Professional Documents
Culture Documents
Dirty Tricks For Statistical Physics 1st Ed by Grant PDF
Dirty Tricks For Statistical Physics 1st Ed by Grant PDF
Dirty Tricks For Statistical Physics 1st Ed by Grant PDF
Martin Grant
Physics Department, McGill University
°
c MG, August 2004, version 0.91
ii
Preface
These are lecture notes for PHYS 559, Advanced Statistical Mechanics,
which I’ve taught at McGill for many years. I’m intending to tidy this up
into a book, or rather the first half of a book. This half is on equilibrium, the
second half would be on dynamics.
These were handwritten notes which were heroically typed by Ryan Porter
over the summer of 2004, and may someday be properly proof-read by me.
Even better, maybe someday I will revise the reasoning in some of the sections.
Some of it can be argued better, but it was too much trouble to rewrite the
handwritten notes. I am also trying to come up with a good book title, amongst
other things. The two titles I am thinking of are “Dirty tricks for statistical
mechanics”, and “Valhalla, we are coming!”. Clearly, more thinking is necessary,
and suggestions are welcome.
While these lecture notes have gotten longer and longer until they are al-
most self-sufficient, it is always nice to have real books to look over. My favorite
modern text is “Lectures on Phase Transitions and the Renormalisation Group”,
by Nigel Goldenfeld (Addison-Wesley, Reading Mass., 1992). This is referred
to several times in the notes. Other nice texts are “Statistical Physics”, by
L. D. Landau and E. M. Lifshitz (Addison-Wesley, Reading Mass., 1970) par-
ticularly Chaps. 1, 12, and 14; “Statistical Mechanics”, by S.-K. Ma (World
Science, Phila., 1986) particularly Chaps. 12, 13, and 28; and “Hydrodynamic
Fluctuations, Broken Symmetry, and Correlation Functions”, by D. Forster (W.
A. Benjamin, Reading Mass., 1975), particularly Chap. 2. These are all great
books, with Landau’s being the classic. Unfortunately, except for Goldenfeld’s
book, they’re all a little out of date. Some other good books are mentioned in
passing in the notes.
Finally, at the present time while I figure out what I am doing, you can read
these notes as many times you like, make as many copies as you like, and refer
to them as much as you like. But please don’t sell them to anyone, or use these
notes without attribution.
Contents
1 Motivation 1
3 Independence of Parts 7
3.1 Self Averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Central Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 Correlations in Space . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Toy Fractals 21
4.1 Von Koch Snowflake . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2 Sierpinskii Triangle . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Correlations in Self-Similar Objects . . . . . . . . . . . . . . . . . 27
5 Molecular Dynamics 31
7 Review of Thermodynamics 49
8 Statistical Mechanics 55
9 Fluctuations 59
9.1 Thermodynamic Fluctuations . . . . . . . . . . . . . . . . . . . . 62
10 Fluctuations of Surfaces 67
10.1 Lattice Models of Surfaces . . . . . . . . . . . . . . . . . . . . . . 70
10.2 Continuum Model of Surface . . . . . . . . . . . . . . . . . . . . 72
10.3 Impossibility of Phase Coexistence in d=1 . . . . . . . . . . . . . 88
10.4 Numbers for d=3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
iii
iv CONTENTS
12 Scattering 99
12.1 Scattering from a Flat Interface . . . . . . . . . . . . . . . . . . . 103
12.2 Roughness and Diffuseness . . . . . . . . . . . . . . . . . . . . . . 106
12.3 Scattering From Many Flat Randomly–Oriented Surfaces . . . . 108
Motivation
0 2 4 6 8 10
h
Figure 1.1: Bias of Flips With Height h
1
2 CHAPTER 1. MOTIVATION
There are a lot of other things one could experimentally measure. In fact, my
cartoon should look exponential, for reasons we will see later on in the course.
Now we have something to think about: Why is the coin toss biased for
flips of small height? In statistical mechanics language we would say, “why is
the coin toss correlated with its initial state?” Remember the coin was always
prepared as heads. (Of course, if it was prepared randomly, why flip it?) There
is a length scale over which those correlations persist. For our example, this
scale is ∼ 2–3cm, which we call ξ, the correlation length. The tailing off of the
graph above follows:
Correlations ∼ e−ξ/h
which we need more background to show. But if we did a careful experiment,
we could show this is true for large h. It is sensible that
ξ ≥ coin diameter
The topics we will cover are as current as any issue of Physical Review E. Just
take a look.
We’ll begin the proper course with a review of, and a reflection on, some
basic ideas of statistical mechanics.
4 CHAPTER 1. MOTIVATION
Chapter 2
where Estate is the energy of the state, and kB is Boltzmann’s Constant. The
probability of being in a state is
1 −Estate /kB T
ρ= e (2.2)
Z
where the partition function is
X
Z= e−Estate /kB T (2.3)
{states}
F = E − TS (2.4)
5
6 CHAPTER 2. REVIEW OF STAT MECH IDEAS
where F = f N . This exponential factor of N arises from the sum over states
which is typically X
= eO(N ) (2.6)
{states}
For example, for a magnetic dipole, which can have each constivent up or down,
there are evidently
1030 m
∼ 1040 (2.9)
10−10 m
This is in no way comparable to (!)
23
1010 (2.10)
In fact, it seems a little odd (and unfair) that a theorist must average over eN
states to explain what an experimentalist does in one experiment.
Chapter 3
Independence of Parts
It is worth thinking about experiments a bit to make this clear. Say an ex-
perimentalist measures the speed of sound in a sample, as shown. The experi-
mentalist obtains some definite value of c, the speed of sound. Afterwards, the
experimentlalist might cut up the sample into many parts so that his or her
friends can do the same experiment. That is as shown in Fig. 3.2. Each of
7
8 CHAPTER 3. INDEPENDENCE OF PARTS
these new groups also measure the sound speed as shown. It would seem that
the statistics improve because 4 experiments are now done?! In fact one can
continue breaking up the system like this, as shown in Fig. 3.3.
ξ ≥ O(Ȧ) (3.2)
and indeed usually ξ ≈ O(Ȧ). Hence for a macroscopic length, the number of
independent points is
N ∼ O(1023 ) (3.3)
3.1. SELF AVERAGING 9
of course. Now, each independent part has a few degrees of freedom, such as
quantum numbers. Hence, the total number of states in a typical system are
R R R
dq1 dq2 ... dqN
quantum quantum quantum
states states states
of part 1 of part 2 of part N
R
Say dqi = 4 for an example, so that the total number of states is
eO(N ) (3.4)
as it should. Now, however, we have a feeling for the origin of some of the
assumptions entering statistical physics.
We can go a little farther with this, and prove the central–limit theorem of
probability theory.
N
X
X= Xi (3.5)
i=1
then
hxi = O(1)
This is not very amazing yet. Before continuing, it is worth giving an example
of what we mean by the average. For example, it might be a microscopically
long, but macroscopically short time average
R
dt X(t)
hXi = τ
τ
as drawn. As also indicated, the quantity h(X − hXi)2 i1/2 is a natural one to
X
h(X − hXi)2 i1/2
hXi
τ t
∆X = X − hXi (3.9)
Then clearly
h∆Xi = hXi − hXi = 0
as is clear from the cartoon. The handier quantity is
h(∆X)2 i = h(X − hXi)2 i
= hX 2 − 2XhXi + hXi2 i (3.10)
2 2
= hX i − hXi
Giving the condition, in passing,
hX 2 i ≥ hXi2 (3.11)
since h(∆X)2 i ≥ 0. Splitting the system into independent parts, again, gives
XN N
X
h(∆X)2 i = h ∆Xi ∆Xj i
i=1 j=1
and,
N
X
h(∆X)2 i = h(∆Xi )2 i
i=1
or
h(∆X)2 i = O(N ) (3.14)
Also, the intensive x = X/N satisfies
1
h(∆x)2 i = O( ) (3.15)
N
This implies that fluctuations in quantities are very small in systems of many
(N → ∞) independent parts. That is, because
h(∆X)2 i1/2
X ≈ hXi{1 + O( )}
hXi (3.16)
1
≈ hXi{1 + O( 1/2 )}
N
So, in essence, X ≈ hXi. The result
h(∆X)2 i 1
= O( ) (3.17)
hXi2 N
that relative fluctuations are very small, if N is large, is called the self-averaging
lemma. Note that, since N = V /ξ 3 ,
h(∆X)2 i ξ3
2
= O( ) (3.18)
hXi V
h(∆X)n i (3.19)
12 CHAPTER 3. INDEPENDENCE OF PARTS
h(∆X)2n+1 i = 0
for intiger n. This means the distrobution ρ is symetric about the mean hXi as
shown. It turns out that one can show that the distrobution function is symetric
ρ ρ ρ
as (N → ∞), but lets just assume it so the ideas in the algebra are clear. It can
be shown (see below) that we can then reconstruct the distrobution ρ(X) if we
know all the even moments
h(∆X)2n i
for positive integer n. We know n = 1, so let’s do n = 2, as we did h(∆X)2 i
above
XN N
X N
X N
X
h(∆X)4 i = h ∆Xi ∆Xj ∆Xk ∆Xl i
i=1 j=1 k=1 l=1
N N
(∆Xi )2 (∆Xj )2 i + Oh∆Xi:
X X 0 (3.20)
=h (∆Xi )4 i + 3h i
i=1 1=1
(i=j=k=l) j=1
(i6=j)
= O(N ) + O(N 2 )
So as (N → ∞)
N
X N
X
h(∆X)4 i = 3h (∆Xi )2 ih (∆Xj )2 i
i=1 j=1
or
h(∆X)4 i = 3h(∆X)2 i2 = O(N 2 ) (3.21)
Note that the largest term (as (N → ∞)) came from pairing up all the ∆Xi ’s
appropriately. A little thought makes it clear that the largest term in h(∆X)2n i
will result from this sort of pairing giving,
There are (2n − 1) ways to pair off the first ∆X. After those two ∆X’s are
gone, there are then (2n − 3) ways to pair off the next free ∆X. And so on,
until finally, we get to the last two ∆X’s which have 1 way to pair off. Hence
or more conventionally,
2n!
h(∆X)2n i = h(∆X)2 in (3.23)
2n n!
Now we know all the even moments in terms of the first two moments hXi and
h(∆X)2 i. We can reconstruct the distribution with a trick. First, let us be
explicit about ρ. By an average we shall mean,
Z ∞
hQ(X)iX = ρ(X)Q(X) (3.24)
−∞
where δ(X) is the Dirac delta function! Recall the Fourrier representation
Z ∞
′ 1 ′
δ(X − X ) = dk e−ik(X−X ) (3.26)
2π −∞
′
Averaging over X gives
Z ∞
′ 1 ′
ρ(X) = hδ(X − X )iX ′ = dk e−ikX heikX iX ′ (3.27)
2π −∞
In statistics, heikX i is called the characteristic function. But we can now expand
the exponential (for convenience, let hXi = 0 for now, one only has to add and
′ ′
subtract it in e−ik(X−X ) = e−ik(∆X−∆X ) but it makes the algebra messier) as
∞
′ X (ik)n ′ n
e−ikX = (X )
n=0
n!
So that we have,
Z ∞ ∞
1 ikX
X (ik)n n
ρ(X) = dk e hX i (3.28)
2π −∞ n=0
n!
14 CHAPTER 3. INDEPENDENCE OF PARTS
where the prime superscript has been dropped. Note that Eq. 3.28 means that
ρ(X) can be reconstructed from all the hX n i, in general. For our case, all the
odd moments vanish, so
Z ∞ ∞
1 X (−k 2 )n 2n
ρ(X) = dk eikX hX i
2π −∞ n=0
(2n)!
(2n)!
But hX 2n i = 2 n
2n n! hX i from above, so,
Z ∞ ∞
1 X 1 −k 2 2 n
ρ(X) = dk eikX ( hX i)
2π −∞ n=0
n! 2
−X 2
Z ∞ khX 2 i1/2
1 2hX −(
21/2
+ iX
2hX 2 i1/2
)2
ρ(X) = e 2i dk e
2π −∞
√
π
−X 2
s Z ∞ *
1 2hX 2 2
= e 2i du e−u
2π hX 2 i −∞
hxi = O(1)
(3.31)
h(∆x)2 i = O(1/N )
Letting
a2
h(∆x)2 i ≡ (3.32)
N
where a2 = O(1) we have, up to normalization,
(x−hxi)2
ρ(x) ∝ e−N 2a2 (3.33)
so the width of the Gaussian is exceedingly small and it most represents a spike.
3.3. CORRELATIONS IN SPACE 15
hxi = O(1) x
Which is to say again that fluctuations are small, and average values are well
hxi = O(1) x
∆x(~r)
~r − ~r′
∆x(~r′)
~r
~r′
o
and, of course,
′ ′
h∆x(~r) ∆x(~r )i = h∆x(~r − ~r ) ∆x(0)i (3.36)
etc.
′
2. If space is isotropic, then the direction of ∆x(~r) from ∆x(~r ) is unim-
′
portant, and only the magnitude of the distance | ~r − ~r | is physically
∆x(~r′)
|~r − ~r′|
∆x(~r)
relevant. Hence ′ ′
C(~r, ~r ) = C(| ~r − ~r |) (3.37)
′
So what was a function of 6 numbers (~r, ~r ) is now a function of only 1,
′
| ~r − ~r |.
3. Usually correlations fall off if points are arbitrarily far apart. That is,
: 0 ∆x(0)i
h∆x(r → ∞) ∆x(0)i = h∆x(r → ∞)ih :0
=0
or
C(r → ∞) = 0 (3.38)
The scale of the correlation is set by the correlation length.
Now we know ′ ′
C(| ~r − ~r |) = h∆x(~r) ∆x(~r )i (3.39)
′
Subject to these conditions. A constraint on C(| ~r − ~r |) is provided by our
previous result
1
h(∆x)2 i = O( )
N
3.3. CORRELATIONS IN SPACE 17
C(r) O(ξ)
2 (~
r − ~r ). The Jacobian of the transformation is unity,
and we obtain
Z Z 2
~′
dR ~ = O( V )
~ C(R)
dR
N
Z (3.42)
=V ~ C(R)
dR ~
~ → ~r,
So letting R Z
V
d~r C(~r) = O( )
N
V Ld
But N = ξ3 , or in d dimensions, N = ξd
, so
Z
d~r C(~r) = O(ξ 3 )
or
Z
d~r h∆x(~r) ∆x(0)i = O(ξ 3 )
(3.43)
18 CHAPTER 3. INDEPENDENCE OF PARTS
and finally
Z
d~k ~
C(~r) = d
e−ik·~r Ĉ(~k)
(2π)
(3.44)
and, Z
~
h|∆x̂(k)|2 i ≡ d~r eik·~r h∆x(~r) ∆x(0)i
or
Many functions satisfy the constraint of Eq. 3.43 (see Fig. 3.11). It turns out
that usually,
h∆x(~r) ∆x(0)i ∼ e−r/ξ (3.46)
for large r, which satisfies Eq. 3.43. This is obtained from analyticity of (Ĉ(k))−1
near k = 0. It can be argued,
1
≈ (const.) + (const.)k 2 + ...
Ĉ(k)
const.
so that h|∆x̂(k)|2 i = const.+k 2 near k = 0. Fourier transformation gives the
or
1 X
∆x = ∆Xi
N i
This means Z
1 1 X
d~r ∆x(~r) = ∆Xi (3.47)
V N i
3.3. CORRELATIONS IN SPACE 19
This is satisfied by
N
V X
∆x(r) = δ(~r − ~ri )∆Xi
N i=1
The purpose of this course is to examine the cases where V /ξ 3 = O(1) and these
results break down! It is worthwhile to give some toy examples now where, in
particular, Eq. 3.43 breaks down.
20 CHAPTER 3. INDEPENDENCE OF PARTS
Chapter 4
Toy Fractals
Toy Fractals are mathamatical constructs that seem to exist in fractional di-
mensions between 1 and 2, or, 2 and 3, or what have you. They can provide a
useful starting point for thinking about real systems where V /ξ 3 = O(1), or if
V = L3 , ξ/L = O(1).
We’ll briefly discuss self-affine, or surface fractals, and volume fractals. This
P ∼ Lds , as L → ∞ (4.1)
Pcircle = πL
Psquare = 4L
21
22 CHAPTER 4. TOY FRACTALS
Usually
ds > d − 1 (4.3)
I’ll give an example below. This is a volume fractal, note all the holes with
missing stuff. The fractal has the propert that its mass M satisfies
M ∼ Ldf as L → ∞ (4.4)
in d dimensions. Usually
df < d (4.6)
Note that since the mass density is M/Ld , this implies that
1
DEN SIT Y OF A F RACT AL ∼ →0 (4.7)
Ld−df
fractal exponent for the perimeter, imagine I had (somehow) drawn one of these
fractals out a very very large number of steps. Say the smallest straight link is
length l. We’ll calculate the perimeter as we go through increasing numbers of
links in a large complex fractal. Hence
Figure 4.4:
Pn = 4n l
when
Ln = 3n l
(4.8)
ln P/l = n ln 4
and
ln L/l = n ln 3
ln P/l ln 4
=
ln L/l ln 3
24 CHAPTER 4. TOY FRACTALS
or
where
ln 4
ds = ≈ 1.26
ln 3
>d−l =1
w(L) ∼ Lx , as L → ∞ (4.11)
where x is the roughening exponent, and clearly, (x ≈ 0.26 for the snowflake),
ds = x + d − 1 (4.12)
exponent for the mass, imagine I had (somehow) drawn one of these fractals
out to a very large number of steps. Say the smallest filled triangle had mass
m and volume equal to v. Then we’ll calculate the mass as we go through
increasing the numbers of triangles. Hence
4.2. SIERPINSKII TRIANGLE 25
Figure 4.6:
Mn = 3n m
when
Vn = 4n v
(4.13)
L2n = 4n l2 or Ln = 2n l
Taking logarithms,
M
ln = n ln 3
m
and
L
ln = n ln 2
l
ln M/m ln 3
=
ln L/l ln 2
and
M L
= ( )df (4.14)
m l
where
ln 3
df = ≈ 1.585 (4.15)
ln 2
So df < d where d = 2 here.
26 CHAPTER 4. TOY FRACTALS
This is not the common notation in the physics literature. Below, but near
a second–order phase transition, the excess density satisfies
n ∝ |T − Tc |β
ξ ∝ |T − Tc |−v
Or, since ξ’s largest value is limited by the edge length L of the entire system
ξ ∼ L ∼ |T − Tc |−v
n ∝ [L−1/v ]β
1
n∝
Lβ/v
1
n∝ d − df
L
so
df = d − β/v (4.16)
where β/v ≈ 0.415 for the Sierpinskii snowflake.
For a real system (in the Ising universality class in two dimensions),
Figure 4.7:
4.3. CORRELATIONS IN SELF-SIMILAR OBJECTS 27
Figure 4.8:
(where the average is, say, over all orientations in space for convenience). Then
the evident self–similarity of fractals implies
C(r) ∝ C(r/b)
or
rk ≡ r/b (4.19)
Then
dC(r) df dC dr∗
= 0 = C(r∗ ) + f ∗
db db dr db
28 CHAPTER 4. TOY FRACTALS
or,
df r∗ dC
C(r∗ ) = + f
db b dr∗
and
d ln f (b) d ln C(r∗ )
= (4.20)
d ln b d ln r∗
Because one side only depends on b, and the other only on r∗ . Hence
d ln f (b)
= const. ≡ −p (4.21)
d ln b
So
f (b) ∝ b−p (4.22)
r
C(r) = (1 + ǫ)−p C( )
1+ǫ
∂C
≈ (1 − pǫ){C(r) − ǫr } + ...
∂r
Terms of order ǫ give
∂C
−p C(r) = r
∂r
The solution of which is
1
C∝ p (4.25)
r
Hence, self similarity implies homogeneous correlation functions
C(r)
1/rp
C(r)
e−r/ξ
ξ
C(r) ∼ e−r/ξ
The restriction Z
d~rC(r) = O(ξ d )
Molecular Dynamics
At this point, one might say, computers are so big and fierce, why not simlpy
solve the microscopic problem completely! This is the approach of molecular
dynamics. It is a very important numerical technique, which is very very con-
servative. It is used by physicists, chemists, and biologists.
Imagine a classical solid, liquid or gas composed of many atoms or molecules.
Each atom has a position ~ri and a velocity ~vi (or momentum m~vi ). The Hamil-
particle i ~vi
~vj
~rj
o
Figure 5.1:
Usually, the total potential can be well approximated as the sum of pairwise
forces, i.e.
N
X
V (r1 , ...rN ) = V (|~ri − ~rj |)
i=1, j=1
(i>j)
which only depends on the distance between atoms. This potential usually has
31
32 CHAPTER 5. MOLECULAR DYNAMICS
a form like that drawn. Where the depth of the well must be
ǫ
r0 r
Figure 5.2:
ǫ ∼ (100′ s K)kB
and
ro ∼ (a f ew Ȧ′ s)
This is because, in a vapor phase, the density is small and the well is unimpor-
tant; when the well becomes important, the atoms condense into a liquid phase.
Hence ǫ ∼ boiling point of a liquid, and ro is the separation of atoms in a liquid.
Of course, say the mass of an atom is
m ∼ 10 × 10−27 kg
ro ∼ 2Ȧ
m 10 × 10−27 kg
ρ∼ 3
=
ro 10Ȧ3
−27
10 × (103 gm)
=
(10−8 cm)3
ρ ∼ 1gm/cm3
where
ro = 21/6 σ ≈ 1.1σ
and
V (ro ) = −ǫ
ro = 3.4Ȧ
ǫ = 120K
(
Lennard – Jones , r < 2ro
V (r) =
0 , r > 2ro
P T
s l and l
v s
v Coexistance
T ρ
But the hard sphere potential has a more simple–minded phase diagram. This
V
∞
, r<σ
V = −ǫ , σ < r < σ′
0 , r > σ′
Figure 5.4:
(
V ∞ , r<σ
V =
0 , r>σ
r
where,
P T
s l
and s
l coex.
T ρ
Figure 5.5:
Clearly, we can simply solve these numerically given all the positions and veloc-
ities initially, i.e. {~ri (t = 0), ~vi (t = 0)}. A very simple (in fact too simple) way
35
to do this is by
d~ri ri (t + ∆t) − ri
=
dt ∆t
for small ∆t. This gives
and
1 ~
~ri (t + ∆t) = ~vi (t) + ∆t Fi
m
where F~i (t) is determined by {~ri (t)} only. A better (but still simple) method
is due to Verlet. It should be remembered that the discretization method has a
little bit of black magic in it.
Valet notes that,
(∆t)2
~ri (t ± ∆t) = ~ri (t) ± ∆t~vi + ~ai (t) + ...
2
d2 ~
where ~ai = F~i /m = ri
dt2 . Add this to itself
1 ~
~ri (t + ∆t) = 2~ri (t) − ~ri (t − ∆t) + (∆t)2 Fi + O(∆t)
m
And us the simple rule
ri (t + ∆t) − ri (t − ∆t)
~vi (t + ∆t) =
2∆t
Thats all. You have to keep one extra set of ri ’s, but it is the common method,
and works well. To do statistical mechanics, you do time averages, after the
system equilibrates. Or ensemble averages over different initial conditions. The
ensemble is microcanonical sine energy is fixed. Usually one solves it for a fixed
volume, using periodic boundary conditions. A common practice is to do a
1 1
mhv 2 i = kB T
2 2
or,
hv 2 i
T =
mkB
This is done by constantly rescaling the magnitude of the velocities to be
p
|v| ≡ hv 2 i1/2 ≡ mkB T
∂~ri
= ~vi
∂t
∂~vi 1 X ∂
= V (|ri − rj |)
∂t m j ∂~ri
(j6=i)
~r → ~r
~v → −~v
How then can entropy increase? This is a deep and uninteresting question.
Entropy does increase; its a law of physics. The real question is how to calculate
how how fast entropy increases; e.g. calculate a viscosity or a diffusion constant,
which set the time scale over which entropy increases.
Nevertheless, here’s a neat way to fool around with this a bit. Take a gas
of N atoms in a box of size Ld where energy is fixed such that the average
equilibrium separation between atoms
where ro is a parameter giving the rough size of the atoms. Now initialize the
atoms in a corner of the box with
s(t = 0) = O(ro )
or smaller (see fig. 5.7). Now, at time t = later, let all ~v → −~v . By microscopic
reversibility you should get fig. 5.8. Try it! You get fig. 5.9. It is round off
37
“Entropy”
later t
Figure 5.7:
later t
Figure 5.8:
“Entropy”
1
later t
Figure 5.9:
error, but try to improve your precision. You will always get the same thing.
Think about it. The second law is a law, not someone’s opinion.
38 CHAPTER 5. MOLECULAR DYNAMICS
Chapter 6
{S} ≡ S (6.4)
As time goes on, the probability of being in state S increases because one makes
transitions into this state from other (presumably nearby in some sence) states
S ′ . Say the probability of going from S → S ′ , in a unit time is
W (S, S ′ ) ≡ W (S ← S ′ ) (6.5)
(Read As)
39
40 CHAPTER 6. MONTE CARLO AND THE MASTER EQUATION
W (S, S ′ )
S′ S
W (S ′ , S)
Figure 6.1:
W (S, S ′ ) P (S ′ ) (6.6)
So that
W (S ′ , S) P (S) (6.8)
are contributions decreasing P (S) as time goes on.
The master equation incorporates both these processes as
∂P (S, t) X
= [W (S, S ′ ) P (S ′ ) − W (S ′ , S) P (S)] (6.9)
∂t ′ S
1. has no memory of events further back in time than the differential element
“dt”. This is called a Markov process
2. is linear.
In equilibrium,
∂Peq (S)
=0 (6.12)
∂t
41
W (S, S ′ ) ′
′
= e−(E(S)−E(S ))/kB T (6.14)
W (S , S)
This gives a surprisingly simple result for the transitions W , it is necessary (but
not sufficient) that they satisfy Eq. 6.14. In fact, if we are only concerned with
equilibrium properties, any W which satisfies detailed balance is allowable. We
need only pick a convenient W !
The two most common W ’s used in numerical work are the metropolis rule
(
e−∆E/kB T , ∆E > 0
WM etropolis (S, S ′ ) = (6.15)
1 , ∆E ≤ 0
′
= = e−∆E/kB T
W (S , S) 1
W (S, S ′ ) 1
= −(E(S)−E(S ′ ))/k T = e−∆E/kB T
W (S ′ , S) e B
Metropolis
1 1 1
∆E ∆E ∆E
Low T Moderate T High T
Figure 6.2:
Glauber
1 1 1
1/2 1/2
∆E ∆E ∆E
Low T Moderate T High T
Figure 6.3:
? or
Outcome with
probability
W ({S}, {S ′ })
Figure 6.4:
43
?
Case 1
prob = e−8J/kB T
?
Case 2
prob = e−4J/kB T
?
Case 3
prob = 1
?
Case 4
prob = 1
?
Case 5
prob = 1
Figure 6.5:
44 CHAPTER 6. MONTE CARLO AND THE MASTER EQUATION
For Case 2:
For Case 3:
For Case 4:
For Case 5:
W = 0.1
You know the spin flips 10% of the time. You can enforce this by comparing
W = 0.1 to a random number uniformly between 0 and 1. If the random number
is less that or equal to W , then you flip the spin.
Here is an outline of a computer code. If you start at a large temperature and
45
Initialize System
- How many spins, How long to average
Grinding Do i = 1 to N
Figure 6.6:
46 CHAPTER 6. MONTE CARLO AND THE MASTER EQUATION
M
1
high T > Tc
Figure 6.7:
M transient
time average to get hM i low T < Tc
1
Figure 6.8:
hM i
M ∼ |T − Tc |−beta , β = 1/8
T
Tc ≈ 2.269J/kB
Figure 6.9:
47
hE/Ji
T
Tc
−2
Figure 6.10:
initially Si = 1, the magnetization quickly relaxes to zero (see fig. 6.7). Below Tc ,
the magnetization quickly relaxes to its equilibrium value (see fig. 6.8). Doing
this for many temperatures, for a large enough latice gives fig. 6.9. Doing this
for the energy per spin gives fig. 6.10. which is pretty boring until one plots the
heat capacity per spin
1 ∂hEi
C≡
J ∂T
Try it. You’ll get very close to Ousager’s results with fairly small systems
C
ln |T − Tc |
T c ≈ 2.269 J/kB
Figure 6.11:
N = 10 × 10, or 100 × 100 over a thousand or so time steps, and you’ll do much
better than the mean field theory we did in class.
As an exercise, think how would you do this for the solid–on–solid model
X
Z= e−EState /kB T
{States}
where
N
X
EState = J |hi − hi+1 |
i=1
where hi are heights at sites i.
48 CHAPTER 6. MONTE CARLO AND THE MASTER EQUATION
Chapter 7
Review of Thermodynamics
There are a lot of good thermodynamics books, and a lot of bad ones too. The
ones I like are “Heat and Thermodynamics” by Zemansky and Dittman (but
beware their mediocre discussion of statistical mechanics, and their confusing
discussion of phase transitions), and Callin’s book, “Thermostatics”, (which is
overly formal).
A thermodynamic state of a system is described by a finite number of vari-
ables, usually at least two. One needs at least two because of conservation
laws for energy E and number of particles N . So one needs at least (E, N ) to
specify a system, or two independent variables which could give (E, N ). Some
systems need more than two, and for some systems, it is a useful approach to
only consider one variable, but for now, we will consider that two are enough.
There is a deep question lurking here. Thermodynamics describes systems
in the thermodynamic limit: N → ∞, volume V → ∞, such that number
density n = N/V = const. Each classical particle has, in three dimensions, 6
independent degrees of freedom describing its position and momentum. So there
are not 2, but ∼ 1023 degrees of freedom! This is a very deep question. Many
people have worried about it. But let’s take the physics view. We need at least
two variables, so let’s use two. Experimentally, this describes thermodynamic
systems on large length and time scales, far from, for example, phase transitions.
49
50 CHAPTER 7. REVIEW OF THERMODYNAMICS
1 2 1 2
Vtotal = V1 + V2
Etotal = E1 + E2
Stotal = S1 + S2
T = T1 = T2
P = P1 = P2
−F dx
where S is entropy. Heat always flows from the hot system to the cold system.
In fig. 7.3, both sides change due to transfer of heat Q. For the hot side
51
THOT TCOLD
Figure 7.3: System with hot and cold parts exchanging heat Q.
−Q
(∆S)HOT =
THOT
for the cold side
−Q
(∆S)COLD =
TCOLD
Hence
1 1
(∆S)T OT AL = Q [ − ]
TCOLD THOT
So
(∆S)T OT AL ≥ 0
This is the usual way the second law is presented. Note that this is the only
law of physics that distinguishes past and future.
As an aside, it is interesting to note that S must be a “good” thermodynam-
ics variable, or else the second law is violated. If we assume the opposite (that
two distinct states have the same S) we can create a perpetual motion machine,
as drawn in fig. 7.4. Blobbing up the 1st and 2nd laws gives
(T, S)
P (T, S) Q=W
isotherm
adiabat
dE = −P dV + T dS
where only 2 out of E, P, V, T and S are independent. This is the most impor-
tant and useful result of thermodynamics.
To get further, we need equations of state, which relate the variables. Whithin
thermodynamics, these are determined experimentally. For example, the ideal
52 CHAPTER 7. REVIEW OF THERMODYNAMICS
P = P (V, T ) = N kB T /V
and
E = E(T )
where usually
E =CT
and C is a constant determined experimentally. Then
N kB T
C dT = − dV + T dS
V
So, for an adiabatic process (wherein no heat is exchanged, so dS = 0), a change
in volume causes a change in temperature.
In general, equations of state are not known. Instead one obtains differential
equations of state
∂P ∂P
dP = ( )T dV + ( )V dT
∂V ∂T
and
∂E ∂E
dE = ( )T dV + ( )V dT
∂V ∂T
Clearly, if we know all the (T, V ) dependence of these derivatives, such as ( ∂E
∂T )V ,
we know the equations of state. In fact, most work focuses on the derivatives.
Some of these are so important they are given their own names.
1 ∂V
Isothermal Compressibility kT = −
( )T
V ∂P
1 ∂V
Adiabatic Compressibility ks = − ( )s
V ∂P
1 ∂V
Volume Expansivity α= ( )P
V ∂T
∂S
Heat Capacity at Constant Volume CV = T ( )V
∂T
∂S
Heat Capacity at Constant Pressure CP = T ( )P
∂T
Of course, kT = kT (T, V ), etc.
These results rest on such firm foundations, and follow from such simple
principles, that any microscopic theory must be consistent with them. Further-
more, a microscopic theory should be able to address the unanswered questions
of thermodynamics, e.g., calculate kT or CV from first principles.
These quantities can have very striking behavior. Consider the phase dia-
gram of a pure substance (fig. 7.5). The lines denote 1st order phase transi-
53
P
s l
s l
Pc
T1 T2 T
Figure 7.6:
tions. The dot denotes a continuous phase transition, often called second order.
Imagine preparing a system at the pressure corresponding to the continuous
transition, and increasing temperature from zero through that transition (see
fig. 7.6).
Figure 7.7 shows the behavior of kT and CV : At the first-order transition,
kT has a delta-function divergence while CV has a cusp. At the continuous
transition, both CV and kT have power-law divergences. We will spend a lot of
time on continuous, so-called second order transitions in the course.
54 CHAPTER 7. REVIEW OF THERMODYNAMICS
kT P ≡ Pc
T1 T2
Cv P ≡ Pc
T1 T2
Figure 7.7:
Chapter 8
Statistical Mechanics
1 2
Figure 8.1:
55
56 CHAPTER 8. STATISTICAL MECHANICS
or
ln ρ(S1 + S2 ) = ln ρ(S1 ) + ln ρ(S2 )
This is of the form
f (S1 + S2 ) = f (S1 ) + f (S2 )
where f is some function. The solution is clearly
f (S) = AS
where A is a constant, so
ρ = eAS
But S is defined up to a constant so
ρ = BeS/kB , A = 1/kB
where the constant B is for normalization, and our earlier carelessness with
normalization has no consequence. This is the probability weight for the micro-
cannonical ensemble since S = S(E, V ) in thermodynamics from
T dS = dE + P dV
That is, it is the ensemble for given (E, V). The more handy ensemble, for given
(T, V) is the canonical one. Split the system into a reservoir and a system. They
are still independent, but the resevoir is much bigger than the system (fig. 8.2).
System
fixed T , V
Resevoir
Figure 8.2:
Then
S(Eresevoir ) = S(Etotal − E)
∂S
≈ S(Dtotal ) − ( )total E
∂E
and
E
S(Eresevoir ) = const −
T
57
Fluctuations
Landan and Lifshitz are responsible for “organizing” the theory of fluctuations.
They attribute the theory to Einstein in their book on statistical physics. I
am following their treatment. First, lets do toy systems, where entropy only
depends on one variable, x.
The probability weight
ρ ∼ eS(x)/kB
so the probability of being in x → x + dx is
const eS(x)/kB dx
hxi x
Figure 9.1:
¯ ¯
∂S ¯¯ 1 ∂ 2 S ¯¯
S(x) = S(hxi) + (x − hxi) + (x + hxi)2 + ...
∂x ¯x=hxi 2 ∂x2 ¯x=hxi
59
60 CHAPTER 9. FLUCTUATIONS
where
σ 2 ≡ −kB /(∂ 2 S/∂hxi2 )
So the probability of being in x → x + dx is
2
/2σ 2
const. e−(x−hxi)
h(x − hxi)2 i = σ 2
i.e.,
−kB
h(x − hxi)2 i =
∂ S/∂hxi2
2
h(∆x)2 i = O(1/N )
−kB
= O(ξ 3 /L3 )
∂ 2 S/∂hxi2
T ∆S = ∆E −
|{z} W
|{z}
≡0 work done
for simple to system
systems
mgx
x = l − l cos θ
θ2
≈l
2
so
mglθ2
W =
2
2
and S = − mglθ
2 + const. so, since hθi = 0, we have
kB T
hθ2 i =
mgl
This is a very small fluctuation for a pendulum like in the physics labs. When
the mass is quite small, the fluctuations are appriciable. This is the origin, for
example, of fluctuations of small particles in water, called Brownian motion,
which was figured out by Einstein. There are many other isothermal toys.
62 CHAPTER 9. FLUCTUATIONS
Time to
Fall
5s
Figure 9.3:
Landan and Lifshitz give several examples with solutions. My favorite, not given
in Landan’s book though, is the pencil balanced on its tip. It is interesting
to estimate how long it takes to fall as a function of temperature (at zero
temperature, one needs to use the uncertainty principle).
Figure 9.4:
∆S = ∆S1 + ∆S2
dE2 + P dV2
= ∆S1 +
| T
{z }
thermodynamics for
the reservoir
But when the reservoir increases its energy, the system decreases its energy.
The same thing for the volume. However, they have the same P (for mechanical
equilibrium) and T (for thermal equilibrium). So we have
∆E1 P
∆S = ∆S1 − − ∆V1
T T
9.1. THERMODYNAMIC FLUCTUATIONS 63
where I have dropped the “1” subscript. There are still only 2 relevant variables,
so, for example:
∆E =∆E(S, V )
∂E ∂E
= ∆S + ∆V +
∂S ∂V
1 ∂2E ∂2E ∂2E
+ [ 2 (∆S)2 + 2 (∆S∆V ) + (∆V )2 ]
2 ∂S ∂S∂V ∂V 2
to second order. By ∂E/∂V , I mean (∂E/∂V )S , but it is already pretty messy
notation. Rearranging:
∂E ∂E 1 ∂E ∂E
∆E = ∆S + ∆V + [∆ ∆S + ∆ ∆V ]
∂S ∂V 2 ∂S ∂V
But
∂E
( )V = T
∂S
and
∂E
( )S = −P
∂V
so we have
1
∆E =T ∆S − P ∆V + (∆T ∆S − ∆P ∆V )
2
The term in brackets gives the first–order correlation to thermodynamics. In
the probability weight
There are still only two independent variables, so, for example
∂S ∂S
∆S = ( )V ∆T + ( )T ∆V
∂T ∂V
CV ∂P
= ∆T + ( )T ∆V
T ∂T
Likewise
∂P ∂P
∆P = ( )V ∆T + ( )T ∆V
∂T ∂V
∂P 1
=( )V ∆T − ∆V
∂T κT V
64 CHAPTER 9. FLUCTUATIONS
Putting these expressions into the probability weight, and simplifying, gives
−CV
[ 2k 2 (∆T )2 − 2k 1
(∆V )2 ]
P robability ∝ e BT B T κT V
kB T 2
h(∆T )2 i =
CV
h(∆V )2 i = kB T κT V
and
h∆T ∆V i = 0
so that
h(∆S)2 i = kB Cp
kB T
h(∆P )2 i =
κS V
h∆S ∆P i = 0
h(∆S)4 i = 3h(∆S)2 i2
= 3(kB CP )2
∆Q = ∆Q(T, V )
∂Q ∂Q
=( )V ∆T + ( )T ∆V
∂T ∂V
so
∂Q 2 : 0 ∂Q 2
h(∆Q)2 i =( ) h(∆T )2 i + ... h∆T ∆V i + ( ) h(∆V )2 i
∂T V ∂V T
9.1. THERMODYNAMIC FLUCTUATIONS 65
and
∂Q 2 kB T 2 ∂Q 2
h(∆Q)2 i =( ) +( ) kB T κT V
∂T V CV ∂V T
Check this, for example, with Q = S or Q = P . Note the form of these relations
is
h(∆x)2 i = O(1/N )
for intensive variables, and
h(∆X)2 i = O(N )
for extensive variables. Also note the explicit relationship between corrections
to thermodynamics (h(∆T )2 i), thermodynamic derivatives (CV = T (∂S/∂T )V ),
and microscopic correlations (1/N = ξ 3 /L3 ).
The most important relationship is
h(∆V )2 i = kB T κT V
although not in this form. Consider the number density n = N/V . Since N is
fixed
N N
∆n = ∆ = − 2 ∆V
V V
Hence
V4
h(∆V )2 i = h(∆n)2 i
N2
so,
N2
h(∆n)2 i = kB T κT V
V4
or
n2 kB T κT
h(∆n)2 i =
V
for the fluctuations in the number density. This corresponds to the total fluc-
tuation of number density in our system. That is, if ∆n(~r) is the fluctuation of
number density at a point in space ~r,
Z
1
∆n = d~r∆n(~r)
V
where the dummy integral “~r” is “~r −~r′” of course. But h(∆n)2 i = n2 kB T κT /V ,
so we have the thermodynamic “sum rule”:
Z Z
d~r C(r) ≡ d~rh∆n(r)∆n(0)i = n2 kB T κT
If Z
~
Ĉ(k) ≡ d~r eik·~r C(~r)
we have
lim Ĉ(k) = n2 kB T κT
k→0
Ĉ(k)
Fluctuations of Surfaces
A striking fact is that, although most natural systems are inhomogeneous, with
complicated form and structure, almost all systems in thermodynamic equilib-
rium must have no structure, by Gibbs’ phase rule.
As an example, consider a pure system which can be in the homogeneous
phases of solid, liquid, or vapour. Each phase can be described by a Gibbs
energy gs (T, P ), gl (T, P ), or gv (T, P ). In a phase diagram, the system chooses
its lowest Gibbs free energy at each point (T, P ). Hence, the only possibility
for inhomogeneous phases are along lines and points:
gs (T, P ) = gl (T, P )
gives the solidification line
gl (T, P ) = gv (T, P )
gives the vaporization line, while
gs (T, P ) = gl (T, P ) = gv (T, P )
gives the triple point of solid–liquid–vapour coexistance.
P
s l
Figure 10.1:
67
68 CHAPTER 10. FLUCTUATIONS OF SURFACES
Pc Vapor
l
v Liquid
Tc T
Figure 10.2:
As every one knows, in a finite volume droplet, droplets are spheres while
much larger droplets are locally flat. This is the same for liquid–vapour systems,
oil–water systems, magnetic domain walls, as well as many other systems. The
reason for this is simple thermodynamics. A system with a surface has extra
free energy proportional to the area of that surface. Using the Helmholtz free
energy now, for convenience,
Z
F = (Bulk Energy) + σ d~x (10.1)
| {z }
Surface Area
where the positive constant σ is the surface tension. Note that the extra free
R
Bulk Energy σ d~x
y
~x
Figure 10.3:
V = Ld
69
to calculate such fluctuations strangely enough, using Eq. 10.1 we can construct
models for Estate in Eq. 10.2! This sounds suspicious since Estate should be
microscopic, but it is a common practice.
We shall make the decision to limit our investigations to long-length-scale
phenomena. So it is natural to expect that phenomena on small length scales,
like the mass of a quark or the number of gluons, do not affect our calculations.
In particle physics they call this renormalizability: if we consider length scales
all the physics on scales below this only gives rise to coupling constants entering
our theory on those larger length sacles. The main such coupling constant we
shall consider is simply the radius of an atom
ro ≈ 5Ȧ (10.3)
r ≥ ro
k≤Λ
~x
Figure 10.4:
i = 1, 2, 3, ..., L (10.6)
All the possible states of the system then correspond to the height of the in-
terface at each point i, called hi , where hi is an integer −L/2 ≤ hi ≥ L/2 as
shown in fig. 10.5. Hence the sum in the partition function is
L L/2
X X Y X
= = (10.7)
{states} {hi } i=1 hi =−L/2
10.1. LATTICE MODELS OF SURFACES 71
h5 = 4
h1 = 0
h2 = 1 h3 = 0 h4 = 1
Figure 10.5:
Figure 10.6:
kB T
Z = Z( , L) (10.12)
J
One can straight forwardly solve Z numerically. Analytically it is a little tough.
It turns out to be easier to solve for the continuum version of Eq. 10.11 which
we shall consider next.
y y = h(~x)
is the interface
h(~x)
~x
Figure 10.7:
Overhang Bubble
Figure 10.8:
“lattice model” (fig. 10.5) also could not have overhangs or bubbles. The states
10.2. CONTINUUM MODEL OF SURFACE 73
the system can have are any value of h(~x) at any point ~x. The continuum
analogue of Eq. 10.7 is
X X YZ
= = dh(~x)
{states} {h~
x} ~
x
Z −Z (10.13)
≡ Dh(·)
−
Q R
where the last step just shows some common notation for ~x dh(~x). Things
are a little subtle because ~x is a continuous variable, but we can work it through.
One
Q R other weaselly point is that Q the sum over states is dimensionless, while
x dh(~
~ x) has dimensions of ( x L). This constant factor of dimensionality
does not affect our results, so we shall ignore it.
For the energy of a state, we shall simply make it proportional to the amount
of area in a state h(~x), as shown in fig. 10.7.
or
Estate = σ(L′ )d−1 (10.14)
′ d−1
where (L ) is the area of the surface y = h(~x), and σ is a positive constant.
It is worth stopping for a second to realize that this is consistent with out
definition of excess free energy ∆F = σLd−1 . In fact, it almost seems a little
too consistent, being essentially equal to the excess free energy!
This blurring of the distinction between microscopic energies and macro-
scopic free energies is common in modern work. It means we are considering a
“coarse–grained” or mesoscopic description. Consider a simple example
X
e−F/kB T ≡ Z = e−Eh /kB T
{h}
where h labels a state. Now say one has a large number of states which have
the same energy Eh . Call these states, with the same Eh , H. Hence we can
rewrite X X
e−Eh /kB T = gH e−EH /kB T
{h} {H}
where
gH = # of states h for state H
Now let the local entropy
S = kB ln gH
So we have X
e−F/kB T = e−FH /kB T
{H}
does a partial sum over states which might have different energies. this long
digression is essentially to rationalize the common notation (instead of Eq. 10.14)
of
Estate = Fstate = σ(L′ )d−1 (10.15)
Let us now calculate (L′ )d−1 for a given surface y = h(~x)
p
(dh)2 + (dx)2
dh
y dx
Figure 10.9:
or,
− σL
d−1 YZ −σ
Z
∂h
Z=e k
BT ( dh(x)) exp dd−1 ~x( ′ )2 (10.21)
2kB T ∂~x
~
x
hh(x)i
and
hh(~x) h(~x′ )i
All properties can be determined from knowledge of these two moments! How-
ever, space is translationally invariant and isotropic in the ~x plane (the plane
of the interface), so properties involving two points do not depend upon their
|x1 − x2 |
x~1
x~2
o
hh(x)i = const.
and
hh(~x1 ) h(~x2 )i = G(|~x1 − ~x2 |) (10.22)
For convenience, we can choose
hhi = 0 (10.23)
To get G(x), we use the partition function from Eq. 10.21, from which the
probability of a state is
−σ
dd−1 ~
x( ∂h 2
R
x)
ρstate ∝ e 2kB T ∂~
(10.25)
The subtle thing here is the gradients which mix the probability of h(~x) with that
of a very near by h(~x′ ). Eventually this means we will use fourier transforms.
To see this mixing, consider
Z Z Z
∂h ∂
d~x( )2 = d~x[ d~x′ h(~x′ )δ(~x − ~x′ )]2
∂~x ∂~x
Z Z Z
′ ∂ ∂
= d~x d~x d~x′′ h(~x)h(~x′′ )( δ(~x − ~x′ )) · ( δ(~x − ~x′′ ))
∂~x ∂~x
x → x′′
x′ → x
x′′ → x′
(you can do this by parts integration, and there will be terms involving surfaces
at ∞, which all vanish.) Hence,
Z Z Z
∂h
dd−1 ~x( )2 = dd−1 ~x dd−1 ~x′ h(~x)M (~x − ~x′ )h(~x′ ) (10.26)
∂~x
∂2
M (~x) = − δ(~x) (10.27)
∂~x2
which has the form shown in fig. 10.11. This “Mexican hat” interaction matrix
appears in many different contexts. It causes h(~x) to interact with h(~x′ ).
In fourier space, gradients ∂/∂~x become numbers, or rather wave numbers ~k.
We will use this to simplify our project. It is convenient to introduce continuous
10.2. CONTINUUM MODEL OF SURFACE 77
2
∂
M (~x) = − ∂~
x2 δ(~
x)
~x
Figure 10.11:
where
Z
~
ĥ(k) = dd−1 ~xe−ik·~x h(~x)
(10.29)
= ĥk
Closure requires R
dd−1~
k i~
(2π)d−1
e k·~x
δ(~x) = 1
P i~
k·~
e x (10.30)
Ld−1
~
k
and Z
dd−1 ~x i~k·~x
δ(~k) = e
(2π)d−1
but Z
1 ~
δ~k, 0 = dd−1 ~xeik·~x (10.31)
Ld−1
is the Kronecker delta (
1 , if ~k = 0
δ~k, 0 = (10.32)
0 , otherwise
not the Dirac delta function.
We have used the density os states in ~k space as
L d−1
( )
2π
so that the discrete → continuum limit is
78 CHAPTER 10. FLUCTUATIONS OF SURFACES
Z
X L d−1
→( ) dd−1~k
2π
~
k
and
L d−1 ~
δ~k, 0 → ( ) δ(k) (10.33)
2π
Using the Fourier transforms we can simplify the form of Estate in Eq. 10.20.
Consider
Z
∂h
dd−1 ~x( )2 =
∂x
Z Z
∂ d~k ~
= d~x( eik·~x ĥ(~k))2
∂~x (2π)d−1
Z Z Z
d~k d~k ′ ~ ~′
= d~x d−1 d−1
(−~k · ~k ′ )ĥ(~k)ĥ(~k ′ )ei(k+k )·~x
(2π) (2π)
Z ~ Z ~ ′ Z
dk dk ~k · ~k ′ )ĥ(~k)ĥ(~k ′ )[ d~xei(~k+~k′ )·x ]
= (−
(2π)d−1 (2π)d−1
Z
d~k
= k 2 ĥ(k)ĥ(−k)
(2π)d−1
after doing ~k ′ integral. But h(~x) is real, so ĥ∗ (~k) = ĥ(−k) (from Eq. 10.29),
and we have
Z Z
d−1 ∂h 2 dd−1~k 2 ~ 2
d ~x( ) = k |ĥ(k)| (Continuum)
∂~x (2π)d−1
or Z
∂h 2 1 X 2
dd−1 x( ) = d−1 k |ĥ~k |2 (Discrete) (10.34)
∂~x L
~
k
Hence modes of ĥ~k are uncoupled in fourier space; the matrix M (~x) has been
replaced by the number k 2 , (see Eqs. 10.26 and 10.27). Therefore we have
σ X 2
Estate = σLd−1 + k |ĥ~k |2 (10.35)
2Ld−1
~
k
P
For the partition function we need to sum out all the states {states} using the
fourier modes ĥk rather than h(~x). This is a little subtle because h(~x) is real
while ĥ~k is complex. Of course not all the components are independent since
ĥk = ĥ∗−k , so we double count each mode. To do the sum properly we have
X YZ Y Z
= dh(~x) = ′ d2 ĥ~k (10.36)
{states} ~
x ~
k
10.2. CONTINUUM MODEL OF SURFACE 79
For the most part, this prime plays no role in the algebra below. Recall that
ky
Q′ Integrating in kg > 0, only, for example
k
kx
Figure 10.12:
Let,
1
~y = ~x − ~x′ ~x = ~y ′ + ~y
2
1 1
~y = (~x + ~x′ ) ′ ′
~x = ~y + ~y
2 2
so that d~xd~x′ = d~y d~y ′ , and
Z Z
~ ~ ′ −i(~
k−~
k′ )· y
~ ~ ~′ ′
hĥ(k) ĥ(k )i = d~y e 2 G(y) × d~y ′ e−i(k+k )·~y
so,
Z
~
hĥ(~k) ĥ(~k ′ )i = d~y e−ik·~y G(y)(2π)d−1 δ(~k + ~k ′ )
Z
~
= [ d~x e−ik·~x G(~x)](2π)d−1 δ(~k + ~k ′ ) , letting ~y → ~x
80 CHAPTER 10. FLUCTUATIONS OF SURFACES
as (
Ĝ(~k) (2π)d−1 δ(~k + ~k ′ ) (Continuum)
hĥ(~k) ĥ(~k ′ )i = (10.37)
Ĝ~k Ld−1 δ~k+~k′ , 0 (Discrete)
Hence, in Fourier space, the only non–zero Fourier cos is
we have,
′′2
σ
|ĥk′′ |2
P
Q R − ~ k
~
k′ ′ d2 ĥk′ |ĥK |2 e 2kB T Ld−1 k′′
h|ĥ~k |i = σ
(10.38)
k′′2 |ĥk′′ |2
P
Q′ R 2 − k~′′
2kB T Ld−1
~
k′ d ĥ k ′ e
Except for the integrals involving ĥ~k , everything on the numerator and denomi-
nator cancels out. We have to be careful with real and imaginary parts though.
Consider X X
k 2 |ĥk |2 = k 2 ĥk ĥ−k
k k
(These steps are given in Goldenfeld’s book in Section 6.3, p.174.) Now let
ĥ~k = Reiθ
so that R∞ 2
2 R dR R2 e−R /a
h|ĥk | i = 0R ∞
0
R dR e−R2 /a
where
kB T 1
a= (10.39)
σk 2 Ld−1
A little fiddling gives
kB T 1
h|ĥk |2 i = a =
σk 2 Ld−1
10.2. CONTINUUM MODEL OF SURFACE 81
W W , width of surface
y
~x
Figure 10.13:
and
kB T
G(0) = L,d=2
2π 2 σ
so r
kB T 1/2
w= L ,d=2 (10.45)
2π 2 σ
Similarly, in d = 3,
Z Λ
kB T 1 k dk
G(0) = 2π
σ 4π 2 2π
L
k2
and
kB T
G(0) = ln (LΛ/2π) , d = 3
2πσ
so, r
kB T LΛ 1/2
w= (ln ) ,d=3 (10.46)
2πσ 2π
It is easy to obtain L dependences as dimension of space varies. One obtaines,
undef ined , d ≤ 1
LX , 1<d<3
w= 1/2
(10.47)
(lnL) , d=3
const. , d>3
where
3−d
X= (10.48)
2
is called the roughening exponent. The fact that w is undefined for d ≤ 1 is
explained below.
Eq. 10.47 implies the interface is rough and increases with L for d ≤ 3.
However, note that the width (for d > 1) is not large compared to the system
Figure 10.14:
itself, as L → ∞.
w 1 1
= 1−X = (d−1)/2 (10.49)
L L L
Hence the width is small for d > 1. In d = 1, it appears the width is comparable
to the system size! This means the interface — and indeed phase coexistance
10.2. CONTINUUM MODEL OF SURFACE 83
The interpretation of g(x) from Eq. 10.54 is simple, it is the (squared) difference
in the heights between two points a distance x apart, as shown in fig. 10.15. Not
surprisingly, we expect (for large L)
g(x → L) = 2w2 (10.56)
* 0h(0)i
since, G(L) = hh(L) h(0)i = hh(L)ih * 0because the points 0 and L are so far
appart so as to be uncorrelated, so G(L) = 0.
In any case, we have
Z i~
k·~
x
kB T 1 d−1~ 1 − e
g(x) = 2 d k( )
σ (2π)d−1 k2
Z (10.57)
kB T 1 ~ x
d−1~ 1 − cos k · ~
=2 d k( )
σ (2π)d−1 k2
84 CHAPTER 10. FLUCTUATIONS OF SURFACES
p
g(~x − ~x′ )
h(~x) h(~x′ )
Figure 10.15:
Consider d = 2. Let
Z Λ
kB T 1 1 − cos kx
g(x) = 2 dk( )
σ 2π −Λ k2
Z
kB T Λ 1 − cos kx
=2 dk
πσ 2π/L k2
let u = kx
Z Λx
kB T 1 − cos u
g(x) = 2 x{ du }
πσ 2πx/L u2
let Λ → ∞ and L → ∞ in the integral (for convenience)
Z ∞
kB T 1 − cos u
g(x) = 2 x( du )
πσ 0 u2
| {z }
π/2
so,
kB T
g(x) = x, d = 2 (10.58)
σ
In d = 3,
Z 2π Z Λ
kB T 1 1 − cos kx cos θ
g(x) = 2 dθ kdk( )
σ 4π 2 0 2π/L k2
This is a little more involved, and I will just quote the answer
kB T
g(x) = ln xΛ , d = 3 (10.59)
4πσ
In general, one obtians
undef ined , d=1
x2X , 1<d<3
g(x) = (10.60)
ln x , d=3
const. , d>3
10.2. CONTINUUM MODEL OF SURFACE 85
where, again X = (3 − d)/2. This confirms our earlier claim that correlations
are power — law like.
(There is a subtlety here involving constraints which deserves a minute of
time. The limit g(x → L) ∼ L2X , but the constraint is not the same as 2w2 as
derived above. Reconsider the integral of Eq. 10.58:
Z Λx
kB T 1 − cos u
g(x) = 2 x{ du }
πσ 2πx
L
u2
or, for d = 2
kB T
g(x, L) = xf (x/L) (10.61)
σ
where Z 2πx
2 1 − cos u
f (x) = 1 − du (10.62)
π 0 u2
Note that
f (x → ∞) = 0
but
f (x → 0) = 1
This is called a scaling function. End of digression.)
f (x/L)
0
x/L
Figure 10.16:
We are pretty well done now, except we shall calculate the “renormalized”
surface tension from the partition function. Recall from Eq. 10.21
Z Z
− σL
d−1 Y
−σ ∂h
Z = e kB T ( dh(x)) exp dd−1 ~x( ′ )2
2kB T ∂~x
~
x
86 CHAPTER 10. FLUCTUATIONS OF SURFACES
~
k
or if ĥ~k = Reiθ
Z Z 2π Z ∞
d2 ĥk = dθ RdR
0 0
(10.65)
Hence
− σL
d−1 Y Z 2π Z ∞
−( 2kσ T k2
)R2
Z=e kBT ′ dθ R dR e B Ld−1
~ 0 0
k
σ k2
let u = ( )R2
2kB T Ld−1
Z
− σL
d−1 Y 1 ∞
:0
Z=e kBT ′2π 2 du e−u
~
2( 2kσB T Lkd−1 ) 0
k
− σL
d−1 Y 2πkB T Ld−1
=e kB T
′
σk 2
~
k
d−1 2πkB T Ld−1
− σL
Q′
=e kB T
eln ~
k σk2
kB T X 2πkB T Ld−1
F = σLd−1 − ln (10.67)
2 σk 2
~
k
is the free energy due to the surface. The second term gives the contribution due
to fluctuations. To take the derivative with respect to ∂/∂Ld−1 , it is important
to note that the ~k’s get shifted with L’s, and that it is the (kL) combination
which is not a function of L. So to take the derivative, we rewrite
L L + dL
Figure 10.17:
kB T X 2πkB T Ld−1
F = σLd−1 − ln
2 σ(kL)2
~
k
or,
kB T X kB T X 2πkB T
F = σLd−1 − ln(Ld−1 )(d+1)/(d−1) − ln
2 2 σ(kL)2
~
k ~
k
or,
kB T d + 1 X kB T X 2πkB T
F = σLd−1 − ( )(ln Ld−1 ) − ln (10.68)
2 d−1 2 σ(kL)2
~
k ~
k
gives,
kB T d + 1 1 X
σ∗ = σ − ( )(Discrete)
2 d − 1 Ld−1
~
k
or,
Z
kB T d + 1 dd−1~k
σ∗ = σ − ( )(Continuum) (10.69)
2 d−1 (2π)d−1
88 CHAPTER 10. FLUCTUATIONS OF SURFACES
is the renormalized surface tension. Note that fluctuations decrease the value
of σ to the lower σ ∗ .
Z Z Λ
dd−1~k 1 Λ
In d = 2, d−1
= dk = (10.70)
(2π) 2π −Λ π
Z Z 2π Z Λ
dd−1~k 1 1 2 Λ2
In d = 3, = dθ kdk = πΛ = (10.71)
(2π)d−1 4π 2 0 0 4π 2 4π
Z
dd−1~k
In d = 1, =1 (10.72)
(2π)d−1
Hence we have,
3kB T
σ∗ = σ − Λ ,d = 2
2π
kB T 2
σ∗ = σ − Λ ,d = 3
4π
kB T
σ∗ = σ − , d = 1 + ǫ , ǫ << 1
ǫ
(10.73)
(It is worth nothing that this is the first term in the d = 1 + ǫ expansion, where
R d−1 q ∂h 2
we have only considered the leading term in EState = σ d ~x 1 + ( ∂~x ) ≈
R
σLd−1 + σ2 dd−1 ~x( ∂h 2
∂x ) + ...)
Ebulk ∝ Ld
Esurf ace ∝ Ld−1
~x
Figure 10.18:
or or or etc.
Figure 10.19:
are all rotations and upside down interfaces, so that in the end there are Ld
ways. The entropy is
∆S = kB ln(numberof states)
∆S = +kB ln Ld
or,
∆S = +kB d(ln L) (10.77)
is the increase in entropy due to having a surface. Hence, the total change in
the free energy due to a surface is
So, for d > 1, L → ∞, the free energy increases if a surface is added, but for
d = 1 (at T > 0) the free energy decreases if a surface is added!
This means coexisting phases are stable for d > 1, but in d = 1 they are
unstable. Since one interface decreases F , we can add another and another and
another, and keep on decreasing F until there are no longer coexisting phases
(of bulk energy E ∼ Ld as L → ∞), just a mishmash of tiny regions of phase
fluctuations. Hence, for systems with short–ranged forces, at T > 0, there can
be no phase coexistance in d = 1. (Short–ranged forces were assumed when we
said the surface energy ESurf ace ∼ Ld−1 ).
Figure 10.20:
or
σ ∗ = σ(1 − T /T ∗ ) (10.80)
where
σ
T∗ ≡ (10.81)
4πkB (Λ/2π)2
Now, reconsider the phase diagram of a pure substance. Our analysis describes,
P
s l
Pc Critical point
triple point v
T
Tc
for example, the behavior along the liquid-vapour coexistance line. Near the
triple point of a typical fluid, the surface tension is such that
r
kB Tt
≈ 1Ȧ (10.82)
2πσ
where Tt is the triple point temperature. Similarly, one roughly has
Λ
≈ 10Ȧ (10.83)
2π
Putting these in Eq. 10.79 above implies
L 1/2
w = 1Ȧ(ln ) (10.84)
10Ȧ
Putting in numbers is amusing. The width divergence is very weak.
2.6Ȧ, L = 10000Ȧ (light wavelength)
4.3Ȧ, L = 10cm (coffee cup)
w=
5.9Ȧ, L = 1000km (ocean)
9.5Ȧ, L = 1030 m (universe)
¡
Of course, in a liquid–vapour system, gravity plays a role. (Certainly it plays
a role if one had a liquid–vapour system the size of the universe.) The extra
energy due to gravity is
Z
1
EGravity = dd−1 ~x(ρl − ρv )gh2 (x)
2
92 CHAPTER 10. FLUCTUATIONS OF SURFACES
where ρ is the mass density of the liquid or vapour, and g ¢is gravity. This
changes our results in a way you should work out for yourself.
Putting in numbers for the renormalized surface tension gives the result
σ ∗ (T = Tt ) ≈ σ(0.3)
T ∗ = Tc
σ∗ = Const.(1 − T /Tc )µ
A major concept in modern work is symmetry. These results for surfaces are
an example of more general properties of correlations in systems with broken
symmetries.
By a broken symmetry, we mean the following. Note that in the partition
function X
Z= e−EStates /kB T
{States}
Since Z
σ dh
EState = σLd−1 + dd−1 ~x( )2
2 ∂~x
P
Naively, since the sum {States} does not break this symmetry, we would expect
it to be respected by any quantity obtained from Z. The two main quantities
we have obtianed are the moments hhi and hhhi. As we found before,
hhi = 0 (11.2)
In fact we could have the average value be any result, without changing our
algebra. Note that h’s expectation value is a definite number. In particular
hh + Const.i =
6 hhi
93
94CHAPTER 11. BROKEN SYMMETRY AND CORRELATION FUNCTIONS
Figure 11.1:
as drawn in fig. 11.1, only has translational invariance and isotropy in the ~x
plane of the interface; There is no invariance in the y direction. Hence the
system with a surface has a lower symmetry (a broken symmetry) compared to
the liquid or vapour system.
This perfect symmetry of a homogeneous system is lost precisely because h,
the broken symmetry variable, has a definite value, like hhi = 0.
It is instructive to see that the fact Z and EState have this symmetry implies
the form of the correlation functions of the broken symmetry variable h. Con-
sider d = 2 only. Since a surface can be anywhere for a given system (it only
has a definite value for a given system), it can be different places for different
systems. In a very large system, part of hhi might have one value, while far far
Figure 11.2:
away on the x axis it might want to have another value. The free energy cost of
such a fluctuation, if it involves very large length scales (equivalent to k → 0)
is infinitesimal.
σ
∆EState (k) = k 2 |ĥk |2
2
as k → 0. Indeed, if one has a very very large system encompassing the four
shown in fig. 11.2, the four subsystems behave like independent systems which
each have a definite value for hhi. Since the energy cost of such fluctuations
is infinitesimal as k → 0, we can set up long–wavelength sloshing modes with
infinitesimal kB T . Roughly,
kB T σ
= k 2 |ĥk |2
2 2
and
kB T 1
h|ĥk |2 i ∼
σ k2
as we recall. These sloshing modes are capillary waves driven by thermal noise.
But we need not know the form of EState to obtain h|ĥk |2 i ∼ 1/k 2 . We
only need to know that it breaks a symmetry of Z and EState . As shown in
fig. 11.3, fluctuations occur, say ±∆y, over a long distance ∆x. The values of
∆y (which could be ∼ Ȧ) and ∆x (which could be ∼ meters) are not important
for the argument. Let us invent a model where, after one goes a distance ∆x
to the right, one goes up or down ∆y, at random, as shown in fig. 11.4. Of
step 2 vapor
y
step 1 liquid
x
yRM S ∼ x1/2
g(x) ∼ x
Fourier transforming, and fiddling via Eqs. 10.55 and 10.56, gives Ĝ(k) ∼ 1/k 2
for small k so that,
h|ĥk |2 i ∼ 1/k 2
as expected. Our argument was only in d = 2, but it can be generalized to other
dimensions.
Note that it was the fact that h was a broken symmetry variable which gave
rise to the 1/k 2 ; no special details of the interaction were necessary.
Before obtaining the general result, note that the system reacts to a sym-
metry breaking variable by “shaking it” in such a way as to try and restore the
96CHAPTER 11. BROKEN SYMMETRY AND CORRELATION FUNCTIONS
A symmetry breaking variable ends up
having its fluctuations try to restore
the symmetry.
w ∼ Lx
Figure 11.5:
broken symmetry. The rough width of a surface is this attempt to restore the
full translational invariance in the y direction, as shown in fig. 11.5. In fact, in
d = 1, the fluctuations are so strong that the full symmetry is restored, and one
has translational invariance in the y drection (since coexisting phases cannot
exist). In, 1 < d ≤ 3, the fluctuations do not restore the symmetry, but they
make the surface rough, or diffuse, as L → ∞. In d > 3, fluctuations do not
affect the flat interface.
Finally, it should be noted that the cartoon on p. 64 is self–similar and self
affine, in the same sense as the von–koch snowflake we discussed below. Note
the important new ingredient: Broken symmetry ⇔ power–law correlations ⇔
self–similar structure.
∂F
But F (0) is just a constant, and ∂B = 0 at equilibrium, so
1 ∂2F 2
F (B) = B + ... (11.3)
2 ∂B 2
Of course, by F we mean EState or FState in the sense of Eq. 10.10. Now, it
turns out that it is important to deal with the local value of B(~r), or the Fourier
component B̂(~k). If we use the local value for the expansion in Eq. 11.3 we have,
Z Z
1 ∂2F
F (B) = d ~r dd~r′
d
B(~r)B(~r′ ) (11.4)
2 ∂B(~r)∂B(~r′ )
2
which looks worse than it is. Since ∂B(~r∂)∂B(~
F
r ′ ) is a thermodynamic derivative of
some kind, in a system which is (we assume) rotationally invariant and transla-
tionally invariant,
∂2F
≡ C(~r, ~r′ ) = C(~r − ~r′ ) (11.5)
∂B(~r)∂B(~r′ )
In fourier space
Z Z
1 dd~k dd~k ∂2F
F (B) = B̂(~k) B̂(~k ′ ) (11.6)
2 (2π)d (2π) ∂ B̂(~k) ∂ B̂(~k ′ )
d
where
1 ~k = 0
Ĝ(k) ∝ ,
k2
as claimed. If Ĉ(k) is not analytic as assumed in Eq. 11.9 we get the weaker
condition
lim Ĝ(k) → ∞
k→0
Scattering
Figure 12.1:
99
100 CHAPTER 12. SCATTERING
where Ê is the electric field. Clearly the electric field variations with ~q are due
to variations in the dielectric properties of the sample, i.e.,
where δǫ̂ is the dielectric variability. Now, if δǫ̂ is a function of only a small
number of local thermodynamic variables, like density and temperature
∂ǫ ∂ǫ
δǫ̂ = ( )δ ρ̂ + ( )δ T̂ (12.5)
∂ρ ∂T
2. The order parameter is not the local density, but instead corresponds to a
sublattice concentration or magnetization, or to a Bragg peak’s intensity.
A cartoon (fig. 12.2) may make this clear. The second case is more subtle. An
example is the sublattice concentration of a binary alloy. Say atoms are A and
B in the alloy, and that locally, A wants to be neighbours with B but not with
another A. This makes the ordered state look like a checker–board as shown in
fig. 12.3.
Assuming the scattering is different from A or B atoms this sort of order can
be seen by looking at S(q). However, it must be realized that the ordering of,
say, the A atoms is on a sublattice, which has twice the lattice spacing of the
original system as shown in fig. 12.4. Hence the density of A atoms is uniform
on a sublattice with twice the lattice constant of the original system, and the
101
y S(q)
Figure 12.2:
Perfectly ordered
A − B alloy
=A
=B
Figure 12.3:
Sublatice 1 (Shaded)
2a
a
(Sublatice 2 is unshaded)
Figure 12.4:
102 CHAPTER 12. SCATTERING
S(q) qy
q0
or qx
−q0 q0
−q0
−q0 q0 qy
Figure 12.5:
q0 q0
or or q0 etc
qx qx qx
Figure 12.6:
ple), except that often one has many different crystallite orientations scattering
simultaneously, so that all the orientations indecated above are smeared into a
ring of radious q0 .
qy
q0 qx
y=0
1
y
is interface
y
x
Figure 12.8:
where θ(y) is the Heaviside step function, and A and B are constants. For
simplicity, we will let A = 1 and B = 0 from now on:
θ =H +C
so that
1
θ̂(ky ) = + C δ(ky )
iky
To determine C, note that
θ(x) + θ(−x) = 1
104 CHAPTER 12. SCATTERING
so
From this we obtain the scattering from a flat interface y = 0 as the modulus,
1
S(k) = |ρ̂(k)|2 ∝ δ(kx ) (12.15)
ky2
where we have neglected the trivial bulk factor delta function. If the unit vector
normal to the interface is n̂ and that along the surface is n̂⊥ , then
1
S(k) ∝ δ(~k · n̂⊥ ) (12.16)
~
(k · n̂)2
n̂
n̂⊥ y
Figure 12.9:
S(k) ky
kx = 0
kx
y=0 ρ
is interface
x y
Figure 12.11:
Lets also consider the scattering from a line defect, which is similar, but has
important differences. In this case,
ρ(x, y) = δ(y) (12.17)
Then
ρ̂ = δ(kx ) (12.18)
and
S(k) ∝ δ(kx ) (12.19)
or if the unit vector normal to the surface is n̂, and that along the surface is
n̂⊥ ,
S(k) ∝ δ(~k · n̂⊥ ) (12.20)
for a line defect. Now, if ρ is not the order parameter (like the second case
S(k)
kx = 0 ky
kx
ky
Grey Scale of S(k)
discussed above) so that the density doesn’t vary across the interface, but an
order parameter corresponding to sublattice concentration or crystalline orien-
tation does, we have to incorporate the wavenumber k~0 corresponding to that
structure. For simplicity, we will assume that all one has to do is modulate the
structure up to that wavenumber. That is, for an interface
1
S(k) ∝ δ((~k − ~k0 ) · n̂⊥ ) (12.21)
~ ~
|(k − k0 ) · n̂|2
106 CHAPTER 12. SCATTERING
y ρ
ξ
x
ξ y
So that
δ(kx )
S(k) = |ρ̂(k)dif f use |2 ∼ (1 + O(ξky )2 ) (12.24)
ky2
for an interface at y = 0. So the diffuseness only effects scattering at kx = 0
ky
kx
extra
scattering
due to diffuseness
Figure 12.14:
and large ky . On the other hand, let us now consider a rough interface, with no
diffuseness, where the interface is given by
y = h(x) (12.25)
and
ρRough (x, y) = θ(y − h(x)) (12.26)
12.2. ROUGHNESS AND DIFFUSENESS 107
kx
Figure 12.16:
geometry δ(k x) 2
ky2 from the diffuseness O(ξky ) , from the roughness 1/kx
2−η
, since
they have different signatures in k space. For excellent samples this is in fact
true, but if there are many randomly oriented interfaces present, these signatures
are smeared out, and washed away.
108 CHAPTER 12. SCATTERING
Figure 12.17:
and
n̂⊥ = cos θx̂ + sin θŷ (12.33)
with
~k = k cos β x̂ + k sin θŷ (12.34)
δ(~
k·n̂⊥ )
Then, for scattering from a surface where |ρ̂k |2 = |~
k·n̂|2
we have (in d = 2)
Z 2π
1 δ[k cos β cos θ + k sin β sin θ]
S= dθ
2π 0 | − k cos β sin θ + k sin β cos θ|2
Z 2π
1 1 δ[cos (θ − β)]
= dθ
2π k 3 0 | sin (θ − β)|2
Z 2π
1 1 δ[cos θ]
= dθ
2π k 3 0 | sin θ|2
for scattering from a surface. For scattering from a line defect which has
|ρ̂(k)|2 ∝ δ(kx ), one similarly obtains
1
Slinedef ects ∝ (12.37)
k d−1
We shall call these results, Porod’s Law. They involve only geometry.
The case where ρ is not the order, and a modulated structure exists is more
subtle. If the wavenumber of that structure is ~k0 then this introduces a new
angle via
~k0 = k0 cos αx̂ + k0 sin αŷ (12.38)
Now there are two possible averages
Figure 12.18:
α or β
average
kx kx
interface n̂ = ŷ
Figure 12.19:
ky ky
k0 θ Average k0
kx kx
Figure 12.20:
If a further average is done over crystallite orientation, we can take these ex-
pressions and angularly average them over the angle between ~k and ~k0 . Clearly
after averaging we will have
|k| − |k0 |
S̄ = S( ) ≡ S(∆k) (12.40)
|k0 |
First, let
φ=β−α (12.41)
so that
(~k − ~k0 )2 = k 2 − 2kk0 cos φ + k02
or, on using
k ≡ k0 (∆k + 1) (12.42)
we have
(~k − ~k0 )2 = k02 {(∆k)2 + 2(1 − cos φ)(∆k + 1)} (12.43)
Hence in two dimensions, for a surface, we have
Z 2π
1 1
S̄(∆k) = dφ (12.44)
2πk03 0 |(∆k)2 + 2(1 − cos φ)(∆k + 1)|3/2
which is some weird integral. We will work out its asymptotic limits ∆k >> 1
and ∆k << 1.
Firstly, if ∆k >> 1, then
Z 2π
1 1
S̄(∆k) = dφ + ...
2πk03 (∆k)3 0
12.3. SCATTERING FROM MANY FLAT RANDOMLY–ORIENTED SURFACES111
or (in arbitrary d)
1
S̄(∆k) ∼ , ∆k >> 1 (12.45)
(∆k)d+1
If ∆k << 1, consider
Z 2π
? 1 1
S̄(∆k → 0) = dφ
2πk03 0 |2(1 − cos φ)|3/2
Z 2π
1 1
S̄(∆k) ≈ dφ + ...
2πk03 0 |(∆k)2 + φ2 |3/2
Z :
2π/∆k
∞
1 1 1
S̄(∆k) = du
2πk03 (∆k)2 0 |1 + u2 |3/2
so,
1
S̄(∆k) ∼ as ∆k → 0 (12.46)
(∆k)2
1
S= (12.47)
|k − ~k0 |γ
~
then
(
1 k−k0
(k−k0 )γ−(d−1)
, k0 << 1
S̄ = 1 k−k0
(12.48)
(k−k0 )γ , k0 >> 1
It should be noted that the actual scattering is given by a weird boring integral
112 CHAPTER 12. SCATTERING
ky 1
, near tails
|k−k0 |d+1
1
|k−k0 |2 , near center
k0
kx
Figure 12.21:
Fluctuations and
Crystalline Order
Density
ρ = ρ(~r)
Atoms on a lattice
Figure 13.1:
113
114 CHAPTER 13. FLUCTUATIONS AND CRYSTALLINE ORDER
~r ~r → ~r′ , with
~r′
~r′ = r0 (+1x̂ − 1ŷ)
r0 shown here
Figure 13.2:
Let us then introduce a variable ~u(~r) which discribes the local variations in the
density. That is, due to some fluctuation,
~r → ~r + ~u(~r)
and
ρ(~r) → ρ(~r + ~u(~r)) (13.3)
where ~u(~r) is the local displacement vector as shown in fig. 13.3.
~u(~r)
Figure 13.4:
~u(~r)
115
A definite value for hu(~r)i = 0, say, breaks space symmetry (translational and
rotational invariance).)
By construction
h~u(~r)i = 0 (13.4)
In Fourier space, if all directions of fluctuation are equally likely,
x
2π/G
More precisely, X
ρ(x) = eiGx âG (13.6)
G
where there might be several wavenumbers G giving the density, and âG is a
Fourier coefficient giving the atomic form. In arbitrary dimension we write
P
ρ(x) = G eiGx âG
x
2π/G
X ~
ρ(~r) = eiG·~r âG
~ (13.7)
~
G
where G~ are called the reciprocal lattice vectors. Then we have a fluctuation
represented by
X ~
ρ(~r + ~u(~r)) = eiG·(~r+~u(~r)) âG
~ (13.8)
~
G
116 CHAPTER 13. FLUCTUATIONS AND CRYSTALLINE ORDER
X ~ ~
hρ(~r + ~u(~r))i = eiG·~r âG
~ he
iG·~
u(~
r)
i (13.9)
~
G
To calculate the average we use the fact that ~u(~r) is a Gaussian variable. There-
fore
~ 1 ~ 2
heiG·~u i = e− 2 hG·~ui
1 ~~
(13.10)
= e− 2 GG:h~u~ui
But ~u is only correlated to ~u if they are in the same direction, we expect. That
is
huα uβ i ∝ δαβ
so
~ 1 2
hu2 i
he−iG·~u i ≈ e 2 G
But, from Eq. 13.5 we have,
Z
dd k 1
hu2 i =
(2π)d k 2
or
Const.
d=3
hu2 i = ln L d=2 (13.11)
L d=1
So in d = 3 X ~
hρi = Const. eiG·~r âG
~ (13.12)
~
G
which is a large effect, much greater than the periodicity as L → ∞. This is the
content of the original Landan–Peierls theorem, which states that a one dimen-
sional variation in density, ρ(x) (whether crystal structure or phase coexistance)
is impossible: The only G which can survive in Eq. 13.14 is G = 0 as L → ∞
which gives ρ(x) → Constant, due to ~u(~r) fluctuations.
It is more revealing to do these things for the scattering intensity I(k), where
or X
ρ̂k = âG δ(~k − G)
~ (13.16)
~
G
so, X
ρ̂∗k = â∗G δ(~k − G)
~
~
G
and X
|ρ̂k |2 = âG â∗G′ δ(~k − G)δ(~k − G
~ ′)
~ G
G, ~′
or, X
|ρ̂k |2 ∝ |âG |2 δ(~k − G)
~ (13.17)
~
G
where |âG |2 is an atomic form factor. This gives well defined peaks where ~k = G
~
as shown in fig. 13.7. When thermal fluctuations are present
are absent.
k
~
various G’s
Figure 13.7:
Z X ~
~ ~
ρ̂k = dd~r e−ik·~r ( eiG·~r âG eiG·u(~r) )
G
or, Z
X ~ ~ ~
ρ̂k = âG dd~r e−i(k−G)·r eiG·u(~r) (13.18)
~
G
so Z
X ~ ~′ ′ ~′ ′
ρ̂∗k = â∗G dd~r′ ei(k−G )·~r e−iG ·u(~r )
~′
G
118 CHAPTER 13. FLUCTUATIONS AND CRYSTALLINE ORDER
Hence
X Z Z
~ ~ ~ ~′ ′ ~ ~′ ′
h|ρ̂k |2 i = âG â∗G dd~r dd~r′ e−i(k−G)·~r ei(k−G )·~r heiG·u(~r)−iG ·u(~r ) i
~ G
G, ~′
(13.19)
This is the same integral as on page so we can simply use that
~ ~′ ′
heiG·u(~r)−iG ·u(~r ) i = f (~r − ~r′ ) (13.20)
and then,
X Z
~ ~~ −G
~ ′ )f (~r)
h|ρ̂k |2 i = âG â∗G dd~r e−i(k−G)·~r (2π)d δ(G (13.21)
~ G
G, ~′
where
~ ~
f (~r) = heiG·~u(r)−iG·u(0) i
G2
r )−u(0))2 i (13.22)
= e− 2 h(u(~
2
= Const. e−G hu(r)u(0)i
so, Z
X ~ ~ 2
h|ρ̂k |2 i = const. ~|
|âG 2
dd r e−i(k−G)·~r e−G hu(~
r )u(0)i
(13.23)
~
G
The correlation function hu(~r)u(0)i can be worked out as before (knowing h|û(k)|2 i ∼
1
k2 ),
Const.,
d=3
hu(~r)u(0)i = ln r, d=2 (13.24)
r, d=1
In d = 3, X
h|ρ̂k |2 i = Const. |âG |2 δ(~k − G)
~
~
G
I(k)
d=3
~
G k
I(k)
d=2
G k
In d = 1, we have
Z
P ~ ~
2
h|ρ̂k | i = Const. ~
G
2
|âG | dd r e−i(k−G)·~r e−(Const.)r
| {z }
this is the Fourier
transform of an
exponential. But
R d i~k′ ·~r −r/ξ
d re e ∝ (k′ )21+ξ−2
for small k ′
So, for d = 1,
X 1
h|ρ̂k |2 i = Const. ~|
|âG 2
(13.26)
~ ~ 2 + ξ −2
(k − G)
~
G
120 CHAPTER 13. FLUCTUATIONS AND CRYSTALLINE ORDER
I(k)
d=1
G G G
Ising Model
The Ising model has a large number of spins which are individually in mi-
croscopic states +1 or −1. The value of a spin at a site is determined by the
spin’s (usually short–range) interaction with the other spins in the model. Since
I have explained it so vaguely, the first, and major import, of the Ising model is
evident: It is the most simple nontrivial model of an interacting system. (The
only simpler model would have only one possible microscopic state for each spin.
I leave it as an exercise for the reader to find the exact solution of this trivial
model.) Because of this, whenever one strips away the complexities of some
interesting problem down to the bone, what is left is usually an Ising model of
some form.
For now, we will think of the Ising model as applied to anisotropic magnets.
The magnet is made up of individual dipoles which can orient up, corresponding
to spin +1 or down corresponding to spin −1. We neglect tilting of the spins,
121
122 CHAPTER 14. ISING MODEL
arrow Spin S = +1
arrow Spin S = −1
Figure 14.1:
or or . . .
X +1
X +1
X +1
X
= ...
{States} s1 =−1 s2 =−1 sN =−1
N X
Y +1
=
i=1 si =−1
where each sum has only two terms, of course. The energy in the exp −
EState /kB T now must be prescribed. Since we want to model a magnet, where
all the spins are +1 or −1, by and large, we make the spin of site i want to
oriented the same as its neighbors. That is, for site i
X
Energy of Site i = (Si − Sneighbors )2
neighbors of i
X
= Si2 − 2 2
Si Sneigh. + Sneigh.
neighbors of i
On a square lattice, the neighbors of “i” are the spin surrounding it (fig. 14.3).
As said before, the idea is that the microscopic constituents -the spins- don’t
know how big the system is, so they interact locally. The choice of the 4 spins
123
above, below and to the right and left as “nearest neighbors” is a convention.
The spins on the diagonals (NW, NE, SE, SW) are called “next–nearest neigh-
bors”. It doesn’t actually matter how many are included, so long as the inter-
action is local. The total energy for all spins is
X
EState = −J Si Sj
hiji
where J is a positive interaction constant, and hiji is a sum over all i and
j = 1, 2, ...N , provided i and j are nearest neighbors, with no double counting.
Note
N X N
X X 4
= =N×
i=1 j=1
2
hiji
| {z }
neighbors no double counting
i=1 S1 =−1
J H
= Z(N, , )
kB T kB T
As described above, the sum over q nearest neighbors is q = 4 for a two–
dimensional square lattice. For three dimensions, q = 6 on a simple cubic
lattice. The one–dimensional Ising model has q = 2.
124 CHAPTER 14. ISING MODEL
The one–dimensional model was solved by Ising for his doctoral thesis (we
will solve it below). The two–dimensional model was solved by Ousager, for H =
0, in one of the premiere works of mathematical physics. We will not give that
solution. you can read it in Landan and Lifshitz’s “Statistical Physics”. They
give a readable, physical, but still difficult treatment. The three–dimensional
Ising model has not been solved.
The most interesting result of the Ising model is that it exhibits broken
symmetry for H = 0. In the absence of a field, the energy of a state, and the
partition function, have an exact symmetry
Si ↔ −Si
hSi i ↔ −hSi i
and
?
hSi i = 0
In fact, for dimension d = 2 and 3 this is not true. The exact result in d = 2 is
surprisingly simple: The magnetization/spin is
(
(1 − cosech2 k2J
BT
)1/8 , T < Tc
m = hSi i =
0, T > Tc
0
Tc T
Figure 14.4: Two–dimensional square lattice Ising model magnetization per spin
in the right ball park. There is no hand–waving explanation for the abrupt
increase of the magnetization per spin m = hSi i at Tc . Note it is not like a
paramagnet which shows a gradual decrease in magnetization with increasing
temperature (in the presence of an external field H). One can think of it (or
Paramagnet No Broken
1
H>0 Symmetry
m = hSi i
Figure 14.5:
T Tc T
H=0 H>0
forbidden region
for single phase
−1 +1 m +1 m
T
H<0
−1 m
Figure 14.6:
symmetry of the Ising model. Hence it is very easy to understand that broken
symmetry is for H = 0, not H 6= 0. Things are more subtle in the lattice–gas
version of the Ising model.
126 CHAPTER 14. ISING MODEL
And
X X
E Ising = −4J Ci Cj − 2(H − Jq ) Ci + Const.
hiji i
and
µ = 2(H − jq )
Note that the plane transition here is not at some convenient point µ = 0, it
is at H = 0, which is the awkward point µ = 2Jq. This is one reason why the
liquid–gas transition is sometimes said — incorrectly — to involve no change
of symmetry (Z2 group) as the Ising model. But the external field must have a
carefully tuned value to get that possibility.
14.2. ONE–DIMENSIONAL ISING MODEL 127
T hci
Tc 0 1
Pc Tc P = Pc
T
Tc Density
Figure 14.7:
Y +1
N X PN PN
J
Sj Sj+1 + k HT Sj
Z= e kB T j=1 B j=1
i=1 Si =−1
P
Note we have simplified the hiji for one dimension. We need, however, a
prescrition for the spin SN +1 . We will use periodic boundary conditions so that
SN +1 ≡ S1 . This is called the Ising ring. You can also solve it, with a little
Ising RingN 1 Periodoc Boundary Conditions
N −1 2
3
. . . .
Figure 14.8:
128 CHAPTER 14. ISING MODEL
more work, with other boundary conditions. Let the coupling constant
k ≡ J/kB T
and
h = H/kB T
be the external field. Then
Y +1
N X PN
Z(k, h, N ) = e j=1 (kSj Sj+1 +hSj )
i=1 Si =−1
i=1 Si =−1
Y N
+1 Y
N X
1
= ekSj Sj+1 + 2 h(Sj +Sj+1 )
i=1 Si =−1 j=1
h+1|P | + 1i = ek+h
h−1|P | − 1i = ek−h
h+1|P | − 1i = e−k
h−1|P | + 1i = e−k
so,
+1
X
Z= hS1 |P N |S1 i
S1 =−1
= T race(P N )
Say the eigenvalues of P are λ+ and λ− . Then,
Z = λN N
+ + λ−
and
F = −N kB T λ+
It remains only to find the eigenvalues of P .
We get the eigenvalues from
det(λ − P ) = 0
This gives
(λ − ek+h )(λ − ek−h ) − e−2k = 0
or,
ek+h + ek−h
λ2 − 2λ ( ) +e2k − e−2k = 0
| 2
{z }
ek cosh h
The solutions are
p
2ek cosh h ± 4e2k cosh2 h − 4e2k + 4e−2k
λ± =
p 2
= ek [cosh h ± cosh2 h − 1 + e−4k ]
So, p
λ+ = ek (cosh h + sinh2 h + e−4k )
is the larger eigenvalue. The free energy is
p
F = −N kB T [k + ln (cosh h + sinh2 h + e−4k )]
where k = J/kB T , and h = H/kB T .
The magnetization per spin is
1 ∂F 1 ∂F
m=− =−
N ∂H N kB T ∂h
Taking the derivative and simplifying, one obtains
sinh h
m(h, k) = p
sinh2 h + e−4k
As anticipated, m(h = 0, k) = 0. There is magnetization for h 6= 0, although
there is no broken symmetry of course. There is a phase transition at T = 0.
130 CHAPTER 14. ISING MODEL
+1 low T
high T
m
−1
and,
+1
Y X H
P
Sj
Z= e kB T j
i Si =−1
X H
S1
X H
SN
=( e kB T )...( e kB T )
S1 =±1 SN =±1
X H
S
=( e kB T )N
S=±1
so,
H N
Z = (2 cosh )
kB T
and
H
F = −N kB T ln 2 cosh
kB T
14.3. MEAN–FIELD THEORY 131
so since,
1
m=− ∂F/∂H
N
then
m = tanh H/kB T
For the interacting Ising model we need to find the effective local field Hi at
+1
1 H>0
m
H m
−1
T
Clearly
∂E
− = Hi
∂Si
Thus, from the Ising model
∂E X
− =J (Sj δk,i + δj,i Sk ) + H
∂Si
hjki
X
=J ·2· Sj + H
j neighbors of i
(no double counting)
Now let’s get rid of the 2 and the double–counting restriction. Then
X
Hi = J Sj + H
j neighbors of i
Hi ≈ hHi i
132 CHAPTER 14. ISING MODEL
where
X
hHi i = J hSj i + H
j neighbors
hHi i = qJhSi + H
where hSi = hS1 i, and contributions due to S2 , S3 ... cancel from top and bottem.
So
hHi
m = tanh
kB T
or
qJm + H
m = tanh
kB T
This gives a critical temperature
kB Tc = qJ
m 1 m3
m= − + ...
T /Tc 3 (T /Tc )3
One solution is m = 0. This is stable above Tc , but unstable for T < Tc .
Consider T < Tc , where m is small, but nonzero. Dividing by m gives
1 1 m2
1= − + ...
T /Tc 3 (T /Tc )3
14.3. MEAN–FIELD THEORY 133
Rearranging
T 2 T
m2 = 3( ) (1 − ) + ...
Tc Tc
Or,
T 1/2 T
m = (3 ) (1 − )1/2
Tc Tc
for small m, T < Tc . Note that this solution does not exist for T > Tc . For
T < Tc , but close to Tc , we have
T β
m ∝ (1 − )
Tc
where the critical exponent, β = 1/2. All this is really saying is that m is a
m
H=0
Close to Tc , mean
field gives m ∝ (1 − T /Tc )1/2
Tc T
Figure 14.11:
parabola near Tc , as drawn in fig. 14.12 (within mean field theory). This is
+1 Parabola near m = 0, T = Tc , H = 0
m
Figure 14.12:
interesting, and useful, and easy to generalize, but only qualitatively correct.
Note
2J
,d = 1
kB Tc = qJ = 4J ,d = 2
6J ,d = 3
The result for d = 1 is wildly wrong, since Tc = 0. In d = 2, kB Tc ≈ 2.27J, so
the estimate is poor. Also, if we take the exact result from Ousageris solution,
134 CHAPTER 14. ISING MODEL
H = kB Tc (tanh−1 m − m)
Expanding giving
1
H = kB Tc (m + m3 + ... − m)
3
so
kB Tc 3
H= m
3
for m close to zero, at T = Tc . This defines another critical exponent
Figure 14.13:
H ∝ mδ
χT
χT , susceptibility
diverges near Tc
Tc
Figure 14.14:
Hence
Jq
hEi = − N m2 − HN m
2
136 CHAPTER 14. ISING MODEL
So, (
−(Const.)(Tc − T ), T < Tc
hEi =
0, T > Tc
and (
Const. T < Tc
c=
0 T > Tc
If we include temperature dependence away from Tc , the specific heat looks
more like a saw tooth, as drawn in fig. 14.15. This discontinuity in c, a second
C
Tc
derivative of the free energy is one reason continuous phase transitions were
called “second order”. Now it is known that this is an artifact of mean field
theory.
It is written as
c ∝ |T − Tc |−α
where the critical exponent
α = o(disc.)
To summarize, near the critical point,
m ∼ (Tc − T )β , T < Tc , H = 0
δ
H∼m , T = Tc
−γ
XT ∼ |T − Tc | , H=0
and
c ∼ |T − Tc |−α , H=0
14.3. MEAN–FIELD THEORY 137
Nowadays, it is well established that, near the critical point, one can either solve
the Ising model
X X
E = −J Si Sj − H Si
hiji i
with
Y X
Z= e−E/kB T
i Si =±1
r ∝ T − Tc
where the kB T factor has been set to kB Tc , and incorporated into k, r, u and
H (this is ok near Tc ).
Look back in the notes for the discussion of the discrete Gaussian model
versus the continuous drumhead model of interfaces. Although it may not look
like it, the field theory model is a very useful simplification of the Ising model,
wherein no important features have been lost.
139
140CHAPTER 15. LANDAN THEORY AND ORNSTIEN–ZERNICKE THEORY
One is simply the continuum limit of the other. The square gradient term is
likewise straightforward, as we now see.
Ising Model ψ 4 Field Theory
N Z
JX X
(Si − Sj )2 ⇐⇒ K dd x(∇ψ)2
2 i=1
j neighbors of i
yusing Si2 = 1
X
const. − J Si Sj
hiji
R
The terms harder to see are the dd ~x ( 2r ψ 2 + u4 ψ 4 ) terms. Let us define
r 2 u 4
f (ψ) = ψ + ψ
2 4
This is the free energy for a homogeneous system in the ψ 4 field theory (that is,
~ ≡ 0), in the absence of an external field, per unit volume. Note
where ∇ψ
r > 0, T > Tc
r < 0,T < Tc
So we have the double well form. This may look familiar, it was originally due
f f
T > Tc T < Tc
r>0 r<0
ψ ψ
~
Figure 15.1: “free energy”/volume. No ∆ψ, no H
to Van der Waals. (In fact, the whole ψ 4 theory is originally due to Van der
Waals.)
Note that, since ψ is determined by the minimum of F , this single or double
well restricts ψ to be within a small range. Otherwise, ψ could have any value,
since Z −Z YZ ∞
Dψ(·) = dψ(~x)
− ~
x −∞
141
These unphysical values of ψ are suppressed in the free energy. In the Ising
model, the restriction is Si = ±1. We can lift this restriction with a kronecker
delta
+1
Y X ∞
Y X
= δSi2 −1
i Si =−1 i Si =−∞
Y X ∞
− ln (1/δS 2 −1)
= e i
i Si =−∞
So, although this is awkward and artificial, it is sort of the same thing. Expand-
ing it out, as Si2 → 0,
the double–well form. This can be done in a very precise way near Tc . Here I
am just trying to give an idea of what is going on.
So the ψ 4 theory is similar to the Ising model, where ψ(~x) is the local
magnetization. We will be concerned with the critical exponents α, β, γ, δ, η,
and ν defined as follows
F (T ) ∼ |T − Tc |2−α ,H = 0
∂2F
So that C = ∂T 2 ∼ |T − Tc |−α
ψ ∼ (Tc − T )β , H = 0 , T < Tc
∂ψ
( )T = XT ∼ |T − Tc |−γ ,H = 0
∂H
ψ ∼ H 1/δ , T = Tc
1
hψ(~x)ψ(0)i ∼ d−2+η , T = Tc , H = 0
x
(
e−x/ξ , T > Tc
hψ(~x)ψ(0)i ∼ 2 −x/ξ
hψi + O(e ) , T < Tc
and
ξ ∼ |T − Tc |−ν ,H = 0
broken at Tc can be quite subtle. This will take us too far afield, so we will
only consider a scalar order parameter, where the symmetry broken is Z2 , that
is, where ψ ↔ −ψ symmetry is spontaneously broken. (Note that a magnetic
spin confined to dimension d = 2, wherein all directions are equivalent, has an
order parameter ψ ~ = ψx x̂ + ψy ŷ. The group in u(1). A spin in d = 3 has
~ = ψx x̂ + ψy ŷ + ψz ẑ, if all directions are equivalent; the group in O(3) I think.)
ψ
So Landan argues that the free energy to describe the transition must be an
analytic function of ψ. Hence
Z
F = dd x f (ψ)
where
r 2 u 4
f (ψ) = ψ + ψ + ...
2 4
No odd terms appear, so that the
ψ ↔ −ψ
r>0
for T > Tc . However, at low temperatures, in the ordered phase, we must have
ψ > 0. Hence
r<0
(To make sure everything is still well defined, we need u > 0. There is no need
for u to have any special dependence in T near Tc , so we take it to be a constant.
Or, if you prefer
: unimportant
u(T ) ≈ u(Tc ) + (Const.)(T − Tc ) + ...
r = r0 (T − Tc ) + ...
R
where K > 0. Similar terms like ψ∇2 ψ turn into this after parts integration.
~ 6= 0) add free energy.
Note that this term means inhomogeneities (∇ψ
Now we’re done. The Landan free energy is
Z Z
k
F = dd x {f (ψ) + |∇ψ|2 } − H dd x ψ
2
with
r 2 u 4
f (ψ) = ψ + ψ
2 4
In high–energy physics, this is called a ψ 4 field theory. The ∇2 term gives
kinetic energy; r ∝ mass of particle; and ψ 4 gives interactions. Van der Waals
wrote down the same free energy (the case with k ≡ 0 is what is given in the
text books). He was thinking of liquid–gas transitions, so ψ = n − nc , where n
is density and nc is the density at the critical point. Finally, from the theory of
thermodynamic fluctuations, we should expect that (in the limit in which the
square gradient and quartic term can be neglected)
r ∝ 1/κT
or equivalently
r ∝ 1/XT
It is worth emphasizing that the whole machinery of quadratic, quartic, and
square gradient terms is necessary. For example, note that the critical point is
defined thermodynamically as
∂P
( )T = 0
∂V c
and
∂2P
( )T = 0
∂V 2 c
in a liquid–gas system. In a magnet:
∂H ∂2H
( )Tc = 0, and( )T = 0
∂m ∂m2 c
Trying to find the equation of state for T ≡ Tc in a liquid we expand:
∂P 1 ∂2P 1 ∂3P
P (n, Tc ) = Pc + ( )Tc (n − nc ) + ( 2 )Tc (n − nc )2 + ( 3 )Tc (n − nc )3 + ...
∂n 2 ∂n 6 ∂n
For a magnet
0 ∂H 0 1 ∂2H 1 ∂3H
H(m, Tc ) = H>
c+( )Tc (m − m*
c) + ( )T m 2
+ ( )T m3 + ...
∂m 2 ∂m2 c 2 ∂m3 c
144CHAPTER 15. LANDAN THEORY AND ORNSTIEN–ZERNICKE THEORY
So, at the critical point for both systems — from thermodynamics and analyticity
only — we have
(P − Pc ) ∝ (n − nc )δ
and
H ∝ mδ
where
δ=3
Experimentally, δ ≈ 5.0... in three dimensions, while δ ≡ 15 in d = 2. So we
have a big subtle problem to solve!
F/V = f (ψ) − Hψ
r0 u
= (T − Tc )ψ 2 + ψ 4 − Hψ
2 4
For H = 0, we find the solution by minimizing F/V (or f ):
∂f
= 0 = r0 (T − Tc )ψ + uψ 3
∂ψ
ψ = 0, (stable f or T > Tc )
r
r0 (Tc − T )
ψ=± , (stable f or T < Tc )
u
(To check stability, consider ∂ 2 f /∂ψ 2 .) for H 6= 0, only one solution is stable
for T < Tc . You can check this yourself.
∂ ∂f
(f (ψ) − Hψ) = 0 = −H
∂ψ ∂ψ
or
H = r0 (T − Tc )ψ + uψ 3
Note that this gives
∂H
1/XT = ( )T = r0 (T − Tc )
∂ψ
close to the critical point, or
XT ∝ |T − Tc |−1
15.2. ORNSTEIN — ZERNICKE THEORY OF CORRELATIONS 145
In summary, these are the same results as we obtained from the mean field
theory of the Ising model. The relationship between analyticity of equations of
state, and mean–field theory, is worth thinking about.
α 0 (disc.)
β 1/2
γ 1
δ 3
∆ψ ≡ ψ − hψi
Then
~ = ∇(∆ψ
∇ψ ~ ~
− hψi) = ∇(∆ψ)
Also,
ψ 2 = (∆ψ − hψi)2 = (∆ψ)2 − 2hψi∆ψ + hψi2
and
ψ 4 = (∆ψ − hψi)4
= (∆ψ)4 − 4hψi(∆ψ)3 + 6hψi2 (∆ψ)2 − 4hψi3 (∆ψ) + hψi4
−r = |r|
Both approximate free energies are pretty much the same as the free energy for
the interface model: Z
σ ∂h
dd−1 x( )2
2 ∂x
They can be solved exactly the same way. The definitions of Fourier transforms
are the same (except for dd−1 x → dd x, and Ld−1 → Ld , for example), and the
arguments concerning translational invariance of space, and isotropy of space
still hold. Let
C(x) ≡ hψ(x) ψ(0)i
−F
Then we have (putting the kB T back in e kB T )
kB T
Ĉ(k) = , T > Tc
Kk 2 + r
15.2. ORNSTEIN — ZERNICKE THEORY OF CORRELATIONS 147
and since
|r|
h∆ψ(x) ∆ψ(0)i = hψ(x) ψ(0)i −
u
using hψi2 = |r|/u, we have
kB T |r|
Ĉ(k) = − (2π)d δ(k) , T < Tc
Kk 2 + 2|r| u
These can be Fourier transformad back to real space.
As an asside, remember the Green function for the Helmholtz equation sat-
isfies
(∇2 − ξ −2 )G(x) = −δ(x)
Hence
Ĝ(k) = 1/(k 2 + ξ −2 )
The Green function is given in many books
−x/ξ
ξ/2 e
, ,d=1
1
G(x) = 2π K0 (x/ξ) ,d=2
1 −x/ξ
4πx e ,d=3
for example. Recall the modified Bessel function of zeroth order satisfies K0 (y →
0) = − ln y, and K0 (y → ∞) = (π/2y)1/2 e−y . This is more than we need. In
discussions of critical phenomena, most books state that if
1
Ĝ(k) ∝
k 2 + ξ −2
then
1
G(x) ∝ e−x/ξ
xd−2
The reason for the seemingly sloppiness is that proportionality constants are
unimportant, and it is only long–length–scale phenomena we are interested in.
Indeed, we should only trust out results for large x, or equivalently small k
~
(since we did a gradient expansion in ∇ψ).
In fact, so far as the Fourier transforms go, here is what we can rely on:
If
1
Ĝ(k) ∝
k 2 + ξ −2
then
G(x) ∝ e−x/ξ
1
G(x) ∝
xd−2
148CHAPTER 15. LANDAN THEORY AND ORNSTIEN–ZERNICKE THEORY
for large x.
Using these results, we finally obtain the correlation function up to a constant
of proportionality,
−x/ξ
e
, , T > Tc
1
C(x) = xd−2
, T = Tc
2
hψi + O(e−x/ξ ) , T < Tc
q
k , T > Tc
ξ = qr
k
2|r| , T < Tc
is a rewriting of
p
In cartoon form: figure 15.2. Note the correlation length satisfies ξ ∼ 1/ |r|
ξ Correlation length ξ
ξ ∼ √1
|r|
or
ξ ∼ |T − Tc |−1/2
Figure 15.2:
ξ ∼ |T − Tc |−1/2
C(x)
T < Tc
T > Tc T = Tc
1
C(x) = , T = Tc , (H = 0)
xd−2
Since
ξ ∼ |T − Tc |ν
and
1
C(x) ∼
xd−2+η
we have ν = 1/2 and η = 0.
This completes the list of critical exponents.
In summary:
Scaling Theory
Historically, the field of phase transitions was in a state of termoil in the 1960’s.
Sensible down–to–earth theories like the mean–field approach, Landan’s ap-
proach, and Ornstein–Zernicke theory were not working. A lot of work was put
into refining these approaches, with limited success.
For example, the Bragg–Williams mean field theory is quite crude. Refine-
ments can be made by incorporating more local structure. Pathria’s book on
Statistical Mechanics reviews some of this work. Consider the specific heat of
the two–dimensional Ising model. Since Ousager’s solution, this was known
exactly. It obeys C ∼ ln |T − Tc |, near Tc . The refinements make for much
better predictions of C away from Tc . At Tc , all mean field theories give C
to be discontinuous. Comparing to the exact result shows this as well. (This
C C C
T T T
Tc Tc Tc
Figure 16.1:
151
152 CHAPTER 16. SCALING THEORY
Figure 16.2:
Figure 16.3:
ξ → ∞ is striking.
Figure 16.4:
Here I show some (quite old) simulations of the Ising model from Stanley’s
book. Note the “fractally” structure near Tc . You can easily program this
yourself if you are interested. Nowadays, a PC is much much faster than one
needs to study the two–dimensional Ising model. Strange things are also seen
experimentally.
For a fluid, imagine we heat it through Tc , fixing the pressure so that it
154 CHAPTER 16. SCALING THEORY
corresponds to phase coexistence (i.e., along the liquid–vapor line in the P–T
diagram as drawn). The system looks quite different depending on temperature.
vapor
Figure 16.5:
It is transparent, except for the meniscus separating liquid and vapour for T <
T < Tc T = Tc T > Tc
Figure 16.6:
Tc , for T < Tc and T > Tc . But at Tc , it is milky looking. This is called “critical
opalescence”. The correlation length ξ, corresponding to fluctuations, is so big
it scatters light.
Rigorous inequalities were established with the six critical exponents. (I
should mention that I am assuming any divergence is the same for T → Tc+ or
T → Tc− , which is nontrivial, but turns out to be true.) For example,
Rushbrooke proved
α + 2β + γ ≥ 2
Griffiths proved
α + β(1 + δ) ≥ 2
These are rigorous. Other inequalities were proved with some plausible assump-
tions. These all work for the exponents quoted previously — as equalities! In
fact, all the thermodynamic inequalities turned out to be obeyed as equalities.
These inequalities were very useful for ruling out bad theories and bad experi-
ments.
16.1. SCALING WITH ξ 155
α + 2β + γ = 2
γ = ν(2 − η)
γ = β(δ − 1)
and
νd = 2 − α
So it turns out there are only two independent exponents. The last equation is
called the “hyperscaling” relation. It is the only one involving spatial dimension
d. The others are called “scaling” relations. This is because the derivation of
them requires scaling arguments.
ξ ∼ |T − Tc |−ν
so
F = |T − Tc |νd
But the singular behavior in F is from the heat capacity exponent α via
F = |T − Tc |2−α
Hence
νd = 2 − α
This back–of–the–envelope argument is the fully rigorous (or as rigorous as it
comes) argument! The hyperscaling relation works for all systems, except the
mean field solutions must have d = 4.
To derive the three scaling relations, we assert that correlations are scale
invariant:
1 ∼
C(x, T, H) = d−2+η C(x/ξ, Hξ y )
x
156 CHAPTER 16. SCALING THEORY
Note that, since blocks of spin are correlated on scale ξ, this means only a little
H is needed to magnetize the system. So the effect of H is magnified by ξ. The
factor is ξ y , where y must be determined. In a finite–size system we would have
1 ∼
C(x, T, H L) = C(x/ξ, Hξ y , L/ξ)
xd−2+η
In any case, in the relation above
ξ ∼ |T − Tc |−ν
1 ∼
C(x, T, H) = C(xtν , H/t∆ )
xd−2+η
where
∆ = yν
is called the “gap exponent”. But the magnetic susceptibility satisfies the sum
rule
Z
XT ∝ dd xC(x, T, H)
Z
1 ∼
= dd x d−2+η C(xtν , H/t∆ )
x
Z ∼
ν ∆
ν −2+η d ν C(xt , H/t )
= (t ) d (xt ) ν d−2+η
(xt )
so,
∼
XT (T, H) = t(−2+η)ν X(H/t∆ )
where Z ∞
1 ∼ ∼ H
dd (xtν ) C(xtν , H/t∆ ) ≡ X( )
−∞ (xtν )d−2+η t∆
But
X(T, H = 0) = t−γ
Hence
γ = (2 − η)ν
called Fisher’s law.
Now note
∂ψ
(T, H) = XT (T, H)
∂H
16.1. SCALING WITH ξ 157
But
ψ(T, H = 0) = tβ
So
β =∆−γ
and the gap exponent
∆=β+γ
or equivalently, y = (β + γ)/ν. So we have
∼ H
ψ(T, H) = tβ ψ( )
tβ+γ
Here is a neat trick for changing the dependence “out front” from t to H. We
∼
H
don’t know what the function ψ is. All we know is that it is a function of tβ+γ
.
∼
Any factor of H/tβ+γ that we multiply or divide into ψ leaves another unknown
function, which is still a function of only H/tβ+γ . In particular, a crafty choice
can cancel the tβ factor:
H β H −β ∼ H
ψ(T, H) = tβ ( ) β+γ ( ) β+γ ψ( )
tβ+γ tβ+γ tβ+γ
∼
∼
We call the last term on the right–hand side ψ, and obtain
∼
β/β+γ
∼ H
ψ(T, H) = H ψ( )
tβ+γ
∼
∼ ∼
Note that H −β/β+γ ψ(H) = ψ(H). Now
ψ(Tc , H) = H 1/δ
γ = β(δ − 1)
yn = 2 − α − n(β + γ)
yn free energy
2
magnetization
1
0 1 2 3
Susceptability
−1
−2
Near Tc , we assert that fluctuations in the order parameter are of the same
order as ψ itself. Hence
h(∆ψ)2 i
= O(1)
hψi2
This is the weak part of the argument, but it is entirely reasonable. Clearly
hψi2 ∼ t2β
h∆ψ(x)∆ψ(0)i ∼ Oh(∆ψ)2 i
Hence Z
dd xh∆ψ(x)∆ψ(0)i = ξ d h(∆ψ)2 i = XT
or,
h(∆ψ)2 i = XT ξ −d
But
XT ∼ t−γ
and
ξ ∼ t−ν
so,
h(∆ψ)2 i = tνd−γ
h(∆ψ)2 i
= tνd−γ−2β
hψi2
As a general feature
h(∆ψ)2 i
hψi2
measures how important fluctuations are. We discussed this at the beginning of
the course. Mean field theory works best when fluctuations are small. So we say
that mean–field theory is correct (or at least self–consistent) if its predictions
give
h(∆ψ)2 i
<< 1
hψi2
This is called the Ginzburg criterion for the correctness of mean–field theory.
But from above
h(∆ψ)2 i
= tνd−γ−2β
hψi2
Putting in the results from mean–field theory (ν = 1/2, γ = 1, β = 1/2) gives
h(∆ψ)2 i 1
= t 2 (d−4)
hψi2
Renormalization Group
{States}
1. Divide the lattice into blocks of size bb (we’ll let b = 2 in the cartoon
below).
2. Each block will have an internal sum done on it leading to a block spin
161
162 CHAPTER 17. RENORMALIZATION GROUP
P
+1, d Si > 0
Piǫb
Si = −1, Piǫb
d Si < 0
±1
, iǫbd Si = 0
randomly
Clearly, if N is the number of spins originally, N/bd is the number of spins after
blocking.
It is easy to see that this is a good idea. Any feature correlated on a finite
length scale l before blocking becomes l/b after blocking once. Blocking m times
means this feature is now on the small scale l/bm . In particular, consider the
correlation length ξ. The renormalization group (RG) equation for ξ is
ξ (m+1) = ξ (m) /b
ξ ∗ = ξ ∗ /b
(
∗ 0, T rivial : T = 0, F ixed P oint : T = ∞
ξ =
∞, N ontrivial f ixed point : T = Tc
The critical fixed point is unstable, while the T = 0 and T = ∞ fixed points
Tc
Figure 17.1:
are stable.
163
Less Fluctuations
(lower T )
Figure 17.2:
(m+1)
(fig. 17.3). We know the states in the new renormalized system, Si = ±1,
Figure 17.3:
i = 1, 2, ..N (m+1)
Note,
N (m+1) N (m) /bd
In the cartoon, N (m+1) = 16, bd = 22 , and N (m) = 64.
This is all pretty clear: we can easily label all the states in the sum for Z.
But how do the spins interact after renormalization? It is clear that, if the
original system has short–ranged interactions, the renormalized system will as
well.
(m)
At this point, historically, Kadanoff simply said that if EState was the Ising
(m+1)
model, so was EState . The coupling constants K (m) and h(m) would depend
on the level of transformation, but that’s all. This is not correct as we will see
shortly, but the idea and approach is dead on target.
164 CHAPTER 17. RENORMALIZATION GROUP
{States}
But
(m)
f (m) (h(m) ,t(m) )
Z (m) = e−N
where t(m) is the reduced temperature obtained from K (m) , and f (m) is the free
energy per unit volume. But we must have Z (m) = Z (m+1) , since we obtain
Z (m+1) by doing some of the sums in Z (m) . Hence
or,
f (m) (h(m) , t(m) ) = b−d f (m+1) (h(m+1) , t(m+1) )
The only reason h and t vary under RG is due to B. We assume
t(m+1) = bA t(m)
and
h(m+1) = bB t(m)
Now we use Kadanoff’s bold, and unfortunately incorrect, assertion that f (m)
and f (m+1) are both simply the Ising model. Hence
This is one form of the scaling assumption, which is due to Widom. Widom
argued that the free energy density was a homogeneous function. This allows
for the possibility of nontrivial exponents (albeit built into the description), and
one can derive the scaling relations.
For example, the factor b is not fixed by our arguments. Let
tbA ≡ 1
Then
f (h, t) = td/A f (ht−B/A , 1)
or,
∼
f (h, t) = td/A f (ht−B/A )
which is of the same form as the scaling relation previously obtained for F .
165
vs.
Figure 17.4:
Figure 17.5:
Kadanoff also presented results for the correlation function, but we’ll skip
that.
Let us consider the interactions between spins again–before and after appli-
cation of RG. For simplicity, let H = 0. The original system has interactions
to its nearest neighbors as drawn in fig. 17.4. The blocked spin treats 4 spins
as interacting together, with identical strength with blocked spins. The nearest
neighbor interaction is as a cartoon, in fig. 17.5. Note that the direct part of
the nearest–neighbor interaction to the right in the renormalized system has 4
spins interacting equivalently with 4 other spins. There are 16 cases, which are
replaced by one global effective interaction. Those 16 cases have: All K (m+1)
contributions treated equally in block spin near neighbor interaction.
2 interactions separated by 1 lattice constant
6 interactions separated by 2 lattice constants
8
interactions separated by 3 lattice constants
2 interactions separated by 4 lattice constants
Let’s compare this to the next–nearest neighbor interaction, along the diagonals.
There are again 16 cases which have: All next nearest neighbor contributions
166 CHAPTER 17. RENORMALIZATION GROUP
(m+1)
Let us call the block–spin next–nearest neighbor coupling Knn , which lumps
all of this into one effective interaction. It is clear from looking at the interac-
(m+1)
tions underlying the block spins that K (m+1) (nearest neighbor) and Knn
(m+1)
(next nearest neighbor) are quite similar, although it looks like Knn <
(m+1) (m+1)
K . However, Kadanoff’s step of Knn ≡ 0 looks very drastic, in fact
with the benifit of hindsight, we know it is too drastic.
One can actually numerically track what happens to the Ising model under
the block–spin renormalization group. It turns out that longer–range interac-
tions between two spins are generatedP (as we have described in detail above)
as well as three spin interactions (like Si Sj Sk ), and other such interactions.
Although the interactions remain short ranged, they become quite complicated.
It is useful to consider the flow of the Energy of a state under application of
RG. For H = 0, for the Ising model, the phase diagram is as in fig. 17.7. For
low T
0 Kc K Recall K = J/kB T
infinite T
the Ising model with only next nearest neighbor interactions (J ≡ 0) one gets
the same phase diagram at H = 0 (fig. 17.8). The phase diagram for an Ising
low T
0 (Knn )c Knn Knn = Jnn /kB T
infinite T
Knn Low T
ordered
(Knn )c 2nd order transitions
disordered
0 Kc K
infinite T
Figure 17.9:
fig. 17.10. Of course there are many other axes on this graph corresponding to
Knn
Critical
fixed point
of RG transformation
K
Figure 17.10:
different interactions. One should also note that the interactions developed by
the RG, depend on the RG. For example, b = 3 would give a different sequence
of interactions, and a different fixed result, than b = 2. But, it turns out,
the exponents would all be the same. This is an aspect of universality: All
models with the same symmetry and spatial dimension share the same critical
exponents.
The formal way to understand these ideas was figured out by Wilson. The
RG transform takes us from one set of coupling constants K (m) to another
K (m+1) via a length scale change of b entering a recursion relation: K (m+1) =
R(k (m) ).
At the fixed point
K ∗ = R(k ∗ )
Let us now imagine we are close enough to the fixed point that we can linearize
around it. In principle this is clear. In practice, because of all the coupling
constants generated on our way to the fixed point, it is a nightmare. Usually it
is done by numerical methods, or by the ǫ = 4 − d expansion described later.
168 CHAPTER 17. RENORMALIZATION GROUP
or, X (m)
δKα(m+1) = Mαβ δKβ
β
where
Mαβ = (∂Rα /∂Kβ )∗
is the linearized RG transform through which “b” appears, and
δK = K − K ∗
somewhat schematically, or
λ(b)λ(b′ ) = λ(bb′ )
This is the homogeneous scaling form, where we now know how to calculate
the yα ’s (in principle), and we know why (in principle) only two exponents are
necessary to describe the Ising model (presumably the remainder are irrelevant).
In practice, the problem is to find the fixed point. A neat trick is the ǫ = 4−d
expansion. From our discussion above about the Ginzburg criteria, mean–field
theory works for d > 4 (it is marginal with logarithmic correlations in d = 4
itself). In fact we then expect there to be a very easy–to–find, easy–to–fiddle–
with, fixed point in d ≥ 4. If we are lucky, this fixed point can be analytically
continued to d < 4 in a controllable ǫ = 4 − d expansion. We will do this below.
I say “lucky” because the fixed point associated with mean–field theory is not
always related to the fixed point we are after (usually it is then a fixed point
related to a new phase transition at even higher dimensions, i.e., it corresponds
to a lower critical dimension).
Finally let us now return to the linear algebra:
X
δKα(m+1) = Mαβ δKα(m)
β
where λ(γ) are eigenvalues and e are the eigenvectors. We expand δK via
X
δKα = δ K̂γ e(γ)
α
γ
or, X
δ K̂γ = δKα e(γ)
α
α
where δ K̂ are the “scaling fields” — a linear combination of the original inter-
P (γ)
actions. Now we act on the linearized RG transform with α eα (...). This
gives X X X
e(γ)
α (δKα
(m+1)
)= e(γ)
α ( Mαβ δKα(m) )
α α β
It is a little cofusing, but the superscripts on the e’s have nothing to do with
the superscripts on the δK’s. We get
X (γ) (m)
δ K̂γ(m+1) = λ(α) eβ δKβ
β
so,
δ K̂γ(m+1) = λ(α) δ K̂γ(m)
This is what we had (for δK’s rather than δ K̂’s) a few pages back. Everything
follows the same from there.
170 CHAPTER 17. RENORMALIZATION GROUP
17.1 ǫ–Expansion RG
Reconsider Z
K r u
F [ψ] = dd x{ (∇ψ)2 + ψ 2 + ψ 4 }
2 2 4
near Tc , where K and u are constants and
r ∝ T − Tc
F
[ ]=1
kB T
or
or,
1
[ψ] = d
L 2 −1
Similarly,
Z
r
[ dd x ψ 2 ] = 1
2
or
1
[Ld r ]=1
Ld−2
17.1. ǫ–EXPANSION RG 171
or
1
[r] =
L2
Finally,
Z
u
[ dd x ψ 4 ] = 1
4
1
so [Ld u 2d−4 ] = 1
L
1
or [u] = 4−d
L
Eventually we will be changing scale by a length rescaling factor, so we need to
find out how these quantities are affected by a simple change in scale. These
are called engineering dimensions.
Note the following,
1
[ψ 2 ] = d−2
L
But at Tc
1
hψ 2 i = d−2+η
x
where η is nonzero! Also, since r ∝ T − Tc , near Tc
1
[T − Tc ] =
L2
but [ξ] = L, so
ξ = (T − Tc )−1/2
near Tc . But
ξ = (T − Tc )−ν
So our results of scaling theory give
1 1 1
[ψ]af ter avg. = = d
L(d−2+η)/2 L 2 −1 Lη/2
and
1 1 1
[r]af ter avg. = = 2 1 −2
L1/ν L Lν
The extra length which appears on averaging is the lattice constant a ≈ 5Ȧ.
Hence
1
hψ 2 i = d−2+η aη
x
and η/2 is called ψ’s anomalous dimension, while 1/η − 2 is the anomalous
dimension of r. By doing the averaging, carefully taking account of changes in
scale by L, we can calculate ν and η as well as other quantities.
172 CHAPTER 17. RENORMALIZATION GROUP
and
x x
x′ = =
[x] L
and
u u
u′ = = = uL4−d
[u] 1/L4−d
so we have Z
1 r′ u′
F = dd x′ { (∇′ ψ ′ )2 + ψ ′2 + ψ ′4 }
2 2 4
Equivalently, since we are interested in the behaviour at Tc , let rL2 ≡ 1 (choos-
ing the scale by the distance from Tc ), then
Z
1 1 u′′ ′4
F = dd x′ { (∇′ ψ ′ )2 + ψ ′2 + ψ }
2 2 4
where
1 4−d
u′′ = u( ) 2
r
Note that u is negligable as L → ∞ for d > 4, and more revealingly, u′′ is
′
′
−L du
dL d>4 initial value of u′ is
desceased as L increases
d=4 u′
Figure 17.11:
′
−L du
dL u′ = 0 fixed pt. d>4
Figure 17.12:
Clearly
du′
−L = (Dim. of u) u′
dL
and similarly
dr′
= (Dim. of r) r′
−L
dL
So the anomalous dimension can be determined by the “flow” near the fixed
point.
174 CHAPTER 17. RENORMALIZATION GROUP
The analysis above is OK for d > 4, and only good for u′ close to zero for
d < 4. To get some idea of behavior below 4 dimensions, we stay close to d = 4
and expand in ǫ = 4 − d. (See note below)
So consider Z
K r u
F = dd x{ (∇ψ)2 + ψ 2 + ψ 4 }
2 2 4
we will find
ǫ = 1 (d = 3) Exact (d = 3)
α = ǫ/6 0.17 0.125
β = 12 − 6ǫ 0.33 0.313
γ = 1 + ǫ/6 4 5.0
δ =3+ǫ 1.17 1.25
ν = 12 + ǫ/12 0.58 0.638
η = 0 + O(ǫ2 ) 0 0.041
But similarly,
[v] = 1/L2(3−d)
2
[w] = 1/L3(2 3 −d)
so,
Z
1 r u vL2(3−d) 6 2
F = dd x′ [ (∇′ ψ ′ )2 + ( L2 )ψ ′2 + ( L4−d )ψ ′4 + ψ + wL3(2 3 −d) ψ 8 ]
2 2 4 6
Hence, v, w look quite unimportant (at least for the Gaussian fixed point) near
d = 4. This is called power counting. For the remainder of these notes we will
simply assume v, w are irrelevant always.)
We will do a scale transformation in Fourier space,
Z
dd k i~k·~x
ψ(x) = d
e ψ̂k
|k|<Λ (2π)
where (before we start) Λ = π/a, since we shall need the lattice constant to
get anomalous dimensions. We will continually apply the RG to integrate out
17.1. ǫ–EXPANSION RG 175
ky ky
Λ ky
Λ/b
kx kx . . . . 0 kx
−Λ Λ
RG Λ/b RG
−Λ
Figure 17.13:
small length scale behavior, which, in Fourier space, means integrating out large
wavenumber behaviour. Schematically shown in fig. 17.13. Let
and Z
dd k i~k·~x
δψ(x) = e ψ̂k
Λ−δΛ<|k|<Λ (2π)d
Schematically, refer to fig. 17.14. This is called the “momentum–shell” RG. We
do this in momentum space rather than real space because it is convenient for
factors of Z Z
dd k 2
dd x(∇ψ)2 = k |ψ̂k |2
| {z } (2π)d
coupling different ψ(x)’s | {z }
fourier modes uncoupled
176 CHAPTER 17. RENORMALIZATION GROUP
ky ky ky
Λ Λ Λ
Λ − δΛ
Λ − δΛ
kx kx kx
Figure 17.14:
Z
dd k
= aΛd−1 δΛ
Λ−δΛ<|k|<Λ (2π)d
where a is a number that we will later set to unity for convenience, although
a(d = 3) = 1/2π 2 , a(d = 2) = 1/2π. Let us now integrate out the modes δψ
X
−F̄ [ψ̄]
|e {z } ≡ e|−F [{z
ψ̄+δψ]
}
scales |k| < Λ − δΛ States δψ scales |k| < Λ
| {z }
shell scales
Z
1 r u
F [ψ̄ + δψ] = dd x{ (∇(ψ̄ + δψ))2 + [ψ̄ + δψ]2 + [ψ̄ + δψ]4 }
2 2 4
We will expand this out, ignoring odd terms in ψ̄ (which must give zero contri-
bution since they change the symmetry of F ), constant terms (which only shift
the zero of the free energy), and terms of order higher than quartic like ψ 6 and
ψ 8 (which are irrelevant near d = 4 by power counting as we did above).
17.1. ǫ–EXPANSION RG 177
In that case
1 1 1
[∇(ψ̄ + δψ)]2 = (∇ψ̄)2 + (∇δψ)2 + odd term
2 2 2
r r r
[ψ̄ + δψ]2 = ψ̄ 2 + (δψ)2 + odd term
2 2 2
u u δψ 4
[ψ̄ + δψ]4 = ψ¯4 [1 + ]
4 4 ψ̄
u δψ (δψ)2 (δψ)3 (δψ)4
= ψ¯4 [1 + 4 +6 2 +4 3 + ]
4 ψ̄ ψ̄ ψ̄ ψ̄ 4
u 3u 2
= ψ̄ 4 + ψ̄ (δψ)2 + O(δψ)4 + odd term
4 2
We will neglect (δψ)4 , since these modes have a small contribution to the k = 0
behavior. This can be self–consistently checked. Hence
Z
1 1
F [ψ̄ + δψ] = F {ψ̄} + dd x[ (∇δψ)2 + (r + 3uψ̄ 2 )(δψ)2 ]
| {z } 2 2
|k|<Λ−δΛ
R dd k i~
but δψ = Λ−δΛ<|~
k|<Λ (2π)d
e k·~x ψ̂k , so
(∇δψ)2 = Λ2 (δψ)2
States δψ
where Z
X Y
= dδψ(x)
States δψ all δψ states
we know the total number of δψ states because we know the total number of
modes in Fourier space in the shell are
Z
dd k
d
= aΛd−1 δΛ
Λ−δΛ<|k|<Λ (2π)
178 CHAPTER 17. RENORMALIZATION GROUP
so Y Y
dx)(aΛd−1 δΛ) ln (...)
R
(...) → e(
x total no.
and
dd x aΛd−1 δΛ 12 ln (Λ2 +r+3uψ̄ 2 )+const.
R
e−F̄ [ψ] = e−F [ψ̄] e−
Now consider
1 1 r 3u
ln [Λ2 + r + 3uψ̄ 2 ] = ln (1 + 2 + 2 ψ̄ 2 ) + const.
2 2 Λ Λ
ǫ2 ǫ3
we expand using ln (1 + ǫ) = ǫ − 2 + 3
1 r 3u 1 r 3u
= { + 2 ψ̄ 2 − ( 2 + 2 ψ̄ 2 )2 + ...}
2 Λ2 Λ 2 Λ Λ
const. const.
1 r7 3u 2 1 r * 3ur 9u2 4
= { 2 + 2 ψ̄ − ( 2 )2 − 2 ψ̄ 2 − ψ̄ + ...}
2 Λ Λ 2 Λ Λ 2Λ4
1 3u 3ur 1 −9u2
= ( 2 − 4 )ψ̄ 2 + ( 4 )ψ̄ 4
2 Λ Λ 4 Λ
So that
Z
1 3u 3ur 1 −9u2
F̄ [ψ̄] = F [ψ̄] + dd x{ ( 2 − 4 )(aΛd−1 δΛ)ψ̄ 2 + ( 4 )(aΛd−1 δΛ)ψ̄ 4 }
2 Λ Λ 4 Λ
so, Z
1 r̄ ū
F̄ [ψ̄] = dd x{ (∇ψ̄)2 + ψ̄ 2 + ψ̄ 4 }
2 2 4
where,
3u 3ur
r̄ = r + ( − 4 )(aΛd−1 δΛ)
Λ2 Λ
−9u2
ū = u + ( 4 )(aΛd−1 δΛ)
Λ
But
r̄ = r(Λ − δΛ) etc,
so
r(Λ − δΛ) − r(Λ) 3u 3ur
= ( 2 − 4 )aΛd−1
δΛ Λ Λ
and
u(Λ − δΛ) − u(Λ) −9u2
= ( 4 )aΛd−1
δΛ Λ
or,
1 ∂r 3u 3ur
− = ( 2 − 4 )Λd−1
a ∂Λ Λ Λ
1 ∂u −9u2 d−1
− = ( 4 )Λ
a ∂Λ Λ
17.1. ǫ–EXPANSION RG 179
L4−d 1
−3urL4−d = −3u′ r′ 4−d 2
= −3u′ r′ 2
L L L
and
L4−d 1
9u2 L4−d = 9u′2 = 9u′2 4−d
(L4−d )2 L
Hence we obtain
dr′
−L = −2r′ − 3u′ + 3u′ r′
dL
and
du′
−L = −(4 − d)u′ + 9u′2
dL
Which are the recursion relations, to first order. Evidently there is still a trivial
Gaussian fixed point r∗ = u+ = 0, but there is a nontrivial “Wilson–Fisher”
fixed point which is stable for d < 4 ! From the u equation (using ǫ = 4 − d)
ǫ
u∗ =
9
and in the r equation (using u∗ ∼ r∗ ∼ ǫ with ǫ << 1), 2r∗ = −3u∗ , so
ǫ
r∗ = −
6
180 CHAPTER 17. RENORMALIZATION GROUP
′
−L du
dL d>4
d=4
Wilson–Fisher fixed pt.
Stable for d > 4
u∗
u′
Gaussian d<4
fixed pt.
Figure 17.15:
Linearlizing around this fixed point gives the dimensions of r and u as that fixed
point is approached.
Let δr = r′ − r∗ , δu = u′ − u∗ . In the u equation,
dδu 2δu
−L = −ǫu∗ − ǫδu + 9(u∗ )2 (1 + ∗ )
dL u
= [18u∗ − ǫ]δu
dδu
−L = ǫδu (Wilson–Fisher fixed pt.)
dL
So the dimension of u is ǫ.
In the r equation,
dδ r
−L = −2r∗ − 3u∗ − 2δr + 3u∗ δr
dL
but δr >> δu as L → ∞, so,
dδr
−L = (−2 + 3u∗ )δr
dL
or,
dδr ǫ
−L = −2(1 − )δr Wilson–Fisher fixed pt. (stable d < 4)
dL 6
So the dimension of r is −2(1 − ǫ/6).
Similarly, near the Gaussian fixed pt. where u∗ = r∗ = 0, we have (as we
did before)
dδu
−L = −ǫδu
dL
dδr
−L = −2δr
dL
Giving the dimensions of r and u there.
17.1. ǫ–EXPANSION RG 181
[r] = 1/L1/ν
Hence
1
ν= , Gaussian fixed pt.
2
1 ǫ
ν = (1 + ), Wilson–Fisher fixed pt.
2 6
To get another exponent we use
1
[ψ] = d
L 2 −1+η/2
or more precisely
1
[ψ] = d f (rL1/ν , uL−yu )
L 2 −1+η/2
where (
−ǫ, Gaussian
yu =
ǫ, Wilson–Fisher
is the dimension of u. Choosing the scale
rL1/ν ≡ 1
gives
ν
[ψ] = r 2 (d−2+η) f (1, urνyu )
for the Wilson–Fisher fixed point this is straight forward. Since νyu > 0, the
term urνyu is irrelavant as r → 0. Hence
ν : const.
[ψ] = r 2 (d−2+η) f (1, 0)
≡ rβ
Now it turns out η = 0 + O(ǫ2 ) (which can be calculated later via scaling
relations), so,
11 ǫ
β= (1 + )(2 − ǫ)
22 6
1 ǫ ǫ
= (1 + )(1 − )
2 6 2
or
1 ǫ
β= (1 − ) Wilson–Fisher fixed pt.
2 3
The Gaussian fixed point is more tricky (although academic since it is for d > 4).
We again obtain
ν
[ψ] = r 2 (d−2+η) f (1, urνyu )
182 CHAPTER 17. RENORMALIZATION GROUP
but now ν = 12 , yu = −ǫ = d − 4, η = 0, so
d−2 d−4
[ψ] = r 4 f (1, ur 2 )
This looks like β = d−2 4 ! Not the well–known β = 1/2. The problem is that
u is irrelevant, but dangerous, for T < Tc (where ψ is nonzero). Without this
quartic term, the free energy is ill defined. For the equation
p above, f (1, u → 0)
is singular. Its form is easy to see. By dimensions ψ ∼ r/u below Tc , so
1
f (1, u → 0) =
u1/2
and
d−2 1
[ψ]u→0 = r 4
d−4
(ur 2 )1/2
d−2 d−4
r 4 − 4
=
u1/2
r 1/2
=( )
u
So, β = 1/2 for the Gaussian fixed point. Hence d ≥ 4, the Gaussian fixed point
is stable and α = 0, β = 1/2, δ = 3, γ = 1, η = 0, ν = 12 (The remainder are
obtained carefully with the scaling relations).
For d < 4, the Wilson–Fisher fixed point is stable and since α + 2β + γ = 2,
γ = ν(2 − η), γ = β(δ − 1), νd = 2 − α we have:
ǫ=1 Ising d = 3
α = 6ǫ 0.17 0.125...
β = 12 (1 − 3ǫ ) 0.33 0.313...
γ = 1 + 6ǫ 1.17 1.25...
δ = 3(1 + 3ǫ ) 4 5.0...
ν = 21 (1 + 12
ǫ
) 0.58 0.638...
η = 0 + O(ǫ2 ) 0 0.041...