Topological Defects in Physics

Essay for the “Differential Geometry in Physics” course

Eugen Rožić
Department of Physics, Faculty of Science,
University of Zagreb
July 31, 2019

Abstract

This essay describes the basics of homotopy theory, its relation to group theory and its use in
several branches of modern physics for characterising topological defects. The goal of the essay
is to show that this is important in condensed matter physics, cosmology and elementary particle
physics, branches of physics not usually related, and to provide a succinct and intuitive introduction
to the subject and understanding of the phenomena. At least basic knowledge of (Lie) group theory
is assumed and basic understanding of quantum field theory is necessary to follow examples from
cosmology and elementary particle physics.

Contents 1 Introduction
1 Introduction 1 Whenever we inspect a physical quantity in space (or
spacetime) we first have to determine the set of val-
2 Homotopy theory 2 ues it can acquire at every point. Let’s call this the
2.1 Homotopy (fundamental) groups . . . 3 parameter space 1 R and let’s assume the parameter
2.1.1 Higher homotopy groups . . . . 4 space to be a manifold. Probably the most typical
2.2 Homotopy on group manifolds . . . . . 4 example is that of spins in matter. If we limit spins
to a plane then we can represent them by unit vec-
3 Application in ordered media 5
tors in a plane and their parameter manifold is S 1 , a
3.1 Spins . . . . . . . . . . . . . . . . . . . 6
circle, and if we allow them to point anywhere then
3.2 Nematics . . . . . . . . . . . . . . . . 7
their parameter manifold is S 2 , a sphere.
3.3 Superfluid helium-3 . . . . . . . . . . . 9
In physics one is usually guided by criteria like min-
4 Application in quantum field theory 10 imization of energy and maximization of entropy to
4.1 ’t Hooft - Polyakov monopole . . . . . 12 find stable configurations. But a question arises if
4.2 Cosmic strings . . . . . . . . . . . . . 13 some topological features of the physical and/or pa-
4.3 Kinks and walls . . . . . . . . . . . . . 14 rameter space (dimensionality being the most basic
4.4 Textures and other wild beasts . . . . 15 one) introduce additional criteria that have to be met
by a configuration, possibly preventing it from relax-
5 Conclusion 15 1 In condensed matter physics the physical quantity would

be called an order parameter and the set of values the order pa-
rameter space, but in general it can be something more abstract
like a wave function in which case a more general terminology
such as manifold of internal states would be fitting, but also
in my opinion too clumsy and vague for our purposes.

1
ing to the globally most favored one. {(s1 , . . . , sn ) ∈ I n | at least one si = 0 or 1}. It should
Topological defects are parts of the domain where the also be clear that such an n-loop is essentially an n-
quantity of interest is not necessarily everywhere well sphere, S n .
defined and the configuration is unable to satisfy the The basis of homotopy theory is the “homotopic to”
usual physical criteria (e.g. minimal energy) because relation between two n-loops α and β at x0 which is
to attain it would cost infinite (or the practical equiva- formally defined as an existence of a continuous map
lent) energy. The reason for the infinite cost in energy F : I × I n → X such that:
is that no local “surgery” (rearrangement of the phys-
ical quantity) would suffice but a change would have i) F0 (s) = α(s), ∀s ∈ I n
to propagate infinitely far. The existence and some ii) F1 (s) = β(s), ∀s ∈ I n (2.1)
limitations to the behaviour of such defects, consider- iii) Ft (0) = Ft (1) = x0 , ∀t ∈ I
ing only the topology of the physical and parameter
spaces, is best studied by a purely mathematical the- The crucial thing is that the map be continuous in
ory called homotopy theory. which case it essentially represents a continuous de-
Considering these special, in a way non-optimal, con- formation of α to β, as is depicted in figure 12 . Such
figurations of a physical parameter the questions of a map is called a homotopy.
their creation and dynamics also arise. Some of the
answers to the question of creation can be found in
phase transitions and symmetry breaking, usually re-
lated to very fundamental problems, but for less en-
ergetic and more common ones creation can result
from simpler things like external forces. Questions
of dynamics are mostly answered by non-topological,
physical considerations but some, like defects meet-
ing, coalescing or crossing, are partially answered by
homotopy theory itself, which we now move on to.

Figure 1: (a) representation of the formal definition,

2 Homotopy theory (b) the idea of a continuous deformation

The basic idea of homotopy theory is to use contin-

uous deformation of objects in a topological space The most important thing to notice is that the de-
to establish their equivalence. Topologically different fined relation is an equivalence relation since a loop is
spaces will then have different classes of such equiv- homotopic to itself, the relation is clearly symmetric
alent objects. Since members from different classes and if there exists a homotopy between α and β and
of those objects cannot be deformed into one another β and γ it’s trivial to construct a homotopy between
that can also tell us something about physics if we α and γ. Being an equivalence relation makes “homo-
map a physical quantity to such objects. topic to” separate (n-)loops naturally into equivalence
The simplest kind of a deformable object is a path, or classes of loops that are topologically the same in an
its closed form - a loop. Mathematically, a path on a intuitive sense3 .
topological space X is a map γ : I → X where I = With loops defined in such a way it is easy to define
[0, 1], and a loop (at x0 ) is a path with γ(0) = γ(1) = operations like the product and the inverse, which are
x0 ∈ X, or put simply it’s end points being the same. 2 This figure is taken from [1] where you can find the basics

It is clear that a loop is essentially a circle, S 1 . This of homotopy and homology theory presented a bit more for-
mally, more in depth and with more instructive figures. I also
definition can be easily generalized to n dimensions by recommend [2] for the same purpose.
3 As is later commented, this is really the property of freely
making a mapping γ n : I n → X where I n = I ×. . .×I
homotopic loops and thus conjugacy classes of the equivalence
and identifying all boundary points as the same point, classes in the framework of homotopy groups. This distinction
is, however, only relevant in noncommutative homotopy groups.
i.e. s = x0 , ∀s ∈ ∂I n where x0 ∈ X and ∂I n =

2
needed to construct algebraic structures like groups. should be said that homotopy groups are topological
The product of two (n-)loops is naturally defined by invariants meaning that two homeomorphic topolog-
just “gluing” them together, first going around one ical spaces must have the same homotopy groups or
and then around the other. Formally it is done by put in other words, a homeomorphism preserves the
reparametrisation, going twice as “faster” around each homotopy groups. Also, an important feature of the
one: homotopy groups is that the homotopy group of a di-

1 rect product of topological spaces is the direct sum of
α(2s)
 0≤s≤ their respective homotopy groups:
(α ∗ β)(s) = 2 (2.2)
β(2s − 1) 1
≤s≤1

2 πn (X × Y ) ' πn (X) ⊕ πn (Y ) (2.4)

The inverse of a loop is defined by going the other

To return to physics for a second, the basic idea of
way around it:
the application of homotopy theory to physics is to
describe defects by looking at the values of a physi-
α−1 (s) := α(1 − s) (2.3)
cal parameter on an n-loop surrounding a defect and
ensuring that a product of a loop with its inverse classifying it by the corresponding element of the nth
doesn’t “surround” anything, i.e. is always homotopic homotopy group.
to a constant path (a point). One of the nuances I’ve left out in the preceding “de-
rivation” of homotopy groups is that I’ve been using
2.1 Homotopy (fundamental) groups loops at a point, anchored at some x0 , all the time, but
in the definition of the homotopy groups there is no
The equivalence classes of homotopic loops, using the reference to any particular point. The reason is that
defined product and inverse, form a group structure. there always exists a natural isomorphism between ho-
To prove it formally we need to define a product of motopy groups at any two points of a path-connected
equivalence classes and an inverse of an equivalence manifold. Another nuance is that the isomorphisms
class and then show that the three conditions for a don’t have to be unique and that is related to the fact
group hold: associativity, the existence of a unit ele- that freely homotopic 4 loops are not necessarily homo-
ment and an inverse element that in a product with topic at a point, which is nicely seen in figure 2. The
the original gives the unit element. Formal proofs can
be found in [1] and [2] but intuitively one only needs
to realise that:

a) a product of two loops α and β is homotopic to

all the loops that can be made by a product of
any two loops homotopic to α and β, i.e. [α ∗
β] = [α] ∗ [β], because if you can deform a loop
to α and another to β then you can also deform
their product to α ∗ β , and

b) a product of a loop and its inverse is not itself Figure 2: Two loops f and g that are freely homotopic
but not homotopic at x
a point but it is homotopic to a point, i.e. [α ∗
α−1 ] = [x0 ] .
relation between freely homotopic and based homo-
From these two realisations it should be easy to see
topic loops is relatively simple with 1D loops where it
why the equivalence classes of loops [α] form a group.
is directly related to the commutativity of π1 (X) and
That group is called the fundamental group or the
its conjugacy classes. I will elaborate on this a bit
first homotopy group π1 (X) if we are using regular
4 The definition of freely homotopic being the natural one
1D loops and the nth homotopy group πn (X) if we of continuous deformation without being anchored at any one
are using n-loops. For the sake of completeness it point.

3
because it has to do with the physical consequences the higher homotopy groups, the only difference be-
of topological defects meeting, crossing or interacting ing that the objects being deformed are n-loops, or
in some other way. equivalently n-spheres.
If π1 (X) is commutative then its conjugacy classes are One of the biggest differences is that higher homo-
made up of single elements and loops that are homo- topy groups are always commutative, but that doesn’t
topic are also freely homotopic. In that case topolog- save them (or us) from the complications of the differ-
ical defects like monopoles can combine into a single ence between homotopic and freely homotopic n-loops
topological defect without any ambiguity and depen- and the resulting complications in the interactions of
dence on the way they are combined, and topological topological defects described by those higher homo-
defects like strings upon crossing just pass through topy groups. The reason is that the said difference
each other without entangling. If, however, π1 (X) fundamentally has to do with the uniqueness of the
is noncommutative then its conjugacy classes are not isomorphisms between homotopy groups at different
trivial and loops are freely homotopic if they belong points, and the connection between points is always
in the same conjugacy class, meaning even if they are made by paths and is thus always related to π1 (X).
not homotopic at a point. The combination of topo- More on that can also be found in [3] (sections III. B.
logical defects is then determined by the product of and IX. A.).
conjugacy classes which can give an ambiguous answer Higher homotopy groups are needed because topolog-
to the question of the conjugacy class of the resulting ical defects can be described by homotopy groups of
topological defect and in fact the conjugacy class of loops that can surround those defects. So for exam-
the resulting defect can depend on the path by which ple, a monopole in 3 spatial dimensions has to be de-
the two defects were joined, relative to other defects in scribed by π2 (X) because it can only be surrounded
the medium. The physical consequence of this is that by a 2-sphere.
some defects can in some cases be “undone” by things
like splitting and combining after traversing a certain 2.2 Homotopy on group manifolds
path around some other defects (like in the biaxial
The determination of homotopy groups is made much
nematics example given later). In the case of cross-
simpler if the topological space one is dealing with
ing string defects the noncommutativity of the homo-
is a group manifold, and even more so a simply con-
topy group results in their entangling if the defects
nected group manifold. The reason for that is a series
belong to elements of the homotopy group that do not
of theorems that I will state without proof or much
commute. The physical consequences are that a new
explanation, one or both of which can be found in the
string defect is created connecting the two and, since
already mentioned review by Mermin [3].
a defect carries some energy, prevents the two orig-
The first very useful theorem says that for a sim-
inal defects from separating. Things like this would
ply connected, continuous group G and any subgroup
be very hard to understand and predict without the
H of G the fundamental group of the coset space
machinery of topology and homotopy theory.
π1 (G/H) is isomorphic to the quotient group H/H0 ,
A more detailed exposition of the nuances of isomor-
where H0 is the connected component of H containing
phisms, noncommutativity and free homotopicity I
the identity, or put simply:
mentioned here is given in [3] using a very intuitive
approach and a lot of insightful figures5 .
π1 (G/H) ' H/H0 (2.5)

2.1.1 Higher homotopy groups So the “only” thing one needs to do to find the fun-
Most of what has been presented so far is applica- damental group of a parameter space is to find a sim-
ble not only to the fundamental group but also to ply connected group and its subgroup whose coset is
5 This isomorphic to the parameter space. This sounds com-
review by N. D. Mermin gives a great introduction to
homotopy theory and topological defects in ordered media and pletely abstract and unpractical but is actually com-
much of my understanding of the subject and the contents of
this essay is derived from it. pletely opposite and can often be done using simple
symmetry arguments.

4
The second theorem concerns the second homotopy the universal covering group of the natural symmetry
group and states simply that if π2 (G) = {e} then: group G since it always exists and is simply connected.
An additional convenience is then given by a theorem
π2 (G/H) ' π1 (H) (2.6) stating that for n ≥ 2 the nth homotopy groups of
a group G and its covering group G̃ are isomorphic.
Both stated theorems can be extracted from a more But even if one manages to find the universal covering
general theorem that says that the sequence of homo-
group, which is very often tricky, additional compli-
morphisms:
cations arise from the need to “lift” the isotropy sub-
group H to the appropriate subgroup of the universal
. . . → πn (G) → πn (G/H) →
covering group.
→ πn−1 (H) → πn−1 (G) → . . . (2.7) A lot more formal and mathematical description of
defects and symmetry with advanced usage of group
is a thing called an exact sequence. In such a se-
theory is given in [4]. The following examples will
quence it holds that if in G1 → G2 → G3 → G4
show how symmetry groups, group theory and homo-
the groups G1 and G4 are trivial (have only the iden-
topy theory are used together to predict and classify
tity) then the homomorphism between G2 and G3 is
topological defects and give constraints on their dy-
an isomorphism. Applied to (2.7) it tells us that if
namics.
πn (G) = πn−1 (G) = {e} then:

πn (G/H) ' πn−1 (H) (2.8) 3 Application in ordered media

As I already pointed out, it seems rather abstract and The domain of condensed matter physics and ordered
unpractical to formulate a parameter space as a sim- media is the natural intuitive area to study defects
ply connected group or a coset of groups. But actu- and apply homotopy theory since the order param-
ally it is the most natural thing because a parameter eters are physical quantities one can easily imagine
space is directly related to the reduction of symmetry by vectors, rods and similar familiar objects. We
of the physical parameter in relation to the symme- will mostly be concerned here by continuous media
try of the whole physical space. For example, a plane although media with broken translational symmetry,
has SO(2) as its symmetry group. The isotropy sub- those possessing a lattice, have an abundance of de-
group of SO(2) for a planar spin is just {e} since fects that are of a topological character. Typical of
any other rotation changes the planar spin. Thus, those are dislocations and disclinations (figure 3). The
the parameter space for a planar spin is isomorphic reason why we will not deal mathematically with me-
1
to SO(2)/{e} = SO(2) ' S . This was a trivial ex- dia with broken translational symmetry here is be-
ample. A more complex and persuasive example is cause the mathematics is even more complicated and
the spin in three dimensions. The whole space has unclear. More on that can be found in [3] (section
SO(3) as its symmetry group, and the isotropy sub- VIII.).
group of SO(3) for a spin is SO(2) because all rota- As already mentioned in the introduction, generation
tions about the axis of the spin leave it unchanged. of defects in continuous media can happen in phase
Thus, the parameter space for a spin is isomorphic to transitions but also by simple influence of external
SO(3)/SO(2) which is known to be ' S 2 . forces like bending of a metal bar. Entanglement of
So basically the only practical obstacle for using the line defects and other interactions, mentioned earlier
theorems given in this section for easy calculation with noncommutativity, can in principle have phys-
of homotopy groups are the conditions on G being ical consequences although no materials having that
n-connected (meaning πn (G) = {e}) because they property have been found so far that I am aware of.
are generally hard to satisfy (SO(3) is not even sim- A simple introduction to and overview of the classi-
ply connected). However, the condition for simple fication of materials and topological defects in them,
connectedness is relatively easy to achieve by taking in just a few pages, is given in [5].
Topological defects in materials are usually divided

5
 physical space play the role of the 3-loop from which
a mapping is made to the parameter space R. That
configuration can be non-trivial but it cannot be said
that it possess a defect, more that it is a “defect” by
itself. It is instead said that the configuration has a
certain texture.
The following sections will present typical order pa-
rameters in condensed matter physics and analyze
what kind of topological defects they admit.

3.1 Spins
Perhaps the simplest model of a parameter space is
◦
Figure 3: (a) a −90 disclination, (b) two disclinations given by planar spins, which were already mentioned
(−90◦ and +90◦ ) making a (1,1) dislocation in the far in the introduction. In this model the spin is repre-
configuration, (c) a (1,0) dislocation
sented by a vector confined to a plane and can thus
be represented by a single angle φ, the mapping from
physical space to parameter space being:
into:

a) monopoles - point defects in 3D described by f (~r) = cos (φ(~r)) û + sin (φ(~r)) v̂ .

π2 (R).
It is then obvious that the parameter space is isomor-
b) vortices - point defects in 2D or, equivalently, phic to S 1 and also fairly straightforward to see that
line defects in 3D, both described by π1 (R). π1 (S 1 ) ' Z because you can “wind” a loop around a
circle as many times as you like in any direction and
c) domain walls - plane defects in 3D, but also
loops with different “winding numbers” would not be
lines in 2D or points in 1D. A structure that
possible to continuously deform into each other (fig-
separates two distinct parts of the domain and
ure 4). But what about higher homotopy groups?
is thus described by π0 (R).
To avoid relying on intuition and being able to easily
d) textures - those are solutions that are not triv- answer further questions is the reason to use group
ial, uniform configurations and cannot be trans- manifolds and the associated theorems presented in
formed into one with finite “surgery”, but are section 2.2.
not strictly defects. They are also sometimes In the case of planar spins the natural group manifold
called solitons and are described by π3 (R). G is SO(2) and the isotropy subgroup H is trivial.
The universal covering group G̃ of SO(2) is the 1D
What defines a topological defect is that it has influ-
translation group T (1) which is isomorphic to R but
ence in the medium far away from the defect itself,
then the isotropy subgroup is lifted to:
in other words it can be detected non-locally. This is
plain to see from the use of homotopy theory because
H̃ = {2πk | k ∈ Z}
the size of the n-loop doesn’t matter, only the class it
belongs to, so one can detect or determine a defect in because all real numbers φ + 2πk correspond to the
a configuration by taking an arbitrarily large n-loop, same element of SO(2). Using theorem (2.5) then
even “at infinity”. This is the reason why textures gives:
are formally not defects. They are defined such that
the parameter takes on the same value everywhere π1 (SO(2)) ' π1 (T (1)/H̃) ' H̃ ' Z, (3.1)
3
at infinity, which is why the R space is practically
compactified to S 3 . A configuration as a whole can
then be classified by π3 (R) by making the whole of

6
 one can guess that π2 (S 2 ) = Z from the analogy with
π1 (S 1 ). But let us use the group-theoretic approach.
As was already explained in section 2.2, the G for
ordinary spins is SO(3) and the isotropy subgroup
H is SO(2). The universal covering group of SO(3)
is the group of unitary 2 × 2 matrices with positive
unit determinant - SU (2), which is commonly found
in physics and the reader may know it is a double
cover of SO(3). The appropriate lift of SO(2) is then
simply to U (1) and since U (1) is connected it imme-
diately follows from (2.5) that:

π1 (S 2 ) = π1 (SU (2)/U (1)) = U (1)/U (1) = {e}.

(3.3)
Figure 4: (a) uniform configuration, (b) a configura- Since SU (2) is 2-connected (it can be represented as
tion that can be brought to a uniform one, meaning a ball) it also follows simply from (2.6) that:
winding number 0 and no defects, (c) a defect with
winding number +2 π2 (S 2 ) = π2 (SU (2)/U (1)) = π1 (U (1)) = Z, (3.4)

as expected. We cannot say much about higher ho-

Also the use of (2.8) simply gives the general result: motopy groups so easily because the condition for the
use of (2.8) is not satisfied, that is πn (SU (2)) 6= {e}
πn (SO(2)) = {e} (3.2)
for n > 2. That the use of (2.8) would be erroneous is
seen from the fact that using it would give the result:
for n ≥ 2 since πn−1 (Z) = {e} because it is discrete.
π3 (S 2 ) = π2 (U (1)) = {e} when in fact the correct
So we conclude that planar spins can make only vor-
result is:
tices, that the vortices can be labeled by integers and
π3 (S 2 ) = Z (3.5)
that, since Z is commutative, two or more vortices
combine into a vortex that is labeled simply by the To summarize, ordinary spins cannot make vortices
sum of their labels (figure 5). but can make monopoles whose nature is the same as
those of vortices for planar spins. Additionally, there
also exists the possibility of textures, solitonic configu-
rations that are in themselves not defects but prevent
the material from relaxing to the uniform configura-
tion.

3.2 Nematics
Nematics are liquid crystals in the nematic phase
which is the most common of thermotropic liquid crys-
Figure 5: Two +1 defects (left) joining into a single tal phases. Other phases include smectic and choleste-
+2 defect (right). ric both of which have interesting but more compli-
cated topological features (see table 2 in [4]).
Nematics are basically consisted of rod-shaped mole-
For the also already mentioned case of ordinary 3D
cules that like to align. Their interesting topological
spins one can also get some results by intuition. For
features come from the fact that the orientation of the
example, since the parameter space is S 2 , a sphere,
molecules doesn’t matter, which enables us to repre-
it is obvious that π1 (S 2 ) = {e} because everybody
sent them by headless vectors called directors. The
knows that you cannot “lasso a sphere”. Similarly

7
“only” thing different from ordinary spins is that the
isotropy subgroup includes rotations by π around axes
perpendicular to the director axis. This makes the
parameter space SO(3)/D∞ ' RP 2 which is the real
projective plane and corresponds to a 2-sphere with
opposite points identified. This matches the intuitive
definition of a director as a headless vector.
In order to use SU (2) instead of SO(3) we need to
find the corresponding lift of D∞ . It is made of two
Figure 6: A vortex in a nematic
sets of elements connected by a π rotation around an
axis perpendicular to the director axis, for example
g = u(x̂, π): then R ' SO(3)/D2 but to apply the theorems from
section 2.2 we need to use SU (2) and the equivalent
u(ẑ, θ) = H0 ' SO(2)
lift of D2 which happens to be the quaternion group
u(x̂, π)u(ẑ, θ) = gH0 Q (more details and intuition in [3]). The quaternion
group is a discrete group consisted of 8 elements: the
unit element, its negative, 3 “imaginary” units i1 , i2
This results in: and i3 , and their negatives. The imaginary units are
connected by the totally antisymmetric Levi-Civita
π1 (SU (2)/H) ' H/H0 ' Z2 = {0, 1}. (3.6)
tensor via ia ib = abc ic . The result is that, by the
theorems of section 2.2, the first and second homo-
which is a known result for π1 (RP 2 ). The conse-
topy groups for biaxial nematics are:
quence of this is that nematics can form vortices, but
only of one kind (figure 6) and whenever two of them
π1 (R) = π1 (SU (2)/Q) = Q (3.7)
meet they annihilate.
Since S 2 is the covering group of RP 2 their higher ho- π2 (R) = π1 (Q) = {e} (3.8)
motopy groups are the same according to a theorem
mentioned at the end of section 2.2. But this doesn’t This tells us that monopoles are unable to form, unlike
mean that the conclusions about monopoles and tex- in normal nematics, and that vortices are determined
tures in ordinary spins can be completely copied to by a very special and unexpected group.
nematics. The reason for caution is stated in section What is especially interesting is that Q is obviously
2.1.1 and can intuitively be related with the isotropy noncommutative so the 8 elements have to be grouped
subgroup H having more than one connected compo- into conjugacy classes to give topologically distinct
nent. So even though the higher homotopy groups of vortices. There are five of those and they define four
H are determined completely by H0 , the law of com- different kinds of defects, those that are 2π rotations
bination of different topological defects is not. One of and those that are π rotations around the three dis-
the consequences for nematics is that monopoles with tinct axes. The law for the combination of those de-
labels +n and −n are topologically the same which fects is very peculiar and gives the possibility of a
leads to the unusual possibility of a monopole with an defect catalyzing the destruction of another defect.
even label 2n separating into two equal n monopoles Again, more details can be found in [3] but the point
and combining back into a defectless configuration. to be taken is that quite unusual topological defects
appear in real life and it would be very hard, if not im-
Another interesting material is the biaxial nematic possible, to understand them, let alone predict their
which aligns by two axes. It can be imagined as elon- existence and behaviour, without the use of homotopy
gated rectangles which are unchanged only by π ro- theory.
tations around the three perpendicular axes, i.e. the
four-element point group D2 . It’s parameter space is

8
3.3 Superfluid helium-3 give:

Although condensed matter can be very complicated

π1 (R) = π1 (SU (2)/{e, −e}) = Z2 (3.10)
and exhibit some very peculiar properties, the dis-
π2 (R) = π1 ({e, −e}) = {e} (3.11)
covery and subsequent theoretical explanation (BCS
theory etc.) of superfluidity and superconductivity
This seems like the poorest homotopy structure so
made it especially hard, probably impossible, for con-
far; it allows only one kind of vortices, like the ne-
densed matter physicists to use their intuition and
matics, and no monopoles. However, using the facts
previous experience in predicting and explaining their
that SO(3) ' RP 3 , that S 3 is the covering space of
behaviour. Both phenomena are described in the fra-
RP 3 and that higher homotopy groups of a space and
mework of quantum field theory but in a very differ-
its covering space are the same we can conclude that:
ent way and context then the examples of the next
section. The consequence is that the order parame- π3 (R) ' π3 (S 3 ) = Z (3.12)
ters are unlike anything used so far, completely un-
intuitive and mathematically abstract. Two of them which means that the dipole-locked A phase of helium-
will be presented here but without any physical mo- 3 admits an infinite variety of textures. Examples of
tivation or explanation because the theories are too such textures are the Anderson-Toulouse and Mermin-
complicated. Ho vortices as well as the Shankar monopole (see [1],
Superfluid helium is basically the trigger that brought section 4.10).
topologists and homotopy theory to condensed matter
physics because condensed matter physicists couldn’t The dipole-free A phase has an order parameter
cope with it using conventional methods they used that doesn’t seem a lot more complicated but results
and developed until then. Two isotopes of helium in a much richer and complicated topology. The order
are superfluid, 3 He and 4 He, but the more interest- parameter is defined as:
ing and complicated one is helium-3 which is theo-
A(~r) = n̂(~r)e(~r) (3.13)
retically very similar to superconductors since both
effects occur with fermions. Helium-3 has two super-
where e is the same as in (3.9) and n̂ is an uncoupled
fluid phases, the A and B phase, and each has various
unit vector.
regimes that are described by different order parame-
The natural first choice for the parameter space would
ters. We will deal here with the dipole-locked and the
be just the direct product of the parameter spaces for
dipole-free A phase.
each of the separate components, which is:
The simpler of the two is the dipole-locked A phase (SO(3)/SO(2))×SO(3), or equivalently SO(3)×SO(3)
whose order parameter is a complex vector field: with the isotropy subgroup SO(2) × 1. But care must
be taken of the fact that if both components simulta-
e(~r) = φˆ1 (~r) + ıφˆ2 (~r) (3.9) neously switch signs it amounts to no change at all.
So alongside (R(ẑ, θ), 1) transformations of the form
with the constraints: |e|2 = 1 and e · e = 0. (R(û, π), R(ẑ, π)) = (R(x̂, π)R(ẑ, θ), R(ẑ, π)) also be-
It can be imagined as a pair of orthonormal axes long to the isotropy subgroup. To understand the
(φˆ1 , φˆ2 ) at each point of space and it is then easy elements of the isotropy group one can imagine the
to see that its parameter space R can be described SO(3)×SO(3) space as two coordinate systems where
by the full group manifold of SO(3) since no element the first is aligned with the ẑ-axis pointing in the n̂
of SO(3) but the identity leaves it unchanged. As direction and the second aligned such that the ẑ-axis
usual, we use SU (2) instead of SO(3). The lift makes is perpendicular to the (φˆ1 , φˆ2 ) plane. The direction
the isotropy group become non trivial since the iden- of û is perpendicular to the ẑ-axis and makes an angle
tity in SO(3) corresponds to both the identity and its θ with the x̂-axis.
negative in SU (2). The theorems of section 2.2 then To use the theorems of section 2.2 we need to use
SU (2) × SU (2) as the group manifold G and lift the

9
isotropy subgroup accordingly. This results in four Lagrangian density:
sets of isotropy transformations that can be related
λ
2
to an element g = (u(x̂, π), u(ẑ, π)) as left cosets: L = ∇µ φ∇µ φ∗ − V = ∇µ φ∇µ φ∗ − |φ|2 − v 2 .
4
(4.1)
(u(ẑ, θ), 1) = H0 ' SO(2) The potential is shown in figure 7 6
and the vacuum
2
(u(ẑ, θ), −1) = g H0 manifold is the circle of points for which |φ|2 = v 2 .
(u(x̂, π)u(ẑ, θ), u(ẑ, π)) = gH0 One can view this as a broken symmetry, the same as
in the ordered media case. For example, in the case
(u(x̂, π)u(ẑ, θ), −u(ẑ, π)) = g 3 H0
of the field described by (4.1) the full internal sym-
Using theorems (2.5) and (2.6) then leads us to the metry group is obviously U (1) and the isotropy sub-
following results: group for a particular vacuum state is just the identity
since all other “rotations” change the vacuum state.
π1 (G/H) ' H/H0 ' Z4 (3.14) Thus, as before, the vacuum manifold is U (1)/1 and
it’s topological properties are the same as for planar
π2 (G/H) ' π1 (H) ' π1 (SO(2)) ' Z. (3.15)
spins. The quantum field can, of course, have more
This leads to the conclusion that there can be three
different types of vortices in the dipole-free A phase
of helium-3 and that they are nicely behaved, as far
as topology is concerned. But, as with nematics, one
should not be quick to draw conclusions on the be-
haviour of monopoles and textures since the isotropy
subgroup contains more than one connected compo-
nent. In fact, monopoles have a similar property to
the ones in nematics (more on that can be found in
[3]).

Figure 7: A “Mexican hat” potential for a complex

4 Application in quantum field quantum field resulting in a manifold of vacua
theory
internal degrees of freedom with complicated symme-
In quantum field theory the quantum field φ is de-
try groups and potentials that break them into highly
scribed by a Lagrangian density from which the field
non-trivial vacuum manifolds. This opens the possi-
dynamics equations and the energy of the field can be
bility for complicated topological defects not readily
derived. The Lagrangian always contains a potential
accessible by intuition and of possible great impor-
V (φ) that usually includes the mass and interaction
tance for elementary particle physics and cosmology.
terms. The field always tends to the minimum of that
Homotopy theory tells us only about the possibili-
potential in every point of physical space and when
ties of topological defects, there are other conditions
the field is in that minimum we say that it is in the
that affect their creation or even the possibility of
vacuum state.
their existence. In fact, energy considerations tell us
The interesting thing that can happen is that there
that the existence of topological defects is impossible
can be more than one available vacuum, meaning that
for theories with Lagrangians of the form like (4.1)
there exists more than one configuration of the field
in more than one spatial dimension (Derrick’s theo-
with the same minimum of the potential and no pref-
rem). What “saves” our standard quantum field the-
erence for the field to be in any particular one. In that
6 This figure is taken from [6] which is an introduction to
case we have a “vacuum manifold” and the vacuum
and an overview of QFT I highly recommend. It contains even
state can be viewed as an order parameter. One sim- short chapters on topology and its applications to QFT like
anyons, Chern-Simons theory, ’t Hooft-Polyakov monopole and
ple case is the single complex field with the following instantons.

10
ories from Derrick’s theorem is local gauge invariance ically depicted in figure 8.
instead of global symmetries. The shift to local gauge
symmetry groups changes nothing as far as homotopy
is concerned which means that our standard quan-
tum field theories admit topological defects, at least
in principle. Another question is the creation and
energy content of eventual topological defects in the
universe and this is were cosmology comes into play.
The theory of the hot big bang says that the universe
was once very hot and dense and that it expanded and
cooled ever since. It also says that when it was very Figure 8: The Kibble mechanism leading to a topolog-
hot that the internal symmetries of the quantum field ically uniform configuration (left) and a vortex defect
(right) [8]
were not broken and that as it cooled the universe
underwent a second-order phase transition that broke
the internal symmetries of the quantum field. For ex- The standard model of physics (SM) is a quantum
ample, the standard Landau theory would expand the field theory with a quantum field that has two com-
potential of the (4.1) Lagrangian with a temperature plex components and (SU (3)×)SU (2) × U (1) as its
dependent part CT 2 |φ|2 + . . . [7] making it: internal gauge symmetry group. The theory says that
the field goes through spontaneous symmetry break-
λv 2
 
2 λ
V (φ) = CT − |φ|2 + |φ|4 + const. ing that breaks the SUL (2) × UY (1) gauge symmetry
2 4
of the electroweak force to the Uem (1) gauge sym-
λv 2 metry of the electromagnetic force through the Higgs
This would give a critical temperature of Tc2 = 2C
over which there would exist only a single vacuum at mechanism giving masses to the W ± , Z 0 and H 0
φ = 0 and under which bosons. The resulting vacuum manifold cannot be
 there2 would exist a vacuum
manifold for |φ|2 = v 2 1 − TT 2 → v 2 . This approach calculated as easily as (SU (2) × U (1))/U (1) ' SU (2)
c
can be easily transferred to quantum fields with a lot because it depends on the embedding of Uem (1) into
more complicated internal structure and gauge sym- the higher symmetry space, which you may know is
metries, such as those of the standard model. not trivial (Weinberg angle etc.). However, it can be
The way that topological defects would occur in case shown [9] that the resulting vacuum manifold indeed
of such a scenario is described by the Kibble mech- is SU (2) in the theory of the Higgs SU (2) doublet
anism. At the point of the phase transition of the field. Since the group manifold of SU (2) is isomor-
universe causal effects between different areas of the phic to S 3 it follows that the homotopy groups for
universe are limited by the speed of light and the age the vacuum manifold M of SM are:
of the universe. The length scale is that of the in-
verse Hubble constant so we can roughly put that π1 (M) ' {e} (4.2)
lcor ∼ H −1 . This means that areas further apart than π2 (M) ' {e} (4.3)
lcor couldn’t have had causal effect on each other and π3 (M) ' Z (4.4)
have no reason to have the same vacuum, since all
vacua are equally likely. This means that the sponta- This means that the SM doesn’t admit cosmic strings
neous symmetry breaking resulted in different vacua or monopoles. However, it is postulated that there
around the universe that, due to energy considera- exists a more fundamental Grand Unification The-
tions, settled into the uniform configuration locally ory (GUT) with a higher gauge symmetry group (for
and expanded. But since all vacua are equally likely example SU (5)) that breaks down to the SM one
and uncorrelated it should have happened that some in a similar way to which the electroweak symmetry
topologically non-trivial configurations occurred that breaks to the electromagnetic. Most of those GUT’s
can’t be settled to the uniform one. This is schemat- admit monopoles as stable configurations, which is
appealing and desired for some reasons but also gen-

11
erates some additional problems. guess, it is just:

4.1 ’t Hooft - Polyakov monopole φ = v (1 − f (r)) r̂

The simplest gauge quantum field theory that admits where f (r → ∞) = 0 and (1 − f (r → 0)) ∝ r. The in-
monopoles is the Georgi-Glashow model with a quan- teresting thing about this solution is that it links spa-
tum field having three real components and a local tial and internal dimensions, because r̂ “lives” in the
SU (2) gauge symmetry. Its Lagrangian density is physical space and φ “lives” in the three-dimensional
given by [6][8][10]: internal space and those two spaces are generally un-
related.
1 λ 2
2 1
L= (Dµ φa )(Dµ φa ) − φ − v 2 − Gaµν Gµν
a . The bigger problem and the reason why finding this
2 4 4
solution was not trivial is finding the gauge fields that
The covariant derivative Dµ and the gauge field tensor keep this solution finite in energy. They correspond
Gaµν are given by the three gauge fields Wµa and the to:
coupling constant q through the following relations:
W0a = 0
Dµ = ∂µ + qabc Wµa ,
1 − a(r) aij j
Wia =  r̂
Gaµν = ∂µ Wνa − ∂ν Wµa + qabc Wµb Wνc . qr

The vacuum state is clearly achieved for φ2 = v 2 where a(r → ∞) = 0 and (1 − a(r → 0)) ∝ r2 .
which, considering that φ has three real components, In the case of a uniform vacuum chosen in (4.5) one
2
makes the vacuum manifold M ' S . This can also can check to see that the electromagnetic tensor is
be seen by breaking the symmetry choosing arbitrar- given by the third component of the field tensor and
ily the vacuum corresponding to: that the others vanish: Fµν = G3µν . It is then easy
  to generalise this to any uniform orientation of the
0 vacuum state since the physics shouldn’t depend on
φ= 0 (4.5)
 
 the choice of the vacuum:
v + h(x)
φa
Fµν = · Gaµν .
Following the calculations through shows that gauge |φ|
field combinations Wµ± = √1 (W 1 ±ıW 2 )
2 µ µ acquire mass Using this equation with the fields of the monopole
mW = qv, that the Higgs field h acquires mass mh = configuration gives:
√
2λv and that the gauge field Wµ3 remains massless
and gives the theory a U (1) gauge symmetry that can 1 ijk k
Fij = −  r̂
qr3
be identified with electromagnetism by Wµ3 ≡ Aµ .
This means that the vacuum manifold should be M ' with all other elements equal to 0. This means that
SO(3)/U (1) ' S 2 which is consistent with our first there is no electric field, only a magnetic field which
2
conclusion. As we already know, π2 (S ) = Z so this is equal to:
theory admits monopoles, an infinite variety of them ~ = 1 r̂ .
B (4.6)
even. q r2
The simplest is the ’t Hooft-Polyakov monopole which This is clearly a field that a magnetic monopole would
belongs to the class with winding number 1 and is produce, completely analogous to the field of an elec-
also called “the hedgehog solution” since the phase tric monopole such as an electron. If we would extend
of the vacuum points in the radial direction at each the Maxwell equations with a magnetic monopole,
point of space far from the monopole. The form of making the second equation:
the quantum field for such a solution is not hard to
∇B = [µ0 ]ρm ,

then for a monopole with magnetic charge qm the field

12
would be: the main reasons why the inflation mechanism, which
~ = [µ0 ]qm r̂ .
B (4.7) seems to provide a way around it for some GUTs at
4π r2
least, was invented.
Comparing (4.7) with the monopole solution (4.6) we
see that the magnetic monopole charge of the ’t Hooft-
4.2 Cosmic strings
Polyakov monopole (ignoring µ0 and 0 ) is:
The simplest model having a vacuum manifold with
4π 2h
qm = → (4.8) π1 (M) 6= 0 after symmetry breaking and so allowing
q q
1D topological defects in 3D space is the abelian Higgs
where the second result is after restoring appropriate model. It is basically the Lagrangian given by (4.1)
factors of ~. One should keep in mind that q is here extended with a U (1) local gauge field. The simplest
the coupling constant of the electromagnetic gauge such topological defect in that model is an infinite
theory which is the quantum of electric charge. straight string. It can be described by the following
The reason why this result is very interesting, and configurations of the quantum and gauge fields:
also why one would like a GUT to have monopole
solutions like this, is a classical result by Dirac that φ = v (1 − f (ρ)) eiN θ (4.10)
says that if we extend Maxwell’s equations with mag- N
Ai = − (1 − a(ρ)) ij rj ; i, j ∈ {1, 2} (4.11)
netic monopoles the magnetic charge must obey the qρ2
relation:
h where ρ is the distance from the string, the z-axis is
qm = n . (4.9)
q pointing in the direction of the string and the f and
We see that the ’t Hooft-Polyakov monopole indeed a functions have appropriate asymptotic behaviour:
obeys that relation, with n = 2, which is a stunning
f (ρ → ∞) = 0 , f (ρ → 0)) = 1 ,
fact considering how vastly different the theories the
two results were obtained with are. The derivation of a(ρ → ∞) = 0 , a(ρ → 0)) = 1 .
the condition in (4.9) and more about the connection
In the center of the string the quantum field is ob-
of the Dirac and ’t Hooft-Polyakov monopoles can be
viously not in a vacuum state which means that the
found in section 10.4 of [10].
string carries energy. For the simplest abelian Higgs
What the classical result doesn’t say is anything else
model used here the estimated tension (energy per
about the hypothetical magnetic monopole, it’s just
unit length) of the string is µ ≈ πv 2 and its thickness
added to the Maxwell’s equations by hand and all its
is ρ ≈ 1/qv which is of the order of 1 fm. The size and
other physical properties, like its mass, are completely
energy of such strings make them very interesting to
arbitrary. The ’t Hooft-Polyakov monopole, on the
cosmology because of their potential implications.
other hand, has all of its other properties determined
An interesting property of the (4.10, 4.11) field con-
by the theory it results from. For example, in the sim-
figurations is that the string core carries a magnetic
plest model presented here the mass of the monopole
flux and that flux is quantized:
is estimated at mM ≈ mW /α ≈ 137mW ≈ 11 TeV.
This is around 7 orders of magnitude heavier than the
Z I
~ 2π
BdS = Aθ dθ = N .
electron. In GUTs, however, the monopoles usually q
have masses of the order of 1017 GeV, which is of the
This same property is found in vortices in type II
order of micrograms, and are formed in a phase tran-
superconductors that “leak” magnetic flux when the
sition that occurs at a critical temperature of roughly
external magnetic field exceeds a critical value. Their
1016 GeV. This results in “the monopole problem”
mathematical origin is similar and in superconductors
which states that standard GUTs lead via the Kibble
they are called Abrikosov vortices while in cosmology
mechanism to an overabundance of monopoles, both
they are called Abrikosov-Nielsen-Olesen strings, or
in number and in energy density, which is completely
simply cosmic strings.
inconsistent with current observational knowledge of
Cosmic strings can be open, horizon-sized objects or
our universe [8]. This “monopole problem” is one of

13
finite-sized, closed loops and both varieties are pre- 4.3 Kinks and walls
dicted to be generated at a phase transition according
The breaking of symmetry can, in principle, also lead
to the Kibble mechanism, irrespective of the details
to a disconnected vacuum manifold, a manifold with
of the theory that supports them.
π0 (M) 6= 0. Such a theory would allow for the exis-
Straight strings wouldn’t gravitate but would affect
tence of topological defects called domain walls which
the geometry of spacetime considerably. The geome-
would present a boundary between two parts of space
try around them would be locally flat but with a so-
with different vacua. The most simple example of
called deficit angle ∆θ ≈ 8πGµ that would make the
such a theory is a single scalar field φ with the follow-
full circle around the string be 2π −∆θ. The effects of
ing Lagrangian:
that would be the creation of double images and non-
Gaussian distortions of the CMB. The observations of 1 λ 2
2
L= (∂µ φ)2 − v − φ2 (4.12)
the CMB thus put a bound on the tension of horizon- 2 4
sized cosmic strings at µ ≤ (0.7·1016 GeV)2 . Another which obviously has two possible vacua: φ(x) = ±v.
effect would be the compression of dust in wakes of If we consider only one spatial dimension a domain
moving cosmic strings which could have perhaps gen- wall is usually called a kink, the reason for which is
erated seeds for structure formation. Some observed clear when looking at figure 9. A static kink solution
1D structures in the universe could be explained in can be easily constructed by any function with the
that way but are inconsistent with some others. Gen- desired asymptotic behaviour, for example:
erally, the hypothesis cannot be easily dismissed.
x 2
Horizon-sized strings needn’t be straight and those φ(x) = v tanh ; l2 = . (4.13)
l λv 2
with “ripples” and singularities like kinks and cusps
would gravitate and dissipate their energy through Since the field is not everywhere in a vacuum state a
gravitational waves which could also be detected. Ad- kink carries energy, but that energy is localized in a
ditionally, such irregular strings could generate smaller region of width l. The energy is easily calculated and
closed strings by self-intersection.
Closed strings, if they existed, would gravitate and
dissipate their energy through gravitational waves.
As it dissipates energy a closed string would get ever
smaller and when its radius would become as small as
its thickness it would look like a high-energy particle.
Such a particle-like topological defect would, in the
simple abelian Higgs theory, have a mass similar to
the monopole but would, unlike a monopole, have a
completely contained magnetic flux and would inter- Figure 9: A “kink” - a domain wall in one dimension
act only gravitationally making it a possible WIMP.
Although the standard model doesn’t allow cosmic is finite. The given solution can be trivially extended
strings various GUTs might and different varieties to two and three spatial dimensions which would pre-
of cosmic strings, but with very similar properties, sume a straight line or a flat plane as the boundary
are allowed in various string and superstring theories. between domains. This, of course, doesn’t have to be
This, coupled with the numerous possible fundamen- the case but if the wall is not strongly curved it is a
tal consequences the existence of cosmic strings would good approximation.
have for our universe, makes them a very interesting The cosmological consequences of horizon-sized do-
topic of study. More details about the various cos- main walls would be considerable and certain condi-
mological consequences of cosmic strings, along with tions can be constructed from existing observations.
some calculations, can be found in [8]. Simple energy density considerations already give an
unlikely satisfiable boundary of surface energy density
η ≤ (10GeV)3 given the usual symmetry breaking en-

14
ergy scales for standard model extensions. An addi- ory went through multiple symmetry breaking phase
tional likely effect of a domain wall would be that it transitions where different kinds of allowed defects
anti-gravitates which would have a big effect on CMB. were produced at each transition. Examples of such
The CMB observations thus put an even tighter upper exotic topological structures are fleece which are cos-
bound on the surface energy density of a horizon-sized mic strings that end on domain walls, and necklaces
domain wall. These results are the reason why the which are cosmic strings containing monopoles [8].
topological possibility of domain walls is undesired in
theories that extend the standard model.
5 Conclusion
4.4 Textures and other wild beasts Physicists dealing with condensed matter have long
ago realised the possibility of defects in materials and
The same as in ordered media, textures are possible,
managed to describe them and deal with them with-
defectless configurations of the vacuum if the vacuum
out using formal mathematics from topology and ho-
manifold has a non-trivial third homotopy group. In
motopy theory. Modern physics has, however, given
quantum field theories of the kind we discussed so far
us theories that are both mathematically and in the
they are also created by the Kibble mechanism dur-
space, time and energy scales they describe far from
ing a symmetry breaking phase transition and they
the possible grasp of our intuition. It is that circum-
are energetically unstable. In broken gauge theories
stance that brought topology and homotopy theory to
textures relax by emitting Nambu-Goldstone bosons
condensed matter physics, but they have since proven
and they decay too fast to still be present. Their
indispensable for the accurate and systematic expla-
effect could be indirectly seen through density per-
nation and prediction of the existence and the dy-
turbations in the early universe reflected in the CMB
namics of topological defects in even classical ordered
but so far observations are inconsistent with textures
media.
being a relevant factor [8].
We have also seen that quantum field theory often
Another kind of a topological defect that depends on gives topologically rich vacuum manifolds. Homotopy
the third homotopy group is an instanton. Instan- theory has here also proven indispensable and has re-
tons are solutions that are localised both in space vealed the possibility of a variety of topological defects
and time and they are studied on the four-dimensional with interesting properties existing in our universe.
Euclidean space (after Wick-rotating the time dimen- The systematic exploration of those possibilities us-
sion) which is why the third homotopy group is rele- ing methods from topology can tell us what to look
vant and the vacuum manifold can facilitate configu- for to confirm the theories we have but also what we
rations with defects. should expect of the theories that are candidates for
Instantons are too complicated, both mathematically their successors.
and physically, to go into any more depth here. They Homotopy theory and topology in general are obvi-
are mentioned only to point out to the reader that ously a necessary tool for modern physics if all impli-
they are possible because of non-trivial topology of cations of the mathematically complicated physical
the vacuum manifold making them also topological theories are to be understood and especially if those
defects. A very elementary introduction to instan- theories are to be meaningfully applied to the totality
tons with some intuitive explanations and elementary of our universe.
calculations can be found in the last chapter of [6] and
a bit more mathematically involved introduction can
References
be found in section 10.5 of [10].
[1] Nakahara M., Geometry, Topology and
The last kind of topological defects that I will men-
Physics, 2003, IOP Publishing
tion in this essay are the so-called hybrid topolog-
ical defects. They are combinations of defects de- [2] Geroch R., Mathematical Physics, 1985, Uni-
scribed so far and are hypothetically possible if a the- versity of Chicago Press

15
[3] Mermin N. D., The topological theory of de-
fects in ordered media, 1979, Reviews of Mod-
ern Physics Vol. 51 No. 3, American Physical So-
ciety, p. 591

[4] Michel L., Symmetry defects and broken

symmetry. Configurations Hidden Symme-
try, 1980, Reviews of Modern Physics Vol. 52 No.
3, American Physical Society, p. 617

[5] Sethna J. P., Order Parameters, Broken

Symmetry, and Topology, 1991 Lectures in
Complex Systems, Proc. Vol. XV, Addison Wes-
ley, p. 243

[6] Blundell S. J, Lancaster T., Quantum Field

Theory for the Gifted Amateur, 2014, Ox-
ford University Press

[7] Carroll S. M., Trodden M., TASI Lectures:

Introduction to Cosmology, 2004, arXiv:
astro-ph/0401547

[8] Gorbunov D. S., Rubakov V. A., Introduction

to the Theory of the Eraly Universe, 2011,
World Scientific Publishing

[9] Gripaios B., Randal-Williams O., Topology of

the Electroweak Vacua, 2016, arXiv:1610.
05623

[10] Ryder L. H., Quantum Field Theory, 1996,

Cambridge University Press

16