( 3 ) ;it is a gauge group for suprafluids. Most often, V is
a quite complicated manifold, but we need to know only its homotopy groups Hi(V)
which can be deduced quite straightforwardly from H (IUCman and Michel 1987b).
All homotopy groups are abelian, except possibly the first homotopy group (also called
the fundamental group) H I .
The topological theory of defects involves mappings from objects homeomorphic
to spheres onto the order parameter space V. Garel (1978) has also considered mappings
from tori onto V. It can be indeed advantageous to surround a closed line defect by
a torus; the resulting mappings fall into classes which have also a natural group
structure. If such a group is non-trivial, this means that the loops belonging to

Final position

Initial position

Figure 48. Change of Burgers vector of a dislocation moving through a disclination in a


hexagonal lattice, and appearance of a defect linking the disclination to the mobile
dislocation.
626 M Kle'man

non-trivial classes do not collapse to nothing by reduction of their size. Such a


phenomenon occurs with some types of loops in nematics, which reduce to singular
points. Klkman (1977b) has obtained apparently similar results by considering the
homotopy classes of the cut surface bound by the loop. The energetical stability of
point defects versus disclination loops in nematics has been studied by Mori and
Nakanishi (1988).

7.2.2. The fundamental group. We have discussed above some specific features of the
interaction between defects of non-commutative symmetry invariants. The same
phenomena receive a quite simple and illuminating description in terms of homotopy
classes. Call a, the class of homotopy of a given defect and let it accomplish a complete
turn around a defect of class c. It is possible to show by methods of algebraic topology
that it has acquired a new class of homotopy az= ca,c-'. If a, and c commute, nothing
is changed. But in the general case the same defect can take any of the classes of
homotopy belonging to the same class of conjugacy of II,. More generally, consider
two defects of homotopy classes a and by belonging to the classes of conjugacy C ( a )
and C ( b )in II,. Note that ai, bi and ci are any homotopy classess belonging to C ( a ) ,
C ( b ) and C(c) respectively. Then:
(i) if a, rotates around c, its class of homotopy changes after a complete turn and
becomes aj = caic-';
(ii) if a and b merge, the resulting defect belongs to a class of homotopy in
C ( a ) C ( b ) . The product C ( a ) C ( b )does not define in general a unique conjugacy
class in I l l . Hence the defect resulting from the coalescence is not uniquely defined
even in the sense of the topological theory of defects;
(iii) if a and 6 cross, they stay connected by a defect belonging to the class of
homotopy c = aba-'b-'. There is no obstruction to crossing if c = 1, i.e. if a and b
commute (Po6naru and Toulouse 1977, KlCman 1977b). The set of all elements of II,
of the form aba-lb-' form the so-called commutator subgroup K, which is the largest
normal abelian subgroup of II,.
The non-trivial defects pertaining to the conjugacy classes of K of II,(V) are
special: they can always be split into two defects which, after some turn of one of
them about a third one, can annhilate (Trebin 1982).
The set of conjugacy classes form a group, the so-called homology group H,( V) =
II,( V)/ K, with K as unit element. H,( V) is abelian.
As an illustration of a non-abelian fundamental group, consider cholesterics, whose
defects have been classified by the Volterra process in 0 6.4, and biaxial nematics
(0 6.5), whose order parameter is the same. The topological classifications are due
respectively to Toulouse for NBs (1977a) and the Mineyev and Volorik (1978) for
cholesterics; the fundamental group is the quaternion group Q = - I * e , , * e 2 , (--e2),
* e 3 , containing eight elements and five conjugacy classes c ~ { I } C0{-1},
, Ck{ek, -ek}
with the combination rules
erez= - e2e1= e3
e1e2e3= - I
( - I ) ~ =I.
Table 2 gives the correspondence between topological classes and Volterra defects in
cholesterics. The correspondence between cholesterics and NBs is in 0 6.5. A very
remarkable fact is that all dislocations whose Burgers' vector b equals 2np enter the
Defects in liquid crystals 627

Table 2,

b = 2np +
b = (2n 1)p b = ( n+ 4 ) p
,y(S=2n) ,y(S=2n+l) ,y ( S = n + f )
7 (S=2n) 7 ( S = 2 n + 1) 7(S=n+$)
A (S=2n) A (S=2n+l) A (S=n+f)

null class, i.e. the core singularity can disappear and the configuration be continuous;
this is evidently reminiscent of the behaviour of the integral lines in nematics.

7.2.3. Conjigurations and the homotopy group I13. The distortion fields whose homotopy
class belongs to a non-trivial element of 113(V) are non-singular topologically stable
fields. These so-called configurations or solitons are not, however, observed in
crystals, rod lattices and lamellar systems, because of obstructions to their presence
which result from the integrability conditions of the distortion fields, as noticed by
Gunn and Ma (1980) and Trebin (1983). More precisely, in a crystal lattice, the
periodic nature of the crystal must be respected in any distortion field; in other words
the distortion field pij, defined in continuum by the differential form

must be derived from a displacement field ui(r ) ; this condition ensures that the lattice
vectors close after distortion, and curl p = 0, i.e. in component rotation:

Conditions (73) are the well known compatibility conditions of a distortion field
(Bilby et a1 1955, Kroner 1981). Trebin (1983) has established by employing general
arguments of differential topology that any field obeying these conditions is diffeomor-
phic to a field without configurations, i.e. to the perfect crystal, in three dimensions.
Reciprocally, this implies that any configuration in a crystal is attended by some sort
of singularity of the distortion field, whose density is measured by one of the quantities
which vanishes in (73); for example
bpk = EpqrPkr.4 (74)
measures a density of dislocations, as shown by Kroner (1981). The identity bpkp = 0
is the expression in continuum theory of the fact that a dislocation cannot terminate.
bpk is therefore a component of the infinitesimal Burgers vector of an infinitesimal
dislocation along the direction k, of components blk,b2k,b3k.
Trebins theorem does not apply to media with continuous symmetries only. For
example, He3-A superfluid has an order parameter whose geometrical representation
is a triad of vectors, but there is no condition of the type (73), since He3-A is a liquid
and can suffer any elastic distortion. Therefore there is no obstruction to topological
configurations in He3-A. The same results hold for biaxial nematics NB,nematics and
cholesterics, and we expect topologically stable configurations in these media. The
quantities bpk # 0 which attend these configurations can be interpreted as frozen
infinitesimal dislocation fields on a virtual lattice superimposed on the medium; this
virtual lattice is defined by the natural local frames in He3-A, NB phases and
cholesterics, in the examples above. However, as noted first by Shankar (1977),
628 M Kle'man

configurations in He3-A and nematics, whose order parameter space is of the type P"
or S", are not stable and shrink spontaneously to zero size. This is evident in the frame
of continuous linear elasticity, since there is no natural macroscopic length in such a
theory. Whether lengths introduced by higher-order terms are enough to provide for
high-energy barriers is still an open question. It has been suggested that configurations
are stable in cholesterics, where the pitch offers a natural length (Pisarski and Stein
1980).

7.2.4. More on singular points. These are classified by the elements of n,(V); this
group is isomorphic to the group of integers 2 in nematics and layered media, but is
restricted to the triGia1 element in biaxial nematics, cholesterics and He3-A. Therefore
singular points are not topologically stable in these media. The corresponding invariant
N (see equation ( 5 5 ) ) changes sign with the sign convention on the director n.
Nonetheless we have to distinguish between different singular point geometries which
can bear the same absolute charge INl; for example for IN1 = 1, the hedgehog (in which
the director points radially with the singular point as centre, in a nematic; or in which
the layers are spherical, in a smectic), and the hyperbolic point defect (this is the
configuration at the centre of a toric focal domain). It is intriguing to notice that if
an IS1 =; disclination rotates around an N point defect, then after completion of the
turn N-. -N. Trebin and Kutka (1981) have proposed a finer classification of singular
points, by the use of the Morse index, which distinguishes between the various
geometries of the same configuration INI. They have also noticed that a focal domain
can be defined as a singular point, with an inner region of hyperbolic nature, and a
hedgehog exterior, separated by a disclination, which is necessary by an argument
akin to the argument used above for the change of sign of N. In the same spirit, it is
worth noting that Lavrentovich and Terentiev (1987) have observed, in a nematic drop
embedded in the isotropic phase, a transition under a temperature change from a radial
hedgehog to a hyperbolic one with the simultaneous formation of a circular non-singular
disclination. The boundary conditions (homeotropic) do not vary during the transition.
They explain this phenomenon as being due to a temperature-induced variation of the
Frank constants.

7.3. Defects with partial topological dejinition (semi-defects)


As emphasised in the case of biaxial nematics, a singular distortion can affect only a
part of the 'order parameter'; for example, only two director fields out of three are
singular. This notion is quite general and has been studied in some detail by Kutka
and Trebin (1984). They have coined the name of semi-defect. An interesting experi-
mental case, to which we have already alluded (0 6.4), occurs with cholesterics, in
which line defects of the type A are of a very small energy because they do not carry
any singularity of the director (the singularity is restricted to the twist axis x and to
the T axis, which are immaterial entities). They can be classified apart by the classes
of the group IT3(P2)= 2 ; 113 because there is no singularity of the director, and P2
because A is a director (Bouligand et a1 1978). To simplify matters, the same absence
of singularity of the director ensures that the director field can be transformed to a
vector field, i.e. the mapping of the director field can be made directly on S', and it
suffices to consider 113(S 2 )= 2. This is the celebrated Hopf mapping S 3+ S 2 , classified
by the Hopf index, which measures the linking number of the inverse images of the
physical space of two points in general position in V = S2. Note that the inverse image
Defects in liquid crystals 629

of a point of S is a closed loop in the 3D physical space. There is therefore no surprise


that the defects in question appear in polarising microscopy observations (Bouligand
1974b) as intertwined lines. The detailed classification of these observations in terms
of Hopf index is given in Bouligand et a1 (1978) (figure 49). In Bouligands original
observations (1974b), these defects are loops which detach from kinks on dislocations.
The topology of linked defects has been developed recently by Monastyrsky and
Retakh (1986) and Nakanishi et a1 (1987).

Figure 49. Non-singular semi-defects in a cholesteric phase with various Hopf indexes
(polarising microscopy; courtesy Y Bouligand).

7.4. Surface defects, cores and the use of the exact homotopy sequence in various problems
involving defects
The so-called exact homotopy sequence (EHS, see Steenrod (1951)) is often used as a
powerful tool to compute homotopy groups. Since its geometrical meaning is closely
related to some physical properties of defects at surfaces, at phase transitions, and to
their core properties, we comment on it in some specific examples.

7.4.1. An interesting relationship (KlCman and Michel 1978a, b) between the defects
of the SmC phase and the SmA phase arises from the fact that the point group H,
(reflections excluded) is a subgroup of the point group HA(reflections excluded). This
is easily seen when considering a continuous phase change SmC + SmA, which consists
in tilting the optical axis towards the normal to the layers. Let us therefore assume
that we know the order parameter space VA= V(SmA). To each point in VA we put
in correspondence a one-parameter family of orientations of the SmC phase, by keeping
the orientation of the layers constant but moving the optical axis on a cone about a
normal to the layer; all these positions define a circle S; in mathematical terms S is
thefibre of the fibre bundle with base VA,and bundle Vc = V(SmC): all the orientations
of the SmC phase are indeed reached once and only once when one performs the
former geometrical operation for all the SmA orientations defined by all the points
of v,.
Reciprocally, let us consider the set of all loops on Vc whose homotopy class is
(0, a), i.e. we consider in VA defect lines of type m (the notation ( n , u p ) means that
the considered defect does contain a dislocation part of Burgers vector b = nd,, and
a disclination of strength S = p ~ )We
. know from former considerations that 2 classifies
the dislocations, and 2, the disclinations, since a4 is the trivial defect, whose core is
630 M Kle'man

able to escape in the third dimension. These loops in Vc clearly deal with the
geometrical operation just considered, and we conclude that the fibres S' are all in
the same homotopy class. The 27r disclination lines disappear when the SmC phase
is transformed to the SmA phase. Therefore V A is obtained from Vc by identifying
on Vc all the points of a family of (0, a') loops fibring V,. In the same way, all the
elements in II,( V,) which are in the conjugacy class of (0, a'), or which pertain to the
smallest subgroup containing this conjugacy class, disappear when Vc + V,. This
subgroup consists of the elements (0, 1) and (0, a 2 ) ;it is isomorphic to 2, and constitutes
the kernel of the mapping II,( V,) + II,( V,). Hence

HI ( VA)2. (
1 vC)/z2 = 02 2 (75)
since HI( Vc)= Z 0 2,. Note further that since (0, a') = (0, a')-', an oriented loop of
this class can change orientation after each of its points has performed a closed loop
on Vc. Therefore Vc is not a direct product VAx S': rather it is a situation analogous
to that one in which one obtains a Mobius ribbon by fibring a circle (the base) by a
line segment (the fibre), letting this line segment suffer a *rr rotation along the base.
Equation (75) which states that 2, is the kernel of the isomorphism HI( Vc)+ II,( V,),
is only an element in the sequence which links the homotopy groups of the base, the
fibre and the bundle, namely
.. ~ . ..
~ ~ 2 ( s 1 ) ~ ~ 2 ( v C ) ~ ~ 2 ( v A ) ~ ~ l ( s ' ) ~ ~(76)
~ ( v C ) ~

Its property of exactness means that if three successive elements of the sequence are
considered
1 1
+A-, B - , C+ (77)
the image of the mapping i is equal to the kernel of the mapping j

im(A, i ) = kerj. (78)


Other results, with their geometrical interpretation, can be obtained from equation
(76). See KlCman and Michel (1978a, b) and KlCman (1983a ch. 3), for more details.

7.4.2. The fact that Vc is not a direct product of the base V, by the fibre S' implies
that mappings of V, onto Vc (take a point P on V,, map it on some point of the fibre
S ' ( P ) attached to P, and proceed similarly for all the points in the neighbourhood of
P in a smooth way, and so on) are not continuous. A striking physical consequence
of that fact is the following: if one starts from a SmA phase which is sufficiently full
of defects, the transformation to the SmC phase which has the same layers is attended
by the appearance of (0, a') disclination loops, along which, precisely, the mapping
V,+ V, which is achieved by the transformation is not continuous. For example, start
from a droplet in which the SmA layers are piled in concentric spheres (this is a
topologically stable point defect, belonging to 112(V,) = Z ) , and cool it to the SmC
phase; at least two (0, a') lines appear, which begin on the singular point. This
transformation, which has been observed by Kurik and Lavrentovich (1983) is reminis-
cent of the singular gauge-induced transition from a t'Hooft-Polyakov monopole
(t'Hooft 1974, Polyakov 1974) into a Dirac monopole (Dirac 1934) with a singular
string. For more on the relation between singular gauge transformations and defects,
see KlCman, in Kroner (1982).
Defects in liquid crystals 63 1

7.4.3. Mermin et a1 (1978) (see also Volovik (1978) and Mermin (1979)) have estab-
lished the classification of defects at boundaries (when some constraint acts on the
order parameter, like a director field forced to be parallel or homeotropic) or when
the order parameter depends on the length on which the medium is considered. In
both cases one has to introduce two-order parameter spaces VI and V,, say, such that
one of them is included in the other ( V2= VI ). The concept of homotopy has then to
be modified to one of relative homotopy.
Call for example V, the order parameter space restricted to the boundary (S' for
a planar nematic, a point I for a homeotropic one). V2 is included in VI, the order
parameter space of the bulk. Consider the case when the surface defect is a point,
and surround it by a loop y. Compute the homotopy classes of y in IIl(Vl) and in
IT,( V,). If the point defect is the termination of a line defect, then clearly the homotopy
classes so computed must be the same, say a. Conversely, isolated point singularities
are represented by elements b of ITl( V,) which vanish in II,( VI) when the loop y
surrounding the point of the surface is transported in a smooth manner in the bulk.
This corresponds to the operation of inclusion i of V2 in VI. This transport of y clearly
defines a homeomorphism between 111(V2) and II,( Vl); all elements b are in the kernel
of this homeomorphism. Finally, some surface point singularities can be generated
by bulk point singularities arriving on the surface; these objects are truly point
singularities and are not in ker {IT,( V,) +IT,( Vl)}, but are classified by the classes of
homotopy of half-sphere X2embedded in the bulk and whose limiting circles are on
the boundary. These are relative homoropy groups IT,( VI, V,); the limiting circle is
mapped in V,. If the image of 8,in the mapping is entirely in V,, we contract it to
a point in V,; its class of homotopy is the identity element in II,( Vl, Vz). In fact, we
have the EHS:

-2
i2 J2 32
+rI2( V2) n2(Vl)
+ + Vl, V,)+ IT,( V,) nl(VI)--+. (79)
In this sequence, 8 is an operation which consists of considering in V2 only its boundary
X2on the specimen surface. It induces a group homeomorphism between It2(VI, V,)
and IT,( V2).
Using (79), it can be proved that II,(V,, V,) is finally the product (direct or
semi-direct according to the case) of two groups of elements; the first is ker i , and
describes true singular surface points which are not terminations of lines; the second
is the quotient group ITz( Vl)/im i2 and describes singular surface points coming from
the bulk. Surface points which are terminations of lines are in im i l .
This discussion (which can be extended to any dimensionality of surface defects)
shows how complex the classification of these defects can be and how powerful the
homotopy method. Let us take the example of nematic phase with parallel boundary
conditions (Volovik 1978). One has V, = P 2 and V, = S'. Here ker il = Z ; these are
the terminations at the boundary of the disclinations of integer strength which are
non-singular in the bulk but cross the surface with a singular point. IT,( Vl)/im i2 is
equal to Z,but it also classifies the bulk point singularities, which are all topologically
stable on the surface. The rest of the surface singularities are the termination of
half-integer lines. Finally, we have
nz(P2, S' ) = zx z. (80)
A detailed study of surface defects in nematic drops with various boundary condi-
tions has been given by Volovik and Lavrentovich (1983), including the dynamics of
their transformation one into another (see also Candau et al (1973)).
632 M Kleman

Mermin (1977) has introduced the term of boojum to denote the surface sin-
gularities in the superfluid phase of 3He-A. The order parameter is an orthogonal
tripod A , , A,, I ; 1 is perpendicular to the surface. The only topologically stable surface
defects have even strength and are multiples of *47r (rotation of the Ais axes), which
limitation results from the fact that there are no topologically stable point defects in
the bulk. Therefore any boundary singular point which has an odd integral strength
(*27r, *67r, . . .) must be a pinning point of a topologically stable line in the bulk.
Eventually, any boojum ( S = 4n7r) can split into two singular points of odd strength,
linked by a singular line.
Booja in cholesterics have also been discussed (Stein er a1 1978). Assume that
the so-called cholesteric axis is perpendicular to the boundary; therefore V, = S / Z , =
SI.We have U,( V,) = II,( V,) = 1; U,( V,) = 2 ; and U,( V,) = Q, where Q is the quater-
nion group. The inclusion of V, in Vl=SU(2)/Q is not easy to figure out. As in
0 7.2.2 let us introduce I, -I, * e l , *e2 and *e3 the elements of Q. To each conjugacy
class of Q corresponds a certain type of loop included in V. Assume that the class
*e3 represents the ~ ( * f ) lines, the class * e , the A(*$) lines and the class *e2 the
T(*$) lines as in table 1. Clearly, the manifold V, is in the mapping in V, of a Burgers
circuit surrounding a x line. Its homotopy class in IT,( V,)(=Z)is *1 and is +e3 in
111(V,). If V, is surrounded twice (elements of homotopy classes *2 in IT,( V,)), its
homotopy class in V , is -I = (*e3). If it is surrounded three times ( 1 3 in U,( V,)),
its homotopy class in V, is again * e 3 . Only elements *4n in U,(V2) map on I in
II,( V,):they constitute the kernel of the inclusion i and correspond to possible booja.
Here, as above in the 3He-A phase, splittings are possible which would divide the
singular points on the surface in two singular points linked by a ( - I ) , or in four
singular points linked by ( * e , ) lines.

7.4.4. In many instances, it is believed that the singular region of the defect core is
filled with material with an order parameter more symmetric than in the bulk, liable
indeed to display the symmetry of the higher temperature phase. The situation can
be easily handled with the help of an EHS if the transition is second order, in which
case one of the order parameter spaces is included in the other, which is a fibre bundle
on the former. Similar problems happen if the order parameter depends on the scale
on which it is observed (for more details, see Klkman (1983a p 306)).

8. Defects in media with frustrated symmetries

8.1. Generalities
As already indicated, the term frustration refers to the fact that the competition
between local interactions in matter leads to structural arrangements which are intrinsi-
cally inhomogeneous, on some typical length 6 much larger than molecular or atomic
lengths. Such a situation arises in solids such as metallic glasses, Frank and Kasper
phases (1958, 1959), quasi-crystals, in liquid crystals such as blue phases, Pp,lamellar
phases, SmI films and probably also in cholesteric solutions of biological polymers.
Some biological structures in vivo might quite well benefit from the same type of concept.
Frustrated media have been approached through various points of view. The
thermodynamical viewpoint has been privileged by Hornreich er a1 (1980, 1988) in
their studies of the blue phase and of the hypothetical icosahedral chiral phase
(Hornreich and Shtrikman 1986); using a mean-field Landau approach with an order
Defects in liquid crystals 633

parameter whose amplitude varies in space, they have established the phase diagram
of the various cubic blue phases, and the rules of stability. In their theory the lines
where the quadrupolar order parameter Q vanishes are most probably lines of defects,
in the sense that the order parameter n rotates by some angle multiple of T when
circumnavigating around those lines. But the description can be made continuous
everywhere and the defects inherent to the structure are not visible at first sight. These
defects are, in contradistinction, the essential ingredients of the structural (rather
than thermodynamical) theories of Kliman and Sadoc (1979) and Rivier (1979) for
metallic glasses and Meiboom et a1 (1981,1983) for blue phases, and of the energetical
theories of Carlson et a1 (1988) for the same materials and for the ripple PpJphase
and for thin SmI films. Finally the theories which involve the construction of
unfrustrated crystals in curved spaces introduce defects which decurve those crystals
and are present in the final euclidean frustrated medium. Our review will stress these
aspects of the frustrated media with defects.
We will briefly review the essential ingredients of geometrical frustration in the
case of metallic glasses, for which the concept was first developed (K16man and Sadoc
1979). We consider a system made of equal atoms of physical diameter d, which like
to pack as densely as possible. Such a requirement is certainly achieved locally if four
spheres occupy the four vertices of a regular tetrahedron of edge a = d. Trying to fill
space with such tetrahedra, we assemble them first in bundles of 5 tetrahedra with a
common edge (but they do not fill entirely the space around the common edge; it
remains an angle of approximately 7, which can be filled only by some elastic distortion
of the tetrahedra) and then in bundles of 20 with a common vertex (figure 50), around
which the other vertices form a regular icosahedron.

Figure 50. Icosahedral close packing of slightly deformed regular tetrahedra

Such a local icosahedral arrangement has a higher density and a smaller energy
than the more classical FCC and HCP arrangements, as first shown by Frank (1952).
This explains why clusters containing a very small number of molecules possess
icosahedral symmetry. However, this is not a crystallographic symmetry; systems with
local icosahedral symmetry present everywhere are apriori disordered on long distances.
Is it possible to describe the structure of this disorder in more detail?
Icosahedral symmetry is frustrated in our usual, euclidean R 3 space; but this
frustration can be relaxed in a curved space of constant curvature, the 3 D sphere S3,
where it is possible to build a 3D crystal of finite extent since S3 is a finite space,
made of 600 regular tetrahedra, 5 meeting around each edge and 20 around each
vertex, where they form an icosahedron. In Schlafli notation, this 3 D spherical crystal
634 M Kliman

is the ( p , q, r ) polytope, with p = 3, meaning that the facets are triangles, q = 3, meaning
that 3 facets meet at a vertex (i.e. the building block is a tetrahedron) and q = 5 meaning
that 5 tetrahedra meet along an edge. This crystal is nothing other than a 3D analogue
of the 2D dodecahedron: regular pentagons do not fill the plane regularly, but build
a 2D crystal on the sphere S 2 .
The relationship between the (3,3,5) polytope and the amorphous structure can
be understood pictorially as follows: introduce disclinations of strengths which are a
multiple of 2 ~ / 5(in order to respect the (3,3,5) symmetry) along some lines. In this
process, one obtains lines of vertices with coordination number 2 = 14,15,16. . . (figure
51). Along these lines of defects, since extra matter has been introduced, the sphere
is decurved. The purpose is to flatten S 3 completely. Most of the flattening can be
achieved by lines of defects; but the strengths of the lines are prescribed, and the lines
are introduced at random; hence it is impossible to reach a perfect flattening without
straining the matter. The reader is invited to transpose this discussion to a 2D sphere
S 2 tiled with regular pentagons, in order to get a more straightforward physical feeling
for the process. Introduction of extra pentagons decurves the sphere too much and
transforms it locally to a surface of negative curvature; but the introduction of strains
(and stresses) brings the material back to zero curvature (IUCman 1983b).
Another way of understanding the relationship between the curved crystal and the
flat one is to consider a mapping between the substrate spaces which conserves lengths
and angles. Such a mapping is possible (Cartan 1928), and consists geometrically in
a rolling along a line of the curved space on the (tangent) flat one, without slipping
(KlCman 1983b). But such a process can only be achieved along the prescribed line.
The failure to a global mapping is computed by choosing for the prescribed line an
infinitesimally small loop, mapping it on the flat space, and measuring the closure
failure which consists in a translation and a rotation. It is the rotation failure (which
is proportional to the Gaussian curvature of the curved space) which is at the origin
of the disclinations we considered.
Finally, the disordered material can be described as a disordered array of lines
along any of which 2 is constant and larger than 12, embedded in a matrix 2 = 12
(see figure 51, with 2 = 14), where local order is best satisfied. The average distance
between lines, 6, depends on a complex balance between the elastic energy of the bulk
and the energy of the lines of defects, but it is expected that it scales with the radius
of s3.

I
Figure 51. A disclination through an icosahedral packing which changes its coordination
number from Z = 12 to Z = 14.
Defects in liquid crystals 635

Carlson et a1 (1989) have discussed similarities between the continuum elastic


theories of a number of frustrated phases with defects and claim that a common nature
of these theories is that the free-energy density shows up total divergence terms, whose
role is to give a negative contribution to the free energy of the defects. We shall
illustrate below, in the case of blue phases, how this property originates in the existence
of a representation in curved space of the frustration.

8.2. Frustrated liquid crystals


8.2.1. Cholesterics and blue phases. Figure 5 ( a ) is a schematic drawing of the classical
cholesteric phase; with one axis of helicity. Figure 5 ( b ) represents a cylindrical
double-twisted configuration. Any radial direction is an axis of helicity, and the
director configuration reads, in cylindrical coordinates:
n, = 0 no = -sin $( r ) n, = cos +( r ) (81)
with $(O) = 0. The integral lines of the director are along helices of chirality opposite
to the chirality of the rotation of the director about the radii. If the director rotates
clockwise about any r axis, the helices are counterclockwise. Note the analogy of this
structure with the curling mode of magneticians (Hornreich et a1 1982). Frustration,
i.e. the fact that double twist is less satisfied, increases with r. There is therefore some
radius 5 above which this geometry breaks, as will now be discussed.
The stability of the double-twist configuration can be explained phenom.enologically
by the presence of a surface term in the free energy (see equation (3)) which, in
cylindrical coordinates, reads:
K24 d
-- (cos2 *).
2r d r
This quantity integrates, between r = 0 and r = r,, to:
-~ ~ -
2 1 4 COS
( $0)

and is negative for any value of ~I/,I # 2 ~ nwhen


, K24 is positive.
Assuming K24 > 0, double-twist cylindrical geometries are therefore favoured
geometries for nucleation of the cholesteric phase in the isotropic phase, if in particular
the molecules tend to be parallel to the isotropic-cholesteric interface, and if K1 is
large compared with K,, since the double-twist cylindrical geometry is splayless
(div n = 0). But this configuration has necessarily a limited radial extent 5, since the
gain in K24 energy is counterbalanced by the bend term K 3 , which soon becomes
prohibitive when r increases. These arguments can explain well the nucleation of the
cholesteric phase in the form of spherulites with either a diametral disclination (which
has the topology of the double-twist cylinder) or a radial disclination (which has the
topology of two double-twist cylinders side by side). Models of such spherulites have
been given by Frank and Pryce (in Robinson (1961)) for cholesteric polypeptide
solutions. The cholesteric defect structure near the smectic A transition also contains
double-twist defects, according to Cladis et a1 (1979).
When K,, is large enough (of the order of K , , K2, K3) new phases with thermody-
namically stable double-twist textures might appear. These are the blue phases first
described by Saupe (1969). They exist in a very small range of temperature between
the isotropic and the cholesteric phase. There are three such phases; BP I is in contact
with the cholesteric phase, BP I1 is in the intermediate temperature domain. BP I and
636 M Kliman

BP I1 are cubic phases; they have been observed in freeze fracture experiments (Costello
et a1 1984); their structures seem to be quite well understood (Meiboom et a1 1983);
the building blocks are finite cylinders of double twist, which can arrange locally as
in figure 5( c ) , three of them along the quaternary axes of the cube. If their lateral size
is limited to $,, = 7r/4, they fit nicely along their contacts, but the intermediary region
along the three-fold axis cannot be filled smoothly with molecular directions: a
disclination of strength S = -4 appears in this region. Meiboom et a1 (1981) proved
that the core energy of such a defect is negative if K24 is positive, and explained in
this way the stability of the blue phases. BP 111, also called blue fog (Marcus 1981),
is seemingly a disordered phase; it has been suggested that BP I11 is a disordered array
of soliton lines with local cylindrical double twist along them, with a characteristic
size 6. If this is the case, BP 111 would be a pretransitional region rather than a true
thermodynamic phase. More recently, it has been argued (Hornreich et a1 1986,
Rokhsar and Sethna 1986) on the basis of Landau theory that B P I I I could be, but
without compelling evidence, a thermodynamically stable phase with long-range quasi-
crystalline ordering and icosahedral symmetry.
In this section, we have discussed the stability of the double-twist unfrustrated
geometry in terms of KZ4. If K24 is very large and positive, we expect some domain
of stability for the blue phases. If K24 is not too large, cylinders of double twist cannot
belong to the ground state and are excitations of the cholesteric phase. In terms of
Landau theory these results are discussed as a function of the ratio K = [ / p where 6
is a coherence length and p the pitch. 6, according to Hornreich et a1 (1982), is the
racemic correlation length at the (virtual) isotropic-nematic transition. K is therefore
a dimensionless measure of the tendency to parallel alignment, as opposed to the
helical alignmeht. Our discussion of the polymeric chiral phases will allow us to
deepen this picture. Landau theory tells us that the larger K , the larger the temperature
range of stability of the blue phases, as shown in figure 52. Therefore, in Landau
theory terms, double twist and parallel alignment are already related in some way.

K =</p

Figure 52. The stability diagram of blue phases in function of K =(/p.

8.2.2. Polymeric chiral phases and the chromosome of dinoflagellates: experimental results
and their interpretation. We first mention a few experimental results which point to
the presence of double twist in solutions of polymeric chiral molecules of biological
origin.
(i) Observations of biological cholesterics in vitro. We have already cited the careful
optical observations of Livolant (0 6.7.3). There is little doubt of the presence of double
twist, either of the type discussed above, or of the type described below for the
chromosome of Prorocentrum micans.
Defects in liquid crystals 637

Note also that some spherulites of collagen display a circular line around which
there is local cylindrical double-twist (Giraud-Guille 1988).
(ii) Optical and electron microscopy observations (Livolant and Bouligand 1980) of
the dinoflagellate chromosome (Prorocentrum micans). In this chromosome, D N A takes
a cholesteric arrangement, with the cholesteric axis globally along the axis of the
chromosome. However, one observes also on the periphery a double-helical furrow
with the same helicity as the cholesteric DNA,which indicates the presence of dis-
crepancies with respect to a perfect cholesteric stacking. These discrepancies have
been analysed by KlCman (1985b) using as a starting point the assumption that locally
the double helices of D N A like to be equidistant (see below). The geometry of the
chromosome can indeed be described as follows: there is a central parent surface
(figure 53) built from rectilinear D N A strands which rotate helically about the long
axis of the chromosome. This surface is a helicoid, and can also be described as a
twisted sheet of constant thickness (the thickness being the width of the D N A double
helix with its attending proteins). Place on this twisted sheet, on both sides, another
twisted sheet of the same thickness in contact with the parent surface, made of D N A
helices which have been rotated by some amount about the normal to the parent
surface. Double twist is perfectly satisfied in this region. Continue this process by
successive dispositions of parallel twisted sheets, with helical twist of the D N A about
their common normal. One obtains a double-twist geometry, whose geometrical size
is limited by the focal surface of the common normals to the sheet: in fact, this focal
surface extends outwards along two focal sheets (which are limited by two cuspidal
edges which are precisely the two observed furrows) which are not physically realised;
the geometry is limited grossly by the cylinder which carries the furrows. Beyond,
double twist would decrease and the frustration would be too large. Note the similarity
of this geometry with that one discussed for the screw dislocations of huge Burgers
vectors in SmAs (see figure 15). The furrows are the analogues of the disclination
lines into which the screw dislocation is split.

Figure 53. The parent surface of the model of the dinoflagellate chromosome. It is a
helicoid, ruled with rectilinear DNA strands. The same parent surface can be used for
globular proteins, with rectilinear P-helices (twisted P-sheets).

The double-twist geometry we have just described is, in some sense, orthogonal
to the double-twist cylindrical geometry observed with short chiral molecules. While
this latter one is favoured if K1> K 3 (since div n = 0), the chromosome geometry is
favoured when K , < K 3 (however, n. curl n does not vanish everywhere). The reason
why in the chromosome there is an effective K1 which is small is not straightforward:
K 1 should be infinite in principle in a polymeric liquid crystal with infinitely long
molecules. However, one can argue that many hair pins (figure 41) exist along the
DNA double helix; these hair pins can be favoured by some interaction with the
neighbouring proteins, which leads to local complex structures having this geometry.
A similar geometry of the chromosome has been proposed by Friedel (1984) indepen-
dently, starting from a different point of view.
638 M Kle'man

(iii) Equidistance of molecules in polymeric cholesteric phases: superstructures. The


double-twist structures just described are ways of achieving a compact stacking of
chiral objects at a local scale. In fact, the molecules are quasi-parallel. This tendency
to compact parallel stacking reflects also in x-ray diffraction patterns: local hexagonal
order has been demonstrated in cholesteric phase solutions of PBLG and other biological
polymers like DNA (Saludjian et a1 1967).
Hexagonal order and cholesteric order are incompatible, but this frustration can
be relieved on a local scale: a very small modification of figure 53 would make it a
helical stacking of small hexagonal fibre domains: in fact, it would be enough that
what we describe as DNA molecules in this figure be already hexagonal packings of
such molecules, at a microscopic scale. This would be a double-twist chiral superstruc-
ture of hexagonal structures. The same phenomenon has been observed at much higher
scales: very concentrated solutions of biological polymers are hexagonal, not
cholesteric, according to x-ray diffusion patterns (Samulski et a1 1975); one observes
also in the optical microscope typical hexagonal textures, in the form of hexagonal
domains, but, at higher scales, these domains often like to pack helically (Livolant
1984, Livolant and Bouligand 1986).
More generally, we expect that a structure which minimises energy at some local
scale 6 can show up some kind of hierarchy of structures at higher and higher scales.

8.2.3. Polymeric chiral phases and the chromosome of dinojlagellates: identity of double-
twisted local order and equidistance of chains in polymers. In fact, double-twist and
compact stacking are of the same nature. This is shown when considering the geometry
of a-helices at a scale of a few strands. These strands assemble to form a string as in
figure 54 where we have represented the case of seven strands: it is clear that such a
geometry is a geometry of close contact for molecules which are not straight. They
are not parallel, but they are equidistant. Clearly also, the number of neighbours z of
the central molecule depends on the degree of torsion; this number tends to z = 6 when
the molecules become parallel (hexagonal stacking) and there is complete continuity
between double-twist (equidistance) and hexagonal order (parallelism).
These considerations shed a new light on the stability theory of the blue phases.
Figure 52 tells us that BPS are stable when 6 is large compared with p. 6 must in the
present context be understood as the length on which double twist, but also equidistance
of the integral lines of the director n ( r ) , is not frustrated. Therefore the integral lines
of n for short molecules appear here as physical objects.

Figure 54. A string made of helicoidal strands.


Defects in liquid crystals 639

8.2.4. Chirality in layers. The transition from the SmC" to the SmI" phase of the chiral
thermotropic liquid crystal HOBACPC (hexyloxybenzylidene p'-amino-2-chloropropyl
cinnamate) shows up, in thin films, the formation of droplets with 'booja', as in figure
5 5 , where the director geometry is pictured. According to Langer et a1 (1986), this
phenomenon is because the unfrustrated state has constant curl
curl c = q (82)
where c is a unit vector which is the projection of the director on the layer, and q the
pitch. Such a state cannot exist in a two-dimensional planar surface. Now the total
divergence term -2q curl c in the free energy amounts to a surface term -2q c dl 5
which is largest and negative when c lies parallel to the boundary of the droplet and
points counterclockwise (if q > 0), The addition of these two effects explains why the
boundary of the droplet is singular (.c jumps by an angle 7~ from one droplet to the
next adjacent) and possesses a point defect (a boojum) where the boundary conditions
are badly satisfied.

Figure55. The stable vector field of a tilted SmI* thin film. The vector field is the projection
on the layer of the director field (redrawn from Carlson et al (1988)).

The SmI" film is a very remarkable example of a chiral situation without twist: it
is of course the chirality of the molecules which imposes the total divergence term
-2q curl c ; however, there is no axis of chirality in the layer. But one might expect
that other cases exist where there is some coupling between the surface curvature and
the chirality. For example if a surface takes the shape of a ruled helicoid, the set of
directions defined by the straight lines which generate the helicoid have a definite
chirality, either left or right, according to the chirality of the helicoid. We have met
such a geometry in the model of the chromosome and when discussing screw disloca-
tions with huge Burgers' vectors in a SmA (see figures 15 and 53). Now the same
geometry can be imposed by the chirality of long polymeric molecules, as it does occur
in the twisted p sheets. The polymeric molecules are polypeptidic chains, and their
chirality yields a left twist for the surface. If such surfaces have to pack parallel to
each other, as happens in some models of proteins (see Lotz et al(l982) for conforma-
tional models of silk and keratin, and Salemme (1983)) frustration should become
intolerable at some distance, as in the geometry of the screw dislocation, and a helical
defect of the disclination type appears and bring a natural limit to the stacking.
640 M Kle'man

It has been proposed recently (Bouligand and KlCman 1989) that the stability of
the M2 phase of the chiral compound MHTAC (1-(methyl)-heptyl terephthalidene bis
amino cinnamate, see Levelut et a1 (1983) for phase diagram and x-ray studies) is also
caused by a frustrated coupling between the chirality of the molecule and the layers.
The M2 phase exists, as the blue phase, in a very small interval of temperature just
below the isotropic phase, and above the so-called M I , which is akin to a SmC* phase,
but to some supplementary anisotropy. The MI is quadratic and can be understood
as a SmD phase (Diele et a1 1972) whose cubic symmetry has been broken by the
frustration effect just mentioned.

8.2.5. Smectic ripple phase. This phase was first observed by Sackman et a1 (1980) with
freeze fracture observations, and consists in corrugated bilayers of surfactant in water,
so that there is a one-dimensional periodic density modulation of the chains. In DPCC
(dipalmitoylphosphatidylcholine)the wavelength of this corrugation is 5 = 130 A; the
frustration comes from the fact that the cross-sectional area of the polar heads is
greater than that of the associated chain; this leads to the presence of frustrated regions
in the form of gaps in the packing of the chains (figure 56). Note that this is not the
only geometrical possibility; for example, the competition between the packing of the
chains and that of the heads can be solved by the formation of spherical micelles, or
curved monolayers. Which solution is chosen depends on the material constants of
the surfactant layer, and a small K1 favours the formation of micelles, while a large
one favours planar layers as in Po,.

Figure 56. Periodic defects (gaps) in the packing of the chains in a Pp. phase (redrawn
from Carlson el al (1988)).

The theory of the ripple phase has been developed by Carlson and Sethna (1989);
in the continuous limit, it is a frustrated q'theory, with the presence of a total divergence
term in the free energy. The gaps are the defects required to relieve the frustration.
Other smectic systems with typical frustration effects have been reported, like the
bi-dimensional fluid antiphase SA(freeze fracture observations by Sigaux et a1 (1984)),
and more generally the incommensurate smectic phases with two periodicities (see
Prost (1984) for a review on Landau theory, and Satna et a1 (1985) for optical
observations).
The one-dimensional modulations of the aliphatic chains along the columns
observed in columnar systems (0 5 ) are 1D examples of a similar type of frustration.

8.2.6. Phases of interfaces. The question of the stability of phases of interfaces (colloids,
micelles, hexagonal and cubic phases of surfactants in water and block copolymers,
etc) and of isolated membranes is today a very active field of research, and it is not
yet really known whether the concept of frustration is totally relevant. This type of
explanation has, however, been advanced in the case of the cubic Qa phases (Sadoc
and Charvolin 1986, Charvolin and Sadoc 1987); in such an instance their stability
would result from a conflict between the tendency of the layers to keep as compact as
possible (this implies, in the case of surfactants in water, that opposite monolayers
Defects in liquid crystals 641

are parallel in the bilayers they form, i.e. there is a tendency to the formation of bilayers
separated by water), and the tendency of each monolayer to curve in opposite directions
in order to relax the difference in cross section between the hydrophobic chains and
the hydrophilic heads. The resulting local geometry of the bilayer, in the unfrustrated
zones (where the conflict is best solved), is saddle-shaped.
It has been argued (Klbman 1988) that such a model should apply most probably
when the rigidity K , of the layers is small (comparable with kBT or smaller). In such
a case configurational entropy of the molecules, in particular of the chains, should be
taken into account. At high temperature this entropy can be large enough to stabilise
the L, phase versus the unfrustrated Qmphase, which will show up only at a lower
temperature. It is then easy to show that, for reasons of isotropy, the mean curvature
of the middle surface
H=u,+u2 (83)
between the two monolayers in conflict should vanish in first approximation. This
middle surface is then a minimal surface. Since there exist minimal surfaces which
extend through euclidean space and form periodic lattices, the unfrustrated solution
is a periodic lattice, of one of the types described by Luzzati er al (Luzzati and Spegt
1967, Luzzati et a1 1987). A systematic approach to the geometry of periodic minimal
surfaces has been given by Hyde et a1 (1984) and Andersson et a1 (1984).
The case above ( K , small) might be relevant to non-ionic surfactants, in which the
Q, phase is of small extent or even non-existant, according to the ratio of the lengths
of the hydrophilic and hydrophobic parts. In fact, since the two monolayers on both
sides of the minimal surface cannot be at constant distance through the whole crystal,
one has to take into account the elastic contribution of the variation of the thickness
of the bilayers, which is large since K 1is small. This might also explain the discrepancies
with perfect minimal surfaces.
In fact, the Q, phase can also be stabilised, as first noticed by Helfrich (1981) (see
also Helfrich (1986)), by the existence of a negative K24 saddle-splay term in the free
energy of the bilayers ( a contribution G, where G = u1u2is the Gaussian cur-
vature). K24 is probably positive (but small, comparable with K , ) in non-ionic surfac-
tants. It is large and negative in lecithin and other systems which show up Quphases.
Then it is not necessary to invoke any frustration phenomenon of the form above. The
formation of a topology like that of the Q, phase, which differs from the topology of
the L , or H , phases, is favoured; the integrated energy -K24 1G d V depends indeed
only on the topology, since it is a quantity which is a line integral on each layer. The
importance of this term has been discussed in various cases, not only cubic phases
but also micellar solutions, colloids, microemulsions, etc (see papers in Physics of
Amphiphilic Layers, Springer Proc. in Physics vol. 21 (1987)).

8.3. Curved space representation of double twist


8.3.1. Short molecules and blue phases. Sethna et a1 (1983) have proposed to relieve
the cholesteric frustration of blue phases by introducing a spherical curved space S 3
decorated with a field of unit vectors. More specifically, if S 3 is represented by the
analytic expression:
x ~ + x ~ + x ~ + xR2
:= (84)
the field of 3D unit vectors:
n ( r ) = 1/ R ( - X I , xo, x3 - x2 1
3 (85)
642 M Kle'man

satisfies double twist and is homogeneous throughout space S 3 ; any field obtained by
applying a constant rotation to all n ( r ) of equation (85) is another homogeneous field
and relieves also frustration everywhere, i.e. is homogeneously double twisted, with
pitch p* = 2n-R. A different choice for the n s , like:
n ( r ) = ( 1 / R ) (- X I 9 xo 9 - x3 9 x2 1 (86)
cannot be obtained by rotating the field of equation (85) and has indeed an opposite
pitch p * = -2n-R.
As noticed by KlCman (1985a) the integral lines of (85) or (86) are equidistant
great circles of S 3 (also called Clifford parallels) which are in mutual skew positions.
The integral lines of (85) form a congruence of right Clifford parallels ( p * > 0), while
those of (86) form a congruence of left Clifford parallels ( p * < O ) . Therefore they
achieve homogeneously the double-twist geometry which we discussed in 0 8.2.2. Since
all the great circles are equivalent in S3, now any great circle can be considered as the
axial line of the double twist geometry. Figure 57 shows two neighbouring Clifford
parallels.

Figure 57. Two close Clifford parallels are equidistant and at a constant angle from each
other. They map on the sphere of figure 58 on two points at an angular distance 28.

Since R is the size on which double twist is unfrustrated, it is convenient to interpret


the radius R of S 3 not as a quantity which scales like the pitch of the cholesteric, but
as a quantity which scales like 6, the correlation length.
In order to find the free energy of the director field or S 3 , we have to compute the
total derivatives of n. This is easily done if one uses the method of the moving frame.
We choose as a moving frame the tripod e , , e 2 , n, with e , , e2 rotating about n with
the pitch p * = 27r/q*. Such frames follow Clifford parallels; when n moves along a
given Clifford parallel, e, and e2 point always towards other Clifford parallels which
Defects in liquid crystals 643

are conserved in the motion. Bring now a point M of S3 to the origin of the euclidean
space E 3 in which we map the crystal which is homogeneously double twisted in S 3 ,
and let S 3 be tangent to E 3 in M. Then e:, e; and no are in E 3 and can be chosen as
the unit vectors of a reference frame in E 3 . It is then easy to see that the derivative
din, of no in S3 writes in this frame
d.n.
1
I
= n1.1. .- q E .Ilk. nk (87)
where ni,j= dnj/axi is the derivative of n in E 3 . We therefore infer that on S 3 the
free-energy density, computed in the tangent space E 3 , gives
K
pf =- ( n] , .I .- E i,kqnk)2
..

in the case of isotropic elasticity.


Equation (87) could have been derived straightforwardly by using the Riemannian
connection rVk = - q E t r k of S 3 (Sethna et al 1983). The advantage of using the method
of the moving frame
d,v= d,(v,e, ) = (&v, )e, + q(&e, 1 (89)
is to stress the nature of the mapping between the unfrustrated crystal in S3 and the
physical space E 3 . This mapping is obtained locally, as pointed out by KlCman and
Sadoc (1979) in the general case of a curved crystal, by letting S3 roll on E 3 without
sliding. Such a mapping (Cartan 1928) conserves locally lengths and angles, but cannot
be extended to all points of S 3 as a one-to-one mapping. This is why one has to
introduce disclihations, which relax the closure failure of a closed line in S 3 mapped
this way on E 3 .
The presence of a total divergence term is visible in equation (88), which gcves
K
pf = 2 [(Vn)+2qn curl n + q div(n div n + n x curl n)]. (90)

However, there is no divergence term in curved space, where pf = (K/2)(dn)2. It is


therefore the frustration itself of the curved crystal in euclidian space which is at the
origin of this phenomenon.
Note also that the free-energy density of (90) carries an excitation mode which has
no equivalent in usual elasticity, and which is obtained by letting the curvature q of
the curved space fluctuate. Physically a local variation 6q of the inverse pitch is liable
to be at the origin of a local helicoidal instability in certain cases, as in the Lequeux
experiment (1986). This is exactly what happens in S3, where it is easy to show that
the Clifford parallels acquire a supplementary longitudinal twist with wavevector
k3 = hl.
The helicoidal instability appears therefore as a localised version, in physical space,
of a global instability in the curved crystal (Lequeux and KlCman 1988).

8.3.2. Some standard results in spherical geometry. KlCman (1985a) has described in
more detail the geometry of the great circles of S 3 . Standard references for spherical
geometry (and elliptic geometry, which is obtained by identifying diametrically opposite
points on S 3 ) are Coxeter (1968) and Sommerville (1958). We stress here some results
that have since been used by other authors in the context of various frustrated media
(Nicolis et al 1986, Pansu et a1 1986).
644 M Kle'man

(i) A euclidean 3D space E 3 can be fibred by the parallel lines which are orthogonal
to a given plane E 2 ; similarly, a spherical 3D space S 3 can be fibred by a set of
equidistant great circles which are above a 2D great sphere S2. This is once more the
celebrated Hopf theorem, which more precisely states that any circle belonging to a
set of Clifford parallels which fibre S 3 , either left or right, can be mapped in a unique
way on a great sphere S2. We have used the Hopf theorem above ( 7.3) in another
context.
The distance between two Clifford parallels (both left or both right) can be measured
directly on S 2 . We have
d=R0
where 0 is half the spherical angle between the two points on which the two Clifford
parallels map on S 2 (figure 58).
(ii) Two Clifford parallels which map diametrically opposite points on S 2 are
conjugate in S 3 and at a distance ( r R ) / 2 . The small circles in figure 58 are mappings
of Clifford parallels equidistant to 1 and L, which cover embedded tori (Clifford
surfaces).

L
Figure 58. Mapping on S 2 of Clifford parallels in S 3 . Each point of the sphere S 2 represents
a Clifford parallel (either right or left). Each small circle represents a Clifford surface.
All the Clifford surfaces of this set of small circles are equidistant to the Clifford parallels
1 and L.

8.3.3. Polymeric chiral molecules: crystallography of strings with strands. These


geometrical properties of Clifford parallels on S3 suggest new considerations on the
local structure of molecules (or strands) in contact. Call do the diameter of a molecule
and let us assume that the molecules are in contact and homogeneously arranged in
S3, along Clifford parallels. Then their mapping on S2 must cover S2 regularly. The
centres of the sections of diameter do are therefore distributed regularly, and there are
only a limited number of cases; the condition of contact requires that the centres are
at the vertices of triangles. Table 3 summarises all the possibilities; z is the number
of neighbours; d 0 / 2 R is the angular width of the molecular section on S,; p is the
density and N is the total number of molecules. In real space, 6 scales like doN. In
the case N = 12, the centres of the molecular sections are at the vertices of a regular
icosahedron on Sz. Similarly N = 3,4,6, represent regular deltahedra on S2.
Note that the limit case z = 6 can be achieved only with R =CO; it is the 2D usual
hexagonal packing. Table 3 stresses also the fact that the problem of close packing
of strands appears as a true 2D problem on a sphere S2.
Defects in liquid crystals 645

Table 3.

Z d,/2R P N

2 0.1666 0.75 3
3 0.1520 0.8453 4
4 0.125 0.8787 6
5 0.0881 0.8961 12
6 0 0.969 m

If we do not restrict ourselves to regular packing, we are led to consider any


arrangement of points on S2. This question has some interest for large packing densities
with 5 < 5 < 6. P is now necessarily an average, since we consider non-regular packings.
The condition of contact between strands is now replaced by the following: given N
points on a sphere, to find the arrangement of these points such that the latest distance
between points be as great as possible. This is a standard problem in palynology (holes
on a pollen grain, through which fertilisation is achieved via the stamen, have to be
as largely dispersed as possible, for a given number of them), and is known as the
Tammes problem (see Fejes Toth 1964). The solutions to this problem are known for
a few values of N ; for each solution there is a specified ratio d o / R , where do is the
least distance between points, and an average number of neighbours z.
As a conclusion to this section, let us stress again that the essential ingredients to
describe close packing of equal strands is either their density, their average number
of neighbours or their local double twist, p*. All these quantities are, according to
this analysis, equivalent.

8.3.4. Some speculations on polymeric systems. The discussion we have so far essentially
concerns chiral molecules. However, there is no conceptual difficulty to extend it to
non-chiral molecules, and in particular molten polymers. We have indeed introduced
two lengths which are of a quite different nature: (i) p* = 6 is a correlation length for
local twist and the fact that the molecules like to be locally twisted does not imply
that they are twisted on long distances; we suggest that some amorphous polymeric
systems might enter this category: their local twist is caused by entropy forces, while
there are forces of contact of van der Waals origin, mostly. The fact that the sign of
p* is not fixed for racemic molecules does not preclude it being a relevant quantity.
(ii) p , the pitch of the cholesteric phase. These two lengths are of course insufficient
to describe the whole family of relevant structural properties; in particular, 5 has to
be larger than the persistence length in order to make our picture coherent. We suggest
the following classification of the structural properties as a function of K = &/p.
Large K : the general trend for large K should be towards disorder of the BP I11
type of disordered polymers. The twisted long-range structure of hexagonal PBLG
alluded to also enter into this category. The effective peffwhich is measured has of
course no relationship with p ; it is imposed by the twisting properties of the hexagonal
domains which become the building entities of the cholesteric order.
Intermediate K : we have double twist as in blue phases BP I and BP 11. The picture
we have given for the chromosome of dinoflagellates also enters this range of values.
Choice between the two double-twist structures depends on other parameters, like K1
or K 3 .
Small K : this is the cholesteric phase.
Very small K : this is the cpxe of nematic order, including biaxial nematic order.
646 M Kliman

References
Alderman N J and Mackley M R 1985 Faraday Discuss. Chem. Soc. 79 149
Allain M 1985 J. Physique 46 225
Allain M and KlCman M 1987 J. Physique 48 1799
Allet C, KlCman M and Vidal P 1978 J. Physique 39 181
Anderson P W and Toulouse G 1977 Phys. Rev. Lett. 38 508
Andersson S, Hyde S T and von Schnering H G 1984 Z. Krist. 168 1
Anisimov S I and Dzyaloshinskii I E 1973 Sou. Phys.-JETP 36 774
Arpin M, Strazielle C and Skoulios A 1977 J. Physique 38 307
Asada T 1982 Polymer Liquid Crystals ed. A Ciferri et al (New York: Academic) p 247
Asher S A and Pershan P S 1979 Biophys. J. 27 393
Atkins E D T, Fulton W S and Miles M J 1980 5fh Int. TAPPI Con$ (Vienna) p 208
Bailin D and Love A 1978 J. Phys. A : Math. Gen. 11 219, 821, 2149
Ballet F, Leong Y S, Wittman J C and Candau F 1981 Mol. Cryst. Liq. Cryst. Lett. 64 305
Barbet-Massin R, Cladis P E and Pieranski P 1984 La Recherche 154 548
Bartolino R and Durand G 1977a Mol. Cryst. Liq. Cryst. 40 117
-1977b Phys. Rev. Lett. 39 1346
-1978 Ann. Phys., Paris 3 257
-1983 J. Physique Left. 44 L79
Ben-Abraham S I 1985 Mol. Cryst. Liq. Cryst. 123 77
Ben-Abraham S I and Oswald P 1983 Mol. Cryst. Liq. Cryst. 94 383
Benton W J and Miller C A 1983 Prog. Colloid Polym. Sei. 68 71
Bema1 J D and Fankuchen L 1941 J. Gen. Physiol. 25 111
Bidaux R, Boccara N, Sarma G, de Seze L, de Gennes P G and Parodi 0 1973 J. Physique 34 661
Bilby B A 1960 Prog. Sol. Mech. 1 329
Bilby B A, Bullough R and Smith E 1955 Proc. R. Soc. A 231 263
Billard J 1980 Liquid Crystals of One- and Two-Dimensional Order (Springer Series in Chemical Physics vol.
11) ed. W Helfrich and G Heppke p 5
Birgeneau R J and Litster J D 1978 J. Physique Lett. 39 L339
Blaha P 1976 Phys. Lett. 36 784
ten Bosch A, Maissa P and Sixou P 1983 J. Chem. Phys. 79 3462
Bouligand Y 1969 J. Physique 30 C4-90
-1972a J. Physique 33 525
- 1912b J. Physique 33 715
- 1973 J. Physique 34 603, 1011
-1974 J. Physique 35 215, 959
-1980 J. Physique 41 1297
- 1981 Physics of Defects ed. R Balian et al (Amsterdam: North-Holland) p 665
Bouligand Y, Derrida B, Poenaru V, Pomeau Y and Toulouse G 1978 J. Physique 39 863
Bouligand Y and KlCman M 1970 J. Physique 31 1041
- 1979 J. Physique 40 79
-1986 in preparation
Bouligand Y and Livolant F 1984 J. Physique 45 1899
Bourdon L 1980 ?%&e University of Orsay
Bourdon L, KlCman M, Lejcek L and Taupin D 1981 J. Physique 42 261
Bourdon L, Sommeria J and KlCman M 1982 J. Physique 43 77
Bragg W 1933 Trans. Faraday Soc. 29 1056
Brand H R and Pleiner H 1982 Phys. Reu. A26 1783
- 1985 J. Physique Lett. 46 L-711
Brezis H 1989 Parfial Diferential Equations and Continuous Phase Transitions ed. M Rascle (Berlin: Springer)
in the press
Brunet-Germain M and Williams C E 1978 Ann. Phys., Paris 3 237
Cagnon M and Durand G 1980 Phys. Rev. Lett. 45 1418
- 1981 J. Physique Lett. 42 L451
Cagnon M, Gharbia M and Durand G 1984 Phys. Rev. Lett. 53 938
Cagnon M, Palierne J F and Durand G 1982 Mol. Cryst. Liq. Cryst. Lett. 82 185
Defects in liquid crystals 647

Candau F, Ballet F, Debeauvais F and Wittman J C 1982 J. Colloid Interface Sci. 87 356
Candau F, Le Roy P and Debeauvais F 1973 Mol. Cryst. Liq. Cryst. 23 283
Cano R 1968 Bull, Soc. Fr. Miner. Crist. 91 20
Carlson J M, Langer S A and Sethna J P 1988 Europhys. Lett. 5 327
Carlson J M and Sethna J P 1989 Phys. Rev. A in the press
Cartan E 1928 Lefons sur la GComPtrie des Espaces de Riemann (Paris: Gauthier-Villars)
Casagrande C, Fabre P, Guedeau M A and Veyssie M 1987 Europhys. Lett. 3 73
Chan W K and Webb W W 1981a J. Physique 42 1007
-1981b Phys. Rev. Lett. 46 39
Chandrasekhar S 1980 Liquid Crystals, Cambridge Monographs on Physics (Cambridge: Cambridge University
Press)
Chandrasekhar S and Ranganath G S 1986 Ado. Phy. 35 507
Chandrasekhar S, Sadashiva B K, Ratna B R, and Raja V N 1988 Pramana 30 U 9 1
Charvolin J and Sadoc J F 1987 J. Physique 48 1559
Chistyakov J G 1963 Sou. Phys.-Crystallogr. 7 619
Cifferi A, Krigbaum R and Meyer R B (ed.) 1982 Polymer Liquid Crystals (New York: Academic)
Cladis P E 1974 Phil. Mag. 29 641
- 1987 I M A vol. 5, Theory and Applications of Liquid Crystals (Berlin: Springer) p 73
Cladis P E, Brand H R and Finn P L 1983 Phys. Rev. A 28 512
Cladis P E, Couder Y and Brandt H R 1985 Phys. Rev. Lett. 55 2945
Cladis P E and KlCman M 1972 J. Physique 33 591
Cladis P E and Torza S 1975 Phys. Rev. Lett. 35 1283
- 1976 Colloid Interface Sci. ConJ: Roc. 4 487
Cladis P E, van Saarloos W, Finn P L and Kortan A R 1987 Phys. Rev. Lett. 58 222
Cladis P E and White A E 1976 J. Appl. Phys. 47 1256
Cladis P E, White A E and Brinkman W F 1979 J. Physique 40 325
Clark N A 1976 Phys. Rev. A 14 1551
-1978 Phys. Rev. Lett. 40 1663
Clark N A and Hurd A J 1982 J. Physique 43 1159
Clark N A, Handschy M A and Lagerwall S T 1983 Mol. Cryst. Liq. Cryst. 94 213
Collett J, Sorensen L B, Pershan P S , Litster J D, Birgeneau R J and Als-Nielsen J 1982 Phys. Reo. Lett. 49
553
Colliex C, Veyssie M and KlCman M 1976 Advances in Chemistry vol. 152 ed. S Friberg (Cleveland, OH:
American Chemical Society) p 71
Costello M J, Meiboom S and Sammon M 1984 Phys. Rev. A 29 2957
Coxeter H S M 1968 Non-Euclidean Geometry (Toronto: University of Toronto Press)
Darboux G 1954 The'orie Ge'ne'rale des Surfaces (New York: Chelsea)
Delaye M, Ribotta R and Durand G 1973 Phys. Lett. 44 139
Demus D 1975 Krist. Tech. 10 933
Demus D and Richter L 1978 Textures of Liquid Crystals (Leipzig: VEB Deutscher)
Demus D, Schiller P and Sharma N K 1984 Cryst. Res. Technol. 19 577
Diele S, Brand P and Sackmann H 1972 Mol. Cryst. Liq. Cryst. 17 163
Dierker S B, Pindak R and Meyer R B 1986 Phys. Rev. Lett. 56 1819
Dirac P A M 1934 Proc. R. Soc. A 133 60
Donald A M, Viney C and Ritter A P 1986 Liq. Cryst. 1 287
Donald A M and Windle 0 H 1984a Polymer 25 1235
-1984b J. Mater. Sci. 19 2085
Doring W 1968 J. Appl. Phys. 39 1006
Dubois-Violette E, Guazelli E and Prost J 1983 Phil. Mag. 48 727
Duke R W and Chapoy L L 1976 Rheol. Acta 15 548
Durand G 1983 J. Chimie Phys. 80 119
Durand G, Bartolino R and Cagnon M 1980 Liquid Crystals of One- and Two-Dimensional Order (Springer
Series in Chemical Physics vol. 11) ed. W Helfrich and G Heppke (Berlin: Springer)
Dzyaloshinskii I E 1970 Sou. Phys.-JETP 31 773
Eshelby J D 1980 Phil. Mag. A 42 359
Fabre P, Casagrande C, Veyssie M and Finkelmann H 1984 Phys. Rev. Lett. 53 993
Fan C 1971 Phys. Lett. A 34 335
648 M Klkman

Fayolle B, Noel C and Billard J 1979 J. Physique 40 C3 485


Fejes Toth L 1964 Regular Figures (Oxford: Pergamon)
Feldtkeller E 1964 2.Ang. Phys. 17 121
-1965 2. Ang. Phys. 19 530
Figueiredo Net0 A M, Galerne Y, Levelut A M and Liebert L 1985a J. Physique Lett. 46 L499
Figueiredo Net0 A M, Galerne Y and Liebert L 1985b J. Phys. Chem. 89 3939
Figueiredo Net0 A M, Liebert L and Galerne Y 1985c J. Phys. Chem. 89 3737
Figueiredo Net0 A M, Liebert L and Levelut A M 1984 J. Physique 45 1505
Filas R W 1978 J. Physique 39 49
Fousek J and Lejcek L 1986 Proc. 6th IEEE Int. Symp. Applications ofFerroelectrics (Bethlehem, USA) p 199
Fraden S, Hurd A J, Meyer R B, Cahoon M and Caspar D L D 1985 J Physique 46 C3 85
Frank F C 1952 Proc. R. Soc. 215 43
-1958 Faraday Disc. Chem. Soc. 25 1
-1979 Faraday Disc. Chem. Soc. 68 1
_. 1983 Phil. Trans. R. Soc. A 309 71
Frank F C 1982 private communication
Frank F C and Chandrasekhar 1980 J. Physique 41 1285
Frank F C and Kasper J 1958 Acta Crystallogr. 11 184
-1959 Acta Crysrallogr. 12 483
Friedel G 1922 Ann. Phys. Paris 18 273
-1925 C.R. Acad. Sci., Paris 180 269
-1927 C.R. Acad. Sci., Paris 185 330
Friedel G and Grandjean F 1910 Bull. Soc. Fr. Miner. Cryst. 192 409
Friedel J 1964 Dislocations (Oxford: Pergamon)
-1975 J. Physique 36 C1 170
-1979 J. Physique Coll. 40 C3 45
-1984 Proc. 6th Gen. Con$ Eur. Phys. Soc. (Prague) ed. J Santa and J PantoflicCk
-1988 Private communication
Friedel J and KlCman M 1969 J. Physique Coll. 30 C4 43
-1970 NBS Special Publication 317 1
Galerne Y 1989 Mol. Cryst. Liq. Cryst. in the press
Galerne Y , Figueiro Net0 A M and Liebert L 1986 J. Physique 84 1732
Galerne Y , Itova J and Liebert L 1988 J. Physique 49 681
Galerne Y and Liebert L 1985 Phys. Rev. Lett. 55 2449
Gallani J L and Martinoty P 1984 Phys. Rev. Lett. 53 1065
Garel A T 1978 J. Physique 39 225
Gathercole L J and Keller A 1982 'The Periodontal Ligament in Health and Disease ed. B K B Berkovits
(Oxford: Pergamon) p 103
de Gennes P G 1968 C.R. Acad. Sci., Paris B 266 571
-1972 C.R. Acad, Sci., Paris B 275 549
-1974a 'The Physics of Liquid Crystals (Oxford: Clarendon)
-1974b Phys. Fluids 17 1645
-1976 Molecular Fluids, Les Houches Summer School 1973 ed. R Balian and G Weil (New York: Gordon
and Breach)
-1977 Mol. Cryst. Liq. Cryst. Lett. 34 177
-1982 Polymer Liquid Crystals ed. A Ciferri et a1 (New York: Academic) ch .5
-1983 J. Physique Lett. 44 L 657
-1984 Mol. Cryst. Liq. Cryst. Lett. 102 95
de Gennes P G and Pincus P 1976 J. Physique 37 1359
de Gennes P G and Taupin C 1982 J. Phys. Chem. 86 2294
Gerritsma C J, Geurst J A and Spruijt A M J 1973 Phys. Lett. A 43 356
Gerritsma C J and van Zanten P 1972 Liquid Crystals vol 3 ed. G H Brown and N M Labes (New York:
Gordon and Breach) p 751
Geurst J A, Spruijt A M J and Gerritsma G 1975 J. Physique 36 653
Gharbia M, Cagnon M and Durand G 1985 J. Physque Lett. 46 L683
Giraud-Guille M M 1988 Calcijed Tissue Int. 42 167
Glogdruva M, Lejcek L, Pave1 J, Janovec V and Fousek J 1983 Mol. Cryst. Liq. Crysf. 91 309
Godreche C and de Seze L 1985 J. Physique Lett. 46 L39
Defects in liquid crystals 649

Goodby J W and Pindak R 1981 Mol. Cryst. Liq. Crysf. 75 233


Gray G W and Goodby J W 1984 Smectic Liquid Crystals (Leonards Hill)
Graziano D J and Mackley M R 1984a Mol. Cryst. Liq. Cryst. 106 73
-1984b Mol. Cryst. Liq. Cryst. 106 103
Groupe des Cristaux Liquides DOrsay 1969 J. Physique 30 C4 38
Gruner S M, Rothschild K J, Degrip W J and Clark N 1985 Preprinf
Gunn J M F and Ma K B 1980 J. Phys. C: Solid State Phys. 13 963
Gunn J M F and Warner M 1987 Phys. Rev. Lett. 58 393
Guyon E and Pidranski P 1975 J. Physique 36 203
Halperin B 1981 Les Houches Summer School on the Physics of Defects ed. R Balian e? al (Amsterdam:
North-Holland) p 665
Halperin B I and Nelson D R 1978 Phys. Rev. Left. 41 121, 519
Handschy M A, Clark N A and Lagerwall S T 1983 Phys. Rev. Lett. 51 471
Harbich W, Servuss R and Helfrich W 1978 2.Naturf 33a 1013
Hardouin F, Sigaud G, Achard M F and Levelut A M 1983 J. Chim. Phys. 80 52
Harris W F 1977 Sci. Am. 237 130
Helfrich W 1969 Phys. Reo. Lett. 23 372
-1974 Z . Natur$ C 29 692
-1981 in Physics ofDefects ed. R Balian et a1 (Amsterdam: North-Holland)
-1986 J. Physique 47 321
Helfrich W and Heppke G (ed.) 1980 Liquid Crystals of One- and Two-Dimensional Order (Springer Series
in Chemical Physics vol. 11) ed. W Helfrich and G Heppke (Berlin: Springer)
Heritier M, Montambaux G and Lederer P 1979 J. Physique Left. 49 L493
Hessel F and Finkelmann H 1986 Polym. Bull. 15 349
Hilbert D and Cohn-Vossen S 1952 Geometry and the Imagination (New York: Chelsea)
Hiltrop K and Fischer F 1976 2. N a f u r j a 31 800
Hinov H P 1984 J. Physique Lett. 45 L185
Hirsch E, Wittman J C and Candau F 1982 J. Dispersion Sci. Technol. 3 351
Hirth J P, Pershan P S, Collett J, Sirota E and Sorensen L B 1984 Phys. Rev. Lett. 53 473
Holmes M C and Charvolin J 1984 J. Phys. Chem. 88 810
tHooft G 1974 Nucl. Phys. B 79 276
Horn R G and Kltman M 1978 Ann. Phys., Paris 3 229
Hornreich R M, Kugler M and Shtrikman S 1982 Phys. Rev. Lett. 48 1404
Hornreich R M and Shtrikman S 1980 Liquid Crystals of One- and Two-Dimensional Order (Springer Series
in Chemical Physics vol. 11) ed. W Helfrich and G Heppke (Berlin: Springer) p 185
-1986 Phys. Rev. Lett. 56 1723
-1988 The Weizmann Institutefor Science Preprint
Hu C R, Ham T E and Saslow W M 1978 J. Low Temp. Phys. 32 301
Hurd A J, Fraden S, Lonberg F and Meyer R B 1985 J. Physique 46 905
Hyde S T, Andersson S, Ericsson B and Larsson K 1984 2. Krist. 168 213
Imura H and Okano K 1973 Phys. Leff.42A 403
Ishikawa K, Uemura T, Takezoe H and Fukuda A 1984 Japan. J. Appl. Phys. 23 L666
Johnson D and Saupe A 1977 Phys. Rev. A 15 2079
Kats E I 1978 Sou. Phys.-JETP 48 916
Kim M G, Park S, Cooper S M and Letcher S V 1976 Mol. Cryst. Liq. Cryst. 36 143
Kiss 0 and Porter R S 1980 Mol. Cryst. Liq. Crysf. 60 267
Kltman M 1970 Phil. Mag. 22 178
-1973 Phil. Mag. 27 1057
-1977a J. Physique Lett. 38 1511
-1977b J. Physique Lett. 38 L199
-1980a in Dislocations in Solids vol. V ed. F R N Nabarro (Amsterdam: North-Holland) ch 22
-1980b J. Physique 41 737
-1980c Liquid Crystals of One- and Two-Dimensional Order (Springer Series in Chemical Physics vol.
11) ed. W Helfrich and G Heppke (Berlin: Springer) p 97
-1982 Magnetism ed. F Cyrot (Amsterdam: North-Holland)
-1983a Point Lines and Walls (Chichester: Wiley)
-1983b J. Physique Lett. 44 L295
-1984 in Dislocations ed. P Veyssibre et al (Paris: Editions du CNRS)
650 M Kle'man

-1985a J. Physique Lett 46 L723


-1985b J. Physique 46 1193
-1987 Phys. Scri. T 19 565
-1988 Liquid Cryst. 3 1355
KlCman M and Friedel J 1969 J. Physique Coll. 30 C4 43
KlCman M and Lejcek L 1980 Phil. Mag. A 42 671
Kl6man M, Liebert L and Strzelecki L 1983 Polymer 25 295
KlCman M and Michel L 1978a J. Physique Lett. 39 L29
-1978b Phys. Rev. Lett. 40 1387
KlCman M and Oswald P 1982 J. Physique 43 655
KlCman M and Ryschenkow G 1976 J. Chem. Phys. 64 413
KlCman M and Sadoc J F 1979 J. Physique Lett. 40 L569
KlCman M and Williams C E 1974 J. Physique Lett. 35 U 9
Kl6man M, Williams C E, Costello M J and Gulik-Krzywicki T 1977 Phil. Mag. 35 33
Kondo K 1955 R.A.A.G. Mem. 1 4 5 8
Kosterlitz J M and Thouless D J 1972 J. Phys. C: Solid State Phys. 5 L124
Kroner E 1981 Physics of Defects ed. R Balian et al (Amsterdam: North-Holland)
-(ed.) 1982 Gauge Field Theories ofDefects in Solids (Stuttgart: MPI)
Kurik M V and Lavrentovich 0 D 1982 Sou. Phys.-JETP Lett. 35 444
-1983 SOU.PhyS.-JETP 58 299
Kutka R and Trebin H R 1984 J. Physique Lett. 45 1119
Landau L D and Lifshitz E M 1962 Quantum Mechanics (New York: Pergamon)
Langer S A and Sethna J P 1986 Phys. Rev. A 34 5035
Lavrentovich 0 D 1986a Sou. Phys.-JETP Lett. 43 382
-1986b Sov. Phys.-JETP Lett. 64 966
Lavrentovich 0 D and Terentiev E M 1987 Sou. Phys.-JETP 64 343
Leger L and Martinet A 1983 J. Physique 37 L303
Lehmann 0 1904 Hussige Kristalle (Leipzig: Engelmann)
Lejcek L 1982 Czech. J. Phys. B 32 767
-1985 Czech. J. Phys. B 35 655
-1986 Liquid Cryst. 1 473
Lequeux F 1986 C.R. Acad. Sci., Paris 303 765
Lequeux F and Kltman M 1988 J. Physique 49 845
Levelut A M 1983 J. Chim. Phys. 80 149
Levelut A M, Doucet J and Lambert M J 1974 J. Physique 35 773
-1977 J. Physique 38 1163
Levelut A M, Germain C, Keller P, Liebert L and Billard J 1983 J. Physique 44 623
Livolant F 1984 Thesis University of Paris
-1987 J. Physique 48 1051
Livolant F and Bouligand Y 1980 Chromosoma 80 97
-1986 J. Physique 47 1813
Loft R and Degrand T A 1987 Phys. Rev. B 35 8528
Lonberg F, Fraden S, Hurd A J and Meyer R B 1984 Phys. Reo. Lett. 52 1903
Lonberg F and Meyer R B 1985 Phys. Rev. Lett. 55 718
Lotz B, Gonthier-Vassal A, Brack A and Magoshi J 1982 J. Mol. Biol. 156 345
Luz 2, Goldfarb D and Zimmerman H 1985 N M R of Liquid Crystals ed. J M Emsley (Dordrecht:
Reidel)
Luzzati V, Mariani P and Gulik-Krzywicki T 1987 Physics of Amphiphilic Layers (Springer Proceedings in
Physics vol. 21) ed. J Meunier (Springer: Berlin) p 131
Luzzati V and Spegt P A 1967 Nature 215 701
Lyuksyutov I F 1978 L D Landau Institute Preprint
Mackley M R, Pinaud F and Siekmann A 1981 Polymer 22 437
Malet G, Marignan J and Parodi 0 1976 J. Physique 37 865
Malozemoff A and Slonczewski J C 1979 Magnetic Domain Walls in Bubble Materials (New York: Academic)
Marcus M 1981 J. Physique 42 61
Marignan J, Malet G and Parodi 0 1978 Ann. Phys., Paris 3 221
Marignan J and Parodi 0 1979 J. Physique 40 C3 78
-1983 J. Physique 44 665
Marignan J, Parodi 0 and Dubois-Violette E 1983 J. Physique 44 263
Martinot-Lagarde P 1976 J. Physique Coll. 31 C3 129
Defects in liquid crystals 65 1

Mauguin C 1911 Bull. Soc. Fr. Miner. 34 71


Mazelet G and Kltman M 1986 Polymer 27 714
-1989 J. Mater. Sci. in the press
Mazenko G F, Ramaswamy S and Toner J 1982 Phys. Rev. A 28 618
Meiboom S, Sammon M and Brinkman W F 1983 Phys. Rev. A 27 438
Meiboom S, Sethna J P, Anderson P W and Brinkman W F 1981 Phys. Rev. 46 1216
Melzer D and Nabarro F R N 1977a Phil. Mag. 35 901
-1977b Phil. Mag. 35 907
Mermin N D 1977 in Quantum Fluids and Solids ed. S B Trickey et a1 (New York: Plenum) p 3
_. 1979 Rev. Mod. Phys. 51 591
Mermin N D and Ho T L 1976 Phys. Rev. Lett. 36 594
Mermin N D, Mineyev V P and Volovik G E 1978 J. Low Temp. Phys. 33 117
Merucci J 1976 Diplom. Essay unpublished
Meunier J P and Billard J 1969 Mol. Cryst. Liq. Cryst. 7 421
Meyer R B 1969 Phys. Rev. Lett. 22 918
-1972 Mol. Cryst. Liq. Cryst. 16 355
-1973 Phil. Mag. 27 405
_. 1976 in Molecular Fluids, Les Houches Summer School 1973 ed. R Balian and G Weil (New York:
Gordon and Breach)
- 1977 Mol. Cryst. Liq. Cryst. 40 33
-1982 in Polymer Liquid Crystals ed. A Cifferi et a/ (New York: Academic) ch. 6
Meyer R B, Liebert L, Strzelecki L and Keller P 1975 J. Physique Lett. 36 69
Meyer R B, Lonberg F, Taratuta V, Fraden S, Lee Sin-Doo and Hurd A J 1985 Faraday Discuss. Chem.
Soc. 79 125
Meyer R B, Stebler B and Lagerwall S T 1978 Phys. Rev. Lett. 41 1393
Michel L 1980 Rev. Mod. Phys. 52 617
Mihailovic M 1987 C.R. Acad. Sci., Paris 303 s6rie I1 875
-1988 Thhse de Doctorat Orsay
Mihailovic M and Oswald P 1989 J. Physique, in the press
Millaud B, Thierry A and Skoulios A 1978 J. Physique 39 1109
Miltat J 1976 Phil. Mag. 33 225
Mineyev V P and Volovik G E 1978 Phys. Rev. B 18 3197
Misner C W, Thorne K S and Wheeler J A 1973 Gravitation (San Francisco: Freeman)
Monastyrsky M I and Retakh V S 1986 Commun. Math. Phys. 103 445
Moncton D E and Pindak R 1979 Phys. Rev. Lett. 43 701
Monge G 1807 Applications de 1Analyse d la Gt!omt!trie (Paris)
Morawietz K and Demus D 1987 Cryst. Res. Technol. 22 1391
Mori H and Nakanishi H 1988 J. Phys. Soc. Japan 57 1281
Mosseri R and Sadoc J F 1983 Structure of Non-Crystalline Materials ed. P H Gaskell (London: Taylor and
Francis)
Nabarro F R N 1948 Bristol Conference, Physical Soc. London
-1967 Theory of Crystal Dislocations (Oxford: Oxford University Press)
-1972 J. Physique 33 1089
Nabarro F R N and Harris W F 1971 Nature 232 423
Nageotte J 1936 Morphologie des Gels Lipoides (Paris: Hermann)
Nakanishi H, Hayashi K and Mori H 1987 Keio University Preprint KSTS/RR-87100
Nallet F and Prost J Europhys. Lett. 4 307
Navard P 1986 J. Polym. Sci.: Polym. Phys. Ed. 24 435
Navard P and Zachariades A E 1987 J. Polym. Sci.: Polym. Phys. Ed. 25 1089
Nehring J 1973 Phys. Rev. A 7 1737
Nelson D R and Peliti L 1987 J. Physique 48 1085
Nicolis S, Mosseri R and Sadoc J F 1986 Europhys. Lett. 1 571
Nityananda R and Ranganath G S 1980 Proc. Int. Conf: Liquid Crystals (Bangalore) ed. S Chandrasekhar
(London: Heyden) p 205
Noel C, Laupretre F, Firedrich C, Fayolle B and Bosio L 1984 Polymer 25 808
Nye J F 1953 Acta Metal/. 1 153
Odijk T 1986 Liquid Cryst. 1 553
Onogi Y, White J L and Fellon J F 1980 J. Non-Newt. Nuid Mech. 7 121
Orsay Group o n Liquid Crystals 1971 Solid State Commun. 9 653
-1975 J. Physique 36 C1 305
652 M Kle'man

Oswald P 1981 J. Physique Lett. 42 L171


-1983 J. Physique Lett. 44 L303
__ 1985 J. Physique 46 1255
-1986a J. Physique 47 1091
-1986b J. Physique 47 1279
-1986c Liquid Cryst. 1 2 2 7
-1987 C.R. Acad. Sci., Paris 304 SBrie I1 1043
Oswald P and Allain M 1985 J. Physique 46 831
Oswald P, Behar J and Kltman M 1982 Phil. Mug. A 46 899
Oswald P and Ben-Abraham S I 1982 J. Physique 43 1193
Oswald P and Kltman M 1981 J. Physique 42 1461
- 1982 J. Physique Lett. 43 L411
-1984 J. Physique Lett. 43 L319
Pansu B, Dandoloff R and Dubois-Violette E 1987 J. Physique 48 297
Pansu B and Dubois-Violette E 1987 J. Physique 48 1861
Parodi 0, Durand G, Malet G and Marignan J 1982 J. Physique Lett. 43 L727
Patel D L and Dupre D B 1979 J. Polymer Sci.: Polym. Lett. Ed. 17 299
Perez A, Brunet M and Parodi 0 1978 J. Physique Lett. 39 L353
-1981 J. Physique 42 1559
Pershan P S 1974 J. Appl. Phys. 45 1590
Pershan P S and Prost J 1975 J. Appl. Phys. 46 2343
Pieranski P and Guyon E 1974 Phys. Rev. A 9 404
Pincus P and de Gennes P G 1977 Polym. Reprints 18 161
Pindak R, Moncton D E, Goodby J N and Davey S C 1981 Phys. Rev. Lett. 46 1135
Pindak R, Young C Y , Meyer R B and Clark N A 1980 Phys. Rev. Lett. 45 1193
Pisarski R D and Stein D L 1980 J. Physique 41 345
Pleiner H 1986a Phil. Mag. A 54 421
-1986b Liq. Cryst. 1 197
-1988a Liq. Cryst. 3 249
-1988b Phys. Rev. A 37 3986
Pleiner H and Brand H R 1985 Phys. Rev. Lett. 54 1817
Poenaru V and Toulouse G 1977 J. Physique 38 887
Polyakov A M 1974 JETP Lett. 20 194
Porte G 1977 J. Physique 38 509
Porter R S and Johnson J F 1983 J. Appl. Phys. 34 55
Powers L and Pershan P S 1977 Biophys. J. 20 137
Press M J and Arrott A S 1974 Phys. Rev. Lett. 33 403
__ 1975 J. Physique 36 C1 177
Prost J 1984 Adv. Phys. 33 1
Ramaswany S and Toner J 1984 Phys. Rev. Lett. 53 477
Ranganath G S 1980 in Liquid Cryszals ed. S Chandrasekhar (London: Heyden) p 219
-1982 Mol. Cryst. Liq. Cryst. 87 187
-1983a Mol. Cryst. Liq. Cryst. 92 201
-1983b Mol. Cryst. Liq. Cryst. 97 77
Rault J 1971 Solid State Commun. 9 1965
-1973 Phil. Mag. 28 11
-1974a Phil. Mag. 30 621
-1974b in Liquid Crystals and Ordered Fluids ed. J F Johnson and R S Porter (New York: Plenum) p 677
-1975 C.R. Acad. Sci., Paris B 280 417
Reek B and Ringsdorf H 1985 Makromol. Chem. Rapid Commun. 6 291
Ribotta R and Durand G 1977 J. Physique 38 179
Ribotta R and Joets A 1984 Cellular Instabilities ed. J E Westfreid and Z Zaleski (Berlin: Springer)
Ribotta R, Meyer R B and Durand G 1974 J. Physique Lett. 35 L161
Ricard L and Prost J 1981 J. Physique 42 861
Rivier N 1979 Phil. Mag. A 40 859
Robinson C 1961 Tetrahedron 13 219
-1966 Mol. Cryst. 1 4 6 7
Rogers J and Winsor P A 1969 J. Colloid Sci. 30
Rokhsar D S and Sethna J P 1986 Cornell University Preprint
Defects in liquid crystals 653

Rosenblatt C, Meyer R B, Pindak R and Clark N A 1980 Phys. Rev. A 21 140


Rosenblatt C S, Pindak R, Clark N A and Meyer R B 1977 J. Physique 38 1105
Ruppel D and Sackmann E 1983 J. Physique 44 1025
Ryschenkow G 1975 J. Physique 36 243
Ryschenkow G and Kltman M 1976 J. Chem. Phys. 64 404
Sackmann E, Ruppel D and Gebhardt C 1980 in Liquid Crystals of One- and Two-Dimensional Order
(Springer Series in Chemical Physics vol. 11) ed. W Helfrich and G Heppke (Berlin: Springer) p309
Sadoc J F 1983 J. Physique Lett. 44 L707
Sadoc J F and Charvolin J 1986 J. Physique 47 683
Sakurai I, Kawamura Y, Sakurai T, Ikegami A and Seto T 1985 Mol. Cryst. Liq. Cryst. 130 203
Salemme F R 1983 Progr. Biophys. Mol. Biol. 42 95
Saludjian P and Luzzati V 1967 Poly-aminoacids ed G D Fasman (New York: Marcel Dekker) p 157
Samulski E T and Tobolsky A V 1975 in Liquid Crystals and Plastic Crystals vol 1 (Chichester: Wiley)
p 174
Sarkar S 1983 Mol. Cryst. Liq. Cryst. 97 111
Saslow W M 1982 Phys. Rev. A 25 3350
Satna B R, Shashidar R and Raja V N 1985 Phys. Rev. Lett. 55 1476
Saupe A 1969 Mol. Cryst. Liq. Cryst. 7 59
-1973 Mol. Cryst. Liq. Cryst. 21 211
-1977 J. Colloid Sci. 58 428
Schneider M B and Webb W W 1984 J. Physique 45 273
Schulze W and Kunz T 1976 Mol. Cryst. Liq. Cryst. 36 223
Schopohl N and Sluckin T J 1987 Phys. Rev. Lett. 59 2582
Scriven L E 1977 in Micellization, Solubilization andMicroemulsions vol I1 ed. K Mittal (New York: Plenum)
p 877
Scudieri F, Ferrari A and Gunduz E 1979 J. Physique 40 C3 90
Sethna J P 1983 Phys. Rev. Lett. 51 2198
-1985 Phys. Rev. B 31 6278
Sethna J P and Kldman M 1982 Phys. Rev. A 26 3037
Sethna J P, Wright D C and Mermin D N 1983 Phys. Rev. Lett. 51 467
Shankar R 1977 J. Physique 38 1405
Shechtman D, Blech I, Gratias D and Cahn J W 1984 Phys. Rev. Lett. 53 1951
Sigaux G, Mercier M and Gasparoux H 1984 Preprint
Sommerville D Y M 1958 Non-Euclidean Geometry (New York: Dover)
Steenrod N 1951 Topology of Fiber Bundles (Princeton, NJ: Princeton University Press)
Steers M, Kldman M and Williams C E 1974 J. Physique Lett. 35 L21
Stein D L, Pisarski R D and Anderson P W 1978 Phys. Rev. Lett. 40 1269
Straley J P 1973 Phys. Rev. A 8 2181
Sun Zheng-Min and Kltman M 1984 Mol. Cryst. Liq. Cryst. 111 321
Taratuta V and Meyer R B 1987 Liq. Cryst. 2 373
Tardieu A and Billard J 1976 J. Physique 37 C3 79
Taylor T, Arora S and Fergason J 1970 Phys. Rev. Lett. 24 359
Thomas E L and Wood B A 1985 Faraday Discuss. Chem. Soc. 79 229
Toulouse G 1977a J. Physique Lett. 38 L37
-1977b Commun. Phys. 2 115
Toulouse G and Kltman M 1976 J. Physique Lett. 37 L149
Trebin H R 1981 J. Physique 42 1573
-1982 Adv. Phys. 31 195
-1983 Phys. Rev. Lett. 50 1381
Trebin H R and Kutka R 1981 J. Physique Lett. 42 L421
Turkevich L A, Safran S A and Pincus P 1984 Surfactants in Solution (New York: Plenum)
Viney C and Windle A H 1982 J. Mater. Sci. 17 2661
Volovik G E 1978 Sou. Phys.-JETP Lett. 28 59
Volovik G E and Lavrentovich 0 D 1983 Sou. Phys.-JETP 58 1159
Volovik G E and Mineyev 1976 JETP Lett. 24 561
-1977 Sou. Phys.-JETP 45 1186
Volterra V 1907 Ann. Ec. Norm. Paris 24 401
Wahl J 1979 2. Naturj a 34 818
Wahl J and Fischer F 1973 Mol. Cryst. Liq. Cryst. 22 359
654 M Klkman

Wang X F and Warner M 1986 J. Phys. A: Math. Gen. 19 2215


Williams C E 1975 Phil. Mag. 32 313
-1976 These Orsay
Williams C E, Cladis P E and Kldman M 1973 Mol. Cryst. Liq. Cryst. 21 355
Williams C E and Kl6man M 1975 J. Physique 36 C315
Williams C E, Pieranski P and Cladis P E 1972 Phys. Rev. Left. 29 90
Windle A H, Viney C, Golombok R, Donald A M and Mitchell G R 1985 Faraday Discuss. Chem. Soc. 79
55
van Winkle D H and Clark N A 1982 Phys. Rev. Lett. 48 1407
-1984 Phys. Rev. Lett. 53 1157
-1985 Bull. Am. Phys. Soc. 30 379
Wissbrun K F 1985 Faraday Discuss. Chem. Soc. 79 161
de Wit R 1971 J. Appl. Phys. 42 3304
Wittman J C, Lotz B, Candau F and Kovacs A J 1982 J. Polym. Sci.: Polym. Phys. 20 1341
Yu L J and Saupe A 1980a Phys. Rev. Lett. 45 1000
-1980b J. Am. Chem. Soc. 1202 4879
Zasadzinski J and Meyer R B 1986 Phys. Rev. Lett. 56 636
Zasadzinski J, Scriven L E and Davis H T 1985 Phil. Mag. A 51 287

You might also like