1

Solitons and Instantons

LECTURE NOTES

Lecture notes largely based on a lectures series given by Csaba Csaki

at Cornell University in 2013

Notes Written by: JEFF ASAF DROR

2016
Contents

1 Introduction 5

2 Scalar solitons 6
2.1 1 + 1 dimensional solitons . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Soliton basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Static solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.3 Explicit solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.4 Particle-like properties . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Complications and topology . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.1 Multiple scalar fields . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.2 Topological charge . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.3 Derrick’s theorem and D >1 . . . . . . . . . . . . . . . . . . . . 16

3 Solitons in gauge theories 18

3.1 Review of gauge fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Non-abelian Higgs theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3 Vacua of spontaneously broken gauge theories . . . . . . . . . . . . . . . 21
3.3.1 Georgi-Glashow model . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.2 SU(5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 Topological solutions with SSB . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Vortex solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5.1 Critical Vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4 Homotopy Theory 32
4.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 Important examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3 Exact homotopic sequence . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5 Magnetic Monopoles 42
5.1 Monopoles in general . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.1.1 Dirac’s derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.1.2 Derivation using quantization of angular momentum . . . . . . . 45

2
CONTENTS 3

5.2 Dyons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3 Duality transformation of Maxwell’s equations . . . . . . . . . . . . . . . 47
5.4 Topological Monopoles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.4.1 ’t Hooft-Polyakov monopole . . . . . . . . . . . . . . . . . . . . . 49
Preface

If you have any corrections please let me know at ajd268@cornell.edu. Useful references
for this course are Aspects of Symmetry by S. Coleman[1], Solitons and Instantons by R.
Rajaraman [2], and ABC of Instantons by M. Shifman et al. [3]

4
Chapter 1

Introduction

This lecture series will be study of non-trivial field configurations in quantum field the-
ories. Conventional quantum field theory (QFT) assumes the classical field we expand
around are independent of space and time. While this is true for the bulk of effects in
QFT, there are some states and effects which are due to non-trivial topologies for the
unperturbed state. These effects are not small perturbations and thus never be extracted
using a perturbative expansion. We begin by studying these effects in Minkowski space,
where we show that additional particles exist in QFT which are not excited states of a
trivial classical configuration. Such particles are collectively known as solitons. We then
move on to studying such solutions in Euclidean space, which provide a description of
tunneling phenomena in QFT through what are known as instantons. Tunneling can
have profound effects producing new effective operators into the Lagrangian.
Solitons and instantons are nonperturbative solutions of the classical nonlinear equa-
tions of motion. We differentiate the two as follows:
Solitons Instantons
Minkowski Euclidean
Finite energy, E < ∞ Finite action, S < ∞
Non-dissapative, remain localized Not a particle, describes tunneling effect
The similarities of the two are:

• Neither of these two involve an expansion in the coupling constant.

• Topological conservation - there will be a different kind of conserved charge but

this current doesn’t follow from Noether’s theorem.

5
Chapter 2

Scalar solitons

2.1 1 + 1 dimensional solitons

2.1.1 Soliton basics
We begin our discussion with classical scalar field theory in Minkowski space. As in
ordinary QFT, the classical solutions will be the vacuum for which we quantize our
theory. Thus understanding the classical solutions is instrumental in understanding the
full quantum theory. To this end we study the Euler-Lagrange equations of motion,
∂L ∂L
∂µ = (2.1)
∂(∂ µ φ) ∂φ
As we will see solutions to this equation can vary in space. In order for this solution to
have physical importance, it needs to have a finite energy,
Z
E ≡ dd xE(x, t) (2.2)

where, E ≡ T 00 . Furthermore, we want vacuum states which are stable over time, instead
of converting to trivial solutions at late times. This condition can be written as,
lim max E(x, t) 6= 0 (2.3)
t→∞ all x

This is a physicists definition of the soliton. This slightly differs from a mathematicians
definition which involves a requirement that the superposition of two solitons remain
a soliton. They would call the less restrictive solutions that satisfy the conditions we
mentioned above, solitary waves.
The crucial aspect of all these solutions is going to be, what is the manifold of vacua
for the theory. If there is no nonlinear term there is nothing nontrivial. We really want
to take the interaction term seriously and add it to the action.
We start with φ4 theory1 ,
1 1
L = ∂µ φ∂ µ φ − λφ4 (2.4)
2 4
1
We use the negative spatial metric throughout and gµν = (1, −1) for two dimensions

6
2.1. 1 + 1 DIMENSIONAL SOLITONS 7

The equation of motion for this theory is ( ≡ ∂µ ∂ µ ),

φ + λφ3 = 0 (2.5)

Our goal is to find solutions this partial differential equation with finite energy. Instead
of attacking this problem head-on, lets first compute the energy. The potential is just a
quartic,

V (φ)

The energy momentum tensor is found by considering the variation of the Lagrangian
under space-time translations which gives,

1 λ
T µν = ∂ µ φ∂ν φ − (∂µ φ)2 δ µν + φ4 δ µν (2.6)
2 4

allowing us to compute the energy density (we use a prime to indicate a spatial derivative)

1 1 λ
E = φ2 + φ02 + φ4 (2.7)
2 2 4
The energy of the given solution is,
Z ∞  
1 2 1 02 λ 4
E [φ] = dx φ + φ + φ (2.8)
−∞ 2 2 4

Here we are working in 1 + 1 dimensions. The energy is clearly positive definite. The
energy is zero only if φ = 0. But we are not requiring zero energy solutions, we are just
requiring finite energy solutions. Can we find other finite energy solutions? In this case,
no. Here we have only 1 classical vacuum. To get a finite energy, we must have φ → 0
for x → ±∞. While field configurations that start and end with the same value of φ are
solutions, it turns out there are no non-trivial solutions with these boundary conditions
(we show this rigorously later on).
Now instead lets consider a theory with two vacua:
8 CHAPTER 2. SCALAR SOLITONS

V (φ)

A sample potential is given by,

1 1 λ
V (φ) = ∂µ φ∂ µ φ + φ2 − φ4 (2.9)
2 2 4
Then the energy density expression then is,
Z ∞  
1 2 1 02 1 2 λ 4
E [φ] = dx φ + φ − φ + φ (2.10)
−∞ 2 2 2 4
√
The minima occur at φ± = ±1/ λ. Then the condition to have finite energy solutions
is,
lim φ → φ± (2.11)
x→±∞

This problem has less trivial boundary conditions and can yield non-trivial solutions.
Instead of study the two previous cases in detail, lets consider an arbitrary potential.
A Lagrangian with a potential, U (φ) is given by,
1
L = (∂µ φ)2 − U (φ) (2.12)
2
with U (φ) being some complicated function,

U (φ)

φ
2.1. 1 + 1 DIMENSIONAL SOLITONS 9

There are many possible positions where the potential is vanishing. If we denote the
minima with φi then we have U (φi ) = 0. The equation of motion is given by,
∂U
φ̈ − φ00 = − (2.13)
∂φ
The energy density is given by,
Z ∞  
1 2 1 02
E [φ] = dx φ + φ + U (φ) (2.14)
−∞ 2 2
E [φ] = 0 implies that the field is constant, φ = φi . Alternatively finite solutions can be
acheived if scalar field asymptotically approaces on of the vacua. Therefore, E [φ] < ∞
implies that,
lim φ = φi , lim φ = φj (2.15)
x→+∞ x→−∞
The finite energy solutions interpolate between two of the zeroes in the potential.
We now want to find explicit form for these solutions. Solving the partial differential
equation is difficult, but we can simplify our analysis using Lorentz invariance. Instead
of solving the full equations we can first look for static solutions and then add the time
dependence by boosting them 2 .

2.1.2 Static solutions

The equation of motion in the static case is,
∂U
φ00 = (2.16)
∂φ
This is an ordinary second order differential equation. This is reminiscent of Newton’s
second law for a 1D particle in a potential with the sign of the potential is reversed,
ẍ = − ∂U
∂x
.

φ
−U (φ)

2
Its not apriori clear whether there are some additional time dependent solutions that we can’t get
from boosting them. While we acknowledge this as a possibility we ignore it for our purposes.
10 CHAPTER 2. SCALAR SOLITONS

As in Newton’s equation we can introduce a conserved quantity, the “total mechanical

energy”,
1
W = φ02 − U (2.17)
2
You know that when x → ±∞ the field has to approach a constant value (its at minimum
of the potential). This implies that φ0 → 0. Futhermore, at a zero of the potential we
know that U = 0. Hence to get the finite field theory energy solutions, the mechanical
energy has to be zero,
lim W = 0 (2.18)
x→±∞

Furthermore, since this quantity is conserved we have,

1 02 p
φ = U ⇒ φ0 = ± 2U (φ) (2.19)
2
What sort of solutions could this equation yield? You are interested in the mechanical
solutions in this potential that go from one maximum (in this inverted potential) to
another maximum. This could be something that goes from φmax 1 → φmax
2 , φmax
2 → φmax
3 ,
max
etc, where we denote the different maxima by φi .
Claim 1. In 1D, you can only have solutions that go between neighbooring vacua.
Proof. If you try to go from φmax
1 → φmax3 then you can only do it by going through φmax
2 .
max max
When you go through φ2 you will have U (φ2 ) = 0. Because this is a maximum the
derivative is also vanishing, ∂U (φmax
2 )/∂φ. This is telling us that,
1 02 p
φ − U (φ) = 0 ⇒ φ0 = 2U (φ) (2.20)
2
which implies that at the minima, φ0 = 0. Now

00 ∂U 
φ = =0 (2.21)
∂φ φ=φmax
2

Furthermore,
d ∂U ∂ 2 U 0 φmax
φ000 = = φ −−2−→ 0 (2.22)
dx ∂φ ∂φ2
and similary for higher derivatives. Thus the particle really gets stuck there. You cannot
roll down on the other side. In higher dimensions you can avoid a vacua by going around
it making this feature only present in the one dimensional case.
Lets now address a claim we made earlier:
Claim 2. Any solution with the same initial and final vacuum is a trivial solution.
Proof. Suppose we had a solution that started and ended at the same vacuum, φ0 , but
moved away from that vacuum away from infinity. The field starts at φ0 and at some
point, φ1 , it needs to turn around to return to φ0 . At the turnaround point we must have
φ01 = 0, but this implies that U (φ1 ) = 0 by equation 2.19 and hence φ1 is a vacuum and
hence the field will be stuck there.
2.1. 1 + 1 DIMENSIONAL SOLITONS 11

For n minima you can have the possible transitions,

n−1 n−1
z }| { z }| {
1 → 2 , 2 → 3 , ... , n − 1 → n n → n − 1 , n − 1 → n − 2 , ... , 2 → 1

Thus in total we have

2(n − 1) (2.23)
possible solutions. The ones that go one direction you call solitons and the ones going in
the other direction are called antisolitons.

2.1.3 Explicit solution

To get an explicit solution we need to integrate over the equation of motion,

1 02
φ = U (φ) (2.24)
2
Separating variables gives,
Z φ(x)
1
x − x0 = ± dφ̃ q (2.25)
φ0 2U (φ̃)

We now consider the kink solution. This is a 1 + 1 dimensional soliton with the
potential,
2
m2

λ 2
U (φ) = φ − (2.26)
4 λ
The higher dimensional version of these are domain walls The way the constant is picked
λ is still the quartic and m is the “mass” of the quantized field. The vacua are at,
r
m2
φ0 = ± (2.27)
λ
We plug this into our general expression,
r Z φ(x)  −1
2 2 m2
x − x0 = ± dφ̃ φ̃ − (2.28)
λ φ0 λ

Lets pick x0 to be the center of the soliton. We can pick where exactly the center of the
soliton is. We pick φ0 = φ(0) = 0.
The integral is easy to perform using Mathematica,
Z x
1 1 x
dx 2 2
= − tanh−1 (2.29)
0 x −a a a
12 CHAPTER 2. SCALAR SOLITONS
√
Using this result we get (with a = m/ λ),
√ √ !
2 −1 λ
x − x0 = ± tanh φ(x) (2.30)
m m
which gives,
 
m m
φ(x) = ± √ tanh √ (x − x0 ) (2.31)
λ 2
This is our first explicit soliton solution. It takes the form,
φ(x)
m
√
λ

kink

pert. vacua x
x0
antikink

− √mλ

Its obvious that we can’t call this field configuration a particle as its not localized in
space. However, nobody really cares about the fields (perturbative vacua aren’t localized
either!). Its only the energy density that needs to be localized. This determines if you
can have a particle-like behavior or not.
The energy density is,
1
E(x) = φ02 + U (φ) (2.32)
2
We have,
m 1
φ0 = √ h i (2.33)
λ cosh2 √m (x − x0 )
2
4
1 02 m 1
φ = h i (2.34)
2 4λ cosh4 √m (x − x )
2 0

The total energy density is just,

m4 1
E(x) = h i (2.35)
2λ cosh √m (x − x )
4
2 0

This is indeed localized as desired,

2.2. COMPLICATIONS AND TOPOLOGY 13

E(x)

1
∆L = m

x
x0

To assign some classical mass we would find the total energy,

Z ∞ √
2 2 m3
M ≡E= dxE(x) = (2.36)
−∞ 3 λ

Note that λ here is not dimensionless since we are not working in 3+1 dimensions, so this
does indeed have dimensions of mass. We see that the mass, M ∝ λ1 . This is something
you will never get in perturbation theory.

2.1.4 Particle-like properties

To get the solution for all times can boost this solution,

x − x0 → γ(x − x0 − vt) , (2.37)

γ ≡ (1 − v 2 )−1/2 . Boosting the field,

 
m γm
φ → ± √ tanh √ (x − x0 − vt) (2.38)
λ 2

with an energy density,

γm4
 
4 γm
E= sech √ (x − x0 − vt) (2.39)
2λ 2

Performing the integral over space gives as expected,

E = γM (2.40)

2.2 Complications and topology

We now consider more complicated soliton solutions.
14 CHAPTER 2. SCALAR SOLITONS

2.2.1 Multiple scalar fields

First suppose we have multiple scalar fields,

φi (x, t) i = 1, ..., N (2.41)

with the Lagrangian and equations of motion:

1 1 ∂U
L = φ̇i 2 − φi 0 2 − U (φi ) ⇒ φi = − (2.42)
2 2 ∂φi
Static solutions can be found from,
∂U
φi 00 = (2.43)
∂φi

The classical mechanical analogue is a N -dimensional motion of particle in −U (φi ) po-

tential. The total energy is,
Z X
φi 0 2 + U (φi )
 i
E φ = dx (2.44)
i

The main difference here is that we can have a non-trivial solutions iwth one minimum
for the potential. For example:

φ2

φ0×

φ1

2.2.2 Topological charge

In both cases we’ve seen so far there is an important quantity which characterizes the
field configuration, the difference in the vacuum at spatial infinity. We claim this is a
conserved quantity suggesting the following definition:
Def. 1. Topologically Equivalent: Two solutions are said to be topologically equivalent
if they cannot be continuously deformed into one another without passing through a barrier
of infinite action.
Def. 2. Continuous Deformation: For two solutions, f1 (x) and f2 (x), a continu-
ous deformation, parameterized by t ∈ [0, 1], between the two solutions is a continuous
function, F (t, x), such that F (0, x) = f1 (x) and F (1, x) = f2 (x).
2.2. COMPLICATIONS AND TOPOLOGY 15

Are the two kink solutions topologically equivalent? To check this we introduce a
generic deformation,
m m
Φ(t, x) = √ [g(t) tanh y + h(t) tanh(−y)] y ≡ √ (x − x0 ) , (2.45)
λ 2
m
= (g − h) √ tanh y , (2.46)
λ
where g, h are arbitrary continuous functions of t with g(0) = h(1) = 1 and g(1) = h(0) =
0. The static action is given by,
2
m2
Z 
1 02 λ 2
S(t) = dx Φ + Φ − (2.47)
2 4 λ
4 Z
m 1
2
= dx(g − h)2 4 + (g − h)2 tanh2 y − 1 (2.48)
4λ cosh y

The first integral is convergent, however the second diverges for |g(t) − h(t)| =
6 1. We
have g(0) − h(0) = 1 and g(1) − h(0) = −1. and since g and h are continuous, there must
be a point in t ∈ (0, 1) for which the integral diverges. Thus we conclude that these two
solutions are topologically inequivalent.
Thus at least for the kink the quantity,

φ(∞, t) − φ(−∞, t) (2.49)

cannot be changed with a finite amount of energy and must be conserved. We will always
assume that the non-trivial topological solutions we find are topologically inequivalent.
Thus we can define a quantity known as the “topological charge” that quantifies the
difference between vacua:
Q ∝ φ(∞, t) − φ(−∞, t) , (2.50)
where the constant of proporitionality is conventional. For the kink one can introduce,
√
λ
Q≡ (φ(∞, t) − φ(−∞, t)) (2.51)
m
Thus, 
2
 (kink)
Q= 0 (trivial) (2.52)

−2 (anti-kink)


Note that this charge does not follow from Noether’s theorem, but instead is conserved
independently from the equations of motion. Since we have a conserved charge, we expect
some associated current. We can find the current by rewriting the charge,
√ Z
λ
Q= dx∂x φ(x, t) (2.53)
m
16 CHAPTER 2. SCALAR SOLITONS
√
Thus we know that j 0 = mλ ∂x φ. We want to make this into a four-vector. The only
tensors we have at our disposal are g µν , µν , and ∂ µ . Thus we must have,
√
µ λ µν
j =  ∂ν φ (2.54)
m
This is divergenceless as required, ∂µ j µ ∝ µν ∂µ ∂ν φ = 0.

2.2.3 Derrick’s theorem and D > 1

We now consider solitons in more than one spatial dimension, D. In such systems there
is a theorem, known as Derrick’s theorem, which restricts the type of systems that
can have topological solutions. Consider a N -component scalar field, φi (x, t) in D spatial
dimensionn. The Lagrangian is given by,
1
L = ∂µ φi ∂ µ φi − U (φi ) (2.55)
2
The static equations of motion are the solutions of the equation,
∂U
φi = (2.56)
∂φi
with static energy, Z Z
1
E = d x ∂a φ ∂a φ + dD xU (φi )
Di i
(2.57)
2
| {z } | {z }
≡E1 [φi ] ≡E2 [φi ]

Now lets assume we find φi1 which is a solution of the equations of motion. Now consider
the class of field configurations (which are generically not solutions),

φiλ (x) ≡ φi1 (λx) (2.58)

The action for a given solution φλ (x) is given by,

Z  
D 1 i i
 i 
S(λ) = − d x ∂a φ1 (λx)∂a φ1 (λx) + U φ1 (λx) (2.59)
2
Z  
D −D 1 2 ∂ i ∂ i
= − d yλ λ φ (y) a φ (y) + U [φ1 (y)] (2.60)
2 ∂y a ∂y
2−D −D
= −(λ E1 + λ E2 ) (2.61)

To have an extremum of the action,

dS(λ) 2−D
= (2 − D)λ1−D E1 − Dλ−D−1 E2 = 0 ⇒ E2 = E1 (2.62)
dλ D
Now the key observation, is that E1 and E2 are both positive definite quantities. This
puts a restriction on the possible values of D. To this let us consider the different cases:
2.2. COMPLICATIONS AND TOPOLOGY 17

1. D = 1: In this case we find, E1 = E2 . This says that in one spatial dimension the
tension in the field is equal to its potential energy. This is an analogue to the viral
theorem.

2. D = 2: This says that E2 = 0 or equivalently, U = 0. The only known possibility

where you have a non-trivial solution is in the non-linear σ model where φi φi = a2 .

3. D > 2: E2 < 0! since E2 is positive definite, there are no non-trivial solutions for
D > 2.

Thus, the somewhat dissapointing conclusion is that in the 3 + 1 dimensions scalar fields
cannot have instanton solutions. Note that its still possible to lift lower dimensional
solutions to 3 + 1 dimensions but these won’t be finite energy and hence do not have a
particle interpretation (e.g., domain walls).
Chapter 3

Solitons in gauge theories

3.1 Review of gauge fields

Before plunging into solitons in gauge theories we quickly review the basics of non-abelian
gauge theories. A group G is charactorized by the structure constants, f abc , which are
related to the generators of the group by,
 a b
T , T = −if abc T c (3.1)

and a = 1, 2, ..., dim(G). The gauge fields are denoted by,

Aµ ≡ Aaµ T a (3.2)

and the fields transform under the gauge symmetry through,

i
Aµ → gAµ g −1 + (∂µ g)g −1 (3.3)
e
where e denotes the coupling constant, and g is a group element in the adjoint represen-
tation,
g ≡ eiαa Ta (3.4)
Explicitly we can write the gauge transformation in terms of the generators:

gAµ g −1 = eiαa Ta T b e−αc Tc Ab (3.5)

= Aµ + iαa (Ta Tb − Tb Ta ) Ab + ... (3.6)
= Aµ + αa fabc Tc Abµ (3.7)

Furthermore,
1 i
(∂µ g)g −1 = (i∂µ αa Ta ) (1 − αa Ta ) (3.8)
2 e
1
= − ∂αa Ta (3.9)
e

18
3.2. NON-ABELIAN HIGGS THEORY 19

Therefore,
1
Aµ → Aµ − ∂αa Ta − a fabc Tc Ab (3.10)
e
Under a gauge transformation the field strength tensor,
a
Fµν = ∂µ Aaν − ∂ν Aaµ − ef abc Abµ Acν (3.11)

transformations in the adjoint representation,

Fµν → gFµν g −1 (3.12)

3.2 Non-abelian Higgs theory

We now consider a non-abelian gauge theory with a Higgs field in a generic representation.
The Lagrangian is given by,
1
L = − TrFµν F µν + |Dµ φ|2 + U [φ] (3.13)
4
with the covariant derivative,

(Dµ φ)m = ∂µ φm + ieAaµ tamn φn (3.14)

Computing the equations of motion is straightforward. The final result is,

Dµ F a µν = ieφta Dν φ + h.c. (3.15)

 ∗
ν ∂U
Dµ D φm = (3.16)
∂φm
We now look for static solutions,

∂0 φ = 0 (static) (3.17)

Claim 3. There exists a gauge such that A0 = 0. This is known as the temporal gauge.
Proof. Consider a generic gauge configuration and apply a gauge transformation,
i
A0 → gA0 g −1 + (∂0 g)g −1 (3.18)
e
We need to show there exists a g such that,
i
gA0 g −1 = − (∂0 g)g −1 (3.19)
e
for any A0 . This condition can be rewritten,
∂g
g −1 = ieA0 (3.20)
∂t
20 CHAPTER 3. SOLITONS IN GAUGE THEORIES

This is a well known diffential equation, familiar from constructing correlation functions
in QFT, and has the solution,
 Z 
0 0
g = T exp ie A0 (t )dt (3.21)

Thus for static solutions we can always work in a gauge such that,

D0 φ = 0 (static) (3.22)

Note that this still does not completely fix the gauge as we can still make t-indepent
gauge transformations.
In this gauge fixing,

F0ia = ∂0 Aai − ∂i Aa0 − ef abc Ab0 Aci = ∂0 Aai (3.23)

which gives the action,

Z   Z  
D 1 a 2 2 D 1 a 2 2
S = d x (∂0 Ai ) + |∂0 φ| + d x (Fij ) + |Di φ| + U [φ] (3.24)
2 4
 2
= 0 + E1 F + E2 [Di φ] + E3 [U ] (static) (3.25)

Using this we can now rebuild the scaling arguement in the theory.
Assume we found a static solution to the equations of motion,

φ1 (x) , Aµ1 (x) (3.26)

Now consider the family of functions which are not solutions to the equations of motion,
except at λ = 1:
φλ (x) = φ1 (λx) , Aµλ (x) = λAµ1 (λx) (3.27)
which implies,
Fij (x) → λ2 Fij (λx) , Di φ → λDi φ(λx) (3.28)
This gives a rescaled action,
Z
1
µ
S [φL , Aλ ] = −λ−D
dD (λx) λ4 Fij (λx)Fij (λx) + λ2 |Di φ|2 + U (φ) (3.29)
4
−D
λ E1 + λ2 E2 + E3
4


= −λ (3.30)

As before we should have an extremum in the action when λ = 1:

− Dλ−D−1 λ4 E1 + λ2 E2 + E3 + λ−D 4λ3 E1 + 2λE2 = 0

(3.31)

or
(4 − D)E1 + (2 − D)E2 − DE3 = 0 (3.32)
As before, lets consider the different cases
3.3. VACUA OF SPONTANEOUSLY BROKEN GAUGE THEORIES 21

1. D = 1: Gauge fields don’t propagate with one spatial dimension, so we don’t expect
any interesting instanton effects here

2. D = 2: Our master equation tells us that E1 = E3 . For this peculiar possibility,

the scalar potential energy is exactly equal to the gauge field energy. This is known
as the “vortex” solution.

3. D = 3: The master equations says that E1 − E2 − 3E3 = 0. This corresponds to a

monopole solution.

4. D = 4: Here the master equations gives, −2E2 − 4E3 = 0. This can only be a
solution if E2 = E3 = 0. In other words in four spatial dimensions soliton solutions
can exist only if the scalar field has vanishing energy ⇒ solution involves only the
gauge field. These will be of fundamental importance when we discuss instantons
but we postpone this discussion until then.

3.3 Vacua of spontaneously broken gauge theories

We now study the vacuum of a gauge theory with a manifold of vacua, instead of finite
set of discrete vacua. The concept of a manifold of vacua is familiar from spontaneously
broken symmetries. Since we are interested in D = 3, we study theories with scalar
and gauge fields. Both of these will have a non-trivial vacuum configurations. The
perturbative vacuum is simmply,

φ0 = const , Aaµ = 0 (3.33)

However, our interest will be in more intricate vacua.

Consider the gauge group, G (with elements, g), broken to a subgroup, H (with
elements h). We denote a vacuum state by φ0 . Since the vacuum is invariant under the
subgroup we have,
hφ0 = φ0 (3.34)
Gauge fields in H will remain massless and the other gauge bosons (those in G/H) pick
up a mass. U (φ) must be invariant under, φ → gφ, which in turn also implies,

U (gφ0 ) = 0 (3.35)

Thus the set of vacua related to φ0 by a full gauge transformation correspond to minima.
While gφ0 always corresponds to a minima its non-trivial to know whether this will
correspond to all minima.

Claim 4. If gφ0 can generate all minima than the minimal set of generators that can
generate all minima, is the “coset”, G/H, defined by g1 = g2 h.
22 CHAPTER 3. SOLITONS IN GAUGE THEORIES

Proof. Let us parameterize all minima by a vector, θ: φ0 (θ). By assumption it is given

by a group element acting on a single vacuum,

φ0 (θ) = g1 φ0 (0) (3.36)

Generically this choice of g will not be unique and so we’ll also have

φ0 (θ) = g2 φ0 (0) (3.37)

This implies that,

g2−1 g1 φ0 (0) = φ0 (0) (3.38)
which is the defining characteristic of the unbroken subgroup. Thus,

g2−1 g1 = h ⇒ g1 = g2 h (3.39)

Thus the elements g which are identified with g1 = g2 h can generate all the minima.

One subtlety to the above is sometimes the invariance of the potential (Ḡ) is a bit
bigger than that of the full gauge group, G. Then subgroup may also be a bit smaller, H̄.
What you are really after is Ḡ/H̄. We will see an example of this shortly when consider
SU(5).

3.3.1 Georgi-Glashow model

The Georgi-Glashow model was a candidate for the weak interactions before the confir-
mation of the Z boson. The model comprises of SU (2) gauge symmetry with a Higgs
triplet, φa and a potential,
λ i i
2
U= φ φ − a2 (3.40)
4
The symmetry of the potential is, SU (2).

Claim 5. SU(2) is isomorphic to S 3 (3D sphere embedded in 4D)

Proof. An arbitrary element of SU(2) can be written as,

 
a b
(3.41)
−b a∗
∗

2
with the condition |a2 | + |b|2 = 1. Writing a = y1 + iy2 and b = y3 + iy4 gives the
condition,
y12 + y22 + y32 + y42 = 1 (3.42)
which is the defining equation for 3-sphere.
3.3. VACUA OF SPONTANEOUSLY BROKEN GAUGE THEORIES 23

The Higgs gains a vacuum expectation value (VEV) which by gauge symmetry can
be taken to be in the first component,
 
h+a
φ =  φ2  (3.43)
φ3

Then,
λ 2
2
V = φ2 + φ23 + 2ha (3.44)
4
The h field is massive and φ2 and φ3 are massless giving 2 goldstone bosons (GB). The
invariant subgroup is H = SO(2) and corresponds to rotations between φ2 and φ3 . The
vacuum manifold is given by SU (2)/U (1) which is a 2-sphere, S 2 (a sphere embedded in
a 3 dimensional space). This is because the vacua are characterized with,

φ21 + φ22 + φ23 = a2 (3.45)

which is the defining equation of a 2-sphere.

φ2

φ21 + φ22 + φ23 = a2

φ3

φ1

3.3.2 SU(5)
Next consider G =SU(5), a candidate for grand unification of the SM, broken by a
scalar adjoint Higgs, φ, to SU (3) × SU (2) × U (1). A scalar adjoint can be formed from a
fundamental and anti-fundamental, forming a hermitian tracless 2×2 matrix transforming
as φ → gφg −1 . Usually the potential which gives rise to this breaking has an accidental
Z2 symmetry under the transformation,
 
−1

 −1 

φ→  −1 φ
 (3.46)
 −1 
−1
24 CHAPTER 3. SOLITONS IN GAUGE THEORIES

This is not an element of SU (5) (it doesn’t have determinant equal to 1). Thus the
potential is invariant under
Ḡ = SU (5) × Z2 (3.47)
An adjoint of the VEV,
 
1

 1 

φ0 = v 
 1 
 (3.48)
 −3/2 
−3/2

can produce the desire symmetry pattern. H is made of,

13×3 e 13×3
      2iθ 
g1
M1 ≡ , M2 ≡ , M3 ≡
12×2 g2 e−3θ/2 12×2
(3.49)
Clearly,
Hφ0 = hφh−1 = φ0 (3.50)
as desired. However, not all of the vacua at distinct, e.g., g1 = e−2π/3 13×3 , g2 = −12×2 ,
θ = π/3. Then
M1 M2 M3 = 15×5 (3.51)
Therefore, the invariance group is not SU (3) × SU (2) × U (1) but smaller. One can show
that the actual subgroup is SU (3) × SU (2) × U (1)/Z6 . This will effect properties of the
monopole!

3.4 Topological solutions with SSB

We now repeat the procedure to find topological solutions. As before we work in the
temporal gauge. Furthermore, it will be convenient to work in polar coordinates. We
claim without proof that one can use the remaining gauge freedom to fix the gauge such
that,
Aar = 0 (3.52)
In order to have a finite energy density we must have,
Z
r→∞
dD x |Di φ|2 −−−→ 0 (3.53)

Due to our gauge fixing condition this implies,

r→∞
∂r φ −−−→ 0 (3.54)

and so φ will asymptotically be r independent,

φ∞ (r, θi , t) = φ∞ (θi , t) (3.55)

3.5. VORTEX SOLUTION 25

The field values at ∞ are called the “sphere at infinity”. We learned from our experience
with solitons that different vacuum configurations of the field at the boundaries, yields
different topological solutions. Thus this sphere at (spatial!) infinity, should somehow be
mapped to the different vacua. In D spatial dimensions this implies a mapping,

S D−1 → G/H (3.56)

The mathematical tool to compute these mappings is known as “homotopy theory”,

which we’ll study in more detail soon. However, in many cases you can use physical
arguments to compute most of what you need to understand the topology, as we did for
SU (2) → U (1) where we found we needed a mapping from S 2 → S 2 .
Our goal is to find non-trivial classical solutions to the equations of motion. Thus we
want the derivatives of the fields to be non-vanishing. However, equation 3.53 combined
with equation 3.54, greatly restrict the possible solutions! We must have,

1 1 ∂φ∞
lim Dθi φ = lim + iAθi φ∞ = 0 (3.57)
r→∞ r→∞ r sin θ... ∂θi

Since φ is asymptotically r-independent, we know how Aθi behaves at large r:

1
lim Aθi ∝ (3.58)
r→∞ r
Recall that we are interested in finite action solutions. For a potential falling off as 1/r
is the energy in the E and B fields finite? Equation 3.24 shows that the energy in our
gauge is, Z Z Z
1 D a 2 D−1 1
d x(Fij ) ∼ d θ drrD−1 4 (3.59)
4 r
which is convergent if D − 5 ≤ −2 or D ≤ 3. Thus we conclude:

Finite action, non-trivial, solutions to the classical equations of motion can only
exist in spontaneously broken gauge theories if D ≤ 3.

3.5 Vortex solution

The vortex is a topological solution of a spontaneously broken abelian gauge symmetry,
also known as “Landau-Ginsburg” theory. We will focus on the case of 2+1 dimensions,
known as the Neilson-Oloson vertex. It has important physical consequences as it forms a
theory of superconductivity through a dynamical breaking of U(1)EM which occurs throgh
condensation of electron pairs giving a non-zero vacuum expectation value,

hΩ|e− e− |Ωi =
6 0 (3.60)
26 CHAPTER 3. SOLITONS IN GAUGE THEORIES

The unbroken group is G = U(1) with a complex scalar, φ, which in superconductivity,

forms the electron pair: φ ≡ f12 e− −
L eR , where f is some dimensionful scale. We will take
the scalar potential to be,

λ
2
V (φ) = |φ|2 − a2 λ>0 (3.61)
4
The vacuum manifold (parametrized by G/H = U (1) ' S 1 ) is given by,

φ0 (σ) = aeiσ

(3.62)

Since we are working in 2 spatial dimensions, the sphere at infinity is also a circle,

φ(∞, θ) : S 1 → S 1 (3.63)

To describe the boundary conditions we need a mapping between θ → σ, i.e., we need

σ(θ). Pictorally,

sphere at ∞
mapping

σ(θ)

We are interested in homotopically inequivalent mappings.

Def. 3. Homotopically Equivalent: Two mappings, σ1 (θi ) and σ2 (θi ) are homotopi-
cally equivalent if there exists a linear combination of the mappings with a parameter t
such that at t = 0 its equal to σ1 and at t = 1 its equal to σ2 while keeping the base-point
fixed.

Def. 4. Based point: A base point of a mapping, σ(θi ) is reference point of the mapping
where σ(θi,0 ) is always equal to some constant, σ0

Before studying this model in detail lets give a few examples. First consider the
mappings,
σ1 (θ) = θ σ2 (θ) = kθ, k ∈ Z (3.64)
The basepoint we have chosen is θ0 = 2πp, σ0 = 2πp, p ∈ Z. Clearly both these solutions
obey this condition. Are these homotopically equivalent? Consider,

tσ1 (θ) + (1 − t)σ2 (θ) = [k + t(1 − k)] θ (3.65)

This homotopy does not keep the basepoint fixed unless k = 1 for all t and hence these
solutions are not homotopically equivalent. We see that solutions which go around the
sphere at infinity a different number of times are homotopically equivalent.
3.5. VORTEX SOLUTION 27

Alternatively, consider two mappings which both go around the sphere at infinity with
different “speeds”:
θ
σ1 (θ) = θ σ2 (θ) = 2π sin (3.66)
4
These are homotopically eqiuvalent since,

tσ1 (2π) + (1 − t)σ2 (2π) = 2π (3.67)

and hence keeps the basepoint fixed.

Lastly consider the trivial mapping and a mapping which doesn’t go once around the
vacuum: 
θ 0≥θ≥π
σ1 (θ) = 0 σ2 (θ) = (3.68)
(2π − θ) π ≥ θ ≥ 2π
The linear combination:
tσ1 (2π) + (1 − t)σ2 (2π) = 0 (3.69)
and so these are homotopically equivalent.
We see that the key concept in finding charatorizing mappings is the number of
times we go around the circle in coset space as we go once around the circle at infintiy.
This is known as the winding number. The above consider show1 that mappings are
homotopically inequivalent if and only if they have the different winding numbers. As
we’ll see there is an infinite number of such solutions, each having a distinct energy.
Explicitly, the winding number can be written,

σ(2π) − σ(0)
n= (3.70)
Z 2π
2π
1 dσ(θ)
= dθ (3.71)
2π 0 dθ

We can compute the action explicitly for the vacuum manifold, φ = aeiθ . The action is
given by, Z  
2 1 ij 2
− d x Fij F + |Di φ| + U (φ) (3.72)
4
For static solutions the energy is always equal to the negative of the action. Thus to
avoid keeping track of a pesky negative sign, we’ll always work with the energy and look
for extrema in the energy instead. Since we are working with an abelian 2D gauge theory
we can simplify the gauge field:

F12 = ∂x Ay − ∂y Ax = ij ∂i Aj = B (3.73)

At r → ∞,
φ(∞, θ) = aeiσ(θ) (3.74)
1
Technically we didn’t actually show that solutions with different winding numbers are homotopically
inequivalent since we only tried one candidate mapping.
28 CHAPTER 3. SOLITONS IN GAUGE THEORIES

Inserting this into equation 3.57 we get,

a 
lim Dθ φ = lim i∂θ σ − ieAθ a eiσ = 0 (3.75)
r→∞ r→∞ r

or,
dσ
= e lim rAθ (3.76)
dθ r→∞

Note that earlier we concluded that Aθ ∝ 1r so this is a finite quantity. This can be used
to compute the winding number:
Z 2π
e
n= dθ lim rAθ (3.77)
2π 0 r→∞
Z
e
= A · ds (3.78)
2π
Z
e
= B · da (3.79)
2π
This is just the magnetic flux passing through the circle at infinity giving the quantization
condition,
2π
Φ= n (3.80)
e
Thus we conclude that for n 6= 0 this theory has quantized magnetic charge. This is
indeed the case in a superconductor.
Lets pause and summarize what happened here. We studied the classical solutions
to a 2+1 dimensional spontaneously broken U(1) gauge theory and looked for solutions
with non-trivial vacua and finite energy. We found that there exists an infinite set of such
solutions, described by a discrete quantum number: the winding number. There will also
be solutions with trivial vacua, such that the quantum theory contains excitations of
an ordinary scalar field, an ordinary massive gauge field, and of a novel ground state,
whose gauge field configuration produces a non-zero magnetic flux. In a quantum theory
there will be small fluctuations about this vacuum. However, these small fluctuations
(by definition of being small) won’t be able to take your out of this vacuum. Thus this
quantity is a conserved quantity often referring to as a topological charge.
Since we have a charge there should be a corresponding current. We want something
which obeys, ∂µ j µ = 0 and which yields,
Z Z
2 0 e
d xj = − d2 xij ∂i Aj (3.81)
2π
Z
e
=− d2 x0µν ∂µ Aν (3.82)
2π
which suggests,
e µνρ
jµ ≡  ∂ν Aρ (3.83)
2π
∂µ j µ = 0 as a consequence of the antisymmetry of the levi-cevita desired.
3.5. VORTEX SOLUTION 29

3.5.1 Critical Vortex

Obtaining solutions to the coupled, second order, partial differential equations for φ and
Aµ is a difficult problem. It would be much nicer if we could reduce these equations to
first order as we did for the kick solution. Unfortunately, this cannot be done in general
but we can recast the solutions into a more convenient form in a certain limit. The idea is
to use that solutions will minimize the action. If we write the action in terms of positive
definite quantities we can have a better idea of when the action will be minimized.
Consider the action of the static solution,
Z  
2 1 2 2 2 λ 2 2 2


E = d x F12 + |D1 φ| + |iD2 φ| + |φ| − a (3.84)
2 4

where we write the kinetic term in a form convenient for the discussion below. The masses
of the radial mode and the gauge boson are, m2A = 2e2 a2 and m2φ = 12 λa2 , respectively.
Shifting terms around in the action gives,
Z " √ !2
1 λ
d2 x (|φ|2 − a2 ) + |(D1 + iD2 )φ|2 − i (D1 φ)† D2 φ − (D2 φ)† D1 φ


E= F12 +
2 2
√ #
λ
− F12 (|φ|2 − a2 ) (3.85)
2

Using,2 Z Z
† †
2
d x(D1 φ) D2 φ − (D2 φ) D1 φ = ie d2 xφ† F12 φ (3.86)

we have,
Z " √ !2 √ ! # √
1 λ λ λ 2
E= d2 x F12 + (|φ|2 − a2 ) + |(D1 + iD2 )φ|2 + e− |φ|2 F12 + aΦ
2 2 2 2
(3.87)

The first two terms are positive definite. The last term is a (quantized) topological term.
The third term gives the coupling between the scalar and the gauge boson and its sign,
which can be positive or negative depending on the relative value of λ and e, has a crucial
2
Note that,
Z Z Z
d2 x(Di φ)† Dj φ = d2 x(−φ† ∂i + ieφ† Ai )Dj φ = − d2 xφ† Di Dj φ

and
i
F12 = [D1 , D2 ]
e
30 CHAPTER 3. SOLITONS IN GAUGE THEORIES

effect on the qualitative features of the model. The relevant ratio is an order parameter
and can be written in terms of the masses,
√
mφ λ/2
= (3.88)
mA e

If mφ < mA then we have a type I supeconductor. Here the vortices attact each other
destropying superconductivity relatively quickly. If mφ > mA then we have a type II
superconductor. Here the vortices repel and don’t combine together. This leads to a
much larger critical magnetic field.
To get the rough behavior of the solutions its interesting to consider the regime that
mφ = mA , known as the critical vortex. In this case the total energy is just,

Z " √ !2 #
1 λ
E= d2 x F12 + (|φ|2 − a2 ) + |(D1 + iD2 )φ|2 + 2πna2 (3.89)
2 2

We can now reduce the corresponding second order equations by making a bold assump-
tion: the solutions to the equations of motion will minimize these positive definite quanti-
ties, lets assume that the terms are identically zero. In this case we get the corresponding
equations known as the BPS equations:

F12 + e |φ|2 − a2 = 0


(3.90)
(D1 + iD2 )φ = 0 (3.91)

For simplicity lets focus on the case where n = 1. We want to use an satz consistent with
our boundary conditions,

φ(∞, θ) = aeiθ (3.92)

1
lim rAθ = (3.93)
r→∞ e
and φ, Aθ finite as r → 0. This gives the generic form,3

φ(r, θ) = aeiθ ϕ(r) (3.94)

1 1 − f (r) 1 ij xj
Aθ (r, θ) = or Ai = (1 − f (r)) (3.95)
e r e 2r2
3
Recall that to convert from cartesian to polar coordinates:
    
Ax cθ −sθ Ar
=
Ay sθ cθ Aθ
    
∂x cθ −sθ ∂r
= 1
∂y sθ cθ r ∂θ
3.5. VORTEX SOLUTION 31

The field strength is given by,

eF12 = eij ∂i Aj (3.96)

 
1 1 − f (r)
= ∂k xk (3.97)
2 r2
1 1 ∂f
=− (3.98)
2 r ∂r
and

(D1 + iD2 )φ = eiθ (Dr + iDθ ) aeiθ ϕ(r) (3.99)

   
2iθ 1
= ae ∂r ϕ(r) − − eAθ ϕ (3.100)
r
 
2iθ f (r)
= ae ∂r ϕ(r) − ϕ (3.101)
r
which leads to the BPS equations:
1 0
f − 2e2 a2 ϕ2 − 1 = 0


(3.102)
r
f
ϕ0 − ϕ = 0 (3.103)
r
The boundary conditions are,
 

 limr→∞ ϕ(r) = 1 

limr→∞ f (r) = 0
 
(3.104)

 limr→0 ϕ(r) = 0 

limr→0 f (r) = 1
 

Pictorially it looks like,

ϕ(r)

f (r)
r
core

At the core of the vortex, the Higgs field vanishes and the gauge symmetry is unbroken.
However, at large distances it will go its usual value, |φ| = a. Roughly we have,

f (r) ∼ e−mφ r , 1 − ϕ(r) ∼ e−mφ r (3.105)

Chapter 4

Homotopy Theory

4.1 Basics
In our discussions of topological solutions in theories with broken continuous symmetries,
we need to know how to map the sphere at infinity to the vacuum manifold, i.e.,

Sn → M (4.1)

As our base point in the n-dimensional sphere(p0 ) we take the north pole, and the cor-
responding base point in the vacuum manifold as, m0 . In other words we need to find a
function,

{f : S n → M|f (p0 ) = m0 } (4.2)

There is a set of such functions which are homotopically equivalent. We denote this set
by [f ].

Def. 5. Homotopy class: The set of all mappings between S n → M which are homo-
topically equivalent.

Def. 6. Homotopy group: The set of homotopy classes. We denote this group by,
π n (M, m0 ).

The fact that the homotopy group forms a group is not obvious. We will show this
soon. In order to prove such statements in homotopy theory pictorally it is convenient
to first define some useful concepts.

32
4.1. BASICS 33

Sphere at infinity (S n )

p0

cut and open

sphere
−−−−−−−→

x21 + x22 + ... + x2D = R2

boundary is p0 (north pole)

The flattened sphere has 1 less coordinate but an inequality instead of an equality as
its defining equation. In particular we can choose xD to be the distance from the south
pole. Then {x1 , x2 , ..., xD−1 } define the direction in flattened sphere. The D-dimensional
coordinates are given by,

1 xD xi
yi = (1 + ) (4.3)
2 R x

where,

 
x1 q
x̄ =  ...  x≡ x21 + x22 + ... + x2D−1 (4.4)
 
xD−1

To simplify the topology of the mapping, we deform this map to be the interior of a
square in 3D, or more generally, a hypercube:
34 CHAPTER 4. HOMOTOPY THEORY

deform
circle
−−−→

boundary is p0 (north pole)

In particular we use the coordinates,

s
1 r  X
zi = 1 + yi r≡ yi2 , i = 1, 2, ..., D − 1 (4.5)
2 M i

where the zi now denote the distances along each direction in this hypercube. In D = 3
(3 spatial dimensions), its the two directions in a square. These are what we will call the
“standard coordinates”. A given map can be represented pictorally as,

f (z)

With this machinery we can define group product f = f2 · f1 as


f1 (z1 , ..., zD−2 , 2zD−1 ) zD−1 ≤ 1/2
f (z1 , z2 , ..., zD−1 ) = (4.6)
f2 (z1 , ..., zD−2 , 2zD−1 ) zD−1 ≥ 1/2
In words: you take two maps and squash them together such that the product map is
equal to the first map for zD−1 ≤ 1/2 and the second map for zD−1 ≥ 1/2. The rescaling
by a constant of zD−1 shows that the maps are “squashed”. Further, we emphasize that
for each map p0 is always on the boundary and so the product map still has p0 on the
boundary. Pictorally we have,

× =
4.2. IMPORTANT EXAMPLES 35

One can show that we can apply this product to equivalence classes:

[f2 ] · [f1 ] = [f2 · f1 ] (4.7)

i.e., [f2 ] · [f1 ] is independent of choice on f2 and f1 . Furthermore, one can show this
product defines a group, i.e., that it contains an identity element, is associative, and has
an inverse. Furthermore, one can show that n ≥ 2, the group is commutative (abelian).
The group is denoted,
π n (M) (4.8)
where for n = D − 1 for the sphere at infinity.

4.2 Important examples

We now present some relevant examples which we will refer back to throughout these
notes. Consider the mappings from a circle to a 2-sphere order higher dimensional space:

π 1 (S m ) m ≥ 2 (4.9)

This can be written as a mapping:

S1 → Sm (4.10)
Claim 6. The only element in π 1 (S m ) is the unit element.
Proof. Since S m is a larger space then S 1 there must at least one point in S m which is not
mapped to in S 1 . Thus we are free to consider the mapping removing this point in S m .
This mapping is homotopically equivalent to a disk (i.e., it can always be continuously
deformed to a disk) so pictorally:

continuously
deform

circle at ∞
36 CHAPTER 4. HOMOTOPY THEORY

Any circle will always map to some closed path on the disk. This path can always be
continuously deformed to a point. This is easy to show explicitly for a circular image
path, charactorized by an angle θ. We denote the point in the image by parameters, x
and y. In this case we can consider the homotopy,
 
t cos θ + (1 − t)
F(t, θ) = r0 (4.11)
t sin θ

where we have chosen as or base point F(t, 2π) = (r0 , 0). For t = 1 this is just a circular
path and for t = 1 this is just a point in the image and this always has the basepoint,
x(t, 2π) = r0 , y(t, 2π) = 0.
While we only show this explicitly for a circular path its easy to convince yourself
this will hold for a generic path as long x and y are valid coordinates (i.e., there are no
topological defects, e.g., a donut shape where clearly x, y coordinates would not make
sense. Thus this mapping is homotopically equivalent to the mapping to a constant and
hence π 1 (S m ) is the unit element.
More generally we have,
Claim 7. For any m > n: π n (S m ) = ∅
We defer proof of this claim to a later lecture once we’ve developed a bit more ma-
chinery.
Claim 8. π 1 (S 1 ) = Z (the homotopy group of the circle at infinity to a circle is the group
integers under addition).
Proof. Recall that in equation 3.65 we considered this exact mapping:

f: S1 → S1
(4.12)
0 ≤ θ ≤ 2π 0 ≤ σ ≤ 2π

We found the mappings, k1 θ and k2 θ, were homotopically equivalent only if k1 = k2 .

Therefore, we conclude that the homotopy group is composed of an infinite set of homo-
topies charactorized by an integer.
Claim 9. We can generalize the previous relation: π n (S n ) = Z
Proof. We can charactorize the sphere at infinity and the coset sphere by two sets of
angles:

θ = (θ1 , θ2 , ...) (4.13)

σ = (σ1 (θ), σ2 (θ), ...) (4.14)

The volume of the domain is given by,

Z p
Vdom = g(θ)dθ1 dθ2 ...dθn (4.15)
Sn
4.2. IMPORTANT EXAMPLES 37

The volume of the range is,

Zp
Vran = g(σ)dσ1 dσ2 ...dσn (4.16)
Sn
Z p  
dσi
= g(σ) det dθ1 dθ2 ...dθn (4.17)
Sn dθj
where the integral runs over the full range of the sphere, i.e., its an integral including the
potential multiple coverings of the group. The ratio of these volumes will be given by the
“covering number” or the number of times the range covers the domain:
Vran
k≡ (4.18)
Vdom
As we’ve seen the σs form an inequivalent homotopy if they cover the θs a different
number of times. Thus, k is an integer.
As an example of the above consider π 1 (S 1 ) where σ = kθ:
Z 2π
Vdom = dθ = 2π (4.19)
0
Z 2πk Z 2π
Vran = dσ = k dθ = 2πk (4.20)
0 0

Lets now consider the cases where the image is “smaller” than the sphere at infinity.
One can show:
Claim 10. π 2 (S 1 ) = ∅
Proof. The generic mapping from the 2-sphere to the circle with two sections, z > 0 and
z < 0:
σ(x, y) (4.21)
p
where z defined as ± R2 − x2 − y 2 . We show that for each half we can shrink the
half-sphere into a point and so we can shrink the whole sphere. Consider the homotopy,
σ(tx + 1 − t, yt) (4.22)
which has the base point, x = 1, y = 0 with the image base point σ(1, 0). Clearly this is
a continuous deformation keeping the basepoint fixed. We can trivially do the same for
the z < 0 half and so this mapping is equivalent to the constant mapping:


σ  = σ(1, 0) (4.23)
t=0

More generally we have,

Claim 11. π n (S 1 ) = ∅, n > 1
Unlike the lasso’ing a basketball these results are not intuitive. One might expect the
logic we used for π 2 (S 1 ) to transfer over to π 3 (S 2 ) however it does not.
38 CHAPTER 4. HOMOTOPY THEORY

4.3 Exact homotopic sequence

4.3.1 Basics
Computing homotopy groups is a difficult task. A very useful tool is called “exact homo-
topy sequence. As usual our goal to find a mapping,
M: G1 → G2 (4.24)
Before we introduce the sequences, we need some definitions:
Def. 7. ker M : the set of all elements in G1 that are mapped to the identity element of
G2 :
ker(M ) = {g ∈ G1 , M (g1 ) = 1} (4.25)
Def. 8. Im M : the set of all elements in G2 which are mapped to by something in G1 :
Im M = {g2 ∈ G2 | ∃g1 ∈ G1 , M (g1 ) = g2 } (4.26)
Def. 9. Exact: A given sequence,
M1 2 M 3 M
G1 −−→ G2 −−→ G2 −−→ ··· (4.27)
is exact if
Im Mi = ker(Mi+1 ) ∀i (4.28)
Claim 12. There exists a sequence,
M M M M0
π n+1 (G/H) → π n (H) −−→
1
π n (G) −−→
2
π n (G/H) −−→
3
π n−1 (H) −−→
1
··· (4.29)
Proving this claim is feasible but somewhat technical. Instead we take this for granted
and proceed to applications of this theorem.

4.3.2 Examples
Perhaps the simplest application of this is for the sequence:
M 1 2M
∅ −−→ G −−→ ∅ (4.30)
Claim 13. If this is an exact sequence (one where Im M1 = ker M2 ) then ⇒ G = ∅.
Proof. An ∅ can only map into ∅ and so Im M1 = ∅. Furthermore, by the definition of
the kernal of M2 (all elements in G which are mapped to the identity, ker M2 = G). Since
we have an exact sequence we also have, ker M2 = ∅. Therefore, ker M2 = G = ∅
Claim 14. If the sequence,
M1 2 M 3 M
∅ −−→ A −−→ B −−→ ∅ (4.31)
is exact then,
A=B (4.32)
4.3. EXACT HOMOTOPIC SEQUENCE 39

Proof. Again we have,

Im M1 = ∅ = ker M2 (4.33)
ker M3 = B = Im M2 (4.34)

M2 is a map from A → B whose kernal is trivial (only identity of A maps to the identity
of B) and whose image is B (the set of elements of A which map to B are equal to B).
Thus there are no “extra elements” in A and the map is one-to-one and so,

A=B (4.35)

Now lets consider some applications of these ideas into actual groups we will be
interested in.
Claim 15. π 3 (S 2 ) = Z
Proof. Recall that for the breaking of SU(2) → U(1), the vacuum manifold is equal to a
2-sphere: SU(2)/U (1) = S 2 . Thus we can instead study π 3 (SU(2)/U(1)). This is much
simpler since now we can use our exact homotopic sequence. We have,

π 3 (H) → π 3 (G) → π 3 (G/H) → π 2 (H) (4.36)

or π 3 (U(1)) → π 3 (SU(2)) → π 3 (SU(2)/U(1)) → π 2 (U(1)) (4.37)

We know that π 3 (U(1)) = π 3 (S 1 ) = ∅ and similarly, π 2 (U(1)) = ∅. Thus we can use the
trick we just learned to say,

π 3 (SU(2)/U(1)) = π 3 (SU(2)) (4.38)

since SU(2) = S 3 , SU(2)/U(1) = S 2 , and earlier we showed that π 3 (S 3 ) = Z so,

π 3 (S 2 ) = Z (4.39)

Claim 16. π 1 (SO(3)) = Z2

Proof. Consider G = SU(2) and H = Z2 with the sequence,

π 1 (SU(2)) → π 1 (SU(2)/Z2 ) → π 0 (Z2 ) → π 0 (SU(2)) (4.40)

| {z } | {z }
∅ ∅

π 0 (G) measures how many connected components G has. Thus π 0 (Z2 ) = Z2 . Further-
more, π 1 (SU(2)/Z2 ) = π 1 (SO(3)). Therefore,

π 1 (SO(3)) = Z2 (4.41)
40 CHAPTER 4. HOMOTOPY THEORY

Claim 17. π 1 (SO(N)) = Z2 for N ≥ 3.

Proof. We can again use our Higgs intuition to simplify the problem. For a general SO(N )
symmetry broken with a scalar φ, a fundamental of SO(N ),
 
v
 0 
hφi =  ..  (4.42)
 
 . 
0
After φ gains a VEV, the unbroken subgroup is SO(N − 1) given a coset,
G/H = SO(N )/SO(N − 1) = S N −1 (4.43)
Now consider the exact sequence,
π 2 (SO(N )/SO(N − 1)) → π 1 (SO(N − 1)) → π 1 (SO(N )) → π 1 (SO(N )/SO(N − 1))
| {z } | {z }
∅ ∅
(4.44)
where the endpoints are both trivial as long as N ≥ 4. This implies that,
π 1 (SO(N )) = π 1 (SO(N − 1)) (4.45)
For N = 4 we have,
π 1 (SO(4)) = π 1 (SO(3)) = Z2 (4.46)
where we have used our previous result. Applying this relation consecutively we get,
π 1 (SO(N )) = Z2 N ≥3 (4.47)

Claim 18. π 1 (SU(N )) = ∅ for N ≥ 2

Proof. Again we use our knowledge of spontaneous symmetry breaking to help us extract
the homotopy group. For a fundamental scalar charged under at SU(N ) we have,
2N
z }| {
|φ|2 = (φR
1 )2
+ (φI 2
1 ) + (φR 2
2 ) + ... (4.48)
Thus the vacuum manifold is a 2N − 1 sphere:
SU(N )/SU(N − 1) = S 2N −1 (4.49)
Now consider the exact sequence,
π 2 (SU(N )/SU(N − 1)) → π 1 (SU(N − 1)) → π 1 (SU(N )) → π 1 (SU(N )/SU(N − 1))
| {z } | {z }
∅ ∅
(4.50)
where the end points are the trivial group if N ≥ 2. Thus we have,
π 1 (SU(N )) = π 1 (SU(N − 1)) (4.51)
since π 1 (SU(2)) = ∅ we can iteratively show that π 1 (SU(N )) = 0 for all N ≥ 2.
4.3. EXACT HOMOTOPIC SEQUENCE 41

One can similarly show two other important homotopy groups:

Claim 19. π 2 (SU(N )) = ∅

This will be used for studying soliton solutions in 3 + 1 dimensions.

Claim 20. π 3 (SU(N )) = Z

This will be important when we discuss instantons since instantons in D dimensions

are equivalent to solitons in D − 1 dimensions.
Chapter 5

Magnetic Monopoles

5.1 Monopoles in general

As we’ll see in 3 + 1 dimensions soliton solutions of spontaneously broken gauge theories

consist of a tower of monopoles, each with different topological charge. However, before
applying our knowledge of topological field theory, let us first discuss the properties of
monopoles in general.
An electric monopole in electromagnetism is simply a charged particle with charge q
whose E field takes the form,1
r
E=e (5.1)
r3

One can form an analogue of this for a magnetic monopole with magnetic charge, g,

r
B=g (5.2)
r3

5.1.1 Dirac’s derivation

How will such a monopole behave? Dirac’s insight: If we can come up with a find a
magnetic field configuration in EM for which there does not exist a test that would dif-
ferentiate it from a monopole than that system should share properties with a monopole.
Consider an half-infinite solenoid:

1
The prefactor here is of course just conventional, but we attempt to keep gaussian units throughtout
this discussion.

42
5.1. MONOPOLES IN GENERAL 43

..
.

This is called a Dirac string and has B field lines that look just like a monopole. The
only difference between the Dirac string and the monopole is the Dirac string has a flux
along the tube,
r3
Z
Φ = B · da = 4πg 3 = 4πg (5.3)
r

where we used that the flux in the tube should also be equal to the flux emitted by the
end. Is this flux an observable quantity? It can be observed through the Abrahav-Bohm
effect.
Recall that the Schrodinger equation with a magnetic field is given by,
2
−~2

ie
∇ + A(r) ψ(r) + V (r)ψ(r) = Eψ(r) (5.4)
2m ~c

If we consider a change of variables,



ie
ψ(r) = exp − f (r) ψ0 (r) (5.5)
~c

with
Z r
f (r) = A(r0 ) · dr0 ⇒ ∇f (r) = A(r) (5.6)
r0

Then ψ0 (r) satisfies,

~2 2
 
− ∇ +V ψ0 (r) = Eψ0 (r) (5.7)
2m

Therefore, the presence of A gives an additional position dependent phase to ψ. This

can be detected through interference. Consider the following setup:
44 CHAPTER 5. MAGNETIC MONOPOLES

r+

flux
e− a
source
r−

where the flux tube represents the center of the Dirac string, far away from the monopole-
like end. The magnetic flux modifies the interference pattern due to the additional phase.
This phase is independent of path (since A is a conservative vector field), and so the phase
difference of the bottom and top wavefunctions (assuming they start in phase) is,
 I 
1 e
exp (r+ − r− ) + A · dr (5.8)
λ ~c

where L is the distance from the slits to the screen, λ is the wavelength, and
p
r± = L2 + (y ± a/2)2 (5.9)

The first term gives the standard formula for the double slit interference and the second
is the new addition due to the fluxtube. The new term can be written,
I
e e
A · dr = Φ (5.10)
~c ~c

For this term to be unphysical we must have,

e
Φ = 2πn n ∈ Z (5.11)
~c

Using our expression for the flux computed earlier we have,

~c
g= n (5.12)
2e

Thus a Dirac string with quantized charge with quantized charge, shares properties with
a monopole. Thus we conclude that a monopole can’t have arbitrarily charged monopoles
but only quanta of ~c/2e.
5.1. MONOPOLES IN GENERAL 45

5.1.2 Derivation using quantization of angular momentum

Now consider an alternative derivation of the same result using angular momentum.
Consider an electric monopole at the origin with a magnetic monopole a distance r0
away. For simplicity we place the magnetic monopole along the ẑ direction:
g

θ
r0
r

The electromagnetic and magnetic fields of the system are,

r r − r0
E=e B=g (5.13)
r3 |r − r0 |3

giving an angular momentum,

Z Z mom density
1 z }| {
dJ = r × (0 E × B dV ) (5.14)
4πc
r × (r × r0 )
Z Z
ge 2
=− drr d cos dφ (5.15)
4πc r3 |r − r0 |3

The numerator can be simplified using the triple product identity,

numerator = (r · r0 )r − r2 r0 (5.16)
= r2 r0 (cos θr̂ − ẑ) (5.17)

which gives,

(cos θr̂ − ẑ)

Z Z Z
ge 0
J=− dr rr d cos dφ (5.18)
4πc (r2 + r02 − 2rr0 cos θ)3/2

Note that since x̂ ∝ sin φ, ŷ ∝ cos φ:

Z
dφf (r, θ)r̂ = 2πf (r, θ) cos θẑ (5.19)

and so we can write the above as,

Z 1
1 − c2θ
Z
ge 0
J = ẑ dr rr dcθ 2 (5.20)
2c −1 (r + r02 − 2rr0 cθ )3/2
46 CHAPTER 5. MAGNETIC MONOPOLES

The integral is easy done using Mathematica:

1
1 − x2 1
r < r0
Z 
4 r03
dx 2 = (5.21)
−1
02 0
(r + r − 2rr x)3/2 3 1
r3
r > r0

and so,
"Z #
r0 ∞
r0
Z
ge 4 r
J= ẑ dr 02 + (5.22)
2c 3 0 r r0 r2
ge
= ẑ (5.23)
c

Since angular momentum is quantized, with a minimum value of ~/2:

~c
g= n (5.24)
2e

in agreement with the result we derived completely independently above.

Both the derivations above required an electric charge to derive the quantization
condition. We have taken the charge to be electron-like, but we know that we have other
objects with smaller charges then electrons, quarks. If we consider the derivations above
with smaller charges, then the lowest magnetic charge must be larger. However, this
conclusion is a bit too quick. If a particle exists, like quarks, that has additional charges
as well as electromagnetic charges the phase from the Aharonov-Bohm effect (or angular
momentum in the second derivation) can be at least partially cancelled canelled by phases
due to additional interactions. For example we can find the phase in the Aharonov-Bohm
effect using a “down-quark Dirac string“ is,
h  e gs i
exp −i − (5.25)
3~c ~c

where the sign is due to color being attractive. This type of cancellation occurs for GUT
monopoles.

5.2 Dyons
Above we considered particles with only electric or only magnetic charge. Dyons are
(hypothetical) particles that carry both charges. What is there quantization condition?
Consider two such particles with charges,

(q1 , g1 ) (q2 , g 2 ) (5.26)

Now consider particle one going around particle two,

5.3. DUALITY TRANSFORMATION OF MAXWELL’S EQUATIONS 47

The phase of −4πiq1 g2 /~c and 4πiq2 g1 ~c respectively giving a relative phase shift of,
 
i4π
exp − (q1 g2 − q2 g1 ) . (5.27)
~c
This phase shift is unobservable if,
1 n
q1 g2 − q2 g1 = (5.28)
~c 2
This is known as the Schwinger-Dyson quantization condition. For q1 = e, q2 = 0, g1 = 0,
g2 = g we recover the earlier result. However, for dyons electric charges are not all integer
multiples of a fundamental charge. In particle consider two particle with charges,
   
1 θ 1 θ
n1 − m1 , m1 , n2 − m2 , m2 (5.29)
~c 2π ~c 2π
for n1 , n2 , m1 , m2 ∈ Z. The Dirac quantization condition gives,
1
q1 g2 − q2 g1 = (n1 m2 − n2 m1 ) (5.30)
~c
(the θ term cancels) satisfying the quantization condition. Hence there is a continuos set
of charges parameterized by a parameter θ that can exist. The Witten effect [4] shows
that this parameter is equal to the θ angle of the gauge theory.

5.3 Duality transformation of Maxwell’s equations

Maxwell’s equations are given in their differential form as,

∇ · E = 4πρe (5.31)
∇·B=0 (5.32)
1
∇ × E = − ∂t B (5.33)
c
1 4π
∇ × B = ∂t E + je (5.34)
c c
48 CHAPTER 5. MAGNETIC MONOPOLES

where ρe and je are the electric charge and current respectively. Generically there is no
symmetry relating the E and B fields. However, when the sources vanish, ρe , je = 0 the
equations are symmetric under an SO(2) transformation:
    
E cθ s θ E
→ (5.35)
B −sθ cθ B

This is known as the duality transformation of Maxwell’s equations. Now suppose we

had electric and/or magnetic charges. You can always redefine your fields by rotating
away one of the charges using a duality transformation:

(q, g) → (qcθ + gsθ , −qsθ + gcθ ) (5.36)

Thus there is nothing special about magnetism. The statement that “magnetic charges
don’t carry charges” is a statement that we can always define our electric and magnetic
fields such that the “one that has a source” is the electric field and the other one is the
magnetic field.
How would Maxwell’s equation look like with a magnetic source? To see this consider
the simplest electromagnetic duality, α = 90o . In this case,

E → B B → −E (5.37)

Recall that Maxwell’s equation can be written in terms of the field strength tensor,
4π
∂ µ Fµν = je,ν ∂ν F̃ µν = 0 (5.38)
c
where F̃αβ ≡ αβµν F µν . In matrix form,

ET −BT
   
0 0
 0 Bz −By   0 Ez −Ey 
F =  −E −Bz
 F̃ = 
 B −Ez
 (5.39)
0 Bx  0 Ex 
By −Bx 0 Ey −Ex 0

Thus under such a rotation we have, F → −F̃ and F̃ → F . Adding in magnetic charges
we expect,
4π e 4π m
∂ µ Fµν = jν ∂ µ F̃µν = j (5.40)
c c ν
We will show that the topological current in monopole-type soolitons will endeed enter
in jµm .

5.4 Topological Monopoles

Thus far we have studied generic properties of monopoles. We now consider a particle
example.
5.4. TOPOLOGICAL MONOPOLES 49

5.4.1 ’t Hooft-Polyakov monopole

Consider the Georgi-Glashow model of an SO(3) symmetry with a triplet φa and poten-
tial,
U (φ) = λ(φ2 − v 2 )2 (5.41)
The symmetry breaking pattern is,
SU(2) → U(1) (5.42)
Thus we have two broken generators (in the toy SM, these would give mass to the W ±
with mass mW = ev). The relevant homotopy group is,
π 2 (SU(2)/U(1)) = π 2 (S 2 ) = Z (5.43)
Thus we expect a conserved integer charge. What is the conserved current? We won’t
derive the form of the current but instead reveal the answer and then show that it is
indeed conserved. The desired current is,
1
Kµ = µνρσ abc (∂ ν φ̂a )(∂ ρ φ̂b )(∂ σ φ̂c ) (5.44)
8π
where φ̂a ≡ φa /v. This is clearly conserved since,
∂ µ Kµ ∝ µνρσ (∂ µ ∂ ν φ̂a )(...) + ... = 0 (5.45)
Note that this current is a non-Noether current, but of topological origin. The conserved
charge is,
Z
Q = d3 xK0 (5.46)
Z
1  
= abc d3 x φ1 φ̂a ∂ 2 φ̂b ∂ 3 φ̂c − ∂ 2 φ̂a ∂ 1 φ̂b ∂ 3 φ̂c + ... (5.47)
8π
Z
1
= abc ijk d3 x∂ i φ̂a ∂ j φ̂b ∂ k φ̂c (5.48)
8π
Z
1 3 i
h
a j b k c
i (( (
= abc ijk d x∂ φ̂ ∂ φ̂ ∂ φ̂ − ( φ̂a((∂(i ∂(j φ̂(b ( k (
)∂( φ̂c + ... (5.49)
8π
Z
1 h i
= abc ijk d3 x∂ i φ̂a ∂ j φ̂b ∂ k φ̂c (5.50)
8π
To simplify this integral note that,
Z Z
D xi
d x∂i f (x) = dΩ lim f (x) (5.51)
r→∞ r

which can be derived as a particular case of Gausses law by considering a spherical

integration region. Furthermore,
 
1 1 r ∂x r ∂x
dΩ = dθ1 dθ2 ... D−1 det − det + ... (5.52)
(D − 1)! r x1 ∂θ 1 x2 ∂θ 2
 
1 1 r ∂xm1 ∂xm2
= dθ1 dθ2 ... D−1 im1 m2 ... p1 p2 ... ... (5.53)
(D − 1)! r xi ∂θp1 ∂θp2
50 CHAPTER 5. MAGNETIC MONOPOLES

or for D = 3:
1 r ∂xm ∂xn
dΩ = dα1 dα2 imn pq (5.54)
2 xi ∂αp ∂αq
where α1 and α2 are the angles (usually denoted θ and φ). Thus we conclude,
Z
1 ∂xm ∂xn
d3 x∂i f (x) = imn pq d2 α lim f (α) (5.55)
2 r→∞ ∂αp ∂αq
and
∂ φ̂b ∂ φ̂c ∂αw ∂αv ∂xm ∂xn
Z
1
Q= abc ijk imn pq d2 α lim φ̂a (5.56)
16π r→∞ ∂αw ∂αv ∂xj ∂xk ∂αp ∂αq
!
Z b c b c
1 ∂ φ̂ ∂ φ̂ ∂ φ̂ ∂ φ̂
= d2 αabc pq φ̂a − (5.57)
16π ∂αp ∂αq ∂αq ∂αp
∂ φ̂b ∂ φ̂c
Z
1
= d2 αabc pq φ̂a (5.58)
8π ∂αp ∂αq

Now convert the integral to coordinates in the internal φ̂a space, (ξ1 , ξ2 ) which gives,

∂ φ̂b ∂ φ̂c a
Z
1 2 1
Q= d ξ abc rs φ̂ (5.59)
4π 2 ∂ξr ∂ξs
The factor in the center is just the differential of the surface in the internal space and so
we can write, Z
1
Q= dV int ∂a φ̂a (5.60)
4π
where the volume integral now runs over the internal (image) space.
We want to argue that this topological current will play the role of the magnetic
current. For these topological solutions the VEV is not pointing in the smae direction
all the time. This implies that the unbroken U(1) direction is changing. Given this how
should you define Fµν ?
’t Hooft showed that,

EM 1
Fµν = φ̂a Fµν
a
− abc φ̂a Dµ φ̂b Dν φ̂c (5.61)
e
is gauge invariant. Note that if φ̂ = (0 , 0 , 1)T then Dµ φ̂a Dν2 φ̂ = 0 and so Fµν
EM 3
= Fµν .
Far out since Dµ φ̂ contributes to the energy density we have,
EM
Dµ φ̂ → 0 ⇒ Fµν ' φ̂a Fµν
a
(5.62)

With lots of algebra one can show that,

1 ρσ 4π
µνρσ ∂ ν FEM = Kµ (5.63)
2 e
Bibliography

[1] S. Coleman. Aspects of Symmetry: Selected Erice Lectures. Cambridge University

Press, 1988.

[2] R. Rajaraman. Solitons and instantons : an introduction to solitons and instantons

in quantum field theory / R. Rajaraman. North-Holland Pub. Co. ; sole distributors
for the USA and Canada, Elsevier Science Pub. Co Amsterdam ; New York : New
York, N.Y, 1982.

[3] A. I. Vainshtein, V. I. Zakharov, V. A. Novikov, and M. A. Shifman. ABC’s of

Instantons. Sov. Phys. Usp., 25:195, 1982. [Usp. Fiz. Nauk136,553(1982)].

[4] E. Witten. Dyons of charge e/2. Physics Letters B, 86(3):283 – 287, 1979.

51