Lectures on Landau Theory

of Phase Transitions

Department of Physics, Georgetown University

Peter D Olmsted

July 9, 2015

Contents

1 Goals of these Lectures 1

2 Phase Transitions 2
2.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Statistical Mechanics and Phase Transitions . . . . . . . . . . . . . . . . . . 3

3 Landau Theory: Fundamentals 5

3.1 The Recipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Relation to Statistical Mechanics . . . . . . . . . . . . . . . . . . . . . . . . 6

4 Examples 7
4.1 Ising Magnet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.2 Heisenberg Ferromagnet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.3 Nematic Liquid Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.4 Crystal Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.4.1 Lamellar Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.4.2 Two and Three Dimensional Crystals . . . . . . . . . . . . . . . . . . 15

i
5 Smectic Liquid Crystals 18
5.1 Mean Field Theory of the Nematic-Smectic–A transition . . . . . . . . . . . 18
5.2 Free Energy and Superconducting Analogy . . . . . . . . . . . . . . . . . . . 19
5.3 Fluctuations and the Halperin-Lubensky-Ma Effect. . . . . . . . . . . . . . . 21
5.4 Layered liquids and the Brazovskii effect . . . . . . . . . . . . . . . . . . . . 22

Distributed under Creative Commons License Attribution 4.0 International.

c Peter D. Olmsted (2015).

ii
1. Goals of these Lectures Lectures on Landau Theory

1 Goals of these Lectures

Phase transitions are ubiquitous in nature. Examples include a magnets, liquid crystals,
superconductors, crystals, amorphous equilibrium solids, and liquid condensation. These
transitions occur between equilibrium states as functions of temperature, pressure, magnetic
field, etc.; and define the nature of the matter we deal with on a day to day basis. Under-
standing how to predict and describe both the existence of these transitions, as well as their
character and consequences for everyday phenemena, is one of the more important roles of
statistical and condensed matter physics.
In this brief set of Lectures I propose to outline one of the basic theoretical tools for
describing phase transitions, the Landau Theory of Phase Transitions. This was de-
veloped by Landau in the 1940’s, originally to describe superconductivity. The procedure is
general, and is one of the most useful tools in condensed matter physics. Not only can we use
Landau Theory to describe and understand the nature of phase transitions among ordered
(and disordered) states, but we can use it as a starting point for understanding the behavior
of ordered states. These lectures are designed to establish the following broad brush strokes:

1. Very few phase transitions can be calculated exactly: nonetheless, there is still much
that can be understood without having to solve the entire problem. These kinds of
questions (order of phase transitions, hydrodynamics and elasticity, fluctuations) are
the domain of Landau theory.

2. I will present (some of) the problems of phase transitions, and introduce Landau Theory
as a way of understanding the behavior (but not the existence) of phase transitions.

3. We will explore, through examples, the profound implications of symmetry for the
nature of ordered states and their associated transitions.

4. We will see how to use the nature of the order parameter to understand deformations
in a broken symmetry state: this often goes by the name of generalized elasticity,
and incorporates elasticity of solids, Frank elasticity in nematic liquid crystals, the
deformation energy of smectic liquid crystals and membrane systems, sound waves in
fluids, etc.

5. Landau theory is a mean field theory, in the sense that the system is assumed to be
adequately described by a single macroscopic state.

6. We will use Landau free energy functionals to calculate observable quantities such as
structure factors; identify the breakdown of Landau theory due to fluctuations.

7. We will examine some interesting paradigms whereby the qualitative nature of phase
transitions, such as the order of the phase transition, is altered by fluctuation effects
and the coupling of different degrees of freedom.

Along the way we will learn how to follow our noses and construct proper free energy
functionals on symmetry grounds; pick up some calculational tools; and examine some fun-
damental ideas in statistical mechanics.

1 c Peter D. Olmsted, 2015.

2. Phase Transitions Lectures on Landau Theory

2 Phase Transitions

A phase transition occurs when the equilibrium state of a system changes qualitatively as
a function of externally imposed constraints. These constraints could be temperature, pres-
sure, magnetic field, concentration, degree of crosslinking, or any number of other physical
quantities. In the following I only consider a transition as a function of temperature, but
note that the idea is, of course, more general than that (physicists strive to be as general as
possible!). In any of these transitions there is some quantity that can be observed to change
qualitatively as a function of temperature. In many cases more than one quantity can be
observed, but it is quite obvious that something is happening. This quantity will be taken
later to be the order parameter of the phase transition.

2.1 Examples

1. Crystals: In a transition to a crystalline solid a disordered liquid with non apparent

structure undergoes a transition to a structure with long range periodic order, usually
in three dimensions. This is most easily parametrized by the mass density, ρ(r):
X 
ρ(r) = ρ̄ + ρ(q)e−iq·r + c.c , (1)
q∈G

where ρ̄ is the mean density and {G} is the sition:

set of reciprocal lattice vectors that char-
acterize the crystal structure. The Fourier First Order Phase Transition
(Nematic LC, 2D/3D crystals, etc)
modes refer to density modulations with
wavenumber q = 2π/λ, with wavelength
λ. The complex conjugate is added to re-
order parameter

tain a real number for the mass density.

Upon cooling a liquid into a crystal the
object that distinguishes the crystal from
the liquid is the set of wavevectors {ρ(G)},
which appear as Bragg peaks in a scat-
tering experiment. Usually these Fourier
modes grow discontinuously from zero, in Tc

what is called a first order phase tran- Temperature

2. Nematic Liquid Crystals: The isotropic-nematic transition occurs in melts (or

solutions) of rigid rod-like molecules. At high temperatures (or dilute concentrations)
the rods are isotropically distributed, and upon cooling an orientational interaction
which is a combination of enthalpic and excluded volume effects encourages the rods
to spontaneously align along a particular direction, denoted the director, nb . A scalar
measure of the order is the anisotropy of the distribution of rods,

hP2 (cos θ)i = hcos2 θ − 31 i, (2)

where the average h·i is taken over the equilibrium distribution for all rods in the
system, and θ is the angle with respect to some fixed direction. Like 2D crystallization,

2 c Peter D. Olmsted, 2015.

2.2 Statistical Mechanics and Phase Transitions Lectures on Landau Theory

the isotropic-nematic is a first order transition, with a discontinuity in hP2 (cos θ)i at the
transition temperature TIN . Since “up” and “down” are the same for such a system (in
liquid crystals the rodlike molecules are typically symmetric under reflection through
the long axis), the order parameter is symmetric under cos θ → − cos θ.

3. Magnets: In a ferromagnet an assembly a critical temperature Tc (often called the

of magnetic spins undergoes a spontaneous Curie Temperature, after its discoverer).
transition from a disordered phase with no Continuous Phase Transition
(Magnet, superconductor, 1D crystal, ...)
net magnetization to a phase with a non-
zero magnetization. The order parameter
is thus the magnetization,
1 X

order parameter
M= hSi i, (3)
N i
where the average is an equilibrium aver-
age over all spins Si in the system. Un-
like the nematic liquid crystal, the ferro-
magnetic phase transition is a continuous Tc
phase transition, in which the magne- Temperature
tization grows smoothly from zero below

4. One Dimensional (layered) Crystals: In a one dimensional crystal, or a layered

system, a one-dimensional modulation develops spontaneously below the critical tem-
perature. Examples include block-copolymers, which can phase separate into layers of
A and B material, to relieve chemical incompatibility but maintain the connectivity
constraint; helical magnets which develop a pitch as the spin twists; cholesteric phases
which develop a twisted nematic conformation; and smectic phases in thermotropic
liquid crystals and lyotropic surfactant solutions. In this case the order parameter is
the Fourier mode of the relevant degree of freedom:

ψ(r) = ψ̄ + ψ(q)e−iq·r + c.c. (4)

This transition is predicted by Landau (mean field) theory to be continuous, but it is

believed to be first order in physical situations, due to fluctuation effects. Hopefully
we will get this far!
5. Phase Separation: Finally, one of the most common phase transitions from every
day life is phase separation, which makes it necessary to shake the salad dressing. In
this case the order parameter is the deviation of the local composition from the mean
value. Usually this is a first order phase transition, but if the concentration is just
right the system can be taken through a critical point, or continuous phase transition.
This system is equivalent to liquid-liquid or liquid-gas phase separation, in which case
the order parameter is a density difference instead of a composition difference.

2.2 Statistical Mechanics and Phase Transitions

One of the uses of statistical mechanics (aside from helping us to understand nature) is to
calculate fundamental properties of matter, including phase transitions. In principle, this

3 c Peter D. Olmsted, 2015.

2.2 Statistical Mechanics and Phase Transitions Lectures on Landau Theory

is a remarkably simple task, thanks to Boltzmann. All of the statistical information of a

system is encoded in the Partition Function, Z:
X
Z= e−H[µ]/kB T , (5)
µ

where µ refers to all the microstates of the system (e.g. all possible configurations spins in
a magnet) and H[µ] is the Hamiltonian (energy). Boltzmann proved that the (Helmholtz)
free energy is given by
F = −kB T ln Z. (6)

If we can calculate the free energy F we can then calculate all desired thermodynamic
quantities by appropriate derivatives.∗ Unfortunately, the free energy can only be evaluated
for a few systems, notably the Ising Model (spins which Pcan point up or down, and interact
with each other by a very simple interaction, H = −J ij Si Sj ). More often than not we
are faced with an impossibly difficult calculation.
In the case of phase transitions, perhaps we can get away with a less rigorous calculation.
A clue is the very nature of a phase transition: at a phase transition a system undergoes a
qualitative change, and develops some order where there was none before. Hence the system
does not vary smoothly as a function of (for example) temperature. This means that the
free energy F (T ) is, mathematically, a non-analytic function of temperature. This is of
fundamental importance in the theory of phase transitions, and forms the starting point of
rigorous studies. A non-analytic function is one for which some derivatives are undefined
at certain points, or singularities. Phase transitions are points (in the parameter space of
field variables such as temperature, pressure, magnetic field) or sets of points which are
singularities in the free energy. The free energy, and hence the behavior of a thermodynamic
system, behaves non-smoothly as it is taken through a phase transition.
It’s interesting to try and understand how to get non-analytic behavior out of a sum of
exponentials, each of which is separately analytic at any finite temperature:
" #
X
F = −kB T ln e−H[µ]/kB T . (9)
µ

The existence of singularities in F is a direct result of the thermodynamic limit, that is, the
presence of an essentially infinite number of degrees of freedom in a thermodynamic system.
Possible singularities we can imagine in F are discontinuities in the first derivative, ∂F/∂T ,
∗
For example, the magnetization can be determined by

∂F
M =− , (7)
∂h h=0

where an additional symmetry-breaking field has been added to the system Hamiltonian. For any system,
it’s only a matter of cleverness to determine the extra field to add to “extract” the desired order parameter
by a suitable derivative. In this case the field enters as
X
H[S] → H[S] − h Si . (8)
i

4 c Peter D. Olmsted, 2015.

3. Landau Theory: Fundamentals Lectures on Landau Theory

or in higher order derivatives ∂ 2 F/∂T 2 , ∂ 3 F/∂T 3 , . . .. These two cases define, respectively,
first order or continuous (often termed “second order”) phase transitions.

First Order Phase Transition Continuous Phase Transition

Free energy F

TC TC
Temperature T Temperature T

From the theory of analytic functions we are familiar with the fact that a surprising
amount of information is contained in singularities: witness the Cauchy Integral Theorem.
So, if we can somehow come to grips with a singularity in the free energy, even in just a
qualitative way, then perhaps some progress can be made in understanding the physics of
phase transitions. This is the goal of Landau Theory.

3 Landau Theory: Fundamentals

3.1 The Recipe

Landau made a series of assumptions to approximate the free energy of a system, in such a
way that it exhibits the non-analyticity of a phase transition and turns out to capture much
of the physics. There are essentially four steps in this procedure: we will visit these steps
first, and then explore them again in terms of the partition function.

1. Define an order parameter ψ: For a given system an order parameter must be con-
structed. This is a quantity which is zero in the disordered phase and non-zero in the
ordered phase. Examples include the magnetization in a ferromagnet, the amplitudes
of the Fourier modes in a crystal, and the degree of orientation of a nematic liquid
crystal.

2. Assume a free energy functional: Assume the free energy is determined by minimizing
the following functional
F̃ = F0 (T ) + FL (T, ψ), (10)

5 c Peter D. Olmsted, 2015.

3.2 Relation to Statistical Mechanics Lectures on Landau Theory

where F0 is an analytic (smooth) function of temperature, and FL (T, ψ) contains all

the information about dependence on the order parameter ψ.
3. Construction of FL (ψ): The Landau Functional is assumed to be an analytic (typically
polynomial expansion) function of ψ that obeys all possible symmetries associated with
ψ; this typically includes translational and rotational invariance, and other “internal”
degrees of dicatated by the nature of the order parameter. This is the most important
part of the theory, wherein most of the physics lies.
4. Temperature Dependence: It is assumed that all the non-trivial temperature depen-
dence resides in the lowest order term in the expansion of FL (T, ψ), typically of the
form Z
 
FL [T, ψ] = dV 21 a0 (T − T∗ )ψ 2 + . . . (11)

Since FL is constructed as an expansion, there will be other unknown constants. In

a physical system these constants have temperature dependence, but if these depen-
dences are smooth then they have negligible effect near the phase transition. This is
rigorous for a continuous phase transition, and an approximation for a first order phase
transition.

Upon constructing the Landau functional, and minimizing it over ψ as a function of tem-
perature, the nature of the phase transition may be determined. The system at this level is
specified as having a uniform, or mean, state; hence Landau theory is really a mean field the-
ory. However, the resulting approximate free energy is a natural starting point for examining
fluctuation effects.

3.2 Relation to Statistical Mechanics

To understand what’s going on, let’s reexamine statistical mechanics. The free energy is
given by X
e−F/kB T = e−H[µ]/kB T . (12)
µ

Landau’s assumption is that we can replace the entire partition function by the following,
Z
−F/kB T −F0 /kB T
e ≃e Dψ e−FL [T, ψ]/kB T , (13)
R
where the integral Dψ is a functional integral over all degrees of freedom associated with
ψ, instead of an integral over all microstates. For example, if ψ is the mean magnetization,
a given value for the magnetization can be determined by many different microstates. It is
assumed that all of this information is contained in FL . This is a non-trivial assumption which
can nonetheless be proven for certain systems. The conversion of the degree of freedom from
µ to ψ is known as coarse-graining, and is at the heart of the relationship between statistical
mechanics and thermodynamics. The next step is to minimize FL [T, ψ], giving:
e−F/kB T ≃ e−F0 /kB T e− min{ψ} FL [T, ψ]/kB T . (14)
This is tantamount to performing a saddle point approximation to the function integral in
Eq. (13).

6 c Peter D. Olmsted, 2015.

4. Examples Lectures on Landau Theory

Here’s a more formal rationalization:

X
e−F/kB T = e−H[µ]/kB T exact Partition Function (15a)
µ
XZ
= Dψ δ[ψ − hµi]e−H[µ]/kB T introduce ψ as an average over µ (15b)
µ
Z X
= Dψ δ[ψ − hµi]e−H[µ]/kB T interchange limits (15c)
µ
• g(ψ) represents the degeneracy of
Z
ψ (number of microstates)
≃ Dψ g(ψ)e−H[ψ]/kB T (15d)
• H[µ] → H[ψ] is generally incor-
rect, but illustrates the idea.
Z
= Dψ e−[H[ψ]−kB T ln g(ψ)]/kB T (15e)

⇒ F ≃ min [H[ψ] − kB T ln g(ψ)] . saddle point approximation (15f)

ψ

Now, the free energy of a system is given by

F = E − T S. (16)

Hence ln g(ψ) is essentially the entropy of the system. To rationalize the assumed form of
the temperature dependence of FL [T, ψ], we write:
F
≃ E0 −E∗ ψ 2 + . . . − T [S0 −aψ 2 + . . . ] (17a)
Volume | {z } | {z }
attraction reduction due to ordering
E∗ 2
= F0 + a(T − a
)ψ + ... (17b)

We assume there is some “attraction” neccessary to induce order in the system, but this oc-
curs at the expense of reducing the entropy; this, in principle, is contained in the degeneracy
g(ψ). The competition of these effects leads to a phase transition.
In a physical system, the steps between Eq. (15c) and Eq. (15d) is exceedingly difficult,
since the relationship between ψ and µ is often not simple, and usually a different functional
form than the appearance of µ in the original Hamiltonian H[µ]. Hence this procedure should
be considered a cartoon; in reality, H[ψ] and g[ψ] are inextricably and inseparably bound
into FL [ψ].†

4 Examples

To make this discussion concrete and useful we must examine some physical systems. We
will do this in order of complexity, and gradually introduce important concepts and gener-
alizations as they appear.
†
P
In the case of polymer blends, where the Hamiltonian is given by H = − α,β φα Vαβ φβ , where φα
denote the mean fractions of species α = A, B, C, . . . in the blend and Vαβ is a matrix of Flory interaction
parameters, the reduction above is essentially correct! This is the RPA (Random Phase Approximation).

7 c Peter D. Olmsted, 2015.

4.1 Ising Magnet Lectures on Landau Theory

4.1 Ising Magnet

Order Parameter —The Ising Model consists of spins which can only point up or down. At
high temperatures the spins are disordered, and at low temperatures the spins spontaneously
choose whether to point up or down. The order parameter is the mean value of the spins,

M = hSi ii+Λ (18)

where the average is within a region Λ about a given spin. Λ is the coarse graining length. To
describe local ordering Λ should be much larger than a lattice spacing a, and to adequately
describe spatial fluctuations (later), Λ should be much smaller than the system size. Since
there are generally 107 orders of magnitude to deal with in a particular system, this separation
of length scales isn’t a problem. For most calculations we don’t need to specify the value of Λ,
but it becomes important when we consider non-uniform terms (later) or the renormalization
group (much later!).
Symmetries—The only symmetry which is relevant for M is that up and down are identical
states, related by a rotation of the sample by π. Since we assume the system is rotationally
invariant (for example, it isn’t in a magnetic field), the free energy in the absence of a
field should be invariant under M → −M. Note that this does not mean that the spins
themselves, or indeed the order parameter, is the same under a flip. This is an important
distinction.
Free Energy—The free energy consistent with this symmetry is
FL
fL ≡ = 21 a(T − Tc )M 2 + 41 cM 4 + . . . (19)
Volume
Higher order terms M 2n are possible, but
we will see that they are not necessary to de-
scribe the transition. As a function of tem-
Free Energy fL

T>Tc
perature, this free energy has minima at
(
0 (T > Tc )
M= q (20)
± a(Tcc−T ) (T < Tc )
for temperatures T < Tc . This is a con-
T<Tc
tinuous transition, since the magnetization
grows smoothly from zero. We shall see that 0

this is a feature of all free energies which, by Order Parameter M

symmetry, have only even powers of the order
parameter.

Thermodynamics—Upon minimizing the free energy, the free energy as a function of tem-
perature is now: 
F0 (T ) (T > Tc )
F = 2 (21)
F0 (T ) − V 1 a|Tc − T | (T < Tc )
2
c

8 c Peter D. Olmsted, 2015.

4.2 Heisenberg Ferromagnet Lectures on Landau Theory

(V is the volume). Hence, we see that ∂F/∂T vanishes at the critical point, while the second
derivative has a discontinuity:

∂2F ∂2F a
− = V. (22)
∂T 2 Tc+ ∂T 2 Tc− c

Hence this is a second order phase transition. This is related to the heat capacity:

∂S ∂2F
CV = T = −T . (23)
∂T V ∂T 2 V

Thus, for the magnet there is a discontinuity (a step decrease) in the heat capacity given by
∆CV = −V T a/c.
The Landau free energy has two free parameters, a and c. In principle, these can be
determined by comparison with experiment from the shape of M(T ) and the heat capacity
jump.

4.2 Heisenberg Ferromagnet

Order Parameter —In the Heisenberg ferromagnet (in, say, 3D), the spin can point anywhere
in space. Hence the magnetization is a vector, defined in the same was as for the Ising Model:

M = hSi ii+Λ (24)

Symmetries—In the magnetic state the system spontaneously chooses a direction in which
to point. In isolation this direction is entirely arbitrary. Hence, the system is rotationally
symmetric with respect to any angle of rotation. Any free energy must be invariant under

M −→ R · M, (25)

where R is a rotation matrix. If M is the only order parameter in the problem, and space is
isotropic, then any free energy must only depend on the invariant M · M. Under a rotation
this invariant transforms as

M · M → M · RT · R · M = M · I · M = M · M, (26a)

where we used the fact that RT · R is the identity matrix I, for any proper rotation.
Free Energy—The free energy consistent with this symmetry is
FL
fL ≡ = 21 a(T − Tc )M · M + 14 c (M · M)2 + . . . (27)
V
Odd order terms are not allowed by symmetry. Hence, like the Ising Magnet, the Heisenberg
Magnet has a second order transition. The primary difference between the two systems
is that, while the Ising magnet breaks up-down symmetry, the Heisenberg magnet breaks
rotational symmetry: M can point anywhere in space. The free energy surface is often
called the “wine bottle” or “Mexican hat” potential. Below the Curie temperature Tc the
system rolls to the bottom of the wine bottle.

9 c Peter D. Olmsted, 2015.

4.3 Nematic Liquid Crystals Lectures on Landau Theory

Mexican Hat Potential

Once the system has broken symmetry, the further behavior at a given temperature
(under, say, the action of a small perturbing field) is much different than that of the Ising
magnet. While the Ising magnet can change only its magnitude, the Heisenberg magnet
can change both magnitude and direction. From the shape of the potential we can see
that changes in magnitude run up the side of the wine bottle, while changes in direction run
around the bottom of the wine bottle and cost no energy! Hence the response, and elasticity,
of the broken symmetry state is dominated by the “soft” modes around the wine bottle.
This is a general feature of all systems which spontaneously break continuous symmetries,
including the big bang! It is arguably one of the most important principles in the universe.
The curvature up the side of the wine bottle is often called a “mass”, and in the context of
the big bang may have something to do with real physical mass.

4.3 Nematic Liquid Crystals

The Isotropic-Nematic transition is an excellent example because, in addition to telling

us about the liquid crystal system, it contains many general features of first order phase
transitions (liquid-gas, liquid-crystal, etc).
Order Parameter —In nematic liquid crystals, rodlike molecules develop an orientation in
b . We have already mentioned that a
space, along an arbitrary direction called the director, n
useful scalar order parameter for this transition is the average hP2 (cos θ)i. However, this is
not a useful order parameter since it is not obvious how to deal with the rotational symmetry.
It is more useful to consider the mean of the orientation vectors ν bi for all rods. Since rods
are assumed to have up-down symmetry (i.e. they look the same [unlike magnetic spins]),
the order parameter itself must be invariant under νb → −b ν . Hence we need something
quadratic in νb. We will use the following tensor order parameter,

Qαβ = hνα νβ − 13 δαβ iΛ , (28)

10 c Peter D. Olmsted, 2015.

4.3 Nematic Liquid Crystals Lectures on Landau Theory

where δαβ is the identity tensor. As before, this is defined within a volume of order Λ3 , much
larger than a molecular volume. For an isotropic distribution of rods all orientations are
uncorrelated, so that hνα νβ i vanishes unless α = β = x, y, z, in which case all directions are
equally probable, να να = 13 . Hence, Q vanishes in the isotropic state and is non-zero in the
nematic state.
Note that the trace of Q, defined as the sum of all diagonal elements, vanishes:
X
Tr Q ≡ Qαα = hbν · νb − 13 · 3i = 0. (29)
α

Hence, although Q has three eigenvalues, only two are independent because of the trace
condition. The most general form of Q is:
 1 
− 3 S1 + S2 0 0
Q= 0 − 31 S1 − S2 0  ≡ S1 (b
nnb − 13 I) + S2 (c
mmc−b
lbl). (30)
2
0 0 S
3 1

If S2 = 0 the system is uniaxial, with principal axis of alignment nb , and S1 = hP2 (cos θ)i.
c and b
For S2 6= 0 the system is biaxial, with m l the major and minor axes of alignment in the
plane normal to n b
b . Note that l = n
b ×mc, so there are two independent axes (hence biaxial).
Symmetries—There are two relevant symmetries to think of for the order parameter.

1. (Lack of) inversion symmetry. Consider a uniaxial state:

 1 
− 3 S1 0 0
Q= 0 − 31 S1 0  (prolate uniaxial) (31)
2
0 0 S
3 1

Under inversion,

1 
S13
0 0
−Q =  0 1
S 0  (oblate uniaxial). (32)
3 1
2
0 0 − 3 S1

Hence the degree of order is qualitatively different under Q → −Q, and odd invariants
are allowed in the free energy.

2. In a homogeneous isolated system the direction of nematic order is arbitrary; that is,
the system and therefore the free energy is rotationally invariant. Q behaves like a
tensor under rotation, because it is a dyad of unit vectors which rotate as usual. Hence
an arbitrary rotation, for which

Qαβ −→ Rαλ Rβρ Qλρ , (33)

must leave the free energy invariant. For a 3D tensor Q, there are two non-trivial
invariants (corresponding to the two independent eigenvalues), Tr Q2 and Tr Q3 . Under

11 c Peter D. Olmsted, 2015.

4.3 Nematic Liquid Crystals Lectures on Landau Theory

a rotation,

Tr Q · Q = Qαβ Qβα → Rαλ Rβρ Qλρ Rβµ Rαν Qµν (34a)

T T
= (R R)νλ (R R)µρ Qλρ Qµν (34b)
= δνλ δµρ Qλρ Qµν (34c)
= Qνµ Qµν (34d)
= Tr Q · Q, (34e)

and similarly for any power Tr Qn .

Free Energy—Respecting the invariants, the free energy is

2
fL = 21 a (T − T∗ ) Tr Q2 + 31 b Tr Q3 + 41 c Tr Q2 . (35)

We can now insert the general form of Q, Eq. (30), into the free energy, and minimize over
S1 and S2 . We need to compute quantities such as
 2      2
1 1
2 2
Tr Q = S1 Tr n bnb − I + 2S1 S2 Tr n bnb− I · m cm c−b lbl + S22 Tr mcmc−b lbl
3 3
 
1 2
= S Tr[b nn 1 2
b + I] − S1 S2 Tr m cm b b
c − l l + S 2 Tr[c mm c+b lb
l] (36a)
3 1 3 3 2

= 32 S12 + 2 S22, (36b)

using properties of unit vectors and the orthonormal system {b c, b

n, m l}. The free energy
reduces to (up to algebraic errors!)
 
fL = 13 a(T − T∗ ) S12 + 27
2
b S13 + 91 c S14 + S22 a(T − T∗ ) − 34 b S1 + 23 c S22 + c S24 . (37)
| {z } | {z }
F1 F2
Free Energy

TIN

Order Parameter S1

Thermodynamics In principle we must look for minima of fL as a function of S1 and S2 . F1

gives, for S2 = 0, a minima at T > T∗ for bS1 negative, due to the cubic term. This will

12 c Peter D. Olmsted, 2015.

4.3 Nematic Liquid Crystals Lectures on Landau Theory

make the term in square brackets in F2 positive, so the system is stable against S2 becoming
non-zero, and is in fact a uniaxial state. So, we consider S2 = 0 and minimize over S1 :

∂F1 0
= 0 −→ S1∗ = 1 h √ i (38)
∂S1 S2 =0  −b ± b2 − 24ac .
4c
By inspection, we take the root with the largest magnitude; now we must change the tem-
perature until the free energy at S1∗ vanishes, at which point the system makes a first order
phase transition to the nematic state. (Somewhat tedious) algebra gives:

b b2
S1∗ = − , ∆T = TIN − T∗ = . (39)
6c 27ac

If want a prolate uniaxial state we must have S1 positive, in which case we must choose
b < 0. Conventionally, then, the Landau Free energy for a nematic liquid crystal is usually
written as fL = . . . − 31 b Tr Q3 + . . ., with b > 0. We’ll continue here, however, with a general
b. Substituting into the free energy, we find
(
0 (T > TIN )
f = f0 + 1 2
(40)
2
a(T − T∗ )S 1∗ + . . . (T ≤ TIN ).

The free energy is continuous at TIN , but has a kink. This kink is related to the entropy
change, for we can calculate
(
S ∂f ∂f0 0 (T > TIN )
=− =− + ∂ 1 2
(41)
V ∂T ∂T ∂T 2
a(T − T∗ )S1∗ + . . . (T ≤ TIN ).

Hence the entropy change upon cooling through the transition temperature is
b4
∆S = ST =T − − ST =T + = − (42)
IN IN 729 a c3
The entropy of the system decreases, which is the latent heat L = TIN ∆S liberated upon
cooling into the nematic state.
Notes:

1. Like the magnet, the liquid crystal spontaneously breaks rotational symmetry, with a
b chosen at random by the system.
direction (director) n

2. The Landau free energy for a first order transition is not strictly correct just below
the transition, since the order parameter grows from a finite value. Hence all results
can only be qualitative at best, even within mean field theory. Conversely, the Landau
energy for a second order transition is exact, within mean field theory, close to the
transition.

3. There are three parameters, a, b, c, T∗ , which can be fitted from experiment by measur-
ing, for example, TIN , S1∗ , ∆T , and the latent heat. ∆T can be measured from light
scattering far above TIN .

13 c Peter D. Olmsted, 2015.

4.4 Crystal Systems Lectures on Landau Theory

4. Other mean field first order phase transitions also typically have a cubic term (e.g.
most crystal phases).

Biaxial Phases—Consider the case where the cubic term vanishes, b = 0. One could imagine
such a system by, for example, mixing molecules of different shapes. In this case the free
energy is all even terms in S1 and S2 :
 
fL = 31 a(T − T∗ ) S12 + 91 c S14 + S22 a(T − T∗ ) + 32 c S22 + c S24 . (43)

This is thus a second order phase transition, and it turns out to have both S1 and S2 non-
zero, so it is a biaxial phase transition. There have been very few examples of biaxial phases;
one can see that the chemistry has to be very special for such a phase.

S1 biaxial

prolate uniaxial oblate uniaxial

−0.4 −0.2 0 0.2 0.4

Cubic Term b

4.4 Crystal Systems

With crystal systems we will round off the most common set of classical phase transitions.
In this case the order parameter is the density (or modulation of some other quantity, such
as composition, charge, etc))
X 
ρ(r) = ρ̄ + ρ(q)e−iq·r + c.c , (44)
q∈G

The Landau free energy is usually expanded in the Fourier coefficients of the density, ρ(q).
Note that these are complex numbers, and hence contain phase information. In particular,
note that ρ∗ (q) = ρ(−q), which can be seen by taking the complex conjugate of ρ(r), which
must be the same thing as ρ(r) (a real number).

4.4.1 Lamellar Systems

Lamellar systems are one-dimensional crystals, and include magnetic systems, smectic liquid
crystals, and lamellar block copolymers. A wavevector is assumed, q = 2π nb /λ, pointing in
b with wavelength λ.
a direction n

14 c Peter D. Olmsted, 2015.

4.4 Crystal Systems Lectures on Landau Theory

Order Parameter —The order parameter is taken to be the single Fourier mode ρ(q). Other
harmonics will be stabilized at lower temperatures. This is an assumption, and in some cases
one could have, due to the microscopic physics, a coincidence of transition temperatures for
several harmonics.
Symmetries—The only relevant symmetry is translational invariance. Although we will see
that the lamellar state breaks translational invariance, the lamellae can be slid anywhere, and
the high temperature state has no preferred point of reference. Under translation r → r + a,
the density transforms to
X 
ρ(r) → ρ(r + a) = ρ̄ + ρ(q)e−iq·r−iq·a + c.c . (45)
q∈G

Hence we have ρ(q) → ρ(q)e−iq·a . That is, all Fourier modes pick up a phase shift. For the
free energy to possess translational invariance, there can be no dependence on the complex
phase of the Fourier modes.
Free Energy—So, we now expand in a single Fourier mode. The lowest order expansion of
the Landau free energy is
1 1
fL = a(T − Tc )ρq ρ−q + b (ρq ρ−q )2 (46a)
2 4
1 1
= a(T − Tc ) |ρq |2 + b |ρq |4 . (46b)
2 4
Since the free energy only depends on the modulus of ρ(q), there is no dependence on
the phase. There is no cubic term, since we always assume an analytic expansion of the
free energy, which precludes terms like |ρ(q)|3/2 . Hence, one-dimensional crystals are, within
mean field theory, continuous transitions, with spontaneously broken translational symmetry.
That is, the direction is arbitrary, as is the phase of the lattice. We will see later that this
transition is ripe for all sorts of interesting fluctuation effects.

4.4.2 Two and Three Dimensional Crystals

Order Parameter — Consider first a 2-dimensional hexagonal crystal. In this case there are
three wavevectors in the first harmonic G∗ = {q1 , q2 , q3 }, of equal magnitude and at relative
angles of 60◦ and 120◦ degrees. They form a star of wavevectors. The Landau free energy
for the transition to a hexagonal crystal must thus be expanded in this star:
X  
ρ(r) = ρ̄ + ρ(q)e−iq·r + c.c , (47)
q∈G∗

Free Energy—As with the lamellar crystal, the free energy must be constructed so that
it is translationally invariant, and doesn’t depend on the absolute position of the lattice.
Consider the transformation of the following terms under a translation, r → r + a:
ρk1 ρk2 −→ ρk1 ρk2 e−i(k1 +k2 )·a (48a)
−i(k1 +k2 +k3 )·a
ρk1 ρk2 ρk3 −→ ρk1 ρk2 ρk3 e (48b)
−i(k1 +k2 +k3 +k4 )·a
ρk1 ρk2 ρk3 ρk4 −→ ρk1 ρk2 ρk3 ρk4 e (48c)

15 c Peter D. Olmsted, 2015.

4.4 Crystal Systems Lectures on Landau Theory

Hexagonal lattice and its star of wavevectors 2 distinct triangles (cubic terms)

Two distinct families

of quartic terms

We can have quadratic terms in the free energy for combinations of Fourier modes for which
k1 + k2 = 0 , cubic terms for combinations for which which k1 + k2 + k3 = 0, and quartic
terms for combinations for which k1 +k2 +k3 +k4 = 0. Note that these are vector sums of the
allowable wavevectors in the star for the relevant crystal, in this case the hexagonal crystal.
Hence it’s easiest to find which terms are allowed by constructing the allowable triangles (for
the cubic terms) and quadrangles (for the quartic terms) from the star of wavevectors.
Hence we have the following terms in the free energy:
ρ1 ρ−1 , ρ2 ρ−2 , ρ3 ρ−3 , 3 quadratic terms (49a)
ρ1 ρ−2 ρ3 , ρ−1 ρ2 ρ−3 2 distinct triangles (49b)
ρ1 ρ2 ρ−1 ρ−2 . . . 3 open quadrangles (49c)
ρ1 ρ−1 ρ1 ρ−1 . . . 3 double quadratic terms. (49d)
Now, we assume a symmetry between all three wavevectors, since the system is rotationally
invariant. That is, we’ll assume the magnitude of each Fourier mode is the same:
ρ1 = ψeiφ1 , ρ2 = ψeiφ2 , ρ3 = ψeiφ3 , (50)
where ψ is the magnitude and φi are arbitrary (so far) phase factors.
Hence all three quadratic terms must be equivalent, and have the same coefficient in
the free energy. Similarly, all three cubic terms are the same, so they must have the same
coefficient. However, the two classes of quartic terms are distinctly different, and each can,
potentially, have its own separate coefficient. So, we are led to the following free energy:
fL = 12 a (T − T∗ ) ψ 2 + 31 b ψ 3 cos (φ1 − φ2 + φ3 ) + 14 (c1 + c2 )ψ 4 . (51)

16 c Peter D. Olmsted, 2015.

4.4 Crystal Systems Lectures on Landau Theory

In a purely phenomenological approach we might as well set c1 + c2 = c, another arbitrary

constant‡ . In some systems (notably block copolymers) c1 and c2 can be calculated from
“first principles”, assuming Gaussian chains. The phase factors are chosen to minimize the
free energy: choosing φ1 − φ2 + φ3 = 0 gives the largest magnitude for the cosine, so the
cubic term can then contribute to the free energy for a non-zero ψ (note that the sign of ψ
doesn’t have any physical significance). Note that the phase factors cannot be determined
to any more precision: higher order terms in the free energy lead to a unique determination
of the relative phases.
Notes—

1. All real three dimensional crystals have first order phase transitions because “triangles”
can be made. Note that a 2D square lattice has, within mean field theory, a continuous
phase transition!

2. One expects single-starred crystals (i.e. square instead of rectangular, hexagonal in-
stead of rhombic) to appear from the melt, since these will typically be triggered by the
first length scale which becomes unstable. For deeper quenches, or strong first order
phase transitions, other harmonics develop.

3. By counting the number of triangles, essentially, one can make a very beautiful and
general argument that most simple substances should become BCC crystals immedi-
ately from the melt§ . This simple rule is obeyed astonishingly often!

4. An excellent example of this approach combined with a derivation of the Landau coef-
ficients is Leibler’s treatment of microphase separation in block copolymers¶ . He pre-
dicted the stability of BCC, hexagonal, and lamellar phases from a single-star theory.
Enlarging the theory to include the second star (harmonic) can predict the stability of
the Gyroid phasek

‡
For higher order terms the different classes of pentangles, sextangles, etc could have different phase
factors, in which case the corresponding coefficients should be kept distinct.
§
S. Alexander and J. McTague, “Should all crystals be bcc? Landau theory of solidification and crystal
nucleation”, Phys. Rev. Lett. 41 (1978) 702-705
¶
L. Leibler, “Theory of Microphase Separation in Block Copolymers”, Macromolecules 13 (1980) 1602-
1617
k
S. T. Milner and P. D. Olmsted, “Analytical weak-segregation theory of bicontinuous phases in diblock
copolymers”, J. Phys. II (France) 7 (1997) 249-255.

17 c Peter D. Olmsted, 2015.

5. Smectic Liquid Crystals Lectures on Landau Theory

5 Smectic Liquid Crystals

5.1 Mean Field Theory of the Nematic-Smectic–A transition

A thermotropic smectic liquid crystal has the symmetry of a one-dimensional lamellar crystal,
but suffers the complications of occuring out of a nematic phase which has broken rotational
symmetry. In the smectic–A phase the rodlike molecules are, on average, parallel to the
layer normals. Hence the director, nb (r), is parallel to the layer normal.
Order Parameter —The order parameter of the smectic transition is taken to be the Fourier
component of the fundamental harmonic of the one-dimensional crystal; i.e. we expand the
density as 
ρ(r) = ρ̄ + ψe−iq nb ·r + ψ ∗ eiq n·r , (52)
b is both the director and the definition of the layer normals.
where n

z θ
n
x n

Symmetries— The layers can be oriented in any direction, as long as the director is also
pointing in that direction. Hence any smectic free energy that we write down must be
invariant under simultaneous rotations of the layers and the director. Under a uniform
rotation of a smectic system about the yb axis by an angle θ, the director changes:

n b′ = n
b −→ n b + δn
b=n
b −θx
b. (53)

Hence the smectic order parameter changes according to:

ψ −→ ψeiq θx = ψe−iqx δnx . (54)

Any free energy we write down my respect this symmetry (rotational invariance).
Free Energy—Now we write down the free energy. First we consider a uniform amplitude of
the density modulation ψ. Since a smectic is a one-dimensional crystal, we can reproduce
what we had before:
fL,hom = 21 a(T − T c) |ψ|2 + 41 c |ψ|4 . (55)
To this we add the energy of deformation of the director fluctuations, the Frank Free Energy:

b )2 + 21 K2 (b
fN = 12 K1 (∇ · δ n b )2 + 13 K3 |b
n · ∇ × δn b )|2 ,
n × (∇ × δ n (56)

18 c Peter D. Olmsted, 2015.

5.2 Free Energy and Superconducting Analogy Lectures on Landau Theory

where K1 , K2 , K3 penalize, respectively, splay, bend, and twist fluctuations.

Finally, we can consider an envelope of spatial modulations in the magnitude of the density
modulation, ψ(r). These fluctuations, in the smectic state, correspond to phonons. If we
consider the rigid body rotation above, transverse spatial derivatives of ψ pick up a term
from the director rotation:

∂x ψ(r) −→ (∂x ψ) e−iq xδnx + iq δnx ψ(r) e−iq xδnx . (57)

| {z }
(rotate about y)
b

Since the free energy must be invariant under a rigid body rotation, such spatial deriva-
tives must not contribute to the free energy. Hence the free energy of deforming the order
parameter is
n · ∇ ψ|2 + 21 g⊥ |(∇⊥ − iqδ n
fL,def = 12 gk |b b ) ψ|2 . (58)
The additional term in parentheses in the second terms subtracts off the unwanted irrelevant
term. This derivative is often called a covariant derivative, and has an analogy in super-
conductivity, in which the superconductor is analogous to the smectic order parameter and
the vector potential is analogous to the director fluctuations. This is an example of a gauge
field theory, which arises when the relevant order parameter has a local internal symmetry
(the smectic order parameter is a complex number with a phase, corresponding to a U(1)
symmetry, which can vary from point to point). There have been many interesting fruits
born out of this analogy between smectics and superconductors! In electrodynamics there
is a gauge symmetry (the vector potential A is defined only up to an additive irrotational
vector field ∇ Θ), while in the smectic liquid crystal the gauge is actually fixed because of
the presence of a director in the nematic state.

5.2 Free Energy and Superconducting Analogy

To compare the liquid crystals and superconductors, let’s look at the free energies. For the
liquid crystal we have

fN A = 12 a |ψ|2 + 41 c |ψ|4 + 21 gk |b
n · ∇ ψ|2 + 12 g⊥ |(∇⊥ − iqδ n
b ) ψ|2 + 21 K (∇α δn̂β )2 , (59)

where we have used the so-called one-constant approximation K1 = K2 = K3 ≡ K.

For a superconductor, we’ll take the order parameter to be ψ again. Landau postulated
that this is a complex number, and in fact it turns out to be closely related to the wavefunc-
tion for Cooper pairs of electrons (which is a complex number). The physics of electrons
is invariant under a local change in the phase of the wavefunction, ψ(r) → ψ(r)eiφ(r) . This
transformation is in fact the gauge invariance which gives rise to the electromagnetic field,
and we usually write R
i2e r A·dr′
ψ(r) → ψ(r)e r0 , (60)
where A is the electromagnetic vector potential (this funny phase factor gives rise to the
Aharonov-Bohm effect), in units where the speed of light is unity. There is a 2 here because
the wavefunction is for a pair of electrons. Now, we can write down the free energy for a
superconducting transition, exactly as for the smectic. When it comes to writing the gradient
terms we must “subtract” off this arbitrary phase, which can’t change the physics, so gradient

19 c Peter D. Olmsted, 2015.

5.2 Free Energy and Superconducting Analogy Lectures on Landau Theory

operators generally enter as ~∇ − i2eA. Finally, there is a free energy associated with gauge
field fluctuations: the energy in the magnetic field is B · B/8π. Writing B = ∇ × A, we
have, for the superconducting free energy,
  2
2 4 2e c2
1 1 1
fsup = 2 a |ψ| + 4 c |ψ| + 2 g ∇ − i A ψ + |∇ × A|2 . (61)
~ 8π
Hence, aside from the fact that the liquid crystal is anisotropic, and the coupling of A to the
b , the transitions
gradient terms is slightly different in tensorial nature to the coupling of δ n
“should” be qualitatively the same. Just as a superconductor expels magnetic flux, a liquid
crystal expels (it turns out) both twist and bend director fluctuations. For example, it was
realized that the analog of the Abrikosov vortex phase of Type II superconductors should
exist in liquid crystals, and this was subsequently predicted and then found experimentally:
the “twisted grain boundary (TGB) phase” in cholesteric smectics.
The phase diagram of a smectic looks something like the following:

Isotropic

first order lines

Nematic
T
tricritical point

??????? Smectic A

(molecular length)-1
The nematic phase intervenes when the molecules are anisotropic enough, and otherwise the
system can go directly from the isotropic to the smectic phase. One can show quite generally
that the following term, which couples liquid crystalline order to density,∗∗

fψQ = Qαβ ∇α ρ ∇β ρ (62)

∗∗
Can you show that this term is allowed by symmetry?

20 c Peter D. Olmsted, 2015.

5.3 Fluctuations and the Halperin-Lubensky-Ma Effect. Lectures on Landau Theory

leads (via nematic fluctuations) to an effective quartic term in the smectic free energy of
the form −χψ 4 , where χ is the susceptibility for fluctuations around the nematic phase.
Hence, it is likely that for molecular parameters with a narrow nematic range, for which the
susceptibility χ is large, the total quartic term in the smectic free energy is negative, leading
to a first order phase transition. For parameters with a wide nematic range the N − A
transition is expected, within mean field theory, to be a continuous transition. In the next
section we will see that this does not, generally, survive inclusion of fluctuations.

5.3 Fluctuations and the Halperin-Lubensky-Ma Effect.

b are fluctuating. If we consider a

In the region of the smectic transition both ψ and δ n
fluctuation in which a smectic blob appears in the nematic, there are two relevant length
scales:
r
gk,⊥
ξk,⊥ = (ψ healing length) (63)
a
s
K
λ= (bend and twist penetration lengths) (64)
gk,⊥ q02 |ψ|2

The Frank constant K is proportional to the magnitude of the nematic order parameter Q,
so λ is large when the nematic phase has a large range and smaller when the nematic range
is smaller. Hence we expect two regimes:
 1

 <√ (Type I)
λ 2
κ≡ 1 (65)
ξ
> √ (Type II)
2
In the Type I case the director fluctuation (or magnetic field) decays quickly compared to
the smectic order parameter, and the system can efficiently expel twist (or magnetic field).
This corresponds to the Meissner phase in superconductors. In the Type II case the twist
can penetrate a longer distance while order parameter has healed, and can in fact penetrate
easier. This corresponds to the Abrikosov vortex phase in superconductors.
If we consider fluctuations in the nematic phase, in the Type I limit we can safely assume
that ψ is uniform relative to director fluctuations (since ψ varies much more smoothly). To
examine the smectic transition in this limit, then, we consider small (uniform) fluctuations
in ψ, and integrate out the director fluctuations. Hence:
Z R 3
Z = Dδ n b Dψ e{ d rfNA /kB T } (66)
Z 
V 1 
≃ Dδ n b Dψ exp − 2
a |ψ|2 + 41 c |ψ|4 + (67)
kB T
Z 
1 3
1 2 1 2
kB T
d r 2 g⊥ |(∇⊥ − iq0 δ n b ) ψ| + 2 K (∇α δn̂β ) , (68)
Z Z
≃ Dψe−V f0NA [ψ]/kB T δ n b e−Fnem /kB T (69)

21 c Peter D. Olmsted, 2015.

5.4 Layered liquids and the Brazovskii effect Lectures on Landau Theory

where f0N A is the free energy in the absence of director fluctuations, and
Z

1
b |2 + K (∇α δn̂β )2 .
Fnem = 2 d3r g⊥ q02 |ψ|2 |δ n (70)

The director fluctuationPintegral is Gaussian, and can be done without too much trouble. In
b (q)eiq·r , we can write
b (r) = q δ n
Fourier space, δ n
X 
Fnem = 1
2
b (q)|2 .
g⊥ q02 |ψ|2 | + Kq 2 |δ n (71)
q

The director integral is a Gaussian integral for each Fourier mode, giving (up to factors of
2π) Z Y − 1 2
b e−Fnem /kB T ≃
δn g⊥ q02 |ψ|2 | + Kq 2 2 (72)
q

(a power of 21 for each independent mode of δ n b ). Hence the effective free energy for the
smectic transition is
Z R 3
ˆ
Z = Dδ n b Dψ e{ d rfNA /kB T } (73)
X  
=⇒fˆN A = f0N A + kB T ln g⊥ q02 |ψ|2 + Kq 2 (74)
q

P R
This integral can be done ( q → 1/(2π)3 d3q), and expanded in the order parameter ψ.
Note that it can’t be expanded first because the integrand is non-analytic at ψ = 0. The
result is  3/2
ˆ 2 1 g⊥ q02 |ψ|2
fN A = f0N A + a1 |ψ| − + a2 ψ 4 + . . . (75)
6π K
Hence, there is a cubic term induced in the magnitude of ψ. This leads to a very weak
fluctuation-induced first order transition. This seems to have been verified experimentally
(most people agree).
In the case of Type II systems, where director and order parameter fluctuations (or mag-
netic field and superconducting order parameter), a renormalization group type of treatment
is necessary. It is believed that the first order transition remains, but results are inconclusive.
In any case if it remains first order it is VERY weak!

5.4 Layered liquids and the Brazovskii effect

To be added!

22 c Peter D. Olmsted, 2015.